Implementation of an SfM-MVS-based photogrammetry approach for detailed 3D reconstruction of plants

Implementation of an SfM-MVS-based photogrammetry approach for detailed 3D reconstruction of plants | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Implementation of an SfM-MVS-based photogrammetry approach for detailed 3D reconstruction of plants Jiří Mach, Zdeněk Svatý, Ondřej Šoupa, Luboš Nouzovský, Martin Halecký This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7178236/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 09 Oct, 2025 Read the published version in Plant Methods → Version 1 posted 12 You are reading this latest preprint version Abstract In recent years, non-destructive and non-invasive methods for 3D plant reconstruction have gained importance in plant phenotyping. Morphological traits reflect a plant’s physiological status and serve as key indicators for precision agriculture, crop protection, and food quality assessment. Accurate and efficient 3D modelling enables objective, repeatable monitoring of plant development and health, supporting data-driven decision-making in agricultural and food research. This study presents a cost-effective and flexible photogrammetric methodology for analysing plant morphological traits under controlled laboratory conditions. The system includes an industrial RGB camera mounted on a robotic arm, a rotating platform with an adjustable plant holder, and stable illumination. The key steps involved camera calibration, exposure optimisation, fine-tuning of evaluation algorithm parameters (tweaks), setting the optimal camera-to-object distance, and reducing computational load for 3D model evaluation. Comparative testing revealed that the most effective calibration strategy integrated simultaneous calibration, pre-calibrated parameters, and adaptive fitting, ensuring high reconstruction accuracy and consistent model quality. The optimal acquisition parameters were a 50 milliseconds exposure time, a tweak value of 0.9, and a 16 cm camera-to-object distance. Using more camera positions with fewer frames per position proved more efficient than the reverse. The optimal configuration consisted of three height levels with 40 frames each. Automation and data reduction led to a 75% decrease in processing time, reducing the scan time from 8 minutes to 2.7 minutes per plant. The developed method proved to be a reliable, reproducible, and affordable tool for routine 3D analysis of plant morphology via close-range photogrammetry. Close-range photogrammetry SfM-MVS-based data processing 3D reconstruction plant phenotyping morphological traits precision agriculture Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Figure 9 Figure 10 Figure 11 Figure 12 Figure 13 Figure 14 Background Recent advances in computational power, combined with the widespread availability of digital cameras, have driven a significant paradigm shift in photogrammetric techniques. This progress has enabled the development of advanced methods such as structure from motion (SfM) and multi-view stereo (MVS), which have been increasingly applied in precision agriculture, particularly for plant phenotyping. These techniques provide non-invasive and non-destructive means to accurately capture a broad range of plant traits through detailed 3D point cloud reconstruction. Moreover, they offer promising potential for the future automation of phenotyping workflows [1]. The methodology of precise plant phenotyping incorporates not only the analysis of plant vitality and prosperity (vegetation indices) but also the characteristics of the plant architecture, namely volume, area, biomass density along the main axis of the plant, leaf inclination, and configuration [2-4]. In contemporary scientific disciplines such as plant and plant pathology, a non-destructive approach to the study of morphological features has emerged as the prevailing standard [5-7]. The use of such a method is widely embraced within the scientific community, signifying its status as a contemporary paradigm in the scientific study of plants. The key indicators enumerated above are typically pivotal in the study of the interaction between a plant and an external factor, which may be an abiotic or biotic agent. As the global average temperature increases, pressure is placed on plant breeders to develop cultivars that are resistant to drought and other adverse soil conditions [8, 9]. These conditions can cause stress to plants, which can have a negative impact on their development and prosperity. High salt concentrations in the soil have also been demonstrated to be deleterious to proper metabolic processes within the plant body, as well as to the development of the plant body in general [10]. Conversely, factors that have been demonstrated or have the potential to exert a favourable influence on plant growth are worthy of mention. These biopreparations are either in conventional use or are undergoing development and laboratory testing. They are composed of cellular or cell-free mixtures of microorganisms and/or their biologically active metabolites. The function of biostimulants is to promote plant growth, either directly or indirectly. These preparations have the capacity to increase the bioavailability of crucial soil elements, such as phosphorus [11-13]. It is evident that microbial producers of auxin phytohormones also belong to this category of biopreparations, which have been demonstrated to exert a positive influence on plant metabolism [14]. The extensive list of biopreparations includes the widely employed Polyversum ® , which is derived from the mycoparasitic oomycete Pythium oligandrum and is effective in the elimination of fungal diseases [15]. The development of potential biopreparations is frequently conducted within a laboratory setting, with subsequent testing being conducted directly on economically significant crops, including but not limited to barley, wheat, maize, rice, and soybean [16]. All the abiotic and biotic factors mentioned above, which exert a negative or positive influence on plants, have one thing in common, namely, the initiation of a morphological response of the plant body by their presence. In the following section, the various technological approaches currently employed or with potential for application are comprehensively delineated. The focus will be on non-destructive methods for investigating the effects of a factor on the morphological characteristics of the affected plant. In accordance with the principle of spatial data acquisition, available 3D reconstruction technologies can be classified into two distinct groups: active and passive systems. The former rely on the emission of active radiation and the subsequent measurement of its reflection from the object surface. A terrestrial, mobile or airborne scanner emits a laser beam for the purpose of distance sensing [17-22]. Time-of-flight depth cameras utilize structured light to determine the distance to an object's surface by measuring the time delay between light emission and its reflection back to the sensor [23, 24]. High-precision X-ray tomography is a reliable method for scanning detailed objects; however, it is relatively expensive [3, 25]. The second group encompasses passive methods that do not rely on light emission. Instead, these methods utilize the ambient electromagnetic radiation that is naturally emitted by the sun or another artificial light source. Photogrammetry is a process that uses images captured from multiple angles; however, it is susceptible to variations in lighting conditions and the texture of the object in question [26]. In the domain of photogrammetric analysis, two principal strategies are employed for the acquisition of images. The first method is based on static-object moving-camera systems, where the object remains stationary while the imaging device moves around it [27, 28]. Second, static-camera moving-object systems are considered, wherein the camera remains stationary while the object is rotated or translated in front of it [29]. The utilization of both configurations in close-range photogrammetry is contingent upon the specific experimental setup, the measured object properties and the spatial constraints inherent to the imaging environment. The combination of the aforementioned methods is indeed feasible. In such cases, the camera is typically affixed to the movable component [30, 31]. The control software deploys the arm in a series of predetermined positions to achieve the desired image configuration. At each position, the camera remains for a predefined period, during which a rotating object is sensed. The process subsequently transitions to the subsequent position, and this sequence is repeated. The acquired images are processed via the structure-from-motion (SfM) and multi-view stereo (MVS) algorithms [32, 33]. SfM is a process that facilitates the recovery of both camera poses and a sparse 3D point cloud. This is achieved by detecting and matching key features on the object surface (keypoints) across overlapping images. In the context of close-range photogrammetry, particularly in the field of plant science, the process of photo alignment is frequently used. It is often facilitated by the strategic placement of an object within the scene that exhibits a distinctive and unique multi-coloured texture. Examples of such objects include a multi-coloured cube, ball or plate [34]. Matched features, termed tie points, are utilized in a subsequent bundle adjustment procedure, which optimise both interior camera parameters and exterior orientations by minimising reprojection error. This stage, known as photo alignment, results in a geometrically consistent configuration of the image set and an initial sparse reconstruction of the scene. Following successful alignment, MVS algorithms are employed to densify the point cloud, thereby computing dense pixel-wise correspondences between images to generate a detailed and accurate surface model. This workflow is a common feature of commercial photogrammetric software, such as Pix4Dmapper, Metashape, ContexCapture or RealityCapture, which automates the majority of these steps while allowing for fine user control over parameters influencing reconstruction quality [35, 36]. When small-scale, highly detailed biological samples such as horticultural plants are targeted, it is imperative to employ high-resolution image acquisition techniques. This involves employing diffuse and uniform lighting to minimise shadows and reflections, applying background separation techniques such as chroma keying or the use of plain, textureless surfaces, and carefully calibrating both interior and exterior camera parameters to reduce distortion and enhance spatial accuracy. At present, a considerable number of research institutes are engaged in the study of close-range photogrammetry as applied to plant science. The focus of this field of study is the analysis of the aerial parts of plants. The objective is to generate the most detailed model, from which leaf position, orientation, and distribution can be accurately determined [37]. The leaf area index (LAI) is defined as the ground area covered by the plant canopy projected onto the soil surface. It serves as a critical qualitative and quantitative descriptor of morphological traits [38]. LAI plays a pivotal role in the optimisation of (bio)preparation dosages for plant treatments [39, 40]. Furthermore, the architecture of the plant root system can be investigated in detail [41-43]. To ensure the effective and consistent implementation of photogrammetric scanning systems, it is imperative to meticulously calibrate and set a number of parameters. These include the determination of the interior and exterior orientations of the cameras, the prevailing lighting conditions, the distance at which the scanning is conducted, and the evaluation pipeline employed for the generation of 3D models [1, 44-46]. For instance, the Perceptron v5 laser scanner is frequently employed as a reference device to evaluate the reliability, accuracy, and usability of photogrammetric systems [33, 47]. The primary aim of previously mentioned studies is to expand the portfolio of precise and cost-effective photogrammetric platforms suitable for high-throughput plant phenotyping applications. This study further contributes to this goal by developing and evaluating an additional low-cost photogrammetric platform. The objective of the present study is to extend the functionality of the existing multifunctional scanning apparatus [48]. The original methodology of plant reconstruction, which relies on the use of a 3D laser scanner, is to be superseded by a more appropriate technique (for this experimental setup). The latter will be subject to an SfM-MVS-based photogrammetric analysis. This will be carried out via an industrial RGB camera mounted on a robotic arm, a turntable, and additional LED illumination. The subsequent analysis of the acquired images will be undertaken with the objective of generating the most accurate 3D model. The subject of interest will be the calibration of the camera and the optimisation and appropriate adjustment of the scanning parameters (e.g., exposure time and scanning distance). The primary evaluation criterion will be the quality and detail of the generated model for each of the three plant species, as indicated by the total area and volume of the digital biomass. The selected statistical indicators of the quality of the spatial reconstruction are subsequently assessed. The apparatus should ideally be capable of producing an image of the plant that allows the creation of a comprehensive spatial model while preserving significant detail. These features will be of particular importance in future research, where the focus will be on a non-destructive way to determine the positive effect of a biocontrol agent or the negative effect of abiotic and biotic stressors on plant growth. Consequently, the accuracy and reliability of the scanning process and back reconstruction will assume a pivotal role in this ensuing phase of research. Methods The subsequent chapter addresses the enhancement of the functionality of the multifunctional scanning apparatus that was previously developed and described. This enhancement is achieved through the application of the photogrammetric method, which allows for improved data acquisition and more detailed 3D reconstruction [48]. Specifically, it provides a comprehensive description of the necessary components, including a detailed description of the object to be scanned, the optimisation procedure within the photogrammetric analysis and computational complexity, and the method of evaluation and interpretation of the resulting data. Photogrammetric system setup Apparatus components The functionality of the multifunctional scanning apparatus was further enhanced by integrating the capability to conduct semi-automatic photogrammetric reconstruction of scanned samples, thereby broadening its range of applications. The enhancement was attributed to the addition of auxiliary illumination (see Figure 1 A and B). The lighting system consisted of two overhead 200 W LED reflectors and two supplementary 30 W side LED LEVE reflectors, providing a combined light intensity of approximately 500 μmol/m²/s across the scene. With this intensity value, the surface of the object was evenly illuminated. Diffused and sufficient illumination was crucial for ensuring correct and accurate photogrammetric reconstruction. It also plays a key role in achieving a high-quality 3D model that preserves morphological details and structural features. The homogenization of object illumination was ensured by a photographic diffusion panel placed above the plant, with emphasis on minimising sharp shadows and uneven light exposure across the scene. Additionally, two undiffused 30 W LED panels provided the sharp side lighting necessary for accurate QR code scanning of the cuvette. The second significant component employed in the process was a MER2-1220-32U3C camera, which was equipped with an LCM-12MP-06MM-F2.4-1.7-ND1 lens (Daheng Imaging, China). The camera generated 24-bit color (RGB) images in JPEG format, which were subsequently used as input for photogrammetric processing to reconstruct 3D models. The technical specifications of the imaging system, including both the camera and lens, are summarized in Table 1. The camera was mounted on an AR4 robotic arm (Annin Robotics, USA) with 6 degrees of freedom to ensure precise positioning during imaging. To enable semi-automatic scanning, the scanned object was always placed in a custom-made holder attached to the turntable. Both of these components were printed from ABS filaments via the standard fused deposition modelling method with an i3 MK3S 3D printer (Original Prusa, Czech Republic). The turntable was driven by an NEMA17 stepper motor with a TMC2130 driver, and the control was mediated by an ArduinoUNO platform with an ATmega328 microcontroller The surface of the object holder and the top of the turntable were covered with a custom-made, randomly coloured splatter texture, forming a unique pattern specifically designed for this setup (see Figure 1 – C). This improved the strength and quality of the alignment of the captured images during the 3D model generation procedure. In addition, sixteen control points (12-bit coded circular targets) were placed on the surface of the turntable, allowing control of the spatial reconstruction accuracy and definition of the scale and coordinate system of the resulting digital model. The original system [48], priced at 8,790 EUR (excluding the 3D scanner and dual-axis turntable), was upgraded by adding a camera and lens costing 388 EUR, bringing the total hardware cost to 9,178 EUR. However, this amount does not include the software license for Agisoft Metashape, which is priced at 540 EUR for the Professional Educational version (3,460 EUR for the full Professional (non-educational) edition). Table 1. List of industrial RGB camera and lens technical parameters (taken from the manufacturer). Camera MER2-1220-32U3C parameters Description or value Communication interface USB3 Resolution 4024 x 3036 (12.2 MPx) Frame rate 32 fps Pixel size 1.85 μm Sensor 1/1.7'' CMOS Pixel bit depth 8bit, 12bit Pixel data format BayerRG8, BayerRG12 Weight 65 g Dimensions 29 x 29 x 29 mm Price* 252 EUR Lens LCM-12MP-06MM-F2.4-1.7-ND1 parameters Lens mount C-mount Optical resolution 12 Mpx Image format 1/1.7'' Focal length 6 mm IR corrected None Aperture (min) F2.4 Iris Manual Working distance 100 mm - Infinity Lens dimensions 29.8 x 37.72 mm Distortion <0.05% Price* 136 EUR * The rounded prices include VAT and correspond to the exchange rate of 18 July 2025 (24.625 CZK per 1 EUR, source: CNB in Prague). Scanned objects The following model plants were selected for optimisation and verification of the reliability of the developed photogrammetric approach: Cucumis sativus L. (cucumber), Solanum lycopersicum L. (tomato), and Lactuca sativa L. var. capitata L. (lettuce). Henceforth, only Latin plant nomenclature will be utilized throughout the remainder of this text. The selection of these plants was driven by their distinctive morphological characteristics, namely their small, thin, and flat features, which serve as a suitable model for the development of an accurate and reliable photogrammetric method. Notably, all three of these plants are significant agricultural and food crops. The plants were cultivated under constant temperature conditions of 25 °C and a light intensity of 160 µmol/m 2 /s, characterized by a long photoperiod of 16 hours of light and 8 hours of darkness. A horticultural substrate was employed, which was placed in 30 ml PP Sterilin cuvettes with a hole in the cap. The use of the cuvette as a cultivation vessel was crucial, as the holder on the rotary turntable was specifically designed to accommodate it, thereby facilitating the handling of the plants. A further advantage of this configuration was the ease with which the camera could access the underside of the rosette leaves, thus facilitating the acquisition of the data necessary to generate a high-quality, full 3D model. Finally, it also enables the utilization of an automatic boundary definition for spatial reconstruction, lowering the computational and time demands. Scanning was conducted on three-week-old plants. Spatial configuration of the scanning process The scanning process was conducted via a static-camera moving-object system, with the camera sequentially positioned at six distinct height levels. The procedure was characterized by the use of discrete, robotically controlled viewpoints, enabling autonomous acquisition of multiple images. The movement of the robotic arm, which is predefined in the basic settings, ensures that the predefined image configuration is achieved (height, distance from the object, angle of the camera) at each of the six locations. During the image acquisition process, the camera was positioned at predefined height levels and held in place long enough to capture up to 60 images at each level (see Figure 2). Notably, the camera did not initiate capture during movement between positions. The scanning process was automatically initiated by a predefined position of the turntable with the object attached to the initial position. This was facilitated by the Hall sensor, which detected the presence of a magnet on the rotating top of the turntable. During imaging, the rotation speed was continuous, with an angular rotation speed of 6°/s (~ 0.105 rad/s). The speed was set to minimise any movement of the scanned object. Acquisition and data processing The acquisition and processing of data was predominantly automated through the utilization of Python-based scripts. Prior to the initiation of the photogrammetric analysis, a Photogram3D program was executed to check the contents of the target folder to which the data were subsequently directed. The evaluation process was initiated automatically upon reaching a predefined number of images in the specified directory. The following section provides a more thorough description of these steps. The image acquisition, incorporating the control of the robotic arm movement, was also facilitated by Python scripts, while the motion of the turntable was managed by a script in Wiring. The timing of camera triggering, robotic arm movement and turntable rotation speed was automated by a wrapper control script. The initial step of the data acquisition process involved automatically scanning the QR code affixed to the surface of the cuvette once the object was correctly positioned and clearly visible to the camera. The code contains the identification data, i.e., the name of the sample and its abbreviation. According to the decoded information, a folder was created and named. The naming process incorporated a date, which was automatically generated by the script. The images were then stored in the designated folder. After a scan of a single object, the folder contained a maximum of 360 images in the dataset. The described process is illustrated in a simplified form in Figure 3. The complete scanning process was completed within a maximum timeframe of 8 minutes, and as part of the optimisation process, an attempt was made to reduce this time. To assess the reliability of the results, each object was subjected to three scanning cycles. The 3D model was generated via a custom-based application called Plant3D, which was developed as part of this study and features its own graphical user interface and console output. Plant3D was launched by a supervisory program (Photogram3D) at the beginning of the analysis to perform photogrammetric processing. The application utilized the imported Metashape module (Agisoft Metashape API) for fully automatic spatial reconstruction [49]. The initial step of data processing involved alignment of the images through the structure-from-motion (SfM) algorithm, with adaptive fitting enabled and automatic filtration of the stationary points. As a part of the alignment phase, the parameters of the internal orientation were also determined. The subsequent step entailed the detection of 16 circular 12-bit control points and the automatic assignment of the corresponding coordinates. This step ensures the definition of the coordinate system and scale of the digital workspace and provides information about the resulting accuracy of the spatial reconstruction. Owing to the size of the samples, processing was performed in millimeters. Following the creation of a sparse point cloud consisting of the detected tie points, a predefined bounding box was automatically set, limiting the reconstruction area for further processing (scanned plant and upper part of the cuvette). This cropping was implemented to reduce the computational time, given that the focus of the study was not on the turntable model with a cuvette. The generation of depth maps was performed via the multi-view stereo (MVS) algorithm, which builds upon previously computed camera positions from SfM. This step was performed with the highest quality settings and medium-level filtering. Additionally, according to the developers' recommendations for the reconstruction of thin-walled structures, customized tweaks ("ooc_surface_blow_up", "ooc_surface_blow_off") were utilized [50]. The selection of a designated filtering mode results in the delicate elimination of extraneous points from a dense point cloud, while ensuring the preservation of the object's finer details and structures. The actual generation of the plant spatial model, represented as a mesh, was performed through depth masks of the images. Post-processing of the generated model involved automatic partial surface smoothing (strength level 1), a combination of manual and partially automated removal of isolated patches from the surroundings, and the closure of holes, e.g., at the bottom of the plant stem. The generation of a report containing all descriptive parameters, e.g., the qualitative and statistical indicators, and the processing and model generation times, was initiated before the termination of the script. Upon completion of the entire evaluation process, a notification was dispatched. This approach resulted in substantial acceleration and streamlining of the evaluation process for many scanned objects. For the entirety of the optimisation procedures, only models without texture and colour information were generated, as the emphasis was exclusively on their morphological characteristics. Consequently, no requirement was placed on the calibration of colour hue or surface reflectance. The HP Z2 Workstation desktop computer (CPU: 13th Gen Intel Core i9, GPU: NVIDIA RTX A4000 16 GB, RAM: 32 GB) was utilized to process the image data and generate a 3D model of the object, owing to the substantial computational demands of photogrammetric reconstruction. All the data were stored and backed up on a Samsung SSD 870 QVO 8TB, with a secondary backup stored in the OneDrive cloud. In this work, all 3D models were visualised in a shaded, artificially coloured form. Photogrammetric reconstruction optimisation Camera calibration Since a non-metric camera with initially unknown interior orientation parameters was used for image acquisition, particular attention was given to assessing its impact on the precision of image alignment during processing. To mitigate potential instability in camera parameters over time, the effectiveness of pre-calibration functionality available in Agisoft Metashape was evaluated. Four calibration strategies were compared. The first used fixed interior orientation parameters obtained through pre-calibration, hereafter referred to as “Fixed”. The remaining three strategies involved simultaneous calibration during image alignment, with varying degrees of prior information. The second approach used pre-calibrated parameters as starting values but allowed adjustment during processing (“Simul.”). The third approach provided only basic physical characteristics of the camera-pixel size and focal length without performing pre-calibration (“No calib.”). The fourth strategy excluded all a priori information (“No info.”). Pre-calibration was conducted in Agisoft Metashape using 18 images of a black-and-white checkerboard pattern, generated by the software and placed at the same position as the scanned object. Importantly, the entire image frame was filled by the pattern and uniformly illuminated to ensure optimal results. The comparative evaluation was carried out on three samples for each plant species. The alignment quality was consistently set to the highest level, with no limit on the number of detected keypoints. Additionally, processing was conducted both with and without the adaptive camera model fitting (ACMF) option, a feature especially recommended for use with uncalibrated cameras, wide-angle lenses, or scenarios involving changes in camera orientation during image acquisition. Coded targets placed on a turntable were used for the assessment, with eight designated as control points and eight as check points. Exposition and tweak setting As part of the optimisation of the imaging parameters, the following exposure times were chosen: 30, 40, 50, 60, and 70 milliseconds. For this and all subsequent experiments, the depth of field was set so that the scanned object was sharp in all images at all 6 different height levels. Scanning thin, small, and detailed objects is generally associated with the problem of producing non-compact, leaky models [50]. This problem can be solved by setting up advanced features called "tweaks". On the basis of the recommendations of the developers of Agisoft Metashape software, advanced settings were applied during the "Build Model" phase, where the model was generated from depth maps. These parameters operate by expanding or contracting the surface contours, analogous to the morphological operations of dilation and erosion known from image processing, but were applied within a three-dimensional context. To find the best setting for each plant species, the following values were tested and evaluated: 0.95, 0.9, 0.7, 0.5, and 0.3. Object and camera position The scanned objects displayed both fine structural detail and morphological diversity, features characteristic of each plant species examined. Consequently, a series of measurements were carried out using varying object-to-camera distances to determine the optimal imaging distance between the camera and plant for each species. The selected distances for this purpose were 12, 14, 16, and 18 centimetres. Each distance was set for all six height levels (P1-P6). The lowest distance was predicted to provide the best detail capture, whereas the highest distance was expected to facilitate the alignment of images during model generation. The lowest distance could not be further reduced, as the minimum focusing distance of the lens is approximately 10 centimetres (plus a margin), according to the technical specifications. The highest distance was determined on the basis of the limitations of the workspace, size of the objects to be scanned and camera field-of-view. Finally, a combined configuration of 12 and 16 centimetres was also defined, with the shorter distance applied at positions P1, P3, and P5 and the longer distance at positions P2, P4, and P6. Optimisation of computational and scanning efficiency The processing of a substantial quantity of the acquired images was both time-consuming and computationally intensive. Consequently, it was imperative to reduce the number of images while preserving the integrity of the 3D model generated. To this end, a reduction protocol was defined. The protocol's fundamental components are delineated in Table 2. The symbol X on the label denotes the plant species. The green square indicates the images taken from a given height level (P1-P6). The marking of all positions (X_2) with a degree of reduction of -1/2 represents the fraction of the images used, specifically a situation in which every second image in each marked position was omitted. In other words, the total number of images (360) for a given object is reduced by a factor of two to 180. Notably, a similar procedure was followed in the remaining cases. Table 2. Summary of image counts aimed at reducing scanning and processing time in 3D model generation. Note: "*" ~ Mark to distinguish duplicate values Evaluation of the suitability of parameter settings The suitability of the chosen settings for the studied parameters defining the photogrammetric reconstruction was evaluated on the basis of several key criteria. The first set of criteria pertained to the qualitative characteristics of the generated mesh model, with a particular emphasis on surface integrity and volume accuracy. The second criterion involved the analysis of statistical indicators. The quality of photo alignment was assessed by qualitative indicators, including the number of tie points (TP), the root-mean-square reprojection error in pixels (RMS RE), the maximum reprojection error in millimetres (Max RE), and the average tie point multiplicity (AVG TP M). The evaluation metrics also included the differences between the known and estimated positions of the control and check points, namely the mean absolute error in millimetres (MAE), the root-mean-square error in millimetres (RMSE), and the standard deviation of the alignment error. In the subsequent stages of the work, all coded markers were treated as control points. Accordingly, the RMSE was referred to as the control point error (CPE) and was used as an indicator of the overall spatial precision reflecting the mean discrepancy between the measured and true coordinates of the 16 circular reference control points that defined the scale of the digital workspace. The quality and detail of the generated 3D model were assessed in terms of the total number of faces and vertices. Table 3 provides a detailed description of the individual statistical indicators. The statistical analysis was also accompanied by a manual image analysis of the model and served as an important concluding criterion in the overall evaluation. This analysis focused primarily on the presence of relics or irregularities on the model surface, imperfections within the model morphology, or the presence of holes even after adjustments by tweak settings had been made. Table 3. Key indicators used in photogrammetry to assess 3D model quality and related processing steps. Workflow stage Indicator* Description Photo alignment TP The key points are automatically detected and matched between the overlapping images to work out the relative orientation. A higher number usually makes alignment more accurate. RMS RE, RMSE It shows the average distance (error) between the reprojected tie point and its observed position in the image. Lower values are better. Max RE The largest observed error in reprojecting tie points across all images. The identification of the worst-case alignment error AVG TP M The mean number of images in which each tie point is visible. Higher values indicate greater redundancy and, by extension, more robust alignment. Coordinate system accuracy CPE The error between the known and calculated coordinates of the control points indicates an error in the georeferencing of the model. This is indicative of the absolute accuracy of the model in real-world coordinates. Model reconstruction quality Faces The total number of polygonal faces (typically triangles) in the 3D mesh corresponds to a specific level of intricacy. Vertices The individual points in 3D space that define the shape of the mesh by forming the corners of faces (typically triangles or polygons) are known as vertices. An increased number of vertices is indicative of greater geometric complexity and detail. Surface The term employed to denote the level of detail, resolution, and noise level of the reconstructed 3D surface. A high-quality surface is characterized by its intricate detail and absence of artefacts. Volume The purpose of this indicator is to provide a quantitative measure of the accuracy of volumetric calculations derived from the model. The accuracy of the results is contingent not only on the correct geometry and scale but also, on the advanced settings of the 3D model generation procedure. * All parameters were computed via Agisoft Metashape and were part of the automatically generated processing report. The reliability and usability of the photogrammetric system were compared with those of equipment that included a previously tested POP 3 3D scanner [48]. To this end, an image analysis of the generated models was conducted. This step was important for evaluating the potential to expand the usability of the multifunctional equipment under development, with the aim of generating high-quality 3D models of plants and enabling non-destructive studies of their morphological traits. Results In the following chapter, the results are clearly presented, visualised and interpreted primarily through graphs and illustrative figures. This chapter offers an overview of the most suitable parameter configurations for optimised photogrammetric analysis, with particular emphasis on the qualitative attributes of the generated 3D model of the scanned object. Camera calibration The alignment quality achieved through different camera calibration strategies is summarized in Figure 4. The results confirmed the anticipated influence of unknown and potentially unstable interior orientation parameters on the quality of spatial reconstruction. The highest accuracy, as measured by residuals at the control and check points, was obtained via simultaneous calibration accompanied by information about the interior parameters derived from pre-calibration. However, the most consistent reconstruction performance across repeated trials was observed with the simultaneous calibration approach using pre-calibrated values as initial estimates, underscoring the robustness of this second strategy. Although the specific effects of the adaptive camera model fitting (ACMF) option are not discussed in depth, the tests indicated a pronounced positive impact when sufficient prior information was available ("Simul.") or when no prior information was provided ("No info."). In contrast, the strategy using only physical camera parameters, focal length and pixel size ("No calib."), resulted in notably reduced accuracy, and in the case of L. sativa , it led to partial image alignment failure. Despite the variability across the scenarios, adaptive fitting generally enhanced both the accuracy and geometric consistency of the reconstructions. As such, its use was considered advantageous and was applied in all subsequent processing steps. Ultimately, the simultaneous calibration approach incorporating pre-calibrated parameters of interior orientation and ACMF was selected for continued use in the photogrammetric workflow. Exposure time and tweak setting The resulting averaged descriptive values of the qualitative features, namely volume and surface area, of the generated 3D models were visualised for each plant species in Figure 5. The visualisation encompasses a full spectrum of predefined tweak settings and corresponding exposure times. The selected optimal settings are highlighted in red (see bold font and dashed lines). To enhance clarity, the volume and surface values have been separated due to their different order levels. Tweaks are not enabled by default in Metashape Agisoft. When scanning thin structures, holes and imperfections frequently occur in the generated models. The resulting data confirmed the positive impact of virtually any tweak setting on model quality. An increase in tweak values was generally associated with an improvement in the quality of the resulting models (see Figure 5). In cases where no tweaks were applied ("No Tweaks"), the negative impact was most pronounced on the surfaces of the C. sativus and S. lycopersicum plant models. Conversely, the volume values remained relatively unaltered by this configuration. The surface and volume of the L. sativa models were found to be relatively resilient to the impact of this setting. When the adjustments were assigned to predefined values, the resulting trends presented marked similarity across all the plant species. An increase in tweak values led to a noticeable reduction in surface holes in the generated models. This improvement was largely attributed to the effectiveness of the applied tweaks, such as dilation and erosion, which successfully compensated for missing data by filling structural gaps. A slight thickening of some plant structures, particularly the leaves, was observed as a side effect of this process, leading to a modest overestimation of the total volume. Nonetheless, the advantages gained through this optimisation step outweighed the minor limitations. In the case of the volume values representing L. sativa , a slightly decreasing trend was observed with increasing exposure time for all adjustments. This could be related to the lighting and resulting lower image quality. The exposure time within the selected test range, was not a significant indicator of the qualitative features of the generated model on the basis of the resulting data. For the subsequent steps of the optimisation process, the median value, specifically 50 milliseconds, was selected. At this value, the images were neither underexposed nor overexposed. The ambiguity of the effect of the tweak settings on the quality of the models precluded the selection of the most suitable option at this stage. Detailed analysis of the generated model images served as an important complementary evaluation of the correctness of both the tweak settings and the exposure time. This image-based assessment helped confirm the validity of the preliminary parameter configuration. As supporting evidence, a set of representative images was produced and is shown in Figure 6. A value of 0.9 was selected as the most effective tweak setting. At this value, no holes were present in the models, making the model complete and compact. A retrospective analysis of the data trends (see Figure 5) revealed that the presence of holes in the models had a negligible effect on the total volume and surface area. In summary, the segment of the model comprising holes was found to be inconsequential in relation to the total mass of the model. This final statement indicates the robustness and relative reliability of the introduced photogrammetric approach in relation to the quality of the generated model, independent of the test plant species. Object and camera position Analysis of the surface data and related trends for the C. sativus and L. sativa models indicated that the object-to-camera distance did not significantly affect the results (see Figure 7). However, in the case of S. lycopersicum , a certain dependency was observed. It is evident from the data that those lower distances, specifically 12, 14, and the combination of 12 and 16 centimetres, tend to overestimate the surface values. From a distance that was too close, even small, thin trichomes on the stem surface were detected, but during the reconstruction of the model, unrealistic, imperfect artefacts were generated, which contributed to a slight increase in the total surface area. With respect to the resulting volume values, no specific trend was observed for any plant species. Consequently, the distance between the object and the camera did not exert a substantial influence on this particular model quality indicator. To determine the optimal distance, supplementary image analysis of the models was conducted (see Figure 8). To provide a clearer context, the images highlight the previously mentioned imperfections identified in the generated models when suboptimal scanning distances were used. The results revealed that the RMS RE, Max RE, AVG TP M, and CPE values exhibited considerable consistency across all the plant species and demonstrated no discernible alterations in response to variations in distance from the object to the camera (see Figure 9). A statistical indicator was employed as an additional complementary measure to assess the influence of distance on model quality, further supporting these findings. The remaining indicators, including TP, faces, and vertices, supported the conclusions from the preceding assessment. Scanning at distances closer than 16 cm led to the generation of artefacts, which consequently caused a slight overestimation of these model quality metrics. Conversely, increasing the distance to 18 centimetres resulted in decreased counts of faces and vertices. This effect can be attributed to the reduced level of detail captured when scanning from distances exceeding 16 centimetres. The analysis of statistical indicators proved to be highly beneficial, providing a valuable complement to the comprehensive assessment of this optimisation step. On the basis of a holistic evaluation and the aforementioned assessment approaches, the optimum distance between the camera and the object for this particular experimental configuration was determined to be 16 centimetres. For scanning larger plants, this optimisation step would need to be repeated, as the optimal scanning distance is likely to increase. Optimisation of computational and scanning efficiency The subsequent trend in the processed data demonstrated a positive correlation between the number of frames and the time required to process the data and generate a compact 3D model. In other words, an increase in the number of frames was associated with an increase in the time required (see Figure 10). The indicators of surface and volume were found to be largely unaffected by the decrease in the number of images, thereby indicating the significant robustness of the photogrammetric approach and the evaluation process. A substantial fluctuation in the trend was observed in the case of "120*". The observed fluctuation can be attributed to the failure to establish sufficient tie points between the two image sets, due to significant differences in camera positions and viewing angles. Consequently, the generation of a model with sufficient quality was rendered unfeasible, and the models contained the most discernible imperfections within this entire study (see Figure 11 – A2, B2, C2). In the case of "180*", an increase in the value of processing time was observed, which did not correspond with the resulting trend. However, no such increase was observed in the case of the values of qualitative features (surface, volume), which indicated that the overestimation was due to software data processing only. The outcome of this optimisation step was the identification of the minimum number of frames and image configuration required to produce a high-quality 3D model. From this perspective, the second lowest tested frame count was selected, specifically 120 images, with 40 images captured at each of the P1, P3, and P5 height levels. It was demonstrated that at this frame count and image configuration, the resulting model quality was practically equivalent to that of the model generated using the full set of 360 original frames from all height levels. The lowest number of images studied, 90 (P1, P3, P6 with 30 images in each height level), was deemed insufficient due to the presence of imperfections and redundant artefacts in the resulting models, particularly in the cases of C. sativus and S. lycopersicum (see Figure 11 – A1, B1). In the case of L. sativa , the quality of the resulting models was relatively comparable to models generated from 120 images (see Figure 11. – C1, C3). Notably, a substantial discrepancy was observed between 120* and 120. In both cases, the initial datasets contained 120 images but differed in image configuration and frame reduction (see Table 2 for a more detailed description). It can be concluded that reducing in the number of frames while preserving a greater number of positions was a more appropriate optimisation procedure than reducing the number of positions while preserving the full number of frames at a given position. This conclusion was based on the resulting quality of the models and is fully in line with the photogrammetric principles. The resulting values of the crucial statistical indicators describing the quality of the generated 3D models corroborated the preceding claim. Specifically, it was determined that 120 images represent the minimum quantity required to generate a plant model of sufficient quality. A similar trend was observed in the values of all the indicators, namely an increase in their value with an increasing number of images for all the plant species (see Figure 12). As the number of frames increased, the software detected a greater number of features, which resulted in an increasing trend in the data for the TP, faces, and vertices. The inadequate degree of image linkage observed in the "120*" case was systematically reflected in all the indicators that were analysed. A decrease in the number of TP was also observed in the case of "180*". It is conceivable that the reduced number of frames had an impact on this value. However, the other indicators represented in this case did not stand out significantly. The most significant outcome of this optimisation stage was the substantial reduction in processing time and computational complexity associated with image processing and model generation. Specifically, this resulted in an approximately 75% reduction across all the plant species. To further contextualise this finding, it is worth noting when the lowest tested number of images (90) was used, the time and energy savings would have reached approximately 82%. However, it should be noted that this would have come at the cost of the aforementioned adverse effects (holes, artefacts). Moreover, a less evident yet noteworthy enhancement pertains to the optimisation of the minimum number of frames utilised. It has been demonstrated that, by theoretically eliminating the need to scan up to 360 images per plant (a process that takes approximately 8 minutes), it would also be possible to reduce the scan time by a third, approximately 2.7 minutes. This would enhance the overall efficiency of the photogrammetric reconstruction. This optimization step substantially extended the applicability of the previously developed multifunctional robotic scanning apparatus [48]. Assessment of plant morphology The subsequent paragraph offers a synopsis of the distinctive characteristics of the morphology of the examined plants with respect to their scanability and the potential for generating a high-quality model. The plant species examined exhibited substantial variation in their morphological characteristics. The size and number of thin parts of the objects posed a challenge for photogrammetric reconstruction. The morphology of C. sativus is characterized by the presence of a relatively thick stem and two types of leaves: cotyledon leaves with smooth edges and true lobed leaves with serrated edges. This specific body constitution was conducive to the scanning process, resulting in complete models devoid of unwanted artefacts. The morphology of L. lycopersicum is characterized by the following traits: the presence of oval-shaped leaves with smooth edges, thin petioles, and a slender stem covered with imperfectly reconstructed thin trichomes. In the majority of cases, the resulting model was composed exclusively of swollen, redundant trichome artefacts. These artefacts slightly overestimate the values of the resulting qualitative indicators. The morphological traits of L. sativa are characterized by the presence of a dense rosette of overlapping and twisted smooth, drop-shaped leaves with smooth edges. The primary issue encountered during the scanning process pertained to the presence of a dense rosette, which obscured the interior from the camera's view, precluding effective penetration and thorough scanning. In this instance, it is plausible that the volume and surface area were also slightly overestimated. The impact of the morphological diversity of the plants under investigation on the quality of the model was the subject of additional evaluation. This was determined by evaluating the confidence level of the reconstruction of points from the original image data. This indicator is intended to express the degree of confidence in the precision and reliability of the reconstruction of the dense point cloud. The visualisation is typically exhibited on a standardised scale ranging from 1-100, which signifies the increasing degree of reliability of the reconstruction (see Figure 13). The image data were acquired with the photogrammetric system set to the optimal settings identified above. In the case of C. sativus , areas on the underside of the cotyledon and true leaves with relatively low confidence levels (1–20%) were observed. This phenomenon may have been caused by the shorter stem (2.5 centimetres) and drooping leaves, which reduced the handling space in the lower area of the plant and may have reduced the accessibility of the camera. The area formed by overlapping leaves was also more difficult to scan (see Figure 13 – A1, red arrow). Conversely, in the case of S. lycopersicum , the longer stem (5.2 centimetres) and leaf petioles formed a more spacious area that was more accessible to the camera, facilitating the scanning process and the acquisition of higher-quality data. In the case of L. sativa , the absence of a stem and curled, drooping leaves made camera accessibility more difficult, which was reflected in the colour map of the model. Once more, there was a manifest decline in confidence level in these pivotal domains. Despite these parts, the majority of the generated dense point cloud points presented values that approximated 100 %. In conclusion, considering the financial investment in the scanning apparatus, the resulting models of the studied plant species proved to be of sufficient detail and quality for future follow-up research, demonstrating a cost-effective approach. A comprehensive scanning procedure was conducted on all the salient morphological traits of the plants, including the stems and leaves. This methodical approach ensured the reliable capture and documentation of plant body features, paving the way for subsequent analysis and research. The upgraded low-cost scanning apparatus, characterized by its proven robustness and reliability, is well suited for use in primary research. Figure 14 shows a comparison of the resulting plant models, which were used to assess the usability of the optimised photogrammetric system. Compared with the previously tested POP 3 3D scanner, the image analysis demonstrated that the photogrammetric system was capable of generating complete and high-quality models. For the evaluation of data from the POP 3 3D scanner, RevoScan 5 software was used. Additionally, it is important to note, that it did not support scripting, thus, it would not have been suitable for the automated processing workflow implemented in this study. The models produced by 3D scanner were often incomplete, particularly due to difficulties in capturing the entire underside of plant leaves. As a result, it would not have been possible to generate reliable values for volume and surface area, or to determine the descriptive quantitative parameters. In the case of S. lycopersicum , thin leaf petioles posed a significant challenge for scanning (Figure 14 – B1). In contrast, the photogrammetric approach allowed for greater detail and completeness. Overall, the combination of an RGB camera and evaluation software based on the Metashape API proved to be a robust and user-friendly solution. This approach effectively enables the generation of high-quality 3D plant models and validates the photogrammetric system’s suitability for non-destructive morphological analysis. Compared with the previously used 3D scanner paired with Revopoint software, this approach offers significantly greater control over the 3D reconstruction algorithms. Such flexibility is typically limited in commercial 3D scanning systems, making the RGB camera and Metashape-based workflow a more adaptable and customizable options. Discussion Non-destructive methods of studying plant morphological features represent a novel and contemporary (bio)technological approach, whose reliability and accuracy are contingent on the appropriate composition and settings of the system. The type, accuracy, and price of the optical equipment used are of particular importance in this context. RGB cameras with CMOS sensors that offer sufficient resolution are commonly used for these purposes (including in this study) [51, 52]. To ensure the overall system’s reliability and accuracy, it is essential to carefully determine the parameters of both the interior and exterior camera orientations. Panels displaying a black-and-white checkerboard pattern or circular reference markers are utilised for this purpose [47, 53-56]. The geometry of the spatial orientation of the system is crucial for acquiring image data from different positions and angles. The high degree of variability of the resulting image dataset allows the generation of high-quality, complete 3D models of plants. As demonstrated in this study, the calibration strategy had a significant effect on the spatial reconstruction quality. The use of pre-calibrated interior parameters produced the highest accuracy, whereas simultaneous calibration with pre-calibrated values as initial estimates resulted in the most stable reconstruction performance across repeated trials. Furthermore, the application of adaptive camera model fitting generally improved the accuracy and consistency of the models, particularly when either no prior information or full calibration data were used. In contrast, the strategy relying solely on basic physical camera parameters, focal length and pixel size, led to reduced accuracy and even partial failure of image alignment in the case of L. sativa . On the basis of these findings, the final configuration employed in this study combined simultaneous calibration with pre-calibrated parameters and adaptive model fitting. A notable advantage of the developed apparatus was the incorporation of a programmable, 6-axis AR4 robotic arm as a camera mount. This innovation enabled the precise positioning of the camera at any point in the workspace, thereby facilitating the definition of the complex spatial arrangement of the captured images (see Figure 2). The simplicity of adjusting and setting up this arrangement was an immense advantage of the established methodology. The ability of the established photogrammetric system to capture images from the underside of the scanned object represents a significant advantage, as this capability is not commonly supported by all systems. A configuration featuring fewer images combined with a greater number of positions proved to be more effective for spatial arrangement than one with fewer images and fewer positions. The optimal setup was identified as three height levels (P1, P3, and P5), each with 40 images taken at 10° spacing. A very similar photogrammetric system setup, including Agisoft Metashape software, was also employed for the reconstruction of detailed archaeological objects. The resulting models achieved sufficient accuracy for the study of morphological features [57]. Competitive devices frequently capture only the upper portion of the object, typically from two positions utilising stereo photogrammetry with two cameras [26, 55, 58-60]. Conversely, there are also systems capable of capturing high-quality images of plants from below to a certain extent [17, 61]. In general, the captured images should provide uniform coverage of the entire scanned object, with sufficient overlap in both the horizontal and vertical directions. Ideally, the angle between adjacent camera height levels should not exceed 45° [62]. In our case, however, a high-quality model was generated even with rather low overlap between images. This was made possible by the favourable spherical arrangement of the captured images, which provided uniform coverage of the scanned plant surface from nearly the full 360°. Another specific solution in photogrammetric system design involves mounting the camera on a tiltable bracket combined with a linear slider. This configuration also yielded high-fidelity models of the S. lycopersicum plant placed on a turntable [63]. The acquisition of substantial image data from multiple vantage points within the workspace facilitates the generation of high-fidelity models, characterised by the preservation of authentic dimensions, a finding that is corroborated by the present study. The quality of 3D reconstruction, as derived from image data, is significantly influenced by lighting conditions. The majority of reconstruction methods, whether geometric or machine learning-based, rely heavily on visual information obtained from pixel intensity, texture, shading, and colour consistency [64]. The selection of appropriate lighting, homogenization, and determination of the optimal exposure time can substantially increase the quality of image data intended for object reconstruction [65]. Ideally, the light should be homogenised. For these purposes, a diffuse photographic panel was utilised, which, however, slightly reduced the intensity of the incident light. This reduction was considered when the exposure time was set. The side LED panels were not homogenised with the intention of intensifying the illumination of the QR code and ensuring its correct reading. The most suitable exposure time of 50 milliseconds was selected for this scanning system. The selection was supported by image analysis of the generated plant models and analysis of chosen statistical indicators. The system demonstrated notable robustness concerning exposure time settings. Consequently, the quality of the resulting models remained largely unaffected by these settings. This finding is consistent with those of other studies [53, 55, 66]. Ensuring even illumination of the scanned object from all directions is widely considered the best practice, as it helps to avoid issues such as strong shadows, overexposure, or underexposure [67]. These effects were also observed to some extent in this study, where the use of underexposed or overexposed images resulted in minor, typically negligible, imperfections and the presence of undesired artefacts in the models. One potential solution to this issue could be the implementation of more thorough post-processing of the model, with the objective of removing these elements. The surface of the plants presented a high degree of reflectance, particularly in the case of L. sativa . The presence of highly reflective surfaces has been shown to complicate 3D reconstruction [68]. In addition to altering the exposure time, the utilisation of matting agents has been posited as a potential solution [69, 70]. However, it is imperative to exercise caution to ensure that the experiment does not have a detrimental effect on plant growth, particularly in the context of studying time-dependent plant growth. In conclusion, close-range photogrammetry under controlled conditions generally results in a reduced degree of fluctuation in lighting conditions compared with outdoor solutions. When the system is transferred from the laboratory scale to the field scale, minimising the influence of variable lighting conditions on the 3D reconstruction of scanned plants. To ensure the robustness of the system and to facilitate the reproducibility and accuracy of the photogrammetric analysis, it would be necessary to make certain adaptations in the methodology. During close-range photogrammetric spatial reconstruction of small and structurally complex objects, such as plant samples, several challenges were encountered in generating accurate 3D models. These included, for example, thin leaves, fine stems, petioles, or trichomes. The presence of these morphological structures was responsible for the holes, imperfections, and extraneous artefacts observed in the models generated in this study. This issue was attributed to the inability of the software to accurately ascertain the surface orientation of thin structures. This resulted in ambiguity when determining whether a particular point belonged to the upper or lower surface of an object. These points were then either wholly excluded from the reconstruction or incorrectly assigned to one of the sides of the leaf, resulting in local errors in the model [60, 71]. For all the tweak values that were tested, the most suitable value (0.9) was selected for all the plant types. At this value, the holes and imperfections in the model were effectively filled but without significantly overestimating the monitored quality indicators (see Figures 5, 6). An alternative approach to standard close-range photogrammetry is the use of multifocus stacking, a technique commonly applied in microphotogrammetry of small and structurally complex objects, such as insects or plant flowers [72, 73]. This method overcomes the depth of field limitations inherent in optical systems by merging multiple images acquired at different focal planes into a single composite image with improved focus throughout the object. However, the implementation of this method necessitates the acquisition of a substantially greater number of images, which has a negative effect on the time and computational complexity of photogrammetric analysis. Another important factor affecting the outcome of close-range photogrammetric reconstruction is the distance between the object and the camera, as it directly impacts the image resolution, depth of field, and precision of feature reconstruction. Within a controlled laboratory environment, where stable conditions are maintained and a camera with adequate resolution and depth of field is employed, the range of values is typically extensive. This observation aligns with the findings of the present study. In the range from 14 to 16 centimetres from the centre of the plant, the models demonstrated a high degree of similarity in quality It was found that only the extreme values of this parameter, whether excessively low or high, had a discernible adverse effect on the quality of the generated model, mainly because of the lack of focus. The selection of 16 as the optimal value was based on a combination of image analysis outcomes and a detailed evaluation of relevant statistical indicators. A notable benefit of the photogrammetric system is its capacity to extend the scanning distance up to a maximum of 30-40 centimetres [48]. In the context of 3D reconstruction of larger plants, increasing the scanning distance is necessary. In certain systems, a greater distance is traversed during the scanning process of plants. Nevertheless, the quality of the resulting models is often found to be inferior [56]. If the utilisation of the developed apparatus was required for the scanning of a plant whose height exceeded the maximum adjustable value, it would be necessary to modify the design. Inspiration can be achieved from systems that scan tall Zea mays L. plants [28, 29]. A plant of such considerable height would be positioned on a rotating table, which would be situated on a vertically adjustable platform that would slide beneath the working area to the lower floor of the apparatus. In this configuration, it would be essential to precisely calibrate the movement of the positioning platform and robotic arm, the rotational speed of the turntable, and the scanning frequency. The proposed design modification would enhance the usability and robustness of the scanning apparatus under development. The ability to precisely position the camera at a specific location and thus create geometrically unique spatial arrangements of images is a significant advantage of the described apparatus. This is also related to the relatively straightforward adjustment of the total number of images and thus the total scanning time. In conventional practice, images are typically captured from only two positions [3, 56]. This paper sets six height levels. During the optimisation process, the number of images was successfully reduced from the original 360 to 120 (40 images at each level P1, P3, and P5) while maintaining the original quality of the generated model. The developed apparatus was then compared with a similar solution in terms of accuracy and reliability. The utilisation of five distinct positions of the RGB camera relative to the S. lycopersicum plant yielded models of inferior quality. The number of detected tie points was found to be significantly lower, ranging from 20000 to 150000. In contrast, our models exhibited a substantially greater number of tie points, ranging from 120000 to 300000 per model. Conversely, the RMS RE was higher, specifically in the range of 0.2-0.8, whereas in our case, the models were of higher quality with values in the range of 0.1-0.22 [74]. Furthermore, the reduction in the number of images necessitated a corresponding reduction in the time required to scan the entire plant, from an original 8 minutes to a mere 2.7 minutes. The time saved by this approach would enable the analysis of a larger number of plants within the same time. This optimisation step resulted in a substantial streamlining of photogrammetric analysis and an expansion of the applicability of the scanning apparatus under development. In subsequent experiments, it would be possible to analyse the influence of biotic and abiotic factors on plant morphological features, as this type of study would require a larger number of plants to ensure the statistical relevance of the results. The ability to scan both treated and untreated plants within a 24-hour period will facilitate a more effective and accurate comparison of these groups, thereby enabling more reliable detection of any potential differences. The scanning photogrammetric apparatus described in this paper, when used with appropriately configured parameters, has expanded the capabilities of commonly employed laboratory-scale equipment [75, 76]. A competing system, comprising two static cameras and a turntable, was found to generate models of similar quality. Furthermore, the model of the generated root system could serve as an inspiration for the future enhancement of the existing apparatus [77]. However, certain apparatuses with this configuration incur significantly higher expenses and necessitate a greater spatial requirements [56]. The presence of a robotic 6-axis arm enabling precise camera positioning was also a feature of competing systems, which demonstrated an acceptable level of accuracy and reliability in 3D plant reconstruction [78]. In certain instances, emphasis has not been placed on comprehensive plant models but rather on specific segments, predominantly leaf models [79]. In such cases, the need to achieve detailed structures was not as pressing as it is in this paper. In this particular instance, the developed apparatus would constitute a suitable alternative, achieving a similar quality of output. The apparatus's autonomous control, the construction design, and modern, reliable components (namely, a robotic arm and a high-resolution camera) render the developed photogrammetric system suitable for deployment at the field scale in the context of precision agriculture. To achieve the desired outcome, it would be necessary to modify the design and readjust the parameters of the photogrammetric system so that it could compensate for negative disturbances from the external environment, such as lighting conditions, humidity, and air temperature. Conclusion This study aimed to develop a non-destructive and cost-effective method for analysing morphological traits in the phenotyping of economically significant plants. In particular, the feasibility of a close-range photogrammetry setup utilizing structure-from-motion and multi-view stereo (SfM-MVS) techniques was evaluated. For this purpose, an existing robotic scanning system based on a static-camera and moving-object configuration was upgraded. An algorithm enabling automated camera positioning, data acquisition, subsequent processing, and model generation has been successfully implemented. The presence of the robotic arm and turntable makes the system robust and offers a variety of possibilities for setting up the spatial arrangement of photogrammetric reconstruction. A comprehensive evaluation of the influence of individual parameters on the quality of the resulting 3D models was conducted via qualitative and statistical indicators, including image analysis of the generated models. Among all the calibration strategies tested, the most robust and accurate results were achieved via simultaneous calibration with pre-calibrated interior orientation parameters combined with adaptive camera model fitting. This configuration significantly improved spatial reconstruction quality, whereas other strategies, particularly those based only on focal length and pixel size, resulted in reduced accuracy or led to partial alignment failure. These findings confirm the critical importance of the camera calibration strategy in ensuring reconstruction reliability, especially when non-metric cameras are used. The optimal settings were identical for all the plant species tested: C. sativus , S. lycopersicum and L. sativa . The optimal exposure time was 50 milliseconds. The ideal scanning distance between the camera and the object was 16 centimetres. The best model generation results from depth maps were achieved when the tweak parameters were set to 0.9. In contrast, the worst models were generated when the tweak parameters were left unset. A configuration characterized by a reduced number of images combined with an increased number of height levels was found to contribute to the generation of higher-quality 3D models of plants compared with a configuration with a less robust image configuration and a greater number of images. The best combination was determined to be three height levels (P1, P3, and P5), with 40 images per position and approximately 10° spacing. The overall evaluation process was successfully streamlined, reducing the number of required images and cutting the total scan cycle time by approximately 75%, from 8 to just 2.7 minutes. The majority of points within the dense point cloud exhibited a confidence level that approached 100%. The introduced methodology enriches the list of available phenotyping platforms on the basis of photogrammetric analysis. The system's main advantages are its low acquisition cost, reliability, ease of use and sufficient accuracy. The ratio between the scanning time and the resulting model quality is highly favourable, indicating that efficient use of time does not compromise, but rather supports, the production of accurate and detailed 3D models. If the plant dimensions exceed the predefined size limit (approximately 30-40 centimetres), a simple structural modification would allow for an extension of the scanning capacity. In conclusion, one possible application of the developed apparatus is its use in the development of microbial or alternative biopreparations aimed at environmentally friendly protection of economically significant crops. Declarations Ethics approval and consent to participate Not applicable: This manuscript does not include human or animal research. Consent for publication Not applicable. Availability of data and materials The datasets used and/or analysed during the current study are available from the corresponding author upon reasonable request. Competing interests The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Funding This work was supported by the Internal Grant Agency of the University of Chemistry and Technology in Prague, Grant No. A1_FPBT_2025_005, Modern Biotechnologies. Authors' contributions JM designed the study, developed the methodology, contributed to the experiments, and created the software tools. He also drafted and edited the manuscript. ZS co-developed the methodology, co-wrote the draft, designed and implemented the evaluation algorithm, worked on software, and supervised parts of the research. OŠ handled the data acquisition and curation, including plant cultivation. LN performed the formal data analysis, supported data curation, and contributed to manuscript editing. MH supervised the project, performed formal analysis, secured funding, and revised the manuscript. All the authors approved the final version and take responsibility for its content. Acknowledgements Not applicable. References J. Hrzich, M. Beck, C. Bidinosti, C. Henry, K. Manawasinghe, and K. Tanino, A low-cost photogrammetry system for 3D plant modeling and phenotyping. 2025. N. An, C.M. Palmer, R.L. Baker, R.J.C. Markelz, J. Ta, M.F. Covington, J.N. Maloof, S.M. Welch, and C. Weinig, Plant high-throughput phenotyping using photogrammetry and imaging techniques to measure leaf length and rosette area, Computers and Electronics in Agriculture. 127 (2016) 376-394. Y. Wang, W. Wen, S. Wu, C. Wang, Z. Yu, X. Guo, and C. Zhao, Maize plant phenotyping: Comparing 3D laser scanning, multi-view stereo reconstruction, and 3D digitizing estimates, Remote Sensing. 11 (2019) 1-17. M.M. Rahaman, D. Chen, Z. Gillani, C. Klukas, and M. Chen, Advanced phenotyping and phenotype data analysis for the study of plant growth and development, Frontiers in Plant Science. 6 (2015) 1-15. Z. Rui, Z. Zhang, M. Zhang, A. Azizi, C. Igathinathane, H. Cen, S. Vougioukas, H. Li, J. Zhang, Y. Jiang, X. Jiao, M. Wang, Y. Ampatzidis, O.I. Oladele, M. Ghasemi-Varnamkhasti, and R. Radi, High-throughput proximal ground crop phenotyping systems - A comprehensive review, Computers and Electronics in Agriculture. 224 (2024) 1-22. C.Q. Xie and C. Yang, A review on plant high-throughput phenotyping traits using UAV-based sensors, Computers and Electronics in Agriculture. 178 (2020) 1-14. S. Dhondt, N. Wuyts, and D. Inzé, Cell to whole-plant phenotyping: The best is yet to come, Trends in Plant Science. 18 (2013) 433-444. M.S.M. Asaari, S. Mertens, L. Verbraeken, S. Dhondt, D. Inzé, K. Bikram, and P. Scheunders, Non-destructive analysis of plant physiological traits using hyperspectral imaging: A case study on drought stress, Computers and Electronics in Agriculture. 195 (2022) 1-13. L. Cattivelli, F. Rizza, F.W. Badeck, E. Mazzucotelli, A.M. Mastrangelo, E. Francia, C. Mare, A. Tondelli, and A.M. Stanca, Drought tolerance improvement in crop plants: An integrated view from breeding to genomics, Field Crops Research. 105 (2008) 1-14. M. Awlia, A. Nigro, J. Faikus, S.M. Schmoeckel, S. Negrao, D. Santelia, M. Trtílek, M. Tester, M.M. Julkowska, and K. Panzarová, High-throughput non-destructive phenotyping of traits that contribute to salinity tolerance in Arabidopsis thaliana , Frontiers in Plant Science. 7 (2016) 1-15. O. Ali, A. Ramsubhag, and J. Jayaraman, Biostimulant properties of seaweed extracts in plants: Implications towards sustainable crop production, Plants. 10 (2021) 1-27. S.S. Mahusook, F.A.A. Rajathi, H.N.S. Maharifa, and R. Sharmila, Comparative study of agarophytes - Gracilaria edulis and Gelidiella acerosa as biostimulant and application of agar for water-holding in soil and plant growth promotion, Agricultural Science Digest - A Research Journal. 41 (2021) 21-27. Y.C. Wu, J.Y. Zhou, C.G. Li, and Y. Ma, Antifungal and plant growth promotion activity of volatile organic compounds produced by Bacillus amyloliquefaciens , Microbiologyopen. 8 (2019) 1-14. L. Pappalettere, S. Bartolini, and A. Toffanin, Auxin-producing bacteria used as microbial biostimulants improve the growth of tomato ( Solanum lycopersicum L.) seedlings in hydroponic systems, Biotech. 13 (2024) 1-14. L. Pellan, C.A.T. Dieye, N. Durand, A. Fontana, C. Strub, and S. Schorr-Galindo, Biocontrol agents: Toolbox for the screening of weapons against mycotoxigenic Fusarium , Journal of Fungi. 7 (2021) 1-29. C.N. Sible, J.R. Seebauer, and F.E. Below, Plant biostimulants: A categorical review, their implications for row crop production, and relation to soil health indicators, Agronomy. 11 (2021) 1-20. S. Paulus, H. Schumann, H. Kuhlmann, and J. Leon, High-precision laser scanning system for capturing 3D plant architecture and analysing growth of cereal plants, Biosystems Engineering. 121 (2014) 1-11. M. Friedli, N. Kirchgessner, C. Grieder, F. Liebisch, M. Mannale, and A. Walter, Terrestrial 3D laser scanning to track the increase in canopy height of both monocot and dicot crop species under field conditions, Plant Methods. 12 (2016) 1-15. C. Todo, H. Ikeno, K. Yamase, T. Tanikawa, M. Ohashi, M. Dannoura, T. Kimura, and Y. Hirano, Reconstruction of conifer root systems mapped with point cloud data obtained by 3D laser scanning compared with manual measurement, Forests. 12 (2021) 1-17. S. Puliti, J. Breidenbach, and R. Astrup, Estimation of forest growing stock volume with UAV laser scanning data: Can it be done without field data?, Remote Sensing. 12 (2020) 1-19. T. Dornbusch, S. Lorrain, D. Kuznetsov, A. Fortier, R. Liechti, I. Xenarios, and C. Fankhauser, Measuring the diurnal pattern of leaf hyponasty and growth in Arabidopsis - a novel phenotyping approach using laser scanning, Functional Plant Biology. 39 (2012) 860-869. A. Šrollerů and M. Potůčková, Evaluating the applicability of high-density UAV LiDAR data for monitoring tundra grassland vegetation, International Journal of Remote Sensing. 46 (2024) 1-36. Z.Y. Ma, R. Rayhana, K. Feng, Z. Liu, G.Z. Xiao, Y.F. Ruan, and J.S. Sangha, A review on sensing technologies for high-throughput plant phenotyping, Ieee Open Journal of Instrumentation and Measurement. 1 (2022) 1-22. W. Kazmi, S. Foix, and G. Alenyà. Plant leaf imaging using time of flight camera under sunlight, shadow and room conditions. in 2012 IEEE International Symposium on Robotic and Sensors Environments Proceedings. 2012. S. Mairhofer, S. Zappala, S.R. Tracy, C. Sturrock, M. Bennett, S.J. Mooney, and T. Pridmore, Rootrak: Automated recovery of three-dimensional plant root architecture in soil from X-ray microcomputed tomography images using visual tracking, Plant Physiology. 158 (2012) 561-569. D. Andújar, M. Calle, C. Fernández-Quintanilla, Á. Ribeiro, and J. Dorado, Three-dimensional modeling of weed plants using low-cost photogrammetry, Sensors. 18 (2018) 1-11. S. Paulus, Measuring crops in 3D: using geometry for plant phenotyping, Plant Methods. 15 (2019) 1-13. B. Daviet, R. Fernandez, L. Cabrera-Bosquet, C. Pradal, and C. Fournier, Phenotrack3D: An automatic high-throughput phenotyping pipeline to track maize organs over time, Plant Methods. 18 (2022) 1-14. S. Wu, W. Wen, Y. Wang, J. Fan, C. Wang, W. Gou, and X. Guo, MVS-pheno: A portable and low-cost phenotyping platform for maize shoots using multiview stereo 3D reconstruction, Plant Phenomics. 2020 (2020) 1-17. R. Rossi, C. Leolini, S. Costafreda-Aumedes, L. Leolini, M. Bindi, A. Zaldei, and M. Moriondo, Performances evaluation of a low-cost platform for high-resolution plant phenotyping, Sensors. 20 (2020) 1-16. T. Luhmann, Precision potential of photogrammetric 6DOF pose estimation with a single camera, Isprs Journal of Photogrammetry and Remote Sensing. 64 (2009) 275-284. W. He, Z.H. Ye, M.S. Li, Y.L. Yan, W. Lu, and G.N. Xing, Extraction of soybean plant trait parameters based on SfM-MVS algorithm combined with GRNN, Frontiers in Plant Science. 14 (2023) 1-16. J.C. Rose, S. Paulus, and H. Kuhlmann, Accuracy analysis of a multi-view stereo approach for phenotyping of tomato plants at the organ level, Sensors. 15 (2015) 9651-9665. W. Wen, S. Wu, X. Lu, X. Liu, S. Gu, and X. Guo, Accurate and semantic 3D reconstruction of maize leaves, Computers and Electronics in Agriculture. 217 (2024) 1-11. S. Krasic and P. Pejic, Comparative analysis of terrestrial semi-automatic and automatic photogrammetry in 3D modeling process, Nexus Network Journal. 16 (2014) 273-283. F.T. Teshome, H.K. Bayabil, G. Hoogenboom, B. Schaffer, A. Singh, and Y. Ampatzidis, Unmanned aerial vehicle (UAV) imaging and machine learning applications for plant phenotyping, Computers and Electronics in Agriculture. 212 (2023) 1-15. N. Ivanov, P. Boissard, M. Chapron, and B. Andrieu, Computer stereo plotting for 3D reconstruction of a maize canopy, Agricultural and Forest Meteorology. 75 (1995) 85-102. M.A. Aguilar, J.L. Pozo, F.J. Aguilar, A.M. García, I. Fernández, J. Negreiros, and J. Sánchez-Hermosilla. Application of close-range photogrammetry and digital photography analysis for the estimation of leaf area index in a greenhouse tomato culture. in ISPRS-Commission V Mid-Term Symposium on Close Range Image Measurement Technique. 2010. Newcastle upon Tyne, ENGLAND. H. Zhu, J. Dorner, D.L. Rowland, R.C. Derksen, and H.E. Ozkan, Spray penetration into peanut canopies with hydraulic nozzle tips, Biosystems Engineering. 87 (2004) 275-283. Y.K. Rui, Y.F. Peng, Z.R. Wang, and J.B. Shen, Stem perimeter, height and biomass of maize ( Zea mays L.) grown under different N fertilization regimes in Beijing, China, International Journal of Plant Production. 3 (2009) 85-90. S.Q. Fang, X.L. Yan, and H. Liao, 3D reconstruction and dynamic modeling of root architecture in situ and its application to crop phosphorus research, Plant Journal. 60 (2009) 1096-1108. L.D. Lopez, D. Shantharaj, L. Liu, H. Bais, and J. Yu, Robust image-based 3D modeling of root architecture. 1 (2011) 1-10. J. Vanderborght, V. Couvreur, F. Meunier, A. Schnepf, H. Vereecken, M. Bouda, and M. Javaux, From hydraulic root architecture models to macroscopic representations of root hydraulics in soil water flow and land surface models, Hydrology and Earth System Sciences. 25 (2021) 4835-4860. A. Junker, M.M. Muraya, K. Weigelt-Fischer, F. Arana-Ceballos, C. Klukas, A.E. Melchinger, R.C. Meyer, D. Riewe, and T. Altmann, Optimizing experimental procedures for quantitative evaluation of crop plant performance in high throughput phenotyping systems, Frontiers in Plant Science. 5 (2015) 1-21. W. Gu, W. Wen, S. Wu, C. Zheng, X. Lu, W. Chang, P. Xiao, and X. Guo, 3D reconstruction of wheat plants by integrating point cloud data and virtual design optimization, Agriculture. 14 (2024) 1-20. J. Martinez-Guanter, Á. Ribeiro, G.G. Peteinatos, M. Pérez-Ruiz, R. Gerhards, J.M. Bengochea-Guevara, J. Machleb, and D. Andújar, Low-cost three-dimensional modeling of crop plants, Sensors. 19 (2019) 1-14. S. Paulus, J. Behmann, A.K. Mahlein, L. Plumer, and H. Kuhlmann, Low-cost 3D systems: Suitable tools for plant phenotyping, Sensors. 14 (2014) 3001-3018. J. Mach, L. Krauz, P. Páta, and M. Halecky, Development of low-cost multifunctional robotic apparatus for high-throughput plant phenotyping, Smart Agricultural Technology. 9 (2024) 1-17. Metashape Python Reference Release 2.2.0. 2024; Available from: https://www.agisoft.com/pdf/metashape_python_api_2_2_0.pdf. How to improve the modeling of thin objects in Metashape. 2024; Available from: https://agisoft.freshdesk.com/support/solutions/articles/31000163220-how-to-improve-the-modeling-of-thin-objects-in-metashape. X. Lu, E. Ono, S. Lu, Y. Zhang, P. Teng, M. Aono, Y. Shimizu, F. Hosoi, and K. Omasa, Reconstruction method and optimum range of camera-shooting angle for 3D plant modeling using a multi-camera photography system, Plant Methods. 16 (2020) 1-16. X. Teng, G. Zhou, Y. Wu, C. Huang, W. Dong, and S. Xu, Three-dimensional reconstruction method of rapeseed plants in the whole growth period using an RGB-D camera, Sensors. 21 (2021) 1-20. K. Bobkowska, P. Burdziakowski, J. Szulwic, and K.M. Zielinska-Dabkowska, Seven different lighting conditions in photogrammetric studies of a 3D urban mock-up, Energies. 14 (2021) 1-27. J. Kolecki, P. Kuras, E. Pastucha, K. Pyka, and M. Sierka, Calibration of industrial cameras for aerial photogrammetric mapping, Remote Sensing. 12 (2020) 1-21. D. Li, L. Xu, X.-s. Tang, S. Sun, X. Cai, and P. Zhang, 3D imaging of greenhouse plants with an inexpensive binocular stereo vision system, Remote Sensing. 9 (2017) 1-27. T.T. Nguyen, D.C. Slaughter, N. Max, J.N. Maloof, and N. Sinha, Structured light-based 3D reconstruction system for plants, Sensors. 15 (2015) 18587-18612. M. Pepe and D. Costantino, Techniques, tools, platforms, and algorithms in close range photogrammetry in building 3D model and 2D representation of objects and complex architectures, Computer-Aided Design and Applications. 18 (2020) 42-65. N. An, S.M. Welch, R.J.C. Markelz, R.L. Baker, C.M. Palmer, J. Ta, J.N. Maloof, and C. Weinig, Quantifying time-series of leaf morphology using 2D and 3D photogrammetry methods for high-throughput plant phenotyping, Computers and Electronics in Agriculture. 135 (2017) 222-232. S. Dandrifosse, A. Bouvry, V. Leemans, B. Dumont, and B. Mercatoris, Imaging wheat canopy through stereo vision: Overcoming the challenges of the laboratory to field transition for morphological features extraction, Frontiers in Plant Science. 11 (2020) 1-15. H. Surmen, Photogrammetry for 3D reconstruction of objects: Effects of geometry, texture and photographing, Image Analysis & Stereology. 42 (2023) 51-63. A. Conn, U.V. Pedmale, J. Chory, and S. Navlakha, High-resolution laser scanning reveals plant architectures that reflect universal network design principles, Cell Systems. 5 (2017) 53-65. A. Bisson-Larrivée and J.-B. LeMoine, Photogrammetry and the impact of camera placement and angular intervals between images on model reconstruction, Digital Applications in Archaeology and Cultural Heritage. 26 (2022) 1-12. R. Rossi, S. Costafreda-Aumedes, L. Leolini, C. Leolini, M. Bindi, and M. Moriondo, Implementation of an algorithm for automated phenotyping through plant 3D-modeling: A practical application on the early detection of water stress, Computers and Electronics in Agriculture. 197 (2022) 1-15. D. Rangelov, S. Waanders, K. Waanders, M.v. Keulen, and R. Miltchev. 3D reconstruction techniques and the impact of lighting conditions on reconstruction quality: A comprehensive review. in Preprints. 2025. Preprints. K. Wu, J. Tan, and C. Liu, A novel approach to obtain optimal exposure for 3D shape reconstruction of high dynamic range objects, Measurement Science and Technology. 32 (2021) 1-16. B. Biskup, H. Scharr, U. Schurr, and U. Rascher, A stereo imaging system for measuring structural parameters of plant canopies, Plant Cell and Environment. 30 (2007) 1299-1308. M. Tancik, V. Casser, X. Yan, S. Pradhan, B. Mildenhall, P. Srinivasan, J. Barron, and H. Kretzschmar, Block-nerf: Scalable large scene neural view synthesis. 2022. 8238-8248. C.-S. Liu, J.-J. Lin, and B.-R. Chen, A novel 3D scanning technique for reflective metal surface based on HDR-like image from pseudo exposure image fusion method, Optics and Lasers in Engineering. 168 (2023) 1-10. J. Franke, T. Koutecký, and D. Koutný, Comparison of sublimation 3D scanning sprays in terms of their effect on the resulting 3d scan, thickness, and sublimation time, Materials. 16 (2023) 1-15. R. Mendřický, Impact of applied anti-reflective material on accuracy of optical 3D digitisation, Materials Science Forum. 919 (2018) 335-344. L.T. De Paolis, V. De Luca, C. Gatto, G. D’Errico, and G.I. Paladini. Photogrammetric 3D reconstruction of small objects for a real-time fruition. in Augmented Reality, Virtual Reality, and Computer Graphics. 2020. Cham: Springer International Publishing. T.-N. Doan and C.V. Nguyen, A low-cost digital 3D insect scanner, Information Processing in Agriculture. 11 (2024) 337-355. B. Ströbel, S. Schmelzle, N. Blüthgen, and M. Heethoff, An automated device for the digitization and 3D modelling of insects, combining extended-depth-of-field and all-side multi-view imaging, Zookeys. (2018) 1-27. T.R. Roshan, M. Jafari, M. Golami, and M. Kazemi, Evaluating geometric measurement accuracy based on 3D model reconstruction of nursery tomato plants by Agisoft photoscan software, Computers and Electronics in Agriculture. 221 (2024) 1-14. G. Bernotas, L.C.T. Scorza, M.F. Hansen, I.J. Hales, K.J. Halliday, L.N. Smith, M.L. Smith, and A.J. McCormick, A photometric stereo-based 3D imaging system using computer vision and deep learning for tracking plant growth, Gigascience. 8 (2019) 1-15. J.Z. Wang, Y. Zhang, and R.R. Gu, Research status and prospects on plant canopy structure measurement using visual sensors based on three-dimensional reconstruction, Agriculture-Basel. 10 (2020) 1-27. S. Wu, W. Wen, W. Gou, X. Lu, W. Zhang, C. Zheng, Z. Xiang, L. Chen, and X. Guo, A miniaturized phenotyping platform for individual plants using multi-view stereo 3D reconstruction, Frontiers in Plant Science. Volume 13 - 2022 (2022) 1-19. J.A. Gibbs, M.P. Pound, A.P. French, D.M. Wells, E.H. Murchie, and T.P. Pridmore, Active vision and surface reconstruction for 3D plant shoot modelling, IEEE/ACM Transactions on Computational Biology and Bioinformatics. 17 (2020) 1907-1917. D. Li, L. Xu, C. Tan, E.D. Goodman, D. Fu, and L. Xin, Digitization and visualization of greenhouse tomato plants in indoor environments, Sensors. 15 (2015) 4019-4051. Additional Declarations No competing interests reported. Cite Share Download PDF Status: Published Journal Publication published 09 Oct, 2025 Read the published version in Plant Methods → Version 1 posted Editorial decision: Revision requested 19 Aug, 2025 Reviews received at journal 19 Aug, 2025 Reviews received at journal 19 Aug, 2025 Reviews received at journal 18 Aug, 2025 Reviewers agreed at journal 15 Aug, 2025 Reviewers agreed at journal 14 Aug, 2025 Reviewers agreed at journal 10 Aug, 2025 Reviewers agreed at journal 10 Aug, 2025 Reviewers invited by journal 29 Jul, 2025 Editor assigned by journal 24 Jul, 2025 Submission checks completed at journal 23 Jul, 2025 First submitted to journal 23 Jul, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7178236","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":494388499,"identity":"4827da19-8ec2-438f-9e55-e4c14108bc16","order_by":0,"name":"Jiří Mach","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAo0lEQVRIiWNgGAWjYHACNoaECiAlQZqWMyAtzKRoYWwjRYtu++FnDx7OO2zPP7v/AMOPP0RoMTuTZm6QuO1w4ow7hxkYe3iI0XIgh00icVtagoFEMgMzUWFgdv4NUMucNHuIFgNitNwA2dJgw7gBrCWBKC3PzCQSjtkkzriRbHCw5wBRDkt+JvmjRsKef0biwwdEhRgKIMaOUTAKRsEoGAXEAABnUjIrgrudpwAAAABJRU5ErkJggg==","orcid":"","institution":"University of Chemistry and Technology","correspondingAuthor":true,"prefix":"","firstName":"Jiří","middleName":"","lastName":"Mach","suffix":""},{"id":494388501,"identity":"13a06377-ae34-49ea-a6db-fdf4a6e450e5","order_by":1,"name":"Zdeněk Svatý","email":"","orcid":"","institution":"Czech Technical University in Prague","correspondingAuthor":false,"prefix":"","firstName":"Zdeněk","middleName":"","lastName":"Svatý","suffix":""},{"id":494388503,"identity":"d062e818-f76b-4713-89ee-8575c1be4ac9","order_by":2,"name":"Ondřej Šoupa","email":"","orcid":"","institution":"University of Chemistry and Technology","correspondingAuthor":false,"prefix":"","firstName":"Ondřej","middleName":"","lastName":"Šoupa","suffix":""},{"id":494388506,"identity":"a4c4fdc7-32aa-492c-926c-6f155bcd5bbf","order_by":3,"name":"Luboš Nouzovský","email":"","orcid":"","institution":"Czech Technical University in Prague","correspondingAuthor":false,"prefix":"","firstName":"Luboš","middleName":"","lastName":"Nouzovský","suffix":""},{"id":494388508,"identity":"eabf5064-05ca-4b87-b3e9-52e940b900f4","order_by":4,"name":"Martin Halecký","email":"","orcid":"","institution":"University of Chemistry and Technology","correspondingAuthor":false,"prefix":"","firstName":"Martin","middleName":"","lastName":"Halecký","suffix":""}],"badges":[],"createdAt":"2025-07-21 13:53:35","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7178236/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7178236/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1186/s13007-025-01445-x","type":"published","date":"2025-10-09T15:57:22+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":88307623,"identity":"00b76e12-62f0-4ca0-8b1a-51af14e40fd4","added_by":"auto","created_at":"2025-08-05 06:12:21","extension":"jpg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":574669,"visible":true,"origin":"","legend":"\u003cp\u003eRepresentation of the scanning apparatus under development: A – The scanning photogrammetric apparatus consists of the following main components: a robotic arm, an RGB camera, a motorized turntable and additional LED illumination with a diffusion membrane; B – The illustrative CAD model; and C – The turntable surface with a coloured splatter pattern and 16 control points.\u003c/p\u003e","description":"","filename":"Picture1.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7178236/v1/d7f20888ce0b2c9a23d6e604.jpg"},{"id":88306442,"identity":"62f272f2-9d19-418f-8eb6-8b7813478458","added_by":"auto","created_at":"2025-08-05 06:04:21","extension":"jpg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":143803,"visible":true,"origin":"","legend":"\u003cp\u003eThe illustrative figure of one of the selected plants (\u003cem\u003eC. sativus\u003c/em\u003e) demonstrates the predefined image configuration, denoted as blue rectangles. The configuration comprised a total of 360 images, providing a comprehensive representation of the final image dataset. At each imaging height level, denoted P1-P6, 60 frames were captured.\u003c/p\u003e","description":"","filename":"Picture2.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7178236/v1/b9c0dabd7c8274f6d002f676.jpg"},{"id":88307864,"identity":"9bdb15f6-dc73-4c1d-8836-c3eee02c0560","added_by":"auto","created_at":"2025-08-05 06:20:21","extension":"jpg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":93333,"visible":true,"origin":"","legend":"\u003cp\u003eThe simplified diagram illustrates the automated data acquisition and evaluation process, followed by subsequent processing. The resulting output was a report that contained a description of the morphological features of the generated model. The key statistical indicators were also included.\u003c/p\u003e","description":"","filename":"Picture3.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7178236/v1/721cff147035e68b7357b01e.jpg"},{"id":88307624,"identity":"4d0c2463-5225-49ec-a488-07538c58aa7a","added_by":"auto","created_at":"2025-08-05 06:12:21","extension":"jpg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":142457,"visible":true,"origin":"","legend":"\u003cp\u003e*Processing of \u003cem\u003eL. sativa\u003c/em\u003e via the adaptive camera model fitting in combination with only the provided focal length and pixel size (\"No calib.\") resulted in partial failure of image alignment. Consequently, this configuration was excluded from the results\u003cem\u003e.\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eFigure 4. Comparison of the four camera calibration strategies across all the plant species. The displayed metrics include accuracy (MAE, RMSE), alignment quality (number of tie points – TP), root-mean-square reprojection error (RMS RE), and maximum reprojection error (Max RE). Accuracy was evaluated via coded markers on the turntable, with 8 GCPs and 8 CPs. The results with adaptive camera model fitting (ACMF) are also shown.\u003c/p\u003e","description":"","filename":"Picture4.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7178236/v1/e9325ec88eb12e1b15c53dbe.jpg"},{"id":88306446,"identity":"a4556f5d-fee2-4149-a173-97e1300e0fe1","added_by":"auto","created_at":"2025-08-05 06:04:21","extension":"jpg","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":452313,"visible":true,"origin":"","legend":"\u003cp\u003eThe following illustration illustrates the combinations of exposure times, tweaks, and plant species concerning the qualitative features of the generated 3D model, specifically its volume and surface. The selected tweak setting (bold text) and exposure time (dashed line) are highlighted in red.\u003c/p\u003e","description":"","filename":"Picture5.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7178236/v1/95b628e15c69b882c7ab1486.jpg"},{"id":88306448,"identity":"2cba712c-8a23-42b7-ab11-e5e98ee3c6cc","added_by":"auto","created_at":"2025-08-05 06:04:21","extension":"jpg","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":370210,"visible":true,"origin":"","legend":"\u003cp\u003eAn illustrative series of additional images represents the effect of tweak settings on the quality and integrity of the model. Each image isaccompanied by a black scale line with a length of 1 centimetre, which is visible in the lower right part. The numerical values are substituted for different tweak settings. Specifically, the number 1 represents the \"No Tweaks\" setting, the number 2 represents a value of 0.3, and the number 3 represents a value of 0.9. For illustrative purposes, a selection was made of only the three most representative tweak settings, including the one that was ultimately considered the most appropriate (0.9). The \u003cem\u003eC. sativus\u003c/em\u003e is designatedA, the \u003cem\u003eS. lycopersicum\u003c/em\u003e is designated B, and the \u003cem\u003eL. sativa\u003c/em\u003e is designatedC. To facilitate orientation in the diagram, red arrows are used to indicate areas in the model where holes and imperfections were observed.\u003c/p\u003e","description":"","filename":"Picture6.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7178236/v1/f4268f33ebe6233abceb49aa.jpg"},{"id":88307626,"identity":"5c59fbad-6541-491c-a67b-b9a064634437","added_by":"auto","created_at":"2025-08-05 06:12:21","extension":"jpg","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":148092,"visible":true,"origin":"","legend":"\u003cp\u003eThe resulting values representing the quality indicators of the models were then visualised for all the plant species depending on the camera-object distance. The red dashed line indicates the selected distance for further processing. This value has been determined through a comprehensive evaluation of all pertinent data and information (see later for details).\u003c/p\u003e","description":"","filename":"Picture7.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7178236/v1/ae06e2e5ca4f1874410daa18.jpg"},{"id":88307625,"identity":"d33f213e-19ff-4ce4-860c-aa8875972f6d","added_by":"auto","created_at":"2025-08-05 06:12:21","extension":"jpg","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":196011,"visible":true,"origin":"","legend":"\u003cp\u003eThe effects of the camera-object distance on the quality and integrity of the model are demonstrated via an illustrative series of auxiliary images. Each image is accompanied by a black scale line representing 1 centimetre, which is visible in the lower right quadrant. The numerical values thus represent specific distances. Specifically, the number 1 represents the shortest distance of 12 centimetres (similar features were also observed in the case of 14 centimetres and a combination of 12 and 16 centimetres), and the number 2 represents a distance of 16 centimetres (similar features were also observed at 18 centimetres). The \u003cem\u003eC. sativus\u003c/em\u003e was denoted as A, \u003cem\u003eS. lycopersicum\u003c/em\u003e as B, and \u003cem\u003eL. sativa\u003c/em\u003eas C. To facilitate orientation, red arrows were used to indicate areas in the model where artefacts and other imperfections were observed.\u003c/p\u003e","description":"","filename":"Picture8.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7178236/v1/a45dc1eb7997b9ec7f841157.jpg"},{"id":88306454,"identity":"f958b831-b194-4001-9f64-e80ce0931f1b","added_by":"auto","created_at":"2025-08-05 06:04:21","extension":"jpg","order_by":9,"title":"Figure 9","display":"","copyAsset":false,"role":"figure","size":166164,"visible":true,"origin":"","legend":"\u003cp\u003eThe resulting trends observed in the values of the key statistical indicators relative to camera-object distance for all the plant species under investigation.\u003c/p\u003e","description":"","filename":"Picture9.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7178236/v1/819832b30b62921b380f1e92.jpg"},{"id":88306455,"identity":"af205c30-c6d1-46bb-ab20-462ed3463398","added_by":"auto","created_at":"2025-08-05 06:04:21","extension":"jpg","order_by":10,"title":"Figure 10","display":"","copyAsset":false,"role":"figure","size":250738,"visible":true,"origin":"","legend":"\u003cp\u003eThe representation of the robustness of the photogrammetric approach in terms of the number of frames and the time required to generate a 3D model from the image data with respect to the quality of the model (expressed in terms of volume and surface area of the model).\u003c/p\u003e","description":"","filename":"Picture10.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7178236/v1/ff4bb93359d15babc4c2ece2.jpg"},{"id":88306452,"identity":"f5fde589-9ef2-43db-b853-aff254560606","added_by":"auto","created_at":"2025-08-05 06:04:21","extension":"jpg","order_by":11,"title":"Figure 11","display":"","copyAsset":false,"role":"figure","size":394633,"visible":true,"origin":"","legend":"\u003cp\u003eThe diagram presents the influence of the number of frames and image configuration on the resulting model quality. Each image is accompanied by a black scale line representing 1 centimetre, which is visible in the lower right quadrant. For illustration purposes, the models generated from the most relevant image datasets in terms of the optimisation process were selected. Specifically, these included 90 (1), 120* (2) and 120 (3) images. Red arrows denote imperfections, such as holes or redundant artefacts, that were created due to either an insufficient number of frames (1) or an improper image configuration (2). The red bold number indicates models that were generated using the lowest number of frames that still maintained sufficient model quality. \u003cem\u003eC. sativus\u003c/em\u003e is denoted as A, \u003cem\u003eS. lycopersicum\u003c/em\u003e is denoted B, and \u003cem\u003eL. sativa\u003c/em\u003e is denoted C. The lower portion of the diagram illustrates the image configuration of the image dataset in space for particular cases. To differentiate between the cases and the images above, Roman numerals are used to identify each. The dataset containing images from height levels P1, P3, and P5, where half of the images in each position were reduced, is designated I. The dataset comprising images from height levels P1 and P6, where there were 60 images per position, is labelled as II. Finally, the dataset containing images from height levels P1, P3, and P5, where a one-third reduction was implemented in each position, is designated III.\u003c/p\u003e","description":"","filename":"Picture11.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7178236/v1/fb77fa2b25b042b7b5e04be9.jpg"},{"id":88307627,"identity":"b750c6d9-ec1d-4a88-963e-8f3ebbd7b849","added_by":"auto","created_at":"2025-08-05 06:12:21","extension":"jpg","order_by":12,"title":"Figure 12","display":"","copyAsset":false,"role":"figure","size":264977,"visible":true,"origin":"","legend":"\u003cp\u003eThe resulting predominantly increasing trends in the values ​​of selected quality indicators relative to the number of processed images (for all plant species).\u003c/p\u003e","description":"","filename":"Picture12.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7178236/v1/0e49825986028ff01aa91ffd.jpg"},{"id":88306449,"identity":"09542d0c-f449-49b9-b5ea-44382810a014","added_by":"auto","created_at":"2025-08-05 06:04:21","extension":"jpg","order_by":13,"title":"Figure 13","display":"","copyAsset":false,"role":"figure","size":276663,"visible":true,"origin":"","legend":"\u003cp\u003eThe following visualisation employs a model with confidence levels for all the plant species. Each image is accompanied by a black scale line representing 1 centimetre, which is visible in the lower right quadrant. The confidence level associated with the points forming the generated model is expressed on a normalized scale ranging from 1 to 100%. The numeral 1 is indicative of the upper side of the plant, whereas the numeral 2 is indicative of the lower side. \u003cem\u003eC. sativus\u003c/em\u003e is designated A, \u003cem\u003eS. lycopersicum\u003c/em\u003e is designated B, and \u003cem\u003eL. sativa\u003c/em\u003e is designated C. The red arrow indicates a location where the degree of certainty is low, due to overlapping leaves and challenging camera access.\u003c/p\u003e","description":"","filename":"Picture13.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7178236/v1/2ed17c8ae034a7cc02e2a875.jpg"},{"id":88306456,"identity":"a291da90-de39-48f5-a371-eee423739adf","added_by":"auto","created_at":"2025-08-05 06:04:21","extension":"jpg","order_by":14,"title":"Figure 14","display":"","copyAsset":false,"role":"figure","size":138675,"visible":true,"origin":"","legend":"\u003cp\u003eIllustrative comparison of models demonstrating two distinct reconstruction approaches. Label 1 refers to gray models generated via the 3D scanning method (POP 3 scanner with RevoScan 5 software), whereas label 2 denotes purple models produced via the optimised photogrammetric workflow described above. Each image is accompanied by a black scale line representing 1 centimetre, which is visible in the lower right quadrant. \u003cem\u003eC. sativus\u003c/em\u003e is designated A, \u003cem\u003eS. lycopersicum\u003c/em\u003e is designated B, and \u003cem\u003eL. sativa\u003c/em\u003e is designated C. Red arrows highlight areas of imperfect reconstruction in thin structures of the \u003cem\u003eS. lycopersicum\u003c/em\u003e plant, particularly in the slender petioles of the leaves. The differing colours of the models result from the distinct colour palettes available in the respective evaluation software.\u003c/p\u003e","description":"","filename":"Picture14.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7178236/v1/4671eab197ee1ab6e1519f9c.jpg"},{"id":93419687,"identity":"d8a7987a-8406-44ad-b0af-e29524fcdfda","added_by":"auto","created_at":"2025-10-13 16:05:54","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":4556360,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7178236/v1/195255c9-0d17-44af-8176-310373941096.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Implementation of an SfM-MVS-based photogrammetry approach for detailed 3D reconstruction of plants","fulltext":[{"header":"Background","content":"\u003cp\u003eRecent advances in computational power, combined with the widespread availability of digital cameras, have driven a significant paradigm shift in photogrammetric techniques. This progress has enabled the development of advanced methods such as structure from motion (SfM) and multi-view stereo (MVS), which have been increasingly applied in precision agriculture, particularly for plant phenotyping. These techniques provide non-invasive and non-destructive means to accurately capture a broad range of plant traits through detailed 3D point cloud reconstruction. Moreover, they offer promising potential for the future automation of phenotyping workflows [1]. The methodology of precise plant phenotyping incorporates not only the analysis of plant vitality and prosperity (vegetation indices) but also the characteristics of the plant architecture, namely volume, area, biomass density along the main axis of the plant, leaf inclination, and configuration [2-4]. In contemporary scientific disciplines such as plant and plant pathology, a non-destructive approach to the study of morphological features has emerged as the prevailing standard [5-7]. The use of such a method is widely embraced within the scientific community, signifying its status as a contemporary paradigm in the scientific study of plants. The key indicators enumerated above are typically pivotal in the study of the interaction between a plant and an external factor, which may be an abiotic or biotic agent. As the global average temperature increases, pressure is placed on plant breeders to develop cultivars that are resistant to drought and other adverse soil conditions [8, 9]. These conditions can cause stress to plants, which can have a negative impact on their development and prosperity. High salt concentrations in the soil have also been demonstrated to be deleterious to proper metabolic processes within the plant body, as well as to the development of the plant body in general [10]. Conversely, factors that have been demonstrated or have the potential to exert a favourable influence on plant growth are worthy of mention. These biopreparations are either in conventional use or are undergoing development and laboratory testing. They are composed of cellular or cell-free mixtures of microorganisms and/or their biologically active metabolites. The function of biostimulants is to promote plant growth, either directly or indirectly. These preparations have the capacity to increase the bioavailability of crucial soil elements, such as phosphorus [11-13]. It is evident that microbial producers of auxin phytohormones also belong to this category of biopreparations, which have been demonstrated to exert a positive influence on plant metabolism [14]. The extensive list of biopreparations includes the widely employed Polyversum\u003csup\u003e®\u003c/sup\u003e, which is derived from the mycoparasitic oomycete \u003cem\u003ePythium oligandrum\u003c/em\u003e and is effective in the elimination of fungal diseases [15]. The development of potential biopreparations is frequently conducted within a laboratory setting, with subsequent testing being conducted directly on economically significant crops, including but not limited to barley, wheat, maize, rice, and soybean [16]. All the abiotic and biotic factors mentioned above, which exert a negative or positive influence on plants, have one thing in common, namely, the initiation of a morphological response of the plant body by their presence. In the following section, the various technological approaches currently employed or with potential for application are comprehensively delineated. The focus will be on non-destructive methods for investigating the effects of a factor on the morphological characteristics of the affected plant.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eIn accordance with the principle of spatial data acquisition, available 3D reconstruction technologies can be classified into two distinct groups: active and passive systems. The former rely on the emission of active radiation and the subsequent measurement of its reflection from the object surface. A terrestrial, mobile or airborne scanner emits a laser beam for the purpose of distance sensing [17-22]. Time-of-flight depth cameras utilize structured light to determine the distance to an object's surface by measuring the time delay between light emission and its reflection back to the sensor [23, 24]. High-precision X-ray tomography is a reliable method for scanning detailed objects; however, it is relatively expensive [3, 25]. The second group encompasses passive methods that do not rely on light emission. Instead, these methods utilize the ambient electromagnetic radiation that is naturally emitted by the sun or another artificial light source. Photogrammetry is a process that uses images captured from multiple angles; however, it is susceptible to variations in lighting conditions and the texture of the object in question [26].\u003c/p\u003e\n\u003cp\u003eIn the domain of photogrammetric analysis, two principal strategies are employed for the acquisition of images. The first method is based on static-object moving-camera systems, where the object remains stationary while the imaging device moves around it [27, 28]. Second, static-camera moving-object systems are considered, wherein the camera remains stationary while the object is rotated or translated in front of it [29]. The utilization of both configurations in close-range photogrammetry is contingent upon the specific experimental setup, the measured object properties and the spatial constraints inherent to the imaging environment. The combination of the aforementioned methods is indeed feasible. In such cases, the camera is typically affixed to the movable component [30, 31]. The control software deploys the arm in a series of predetermined positions to achieve the desired image configuration. At each position, the camera remains for a predefined period, during which a rotating object is sensed. The process subsequently transitions to the subsequent position, and this sequence is repeated. The acquired images are processed via the structure-from-motion (SfM) and multi-view stereo (MVS) algorithms [32, 33]. SfM is a process that facilitates the recovery of both camera poses and a sparse 3D point cloud. This is achieved by detecting and matching key features on the object surface (keypoints) across overlapping images. In the context of close-range photogrammetry, particularly in the field of plant science, the process of photo alignment is frequently used. It is often facilitated by the strategic placement of an object within the scene that exhibits a distinctive and unique multi-coloured texture. Examples of such objects include a multi-coloured cube, ball or plate [34]. Matched features, termed tie points, are utilized in a subsequent bundle adjustment procedure, which optimise both interior camera parameters and exterior orientations by minimising reprojection error. This stage, known as photo alignment, results in a geometrically consistent configuration of the image set and an initial sparse reconstruction of the scene. Following successful alignment, MVS algorithms are employed to densify the point cloud, thereby computing dense pixel-wise correspondences between images to generate a detailed and accurate surface model. This workflow is a common feature of commercial photogrammetric software, such as Pix4Dmapper, Metashape, ContexCapture or RealityCapture, which automates the majority of these steps while allowing for fine user control over parameters influencing reconstruction quality [35, 36]. When small-scale, highly detailed biological samples such as horticultural plants are targeted, it is imperative to employ high-resolution image acquisition techniques. This involves employing diffuse and uniform lighting to minimise shadows and reflections, applying background separation techniques such as chroma keying or the use of plain, textureless surfaces, and carefully calibrating both interior and exterior camera parameters to reduce distortion and enhance spatial accuracy.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAt present, a considerable number of research institutes are engaged in the study of close-range photogrammetry as applied to plant science. The focus of this field of study is the analysis of the aerial parts of plants. The objective is to generate the most detailed model, from which leaf position, orientation, and distribution can be accurately determined [37]. The leaf area index (LAI) is defined as the ground area covered by the plant canopy projected onto the soil surface. It serves as a critical qualitative and quantitative descriptor of morphological traits [38]. LAI plays a pivotal role in the optimisation of (bio)preparation dosages for plant treatments [39, 40]. Furthermore, the architecture of the plant root system can be investigated in detail [41-43]. To ensure the effective and consistent implementation of photogrammetric scanning systems, it is imperative to meticulously calibrate and set a number of parameters. These include the determination of the interior and exterior orientations of the cameras, the prevailing lighting conditions, the distance at which the scanning is conducted, and the evaluation pipeline employed for the generation of 3D models [1, 44-46]. For instance, the Perceptron v5 laser scanner is frequently employed as a reference device to evaluate the reliability, accuracy, and usability of photogrammetric systems [33, 47]. The primary aim of previously mentioned studies is to expand the portfolio of precise and cost-effective photogrammetric platforms suitable for high-throughput plant phenotyping applications. This study further contributes to this goal by developing and evaluating an additional low-cost photogrammetric platform.\u003c/p\u003e\n\u003cp\u003eThe objective of the present study is to extend the functionality of the existing multifunctional scanning apparatus [48]. The original methodology of plant reconstruction, which relies on the use of a 3D laser scanner, is to be superseded by a more appropriate technique (for this experimental setup). The latter will be subject to an SfM-MVS-based photogrammetric analysis. This will be carried out via an industrial RGB camera mounted on a robotic arm, a turntable, and additional LED illumination. The subsequent analysis of the acquired images will be undertaken with the objective of generating the most accurate 3D model. The subject of interest will be the calibration of the camera and the optimisation and appropriate adjustment of the scanning parameters (e.g., exposure time and scanning distance). The primary evaluation criterion will be the quality and detail of the generated model for each of the three plant species, as indicated by the total area and volume of the digital biomass. The selected statistical indicators of the quality of the spatial reconstruction are subsequently assessed. The apparatus should ideally be capable of producing an image of the plant that allows the creation of a comprehensive spatial model while preserving significant detail. These features will be of particular importance in future research, where the focus will be on a non-destructive way to determine the positive effect of a biocontrol agent or the negative effect of abiotic and biotic stressors on plant growth. Consequently, the accuracy and reliability of the scanning process and back reconstruction will assume a pivotal role in this ensuing phase of research.\u003c/p\u003e"},{"header":"Methods","content":"\u003cp\u003eThe subsequent chapter addresses the enhancement of the functionality of the multifunctional scanning apparatus that was previously developed and described. This enhancement is achieved through the application of the photogrammetric method, which allows for improved data acquisition and more detailed 3D reconstruction [48]. Specifically, it provides a comprehensive description of the necessary components, including a detailed description of the object to be scanned, the optimisation procedure within the photogrammetric analysis and computational complexity, and the method of evaluation and interpretation of the resulting data.\u003c/p\u003e\n\u003ch2\u003ePhotogrammetric system setup\u003c/h2\u003e\n\u003ch3\u003eApparatus components\u003c/h3\u003e\n\u003cp\u003eThe functionality of the multifunctional scanning apparatus was further enhanced by integrating the capability to conduct semi-automatic photogrammetric reconstruction of scanned samples, thereby broadening its range of applications. The enhancement was attributed to the addition of auxiliary illumination (see Figure 1 A and B). The lighting system consisted of two overhead 200 W LED reflectors and two supplementary 30 W side LED LEVE reflectors, providing a combined light intensity of approximately 500 \u0026mu;mol/m\u0026sup2;/s across the scene. With this intensity value, the surface of the object was evenly illuminated. Diffused and sufficient illumination was crucial for ensuring correct and accurate photogrammetric reconstruction. It also plays a key role in achieving a high-quality 3D model that preserves morphological details and structural features. The homogenization of object illumination was ensured by a photographic diffusion panel placed above the plant, with emphasis on minimising sharp shadows and uneven light exposure across the scene. Additionally, two undiffused 30 W LED panels provided the sharp side lighting necessary for accurate QR code scanning of the cuvette. The second significant component employed in the process was a MER2-1220-32U3C camera, which was equipped with an LCM-12MP-06MM-F2.4-1.7-ND1 lens (Daheng Imaging, China). The camera generated 24-bit color (RGB) images in JPEG format, which were subsequently used as input for photogrammetric processing to reconstruct 3D models. The technical specifications of the imaging system, including both the camera and lens, are summarized in Table 1. The camera was mounted on an AR4 robotic arm (Annin Robotics, USA) with 6 degrees of freedom to ensure precise positioning during imaging. To enable semi-automatic scanning, the scanned object was always placed in a custom-made holder attached to the turntable. Both of these components were printed from ABS filaments via the standard fused deposition modelling method with an i3 MK3S 3D printer (Original Prusa, Czech Republic). The turntable was driven by an NEMA17 stepper motor with a TMC2130 driver, and the control was mediated by an ArduinoUNO platform with an ATmega328 microcontroller The surface of the object holder and the top of the turntable were covered with a custom-made, randomly coloured splatter texture, forming a unique pattern specifically designed for this setup (see Figure 1 \u0026ndash; C). This improved the strength and quality of the alignment of the captured images during the 3D model generation procedure. In addition, sixteen control points (12-bit coded circular targets) were placed on the surface of the turntable, allowing control of the spatial reconstruction accuracy and definition of the scale and coordinate system of the resulting digital model. The original system [48], priced at 8,790 EUR (excluding the 3D scanner and dual-axis turntable), was upgraded by adding a camera and lens costing 388 EUR, bringing the total hardware cost to 9,178 EUR. However, this amount does not include the software license for Agisoft Metashape, which is priced at 540 EUR for the Professional Educational version (3,460 EUR for the full Professional (non-educational) edition). \u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTable 1. List of industrial RGB camera and lens technical parameters (taken from the manufacturer).\u003c/p\u003e\n\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\" width=\"576\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 350px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eCamera MER2-1220-32U3C parameters\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 227px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eDescription or value\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 350px;\"\u003e\n \u003cp\u003eCommunication interface\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 227px;\"\u003e\n \u003cp\u003eUSB3\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 350px;\"\u003e\n \u003cp\u003eResolution\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 227px;\"\u003e\n \u003cp\u003e4024 x 3036 (12.2 MPx)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 350px;\"\u003e\n \u003cp\u003eFrame rate\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 227px;\"\u003e\n \u003cp\u003e32 fps\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 350px;\"\u003e\n \u003cp\u003ePixel size\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 227px;\"\u003e\n \u003cp\u003e1.85 \u0026mu;m\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 350px;\"\u003e\n \u003cp\u003eSensor\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 227px;\"\u003e\n \u003cp\u003e1/1.7\u0026apos;\u0026apos; CMOS\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 350px;\"\u003e\n \u003cp\u003ePixel bit depth\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 227px;\"\u003e\n \u003cp\u003e8bit, 12bit\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 350px;\"\u003e\n \u003cp\u003ePixel data format\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 227px;\"\u003e\n \u003cp\u003eBayerRG8, BayerRG12\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 350px;\"\u003e\n \u003cp\u003eWeight\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 227px;\"\u003e\n \u003cp\u003e65 g\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 350px;\"\u003e\n \u003cp\u003eDimensions\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 227px;\"\u003e\n \u003cp\u003e29 x 29 x 29 mm\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 350px;\"\u003e\n \u003cp\u003ePrice*\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 227px;\"\u003e\n \u003cp\u003e252 EUR\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 350px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eLens LCM-12MP-06MM-F2.4-1.7-ND1 parameters\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 227px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 350px;\"\u003e\n \u003cp\u003eLens mount\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 227px;\"\u003e\n \u003cp\u003eC-mount\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 350px;\"\u003e\n \u003cp\u003eOptical resolution\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 227px;\"\u003e\n \u003cp\u003e12 Mpx\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 350px;\"\u003e\n \u003cp\u003eImage format\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 227px;\"\u003e\n \u003cp\u003e1/1.7\u0026apos;\u0026apos;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 350px;\"\u003e\n \u003cp\u003eFocal length\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 227px;\"\u003e\n \u003cp\u003e6 mm\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 350px;\"\u003e\n \u003cp\u003eIR corrected\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 227px;\"\u003e\n \u003cp\u003eNone\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 350px;\"\u003e\n \u003cp\u003eAperture (min)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 227px;\"\u003e\n \u003cp\u003eF2.4\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 350px;\"\u003e\n \u003cp\u003eIris\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 227px;\"\u003e\n \u003cp\u003eManual\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 350px;\"\u003e\n \u003cp\u003eWorking distance\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 227px;\"\u003e\n \u003cp\u003e100 mm - Infinity\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 350px;\"\u003e\n \u003cp\u003eLens dimensions\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 227px;\"\u003e\n \u003cp\u003e29.8 x 37.72 mm\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 350px;\"\u003e\n \u003cp\u003eDistortion\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 227px;\"\u003e\n \u003cp\u003e\u0026lt;0.05%\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 350px;\"\u003e\n \u003cp\u003ePrice*\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 227px;\"\u003e\n \u003cp\u003e136 EUR\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e* The rounded prices include VAT and correspond to the exchange rate of 18 July 2025 (24.625 CZK per 1 EUR, source: CNB in Prague).\u003c/p\u003e\n\u003ch3\u003eScanned objects\u003c/h3\u003e\n\u003cp\u003eThe following model plants were selected for optimisation and verification of the reliability of the developed photogrammetric approach: \u003cem\u003eCucumis sativus\u003c/em\u003e L. (cucumber), \u003cem\u003eSolanum lycopersicum\u003c/em\u003e L. (tomato), and \u003cem\u003eLactuca sativa\u003c/em\u003e L. var. \u003cem\u003ecapitata\u003c/em\u003e L. (lettuce). Henceforth, only Latin plant nomenclature will be utilized throughout the remainder of this text. The selection of these plants was driven by their distinctive morphological characteristics, namely their small, thin, and flat features, which serve as a suitable model for the development of an accurate and reliable photogrammetric method. Notably, all three of these plants are significant agricultural and food crops. The plants were cultivated under constant temperature conditions of 25 \u0026deg;C and a light intensity of 160 \u0026micro;mol/m\u003csup\u003e2\u003c/sup\u003e/s, characterized by a long photoperiod of 16 hours of light and 8 hours of darkness. A horticultural substrate was employed, which was placed in 30 ml PP Sterilin cuvettes with a hole in the cap. The use of the cuvette as a cultivation vessel was crucial, as the holder on the rotary turntable was specifically designed to accommodate it, thereby facilitating the handling of the plants. A further advantage of this configuration was the ease with which the camera could access the underside of the rosette leaves, thus facilitating the acquisition of the data necessary to generate a high-quality, full 3D model. Finally, it also enables the utilization of an automatic boundary definition for spatial reconstruction, lowering the computational and time demands. Scanning was conducted on three-week-old plants. \u0026nbsp;\u003c/p\u003e\n\u003ch3\u003eSpatial configuration of the scanning process\u003c/h3\u003e\n\u003cp\u003eThe scanning process was conducted via a static-camera moving-object system, with the camera sequentially positioned at six distinct height levels. The procedure was characterized by the use of discrete, robotically controlled viewpoints, enabling autonomous acquisition of multiple images. The movement of the robotic arm, which is predefined in the basic settings, ensures that the predefined image configuration is achieved (height, distance from the object, angle of the camera) at each of the six locations. During the image acquisition process, the camera was positioned at predefined height levels and held in place long enough to capture up to 60 images at each level (see Figure 2). Notably, the camera did not initiate capture during movement between positions. The scanning process was automatically initiated by a predefined position of the turntable with the object attached to the initial position. This was facilitated by the Hall sensor, which detected the presence of a magnet on the rotating top of the turntable. During imaging, the rotation speed was continuous, with an angular rotation speed of 6\u0026deg;/s (~ 0.105 rad/s). The speed was set to minimise any movement of the scanned object.\u0026nbsp;\u003c/p\u003e\n\u003ch3\u003eAcquisition and data processing \u0026nbsp;\u003c/h3\u003e\n\u003cp\u003eThe acquisition and processing of data was predominantly automated through the utilization of Python-based scripts. Prior to the initiation of the photogrammetric analysis, a Photogram3D program was executed to check the contents of the target folder to which the data were subsequently directed. The evaluation process was initiated automatically upon reaching a predefined number of images in the specified directory. The following section provides a more thorough description of these steps. The image acquisition, incorporating the control of the robotic arm movement, was also facilitated by Python scripts, while the motion of the turntable was managed by a script in Wiring. The timing of camera triggering, robotic arm movement and turntable rotation speed was automated by a wrapper control script. The initial step of the data acquisition process involved automatically scanning the QR code affixed to the surface of the cuvette once the object was correctly positioned and clearly visible to the camera. The code contains the identification data, i.e., the name of the sample and its abbreviation. According to the decoded information, a folder was created and named. The naming process incorporated a date, which was automatically generated by the script. The images were then stored in the designated folder. After a scan of a single object, the folder contained a maximum of 360 images in the dataset. The described process is illustrated in a simplified form in Figure 3. The complete scanning process was completed within a maximum timeframe of 8 minutes, and as part of the optimisation process, an attempt was made to reduce this time. To assess the reliability of the results, each object was subjected to three scanning cycles. The 3D model was generated via a custom-based application called Plant3D, which was developed as part of this study and features its own graphical user interface and console output. Plant3D was launched by a supervisory program (Photogram3D) at the beginning of the analysis to perform photogrammetric processing. The application utilized the imported Metashape module \u0026nbsp;(Agisoft Metashape API) for fully automatic spatial reconstruction [49]. The initial step of data processing involved alignment of the images through the structure-from-motion (SfM) algorithm, with adaptive fitting enabled and automatic filtration of the stationary points. As a part of the alignment phase, the parameters of the internal orientation were also determined. The subsequent step entailed the detection of 16 circular 12-bit control points and the automatic assignment of the corresponding coordinates. This step ensures the definition of the coordinate system and scale of the digital workspace and provides information about the resulting accuracy of the spatial reconstruction. Owing to the size of the samples, processing was performed in millimeters. Following the creation of a sparse point cloud consisting of the detected tie points, a predefined bounding box was automatically set, limiting the reconstruction area for further processing (scanned plant and upper part of the cuvette). This cropping was implemented to reduce the computational time, given that the focus of the study was not on the turntable model with a cuvette. The generation of depth maps was performed via the multi-view stereo (MVS) algorithm, which builds upon previously computed camera positions from SfM. This step was performed with the highest quality settings and medium-level filtering. Additionally, according to the developers\u0026apos; recommendations for the reconstruction of thin-walled structures, customized tweaks (\u0026quot;ooc_surface_blow_up\u0026quot;, \u0026quot;ooc_surface_blow_off\u0026quot;) were utilized [50]. The selection of a designated filtering mode results in the delicate elimination of extraneous points from a dense point cloud, while ensuring the preservation of the object\u0026apos;s finer details and structures. The actual generation of the plant spatial model, represented as a mesh, was performed through depth masks of the images. Post-processing of the generated model involved automatic partial surface smoothing (strength level 1), a combination of manual and partially automated removal of isolated patches from the surroundings, and the closure of holes, e.g., at the bottom of the plant stem. The generation of a report containing all descriptive parameters, e.g., the qualitative and statistical indicators, and the processing and model generation times, was initiated before the termination of the script. Upon completion of the entire evaluation process, a notification was dispatched. This approach resulted in substantial acceleration and streamlining of the evaluation process for many scanned objects. For the entirety of the optimisation procedures, only models without texture and colour information were generated, as the emphasis was exclusively on their morphological characteristics. Consequently, no requirement was placed on the calibration of colour hue or surface reflectance. The HP Z2 Workstation desktop computer (CPU: 13th Gen Intel Core i9, GPU: NVIDIA RTX A4000 16 GB, RAM: 32 GB) was utilized to process the image data and generate a 3D model of the object, owing to the substantial computational demands of photogrammetric reconstruction. All the data were stored and backed up on a Samsung SSD 870 QVO 8TB, with a secondary backup stored in the OneDrive cloud. In this work, all 3D models were visualised in a shaded, artificially coloured form.\u003c/p\u003e\n\u003ch2\u003ePhotogrammetric reconstruction optimisation\u003c/h2\u003e\n\u003ch3\u003eCamera calibration\u003c/h3\u003e\n\u003cp\u003eSince a non-metric camera with initially unknown interior orientation parameters was used for image acquisition, particular attention was given to assessing its impact on the precision of image alignment during processing. To mitigate potential instability in camera parameters over time, the effectiveness of pre-calibration functionality available in Agisoft Metashape was evaluated. Four calibration strategies were compared. The first used fixed interior orientation parameters obtained through pre-calibration, hereafter referred to as \u0026ldquo;Fixed\u0026rdquo;. The remaining three strategies involved simultaneous calibration during image alignment, with varying degrees of prior information. The second approach used pre-calibrated parameters as starting values but allowed adjustment during processing (\u0026ldquo;Simul.\u0026rdquo;). The third approach provided only basic physical characteristics of the camera-pixel size and focal length without performing pre-calibration (\u0026ldquo;No calib.\u0026rdquo;). The fourth strategy excluded all a priori information (\u0026ldquo;No info.\u0026rdquo;). Pre-calibration was conducted in Agisoft Metashape using 18 images of a black-and-white checkerboard pattern, generated by the software and placed at the same position as the scanned object. Importantly, the entire image frame was filled by the pattern and uniformly illuminated to ensure optimal results. The comparative evaluation was carried out on three samples for each plant species. The alignment quality was consistently set to the highest level, with no limit on the number of detected keypoints. Additionally, processing was conducted both with and without the adaptive camera model fitting (ACMF) option, a feature especially recommended for use with uncalibrated cameras, wide-angle lenses, or scenarios involving changes in camera orientation during image acquisition. Coded targets placed on a turntable were used for the assessment, with eight designated as control points and eight as check points. \u0026nbsp;\u003c/p\u003e\n\u003ch3\u003eExposition and tweak setting\u003c/h3\u003e\n\u003cp\u003eAs part of the optimisation of the imaging parameters, the following exposure times were chosen: 30, 40, 50, 60, and 70 milliseconds. For this and all subsequent experiments, the depth of field was set so that the scanned object was sharp in all images at all 6 different height levels. Scanning thin, small, and detailed objects is generally associated with the problem of producing non-compact, leaky models [50]. This problem can be solved by setting up advanced features called \u0026quot;tweaks\u0026quot;. On the basis of the recommendations of the developers of Agisoft Metashape software, advanced settings were applied during the \u0026quot;Build Model\u0026quot; phase, where the model was generated from depth maps. These parameters operate by expanding or contracting the surface contours, analogous to the morphological operations of dilation and erosion known from image processing, but were applied within a three-dimensional context. To find the best setting for each plant species, the following values were tested and evaluated: 0.95, 0.9, 0.7, 0.5, and 0.3.\u0026nbsp;\u003c/p\u003e\n\u003ch3\u003eObject and camera position\u003c/h3\u003e\n\u003cp\u003eThe scanned objects displayed both fine structural detail and morphological diversity, features characteristic of each plant species examined. Consequently, a series of measurements were carried out using varying object-to-camera distances to determine the optimal imaging distance between the camera and plant for each species. The selected distances for this purpose were 12, 14, 16, and 18 centimetres. Each distance was set for all six height levels (P1-P6). The lowest distance was predicted to provide the best detail capture, whereas the highest distance was expected to facilitate the alignment of images during model generation. The lowest distance could not be further reduced, as the minimum focusing distance of the lens is approximately 10 centimetres (plus a margin), according to the technical specifications. The highest distance was determined on the basis of the limitations of the workspace, size of the objects to be scanned and camera field-of-view. Finally, a combined configuration of 12 and 16 centimetres was also defined, with the shorter distance applied at positions P1, P3, and P5 and the longer distance at positions P2, P4, and P6.\u003c/p\u003e\n\u003ch2\u003eOptimisation of computational and scanning efficiency\u003c/h2\u003e\n\u003cp\u003eThe processing of a substantial quantity of the acquired images was both time-consuming and computationally intensive. Consequently, it was imperative to reduce the number of images while preserving the integrity of the 3D model generated. To this end, a reduction protocol was defined. The protocol\u0026apos;s fundamental components are delineated in Table 2. The symbol X on the label denotes the plant species. The green square indicates the images taken from a given height level (P1-P6). The marking of all positions (X_2) with a degree of reduction of -1/2 represents the fraction of the images used, specifically a situation in which every second image in each marked position was omitted. In other words, the total number of images (360) for a given object is reduced by a factor of two to 180. Notably, a similar procedure was followed in the remaining cases.\u003c/p\u003e\n\u003cp\u003eTable 2. Summary of image counts aimed at reducing scanning and processing time in 3D model generation.\u003c/p\u003e\n\u003cp\u003e\u003cimg src=\"data:image/png;base64,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\"\u003e\u003c/p\u003e\n\u003cp\u003eNote: \u0026quot;*\u0026quot; ~ Mark to distinguish duplicate values\u003c/p\u003e\n\u003ch2\u003eEvaluation of the suitability of parameter settings\u003c/h2\u003e\n\u003cp\u003eThe suitability of the chosen settings for the studied parameters defining the photogrammetric reconstruction was evaluated on the basis of several key criteria. The first set of criteria pertained to the qualitative characteristics of the generated mesh model, with a particular emphasis on surface integrity and volume accuracy. The second criterion involved the analysis of statistical indicators. The quality of photo alignment was assessed by qualitative indicators, including the number of tie points (TP), the root-mean-square reprojection error in pixels (RMS RE), the maximum reprojection error in millimetres (Max RE), and the average tie point multiplicity (AVG TP M). The evaluation metrics also included the differences between the known and estimated positions of the control and check points, namely the mean absolute error in millimetres (MAE), the root-mean-square error in millimetres (RMSE), and the standard deviation of the alignment error. In the subsequent stages of the work, all coded markers were treated as control points. Accordingly, the RMSE was referred to as the control point error (CPE) and was used as an indicator of the overall spatial precision reflecting the mean discrepancy between the measured and true coordinates of the 16 circular reference control points that defined the scale of the digital workspace. The quality and detail of the generated 3D model were assessed in terms of the total number of faces and vertices. Table 3 provides a detailed description of the individual statistical indicators. The statistical analysis was also accompanied by a manual image analysis of the model and served as an important concluding criterion in the overall evaluation. This analysis focused primarily on the presence of relics or irregularities on the model surface, imperfections within the model morphology, or the presence of holes even after adjustments by tweak settings had been made.\u003c/p\u003e\n\u003cp\u003eTable 3. Key indicators used in photogrammetry to assess 3D model quality and related processing steps.\u003c/p\u003e\n\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\" width=\"614\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eWorkflow stage\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 85px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eIndicator*\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 416px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eDescription\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"4\" valign=\"top\" style=\"width: 113px;\"\u003e\n \u003cp\u003ePhoto alignment\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003eTP\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 416px;\"\u003e\n \u003cp\u003eThe key points are automatically detected and matched between the overlapping images to work out the relative orientation. A higher number usually makes alignment more accurate.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003eRMS RE, RMSE\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 416px;\"\u003e\n \u003cp\u003eIt shows the average distance (error) between the reprojected tie point and its observed position in the image. Lower values are better.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003eMax RE\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 416px;\"\u003e\n \u003cp\u003eThe largest observed error in reprojecting tie points across all images. The identification of the worst-case alignment error\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003eAVG TP M\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 416px;\"\u003e\n \u003cp\u003eThe mean number of images in which each tie point is visible. Higher values indicate greater redundancy and, by extension, more robust alignment.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 113px;\"\u003e\n \u003cp\u003eCoordinate system accuracy\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003eCPE\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 416px;\"\u003e\n \u003cp\u003eThe error between the known and calculated coordinates of the control points indicates an error in the georeferencing of the model. This is indicative of the absolute accuracy of the model in real-world coordinates.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"4\" valign=\"top\" style=\"width: 113px;\"\u003e\n \u003cp\u003eModel reconstruction quality\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003eFaces\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 416px;\"\u003e\n \u003cp\u003eThe total number of polygonal faces (typically triangles) in the 3D mesh corresponds to a specific level of intricacy.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003eVertices\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 416px;\"\u003e\n \u003cp\u003eThe individual points in 3D space that define the shape of the mesh by forming the corners of faces (typically triangles or polygons) are known as vertices. An increased number of vertices is indicative of greater geometric complexity and detail.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003eSurface\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 416px;\"\u003e\n \u003cp\u003eThe term employed to denote the level of detail, resolution, and noise level of the reconstructed 3D surface. A high-quality surface is characterized by its intricate detail and absence of artefacts.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003eVolume\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 416px;\"\u003e\n \u003cp\u003eThe purpose of this indicator is to provide a quantitative measure of the accuracy of volumetric calculations derived from the model. The accuracy of the results is contingent not only on the correct geometry and scale but also, on the advanced settings of the 3D model generation procedure.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e* All parameters were computed via Agisoft Metashape and were part of the automatically generated processing report.\u003c/p\u003e\n\u003cp\u003eThe reliability and usability of the photogrammetric system were compared with those of equipment that included a previously tested POP 3 3D scanner [48]. To this end, an image analysis of the generated models was conducted. This step was important for evaluating the potential to expand the usability of the multifunctional equipment under development, with the aim of generating high-quality 3D models of plants and enabling non-destructive studies of their morphological traits.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003eIn the following chapter, the results are clearly presented, visualised and interpreted primarily through graphs and illustrative figures. This chapter offers an overview of the most suitable parameter configurations for optimised photogrammetric analysis, with particular emphasis on the qualitative attributes of the generated 3D model of the scanned object.\u003c/p\u003e\n\u003ch2\u003eCamera calibration\u003c/h2\u003e\n\u003cp\u003eThe alignment quality achieved through different camera calibration strategies is summarized in Figure 4. The results confirmed the anticipated influence of unknown and potentially unstable interior orientation parameters on the quality of spatial reconstruction. The highest accuracy, as measured by residuals at the control and check points, was obtained via simultaneous calibration accompanied by information about the interior parameters derived from pre-calibration. However, the most consistent reconstruction performance across repeated trials was observed with the simultaneous calibration approach using pre-calibrated values as initial estimates, underscoring the robustness of this second strategy. Although the specific effects of the adaptive camera model fitting (ACMF) option are not discussed in depth, the tests indicated a pronounced positive impact when sufficient prior information was available (\u0026quot;Simul.\u0026quot;) or when no prior information was provided (\u0026quot;No\u003cem\u003e\u0026nbsp;\u003c/em\u003einfo.\u0026quot;). In contrast, the strategy using only physical camera parameters, focal length and pixel size (\u0026quot;No calib.\u0026quot;), resulted in notably reduced accuracy, and in the case of \u003cem\u003eL. sativa\u003c/em\u003e, it led to partial image alignment failure. Despite the variability across the scenarios, adaptive fitting generally enhanced both the accuracy and geometric consistency of the reconstructions. As such, its use was considered advantageous and was applied in all subsequent processing steps. Ultimately, the simultaneous calibration approach incorporating pre-calibrated parameters of interior orientation and ACMF was selected for continued use in the photogrammetric workflow.\u003c/p\u003e\n\u003ch2\u003eExposure time and tweak setting\u003c/h2\u003e\n\u003cp\u003eThe resulting averaged descriptive values of the qualitative features, namely volume and surface area, of the generated 3D models were visualised for each plant species in Figure 5. The visualisation encompasses a full spectrum of predefined tweak settings and corresponding exposure times. The selected optimal settings are highlighted in red (see bold font and dashed lines). To enhance clarity, the volume and surface values have been separated due to their different order levels.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTweaks are not enabled by default in Metashape Agisoft. When scanning thin structures, holes and imperfections frequently occur in the generated models. The resulting data confirmed the positive impact of virtually any tweak setting on model quality. An increase in tweak values was generally associated with an improvement in the quality of the resulting models (see Figure 5). In cases where no tweaks were applied (\u0026quot;No Tweaks\u0026quot;), the negative impact was most pronounced on the surfaces of the \u003cem\u003eC. sativus\u003c/em\u003e and \u003cem\u003eS. lycopersicum\u003c/em\u003e plant models. Conversely, the volume values remained relatively unaltered by this configuration. The surface and volume of the \u003cem\u003eL. sativa\u003c/em\u003e models were found to be relatively resilient to the impact of this setting. When the adjustments were assigned to predefined values, the resulting trends presented marked similarity across all the plant species. An increase in tweak values led to a noticeable reduction in surface holes in the generated models. This improvement was largely attributed to the effectiveness of the applied tweaks, such as dilation and erosion, which successfully compensated for missing data by filling structural gaps. A slight thickening of some plant structures, particularly the leaves, was observed as a side effect of this process, leading to a modest overestimation of the total volume. Nonetheless, the advantages gained through this optimisation step outweighed the minor limitations. In the case of the volume values representing \u003cem\u003eL. sativa\u003c/em\u003e, a slightly decreasing trend was observed with increasing exposure time for all adjustments. This could be related to the lighting and resulting lower image quality. The exposure time within the selected test range, was not a significant indicator of the qualitative features of the generated model on the basis of the resulting data. For the subsequent steps of the optimisation process, the median value, specifically 50 milliseconds, was selected. At this value, the images were neither underexposed nor overexposed. The ambiguity of the effect of the tweak settings on the quality of the models precluded the selection of the most suitable option at this stage. Detailed analysis of the generated model images served as an important complementary evaluation of the correctness of both the tweak settings and the exposure time. This image-based assessment helped confirm the validity of the preliminary parameter configuration. As supporting evidence, a set of representative images was produced and is shown in Figure 6. A value of 0.9 was selected as the most effective tweak setting. At this value, no holes were present in the models, making the model complete and compact. A retrospective analysis of the data trends (see Figure 5) revealed that the presence of holes in the models had a negligible effect on the total volume and surface area. In summary, the segment of the model comprising holes was found to be inconsequential in relation to the total mass of the model. This final statement indicates the robustness and relative reliability of the introduced photogrammetric approach in relation to the quality of the generated model, independent of the test plant species.\u003c/p\u003e\n\u003ch2\u003eObject and camera position\u003c/h2\u003e\n\u003cp\u003eAnalysis of the surface data and related trends for the \u003cem\u003eC. sativus\u003c/em\u003e and \u003cem\u003eL. sativa\u003c/em\u003e models indicated that the object-to-camera distance did not significantly affect the results (see Figure 7). However, in the case of \u003cem\u003eS. lycopersicum\u003c/em\u003e, a certain dependency was observed. It is evident from the data that those lower distances, specifically 12, 14, and the combination of 12 and 16 centimetres, tend to overestimate the surface values. From a distance that was too close, even small, thin trichomes on the stem surface were detected, but during the reconstruction of the model, unrealistic, imperfect artefacts were generated, which contributed to a slight increase in the total surface area. With respect to the resulting volume values, no specific trend was observed for any plant species. Consequently, the distance between the object and the camera did not exert a substantial influence on this particular model quality indicator. To determine the optimal distance, supplementary image analysis of the models was conducted (see Figure 8). To provide a clearer context, the images highlight the previously mentioned imperfections identified in the generated models when suboptimal scanning distances were used.\u003c/p\u003e\n\u003cp\u003eThe results revealed that the RMS RE, Max RE, AVG TP M, and CPE values exhibited considerable consistency across all the plant species and demonstrated no discernible alterations in response to variations in distance from the object to the camera (see Figure 9). A statistical indicator was employed as an additional complementary measure to assess the influence of distance on model quality, further supporting these findings. The remaining indicators, including TP, faces, and vertices, supported the conclusions from the preceding assessment. Scanning at distances closer than 16 cm led to the generation of artefacts, which consequently caused a slight overestimation of these model quality metrics. Conversely, increasing the distance to 18 centimetres resulted in decreased counts of faces and vertices. This effect can be attributed to the reduced level of detail captured when scanning from distances exceeding 16 centimetres. The analysis of statistical indicators proved to be highly beneficial, providing a valuable complement to the comprehensive assessment of this optimisation step. On the basis of a holistic evaluation and the aforementioned assessment approaches, the optimum distance between the camera and the object for this particular experimental configuration was determined to be 16 centimetres. For scanning larger plants, this optimisation step would need to be repeated, as the optimal scanning distance is likely to increase.\u003c/p\u003e\n\u003ch2\u003eOptimisation of computational and scanning efficiency\u003c/h2\u003e\n\u003cp\u003eThe subsequent trend in the processed data demonstrated a positive correlation between the number of frames and the time required to process the data and generate a compact 3D model. In other words, an increase in the number of frames was associated with an increase in the time required (see Figure 10). The indicators of surface and volume were found to be largely unaffected by the decrease in the number of images, thereby indicating the significant robustness of the photogrammetric approach and the evaluation process. A substantial fluctuation in the trend was observed in the case of \u0026quot;120*\u0026quot;. \u0026nbsp;The observed fluctuation can be attributed to the failure to establish sufficient tie points between the two image sets, due to significant differences in camera positions and viewing angles. Consequently, the generation of a model with sufficient quality was rendered unfeasible, and the models contained the most discernible imperfections within this entire study (see Figure 11 \u0026ndash; A2, B2, C2). In the case of \u0026quot;180*\u0026quot;, an increase in the value of processing time was observed, which did not correspond with the resulting trend. However, no such increase was observed in the case of the values of qualitative features (surface, volume), which indicated that the overestimation was due to software data processing only.\u003c/p\u003e\n\u003cp\u003eThe outcome of this optimisation step was the identification of the minimum number of frames and image configuration required to produce a high-quality 3D model. From this perspective, the second lowest tested frame count was selected, specifically 120 images, with 40 images captured at each of the P1, P3, and P5 height levels. It was demonstrated that at this frame count and image configuration, the resulting model quality was practically equivalent to that of the model generated using the full set of 360 original frames from all height levels. The lowest number of images studied, 90 (P1, P3, P6 with 30 images in each height level), was deemed insufficient due to the presence of imperfections and redundant artefacts in the resulting models, particularly in the cases of \u003cem\u003eC. sativus\u0026nbsp;\u003c/em\u003eand \u003cem\u003eS. lycopersicum\u003c/em\u003e (see Figure 11 \u0026ndash; A1, B1). In the case of \u003cem\u003eL. sativa\u003c/em\u003e, the quality of the resulting models was relatively comparable to models generated from 120 images (see Figure 11. \u0026ndash; C1, C3). Notably, a substantial discrepancy was observed between 120* and 120. In both cases, the initial datasets contained 120 images but differed in image configuration and frame reduction (see Table 2 for a more detailed description). It can be concluded that reducing in the number of frames while preserving a greater number of positions was a more appropriate optimisation procedure than reducing the number of positions while preserving the full number of frames at a given position. This conclusion was based on the resulting quality of the models and is fully in line with the photogrammetric principles.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe resulting values of the crucial statistical indicators describing the quality of the generated 3D models corroborated the preceding claim. Specifically, it was determined that 120 images represent the minimum quantity required to generate a plant model of sufficient quality. A similar trend was observed in the values of all the indicators, namely an increase in their value with an increasing number of images for all the plant species (see Figure 12). As the number of frames increased, the software detected a greater number of features, which resulted in an increasing trend in the data for the TP, faces, and vertices. The inadequate degree of image linkage observed in the \u0026quot;120*\u0026quot; case was systematically reflected in all the indicators that were analysed. A decrease in the number of TP was also observed in the case of \u0026quot;180*\u0026quot;. It is conceivable that the reduced number of frames had an impact on this value. However, the other indicators represented in this case did not stand out significantly.\u003c/p\u003e\n\u003cp\u003eThe most significant outcome of this optimisation stage was the substantial reduction in processing time and computational complexity associated with image processing and model generation. Specifically, this resulted in an approximately 75% reduction across all the plant species. To further contextualise this finding, it is worth noting when the lowest tested number of images (90) was used, the time and energy savings would have reached approximately 82%. However, it should be noted that this would have come at the cost of the aforementioned adverse effects (holes, artefacts). Moreover, a less evident yet noteworthy enhancement pertains to the optimisation of the minimum number of frames utilised. It has been demonstrated that, by theoretically eliminating the need to scan up to 360 images per plant (a process that takes approximately 8 minutes), it would also be possible to reduce the scan time by a third, approximately 2.7 minutes. This would enhance the overall efficiency of the photogrammetric reconstruction. This optimization step substantially extended the applicability of the previously developed multifunctional robotic scanning apparatus [48].\u003c/p\u003e\n\u003ch2\u003eAssessment of plant morphology\u003c/h2\u003e\n\u003cp\u003eThe subsequent paragraph offers a synopsis of the distinctive characteristics of the morphology of the examined plants with respect to their scanability and the potential for generating a high-quality model.\u003c/p\u003e\n\u003cp\u003eThe plant species examined exhibited substantial variation in their morphological characteristics. The size and number of thin parts of the objects posed a challenge for photogrammetric reconstruction. The morphology of \u003cem\u003eC. sativus\u003c/em\u003e is characterized by the presence of a relatively thick stem and two types of leaves: cotyledon leaves with smooth edges and true lobed leaves with serrated edges. This specific body constitution was conducive to the scanning process, resulting in complete models devoid of unwanted artefacts. The morphology of \u003cem\u003eL. lycopersicum\u003c/em\u003e is characterized by the following traits: the presence of oval-shaped leaves with smooth edges, thin petioles, and a slender stem covered with imperfectly reconstructed thin trichomes. In the majority of cases, the resulting model was composed exclusively of swollen, redundant trichome artefacts. These artefacts slightly overestimate the values of the resulting qualitative indicators. The morphological traits of \u003cem\u003eL. sativa\u003c/em\u003e are characterized by the presence of a dense rosette of overlapping and twisted smooth, drop-shaped leaves with smooth edges. The primary issue encountered during the scanning process pertained to the presence of a dense rosette, which obscured the interior from the camera\u0026apos;s view, precluding effective penetration and thorough scanning. In this instance, it is plausible that the volume and surface area were also slightly overestimated.\u003c/p\u003e\n\u003cp\u003eThe impact of the morphological diversity of the plants under investigation on the quality of the model was the subject of additional evaluation. This was determined by evaluating the confidence level of the reconstruction of points from the original image data. This indicator is intended to express the degree of confidence in the precision and reliability of the reconstruction of the dense point cloud. The visualisation is typically exhibited on a standardised scale ranging from 1-100, which signifies the increasing degree of reliability of the reconstruction (see Figure 13). The image data were acquired with the photogrammetric system set to the optimal settings identified above. In the case of \u003cem\u003eC. sativus\u003c/em\u003e, areas on the underside of the cotyledon and true leaves with relatively low confidence levels (1\u0026ndash;20%) were observed. This phenomenon may have been caused by the shorter stem (2.5 centimetres) and drooping leaves, which reduced the handling space in the lower area of the plant and may have reduced the accessibility of the camera. The area formed by overlapping leaves was also more difficult to scan (see Figure 13 \u0026ndash; A1, red arrow). Conversely, in the case of \u003cem\u003eS. lycopersicum\u003c/em\u003e, the longer stem (5.2 centimetres) and leaf petioles formed a more spacious area that was more accessible to the camera, facilitating the scanning process and the acquisition of higher-quality data. In the case of \u003cem\u003eL. sativa\u003c/em\u003e, the absence of a stem and curled, drooping leaves made camera accessibility more difficult, which was reflected in the colour map of the model. Once more, there was a manifest decline in confidence level in these pivotal domains. Despite these parts, the majority of the generated dense point cloud points presented values that approximated 100\u0026nbsp;%.\u003c/p\u003e\n\u003cp\u003eIn conclusion, considering the financial investment in the scanning apparatus, the resulting models of the studied plant species proved to be of sufficient detail and quality for future follow-up research, demonstrating a cost-effective approach. A comprehensive scanning procedure was conducted on all the salient morphological traits of the plants, including the stems and leaves. This methodical approach ensured the reliable capture and documentation of plant body features, paving the way for subsequent analysis and research. The upgraded low-cost scanning apparatus, characterized by its proven robustness and reliability, is well suited for use in primary research.\u003c/p\u003e\n\u003cp\u003eFigure 14 shows a comparison of the resulting plant models, which were used to assess the usability of the optimised photogrammetric system. Compared with the previously tested POP 3 3D scanner, the image analysis demonstrated that the photogrammetric system was capable of generating complete and high-quality models. For the evaluation of data from the POP 3 3D scanner, RevoScan 5 software was used. Additionally, it is important to note, that it did not support scripting, thus, it would not have been suitable for the automated processing workflow implemented in this study. The models produced by 3D scanner were often incomplete, particularly due to difficulties in capturing the entire underside of plant leaves. As a result, it would not have been possible to generate reliable values for volume and surface area, or to determine the descriptive quantitative parameters. In the case of \u003cem\u003eS. lycopersicum\u003c/em\u003e, thin leaf petioles posed a significant challenge for scanning (Figure 14 \u0026ndash; B1). In contrast, the photogrammetric approach allowed for greater detail and completeness. Overall, the combination of an RGB camera and evaluation software based on the Metashape API proved to be a robust and user-friendly solution. This approach effectively enables the generation of high-quality 3D plant models and validates the photogrammetric system\u0026rsquo;s suitability for non-destructive morphological analysis. Compared with the previously used 3D scanner paired with Revopoint software, this approach offers significantly greater control over the 3D reconstruction algorithms. Such flexibility is typically limited in commercial 3D scanning systems, making the RGB camera and Metashape-based workflow a more adaptable and customizable options.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eNon-destructive methods of studying plant morphological features represent a novel and contemporary (bio)technological approach, whose reliability and accuracy are contingent on the appropriate composition and settings of the system. The type, accuracy, and price of the optical equipment used are of particular importance in this context. RGB cameras with CMOS sensors that offer sufficient resolution are commonly used for these purposes (including in this study) [51, 52]. To ensure the overall system’s reliability and accuracy, it is essential to carefully determine the parameters of both the interior and exterior camera orientations. Panels displaying a black-and-white checkerboard pattern or circular reference markers are utilised for this purpose [47, 53-56]. The geometry of the spatial orientation of the system is crucial for acquiring image data from different positions and angles. The high degree of variability of the resulting image dataset allows the generation of high-quality, complete 3D models of plants. As demonstrated in this study, the calibration strategy had a significant effect on the spatial reconstruction quality. The use of pre-calibrated interior parameters produced the highest accuracy, whereas simultaneous calibration with pre-calibrated values as initial estimates resulted in the most stable reconstruction performance across repeated trials. Furthermore, the application of adaptive camera model fitting generally improved the accuracy and consistency of the models, particularly when either no prior information or full calibration data were used. In contrast, the strategy relying solely on basic physical camera parameters, focal length and pixel size, led to reduced accuracy and even partial failure of image alignment in the case of \u003cem\u003eL. sativa\u003c/em\u003e. On the basis of these findings, the final configuration employed in this study combined simultaneous calibration with pre-calibrated parameters and adaptive model fitting. A notable advantage of the developed apparatus was the incorporation of a programmable, 6-axis AR4 robotic arm as a camera mount. This innovation enabled the precise positioning of the camera at any point in the workspace, thereby facilitating the definition of the complex spatial arrangement of the captured images (see Figure 2). The simplicity of adjusting and setting up this arrangement was an immense advantage of the established methodology. The ability of the established photogrammetric system to capture images from the underside of the scanned object represents a significant advantage, as this capability is not commonly supported by all systems. A configuration featuring fewer images combined with a greater number of positions proved to be more effective for spatial arrangement than one with fewer images and fewer positions. The optimal setup was identified as three height levels (P1, P3, and P5), each with 40 images taken at 10° spacing. A very similar photogrammetric system setup, including Agisoft Metashape software, was also employed for the reconstruction of detailed archaeological objects. The resulting models achieved sufficient accuracy for the study of morphological features [57]. Competitive devices frequently capture only the upper portion of the object, typically from two positions utilising stereo photogrammetry with two cameras [26, 55, 58-60]. Conversely, there are also systems capable of capturing high-quality images of plants from below to a certain extent [17, 61]. In general, the captured images should provide uniform coverage of the entire scanned object, with sufficient overlap in both the horizontal and vertical directions. Ideally, the angle between adjacent camera height levels should not exceed 45° [62]. In our case, however, a high-quality model was generated even with rather low overlap between images. This was made possible by the favourable spherical arrangement of the captured images, which provided uniform coverage of the scanned plant surface from nearly the full 360°. Another specific solution in photogrammetric system design involves mounting the camera on a tiltable bracket combined with a linear slider. This configuration also yielded high-fidelity models of the \u003cem\u003eS. lycopersicum\u003c/em\u003e plant placed on a turntable [63]. The acquisition of substantial image data from multiple vantage points within the workspace facilitates the generation of high-fidelity models, characterised by the preservation of authentic dimensions, a finding that is corroborated by the present study.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp;The quality of 3D reconstruction, as derived from image data, is significantly influenced by lighting conditions. The majority of reconstruction methods, whether geometric or machine learning-based, rely heavily on visual information obtained from pixel intensity, texture, shading, and colour consistency [64]. The selection of appropriate lighting, homogenization, and determination of the optimal exposure time can substantially increase the quality of image data intended for object reconstruction [65]. Ideally, the light should be homogenised. For these purposes, a diffuse photographic panel was utilised, which, however, slightly reduced the intensity of the incident light. This reduction was considered when the exposure time was set. The side LED panels were not homogenised with the intention of intensifying the illumination of the QR code and ensuring its correct reading. The most suitable exposure time of 50 milliseconds was selected for this scanning system. The selection was supported by image analysis of the generated plant models and analysis of chosen statistical indicators. The system demonstrated notable robustness concerning exposure time settings. Consequently, the quality of the resulting models remained largely unaffected by these settings. This finding is consistent with those of other studies [53, 55, 66]. Ensuring even illumination of the scanned object from all directions is widely considered the best practice, as it helps to avoid issues such as strong shadows, overexposure, or underexposure [67]. These effects were also observed to some extent in this study, where the use of underexposed or overexposed images resulted in minor, typically negligible, imperfections and the presence of undesired artefacts in the models. One potential solution to this issue could be the implementation of more thorough post-processing of the model, with the objective of removing these elements. The surface of the plants presented a high degree of reflectance, particularly in the case of \u003cem\u003eL. sativa\u003c/em\u003e. The presence of highly reflective surfaces has been shown to complicate 3D reconstruction [68]. In addition to altering the exposure time, the utilisation of matting agents has been posited as a potential solution [69, 70]. However, it is imperative to exercise caution to ensure that the experiment does not have a detrimental effect on plant growth, particularly in the context of studying time-dependent plant growth. In conclusion, close-range photogrammetry under controlled conditions generally results in a reduced degree of fluctuation in lighting conditions compared with outdoor solutions. When the system is transferred from the laboratory scale to the field scale, minimising the influence of variable lighting conditions on the 3D reconstruction of scanned plants. To ensure the robustness of the system and to facilitate the reproducibility and accuracy of the photogrammetric analysis, it would be necessary to make certain adaptations in the methodology.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp;During close-range photogrammetric spatial reconstruction of small and structurally complex objects, such as plant samples, several challenges were encountered in generating accurate 3D models. These included, for example, thin leaves, fine stems, petioles, or trichomes. The presence of these morphological structures was responsible for the holes, imperfections, and extraneous artefacts observed in the models generated in this study. This issue was attributed to the inability of the software to accurately ascertain the surface orientation of thin structures. This resulted in ambiguity when determining whether a particular point belonged to the upper or lower surface of an object. These points were then either wholly excluded from the reconstruction or incorrectly assigned to one of the sides of the leaf, resulting in local errors in the model [60, 71]. For all the tweak values that were tested, the most suitable value (0.9) was selected for all the plant types. At this value, the holes and imperfections in the model were effectively filled but without significantly overestimating the monitored quality indicators (see Figures 5, 6). An alternative approach to standard close-range photogrammetry is the use of multifocus stacking, a technique commonly applied in microphotogrammetry of small and structurally complex objects, such as insects or plant flowers [72, 73]. This method overcomes the depth of field limitations inherent in optical systems by merging multiple images acquired at different focal planes into a single composite image with improved focus throughout the object. However, the implementation of this method necessitates the acquisition of a substantially greater number of images, which has a negative effect on the time and computational complexity of photogrammetric analysis.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp;Another important factor affecting the outcome of close-range photogrammetric reconstruction is the distance between the object and the camera, as it directly impacts the image resolution, depth of field, and precision of feature reconstruction. Within a controlled laboratory environment, where stable conditions are maintained and a camera with adequate resolution and depth of field is employed, the range of values is typically extensive. This observation aligns with the findings of the present study. In the range from 14 to 16 centimetres from the centre of the plant, the models demonstrated a high degree of similarity in quality It was found that only the extreme values of this parameter, whether excessively low or high, had a discernible adverse effect on the quality of the generated model, mainly because of the lack of focus. The selection of 16 as the optimal value was based on a combination of image analysis outcomes and a detailed evaluation of relevant statistical indicators. A notable benefit of the photogrammetric system is its capacity to extend the scanning distance up to a maximum of 30-40 centimetres [48]. In the context of 3D reconstruction of larger plants, increasing the scanning distance is necessary. In certain systems, a greater distance is traversed during the scanning process of plants. Nevertheless, the quality of the resulting models is often found to be inferior [56]. If the utilisation of the developed apparatus was required for the scanning of a plant whose height exceeded the maximum adjustable value, it would be necessary to modify the design. Inspiration can be achieved from systems that scan tall \u003cem\u003eZea mays\u003c/em\u003e L. plants [28, 29]. A plant of such considerable height would be positioned on a rotating table, which would be situated on a vertically adjustable platform that would slide beneath the working area to the lower floor of the apparatus. In this configuration, it would be essential to precisely calibrate the movement of the positioning platform and robotic arm, the rotational speed of the turntable, and the scanning frequency. The proposed design modification would enhance the usability and robustness of the scanning apparatus under development.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp;The ability to precisely position the camera at a specific location and thus create geometrically unique spatial arrangements of images is a significant advantage of the described apparatus. This is also related to the relatively straightforward adjustment of the total number of images and thus the total scanning time. In conventional practice, images are typically captured from only two positions [3, 56]. This paper sets six height levels. During the optimisation process, the number of images was successfully reduced from the original 360 to 120 (40 images at each level P1, P3, and P5) while maintaining the original quality of the generated model. The developed apparatus was then compared with a similar solution in terms of accuracy and reliability. The utilisation of five distinct positions of the RGB camera relative to the \u003cem\u003eS. lycopersicum\u003c/em\u003e plant yielded models of inferior quality. The number of detected tie points was found to be significantly lower, ranging from 20000 to 150000. In contrast, our models exhibited a substantially greater number of tie points, ranging from 120000 to 300000 per model. Conversely, the RMS RE was higher, specifically in the range of 0.2-0.8, whereas in our case, the models were of higher quality with values in the range of 0.1-0.22 [74]. Furthermore, the reduction in the number of images necessitated a corresponding reduction in the time required to scan the entire plant, from an original 8 minutes to a mere 2.7 minutes. The time saved by this approach would enable the analysis of a larger number of plants within the same time. This optimisation step resulted in a substantial streamlining of photogrammetric analysis and an expansion of the applicability of the scanning apparatus under development. In subsequent experiments, it would be possible to analyse the influence of biotic and abiotic factors on plant morphological features, as this type of study would require a larger number of plants to ensure the statistical relevance of the results. The ability to scan both treated and untreated plants within a 24-hour period will facilitate a more effective and accurate comparison of these groups, thereby enabling more reliable detection of any potential differences.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp;The scanning photogrammetric apparatus described in this paper, when used with appropriately configured parameters, has expanded the capabilities of commonly employed laboratory-scale equipment [75, 76]. A competing system, comprising two static cameras and a turntable, was found to generate models of similar quality. Furthermore, the model of the generated root system could serve as an inspiration for the future enhancement of the existing apparatus [77]. However, certain apparatuses with this configuration incur significantly higher expenses and necessitate a greater spatial requirements [56]. The presence of a robotic 6-axis arm enabling precise camera positioning was also a feature of competing systems, which demonstrated an acceptable level of accuracy and reliability in 3D plant reconstruction [78]. In certain instances, emphasis has not been placed on comprehensive plant models but rather on specific segments, predominantly leaf models [79]. In such cases, the need to achieve detailed structures was not as pressing as it is in this paper. In this particular instance, the developed apparatus would constitute a suitable alternative, achieving a similar quality of output. The apparatus's autonomous control, the construction design, and modern, reliable components (namely, a robotic arm and a high-resolution camera) render the developed photogrammetric system suitable for deployment at the field scale in the context of precision agriculture. To achieve the desired outcome, it would be necessary to modify the design and readjust the parameters of the photogrammetric system so that it could compensate for negative disturbances from the external environment, such as lighting conditions, humidity, and air temperature.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eThis study aimed to develop a non-destructive and cost-effective method for analysing morphological traits in the phenotyping of economically significant plants. In particular, the feasibility of a close-range photogrammetry setup utilizing structure-from-motion and multi-view stereo (SfM-MVS) techniques was evaluated.\u0026nbsp;For this purpose, an existing robotic scanning system based on a static-camera and moving-object configuration was upgraded. An algorithm enabling automated camera positioning, data acquisition, subsequent processing, and model generation has been successfully implemented. The presence of the robotic arm and turntable makes the system robust and offers a variety of possibilities for setting up the spatial arrangement of photogrammetric reconstruction. A comprehensive evaluation of the influence of individual parameters on the quality of the resulting 3D models was conducted via qualitative and statistical indicators, including image analysis of the generated models. Among all the calibration strategies tested, the most robust and accurate results were achieved via simultaneous calibration with pre-calibrated interior orientation parameters combined with adaptive camera model fitting. This configuration significantly improved spatial reconstruction quality, whereas other strategies, particularly those based only on focal length and pixel size, resulted in reduced accuracy or led to partial alignment failure. These findings confirm the critical importance of the camera calibration strategy in ensuring reconstruction reliability, especially when non-metric cameras are used. The optimal settings were identical for all the plant species tested: \u003cem\u003eC. sativus\u003c/em\u003e, \u003cem\u003eS. lycopersicum\u003c/em\u003e and \u003cem\u003eL. sativa\u003c/em\u003e. The optimal exposure time was 50 milliseconds. The ideal scanning distance between the camera and the object was 16 centimetres. The best model generation results from depth maps were achieved when the tweak parameters were set to 0.9. In contrast, the worst models were generated when the tweak parameters were left unset. A configuration characterized by a reduced number of images combined with an increased number of height levels was found to contribute to the generation of higher-quality 3D models of plants compared with a configuration with a less robust image configuration and a greater number of images. The best combination was determined to be three height levels (P1, P3, and P5), with 40 images per position and approximately 10° spacing. The overall evaluation process was successfully streamlined, reducing the number of required images and cutting the total scan cycle time by approximately 75%, from 8 to just 2.7 minutes. The majority of points within the dense point cloud exhibited a confidence level that approached 100%. The introduced methodology enriches the list of available phenotyping platforms on the basis of photogrammetric analysis. The system's main advantages are its low acquisition cost, reliability, ease of use and sufficient accuracy. The ratio between the scanning time and the resulting model quality is highly favourable, indicating that efficient use of time does not compromise, but rather supports, the production of accurate and detailed 3D models. If the plant dimensions exceed the predefined size limit (approximately 30-40 centimetres), a simple structural modification would allow for an extension of the scanning capacity. In conclusion, one possible application of the developed apparatus is its use in the development of microbial or alternative biopreparations aimed at environmentally friendly protection of economically significant crops.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003eEthics approval and consent to participate\u003c/p\u003e\n\u003cp\u003eNot applicable: This manuscript does not include human or animal research.\u003c/p\u003e\n\u003cp\u003eConsent for publication\u003c/p\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e\n\u003cp\u003eAvailability of data and materials\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe datasets used and/or analysed during the current study are available from the corresponding author upon reasonable request.\u003c/p\u003e\n\u003cp\u003eCompeting interests\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.\u003c/p\u003e\n\u003cp\u003eFunding\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThis work was supported by the Internal Grant Agency of the University of Chemistry and Technology in Prague, Grant No. A1_FPBT_2025_005, Modern Biotechnologies.\u003c/p\u003e\n\u003cp\u003eAuthors\u0026apos; contributions\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eJM designed the study, developed the methodology, contributed to the experiments, and created the software tools. He also drafted and edited the manuscript. ZS co-developed the methodology, co-wrote the draft, designed and implemented the evaluation algorithm, worked on software, and supervised parts of the research. O\u0026Scaron; handled the data acquisition and curation, including plant cultivation. LN performed the formal data analysis, supported data curation, and contributed to manuscript editing. MH supervised the project, performed formal analysis, secured funding, and revised the manuscript. All the authors approved the final version and take responsibility for its content.\u003c/p\u003e\n\u003cp\u003eAcknowledgements\u003c/p\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n \u003cli\u003eJ. Hrzich, M. Beck, C. Bidinosti, C. Henry, K. Manawasinghe, and K. Tanino, A low-cost photogrammetry system for 3D plant modeling and phenotyping. 2025.\u003c/li\u003e\n \u003cli\u003eN. An, C.M. Palmer, R.L. Baker, R.J.C. Markelz, J. Ta, M.F. Covington, J.N. Maloof, S.M. Welch, and C. Weinig, Plant high-throughput phenotyping using photogrammetry and imaging techniques to measure leaf length and rosette area, Computers and Electronics in Agriculture. 127 (2016) 376-394.\u003c/li\u003e\n \u003cli\u003eY. Wang, W. Wen, S. Wu, C. Wang, Z. Yu, X. Guo, and C. Zhao, Maize plant phenotyping: Comparing 3D laser scanning, multi-view stereo reconstruction, and 3D digitizing estimates, Remote Sensing. 11 (2019) 1-17.\u003c/li\u003e\n \u003cli\u003eM.M. Rahaman, D. Chen, Z. Gillani, C. Klukas, and M. Chen, Advanced phenotyping and phenotype data analysis for the study of plant growth and development, Frontiers in Plant Science. 6 (2015) 1-15.\u003c/li\u003e\n \u003cli\u003eZ. Rui, Z. Zhang, M. Zhang, A. Azizi, C. Igathinathane, H. Cen, S. Vougioukas, H. Li, J. Zhang, Y. Jiang, X. Jiao, M. Wang, Y. Ampatzidis, O.I. Oladele, M. Ghasemi-Varnamkhasti, and R. Radi, High-throughput proximal ground crop phenotyping systems - A comprehensive review, Computers and Electronics in Agriculture. 224 (2024) 1-22.\u003c/li\u003e\n \u003cli\u003eC.Q. Xie and C. Yang, A review on plant high-throughput phenotyping traits using UAV-based sensors, Computers and Electronics in Agriculture. 178 (2020) 1-14.\u003c/li\u003e\n \u003cli\u003eS. Dhondt, N. Wuyts, and D. Inz\u0026eacute;, Cell to whole-plant phenotyping: The best is yet to come, Trends in Plant Science. 18 (2013) 433-444.\u003c/li\u003e\n \u003cli\u003eM.S.M. Asaari, S. Mertens, L. Verbraeken, S. Dhondt, D. Inz\u0026eacute;, K. Bikram, and P. Scheunders, Non-destructive analysis of plant physiological traits using hyperspectral imaging: A case study on drought stress, Computers and Electronics in Agriculture. 195 (2022) 1-13.\u003c/li\u003e\n \u003cli\u003eL. Cattivelli, F. Rizza, F.W. Badeck, E. Mazzucotelli, A.M. Mastrangelo, E. Francia, C. Mare, A. Tondelli, and A.M. Stanca, Drought tolerance improvement in crop plants: An integrated view from breeding to genomics, Field Crops Research. 105 (2008) 1-14.\u003c/li\u003e\n \u003cli\u003eM. Awlia, A. Nigro, J. Faikus, S.M. Schmoeckel, S. Negrao, D. Santelia, M. Trt\u0026iacute;lek, M. Tester, M.M. Julkowska, and K. Panzarov\u0026aacute;, High-throughput non-destructive phenotyping of traits that contribute to salinity tolerance in \u003cem\u003eArabidopsis thaliana\u003c/em\u003e, Frontiers in Plant Science. 7 (2016) 1-15.\u003c/li\u003e\n \u003cli\u003eO. Ali, A. Ramsubhag, and J. Jayaraman, Biostimulant properties of seaweed extracts in plants: Implications towards sustainable crop production, Plants. 10 (2021) 1-27.\u003c/li\u003e\n \u003cli\u003eS.S. Mahusook, F.A.A. Rajathi, H.N.S. Maharifa, and R. Sharmila, Comparative study of agarophytes - \u003cem\u003eGracilaria edulis\u0026nbsp;\u003c/em\u003eand \u003cem\u003eGelidiella acerosa\u0026nbsp;\u003c/em\u003eas biostimulant and application of agar for water-holding in soil and plant growth promotion, Agricultural Science Digest - A Research Journal. 41 (2021) 21-27.\u003c/li\u003e\n \u003cli\u003eY.C. Wu, J.Y. Zhou, C.G. Li, and Y. Ma, Antifungal and plant growth promotion activity of volatile organic compounds produced by \u003cem\u003eBacillus amyloliquefaciens\u003c/em\u003e, Microbiologyopen. 8 (2019) 1-14.\u003c/li\u003e\n \u003cli\u003eL. Pappalettere, S. Bartolini, and A. Toffanin, Auxin-producing bacteria used as microbial biostimulants improve the growth of tomato (\u003cem\u003eSolanum lycopersicum\u0026nbsp;\u003c/em\u003eL.) seedlings in hydroponic systems, Biotech. 13 (2024) 1-14.\u003c/li\u003e\n \u003cli\u003eL. Pellan, C.A.T. Dieye, N. Durand, A. Fontana, C. Strub, and S. Schorr-Galindo, Biocontrol agents: Toolbox for the screening of weapons against mycotoxigenic \u003cem\u003eFusarium\u003c/em\u003e, Journal of Fungi. 7 (2021) 1-29.\u003c/li\u003e\n \u003cli\u003eC.N. Sible, J.R. Seebauer, and F.E. Below, Plant biostimulants: A categorical review, their implications for row crop production, and relation to soil health indicators, Agronomy. 11 (2021) 1-20.\u003c/li\u003e\n \u003cli\u003eS. Paulus, H. Schumann, H. Kuhlmann, and J. Leon, High-precision laser scanning system for capturing 3D plant architecture and analysing growth of cereal plants, Biosystems Engineering. 121 (2014) 1-11.\u003c/li\u003e\n \u003cli\u003eM. Friedli, N. Kirchgessner, C. Grieder, F. Liebisch, M. Mannale, and A. Walter, Terrestrial 3D laser scanning to track the increase in canopy height of both monocot and dicot crop species under field conditions, Plant Methods. 12 (2016) 1-15.\u003c/li\u003e\n \u003cli\u003eC. Todo, H. Ikeno, K. Yamase, T. Tanikawa, M. Ohashi, M. Dannoura, T. Kimura, and Y. Hirano, Reconstruction of conifer root systems mapped with point cloud data obtained by 3D laser scanning compared with manual measurement, Forests. 12 (2021) 1-17.\u003c/li\u003e\n \u003cli\u003eS. Puliti, J. Breidenbach, and R. Astrup, Estimation of forest growing stock volume with UAV laser scanning data: Can it be done without field data?, Remote Sensing. 12 (2020) 1-19.\u003c/li\u003e\n \u003cli\u003eT. Dornbusch, S. Lorrain, D. Kuznetsov, A. Fortier, R. Liechti, I. Xenarios, and C. Fankhauser, Measuring the diurnal pattern of leaf hyponasty and growth in \u003cem\u003eArabidopsis\u003c/em\u003e - a novel phenotyping approach using laser scanning, Functional Plant Biology. 39 (2012) 860-869.\u003c/li\u003e\n \u003cli\u003eA. \u0026Scaron;rollerů and M. Potůčkov\u0026aacute;, Evaluating the applicability of high-density UAV LiDAR data for monitoring tundra grassland vegetation, International Journal of Remote Sensing. 46 (2024) 1-36.\u003c/li\u003e\n \u003cli\u003eZ.Y. Ma, R. Rayhana, K. Feng, Z. Liu, G.Z. Xiao, Y.F. Ruan, and J.S. Sangha, A review on sensing technologies for high-throughput plant phenotyping, Ieee Open Journal of Instrumentation and Measurement. 1 (2022) 1-22.\u003c/li\u003e\n \u003cli\u003eW. Kazmi, S. Foix, and G. Aleny\u0026agrave;. Plant leaf imaging using time of flight camera under sunlight, shadow and room conditions. in 2012 IEEE International Symposium on Robotic and Sensors Environments Proceedings. 2012.\u003c/li\u003e\n \u003cli\u003eS. Mairhofer, S. Zappala, S.R. Tracy, C. Sturrock, M. Bennett, S.J. Mooney, and T. Pridmore, Rootrak: Automated recovery of three-dimensional plant root architecture in soil from X-ray microcomputed tomography images using visual tracking, Plant Physiology. 158 (2012) 561-569.\u003c/li\u003e\n \u003cli\u003eD. And\u0026uacute;jar, M. Calle, C. Fern\u0026aacute;ndez-Quintanilla, \u0026Aacute;. Ribeiro, and J. Dorado, Three-dimensional modeling of weed plants using low-cost photogrammetry, Sensors. 18 (2018) 1-11.\u003c/li\u003e\n \u003cli\u003eS. Paulus, Measuring crops in 3D: using geometry for plant phenotyping, Plant Methods. 15 (2019) 1-13.\u003c/li\u003e\n \u003cli\u003eB. Daviet, R. Fernandez, L. Cabrera-Bosquet, C. Pradal, and C. Fournier, Phenotrack3D: An automatic high-throughput phenotyping pipeline to track maize organs over time, Plant Methods. 18 (2022) 1-14.\u003c/li\u003e\n \u003cli\u003eS. Wu, W. Wen, Y. Wang, J. Fan, C. Wang, W. Gou, and X. Guo, MVS-pheno: A portable and low-cost phenotyping platform for maize shoots using multiview stereo 3D reconstruction, Plant Phenomics. 2020 (2020) 1-17.\u003c/li\u003e\n \u003cli\u003eR. Rossi, C. Leolini, S. Costafreda-Aumedes, L. Leolini, M. Bindi, A. Zaldei, and M. Moriondo, Performances evaluation of a low-cost platform for high-resolution plant phenotyping, Sensors. 20 (2020) 1-16.\u003c/li\u003e\n \u003cli\u003eT. Luhmann, Precision potential of photogrammetric 6DOF pose estimation with a single camera, Isprs Journal of Photogrammetry and Remote Sensing. 64 (2009) 275-284.\u003c/li\u003e\n \u003cli\u003eW. He, Z.H. Ye, M.S. Li, Y.L. Yan, W. Lu, and G.N. Xing, Extraction of soybean plant trait parameters based on SfM-MVS algorithm combined with GRNN, Frontiers in Plant Science. 14 (2023) 1-16.\u003c/li\u003e\n \u003cli\u003eJ.C. Rose, S. Paulus, and H. Kuhlmann, Accuracy analysis of a multi-view stereo approach for phenotyping of tomato plants at the organ level, Sensors. 15 (2015) 9651-9665.\u003c/li\u003e\n \u003cli\u003eW. Wen, S. Wu, X. Lu, X. Liu, S. Gu, and X. Guo, Accurate and semantic 3D reconstruction of maize leaves, Computers and Electronics in Agriculture. 217 (2024) 1-11.\u003c/li\u003e\n \u003cli\u003eS. Krasic and P. Pejic, Comparative analysis of terrestrial semi-automatic and automatic photogrammetry in 3D modeling process, Nexus Network Journal. 16 (2014) 273-283.\u003c/li\u003e\n \u003cli\u003eF.T. Teshome, H.K. Bayabil, G. Hoogenboom, B. Schaffer, A. Singh, and Y. Ampatzidis, Unmanned aerial vehicle (UAV) imaging and machine learning applications for plant phenotyping, Computers and Electronics in Agriculture. 212 (2023) 1-15.\u003c/li\u003e\n \u003cli\u003eN. Ivanov, P. Boissard, M. Chapron, and B. Andrieu, Computer stereo plotting for 3D reconstruction of a maize canopy, Agricultural and Forest Meteorology. 75 (1995) 85-102.\u003c/li\u003e\n \u003cli\u003eM.A. Aguilar, J.L. Pozo, F.J. Aguilar, A.M. Garc\u0026iacute;a, I. Fern\u0026aacute;ndez, J. Negreiros, and J. S\u0026aacute;nchez-Hermosilla. Application of close-range photogrammetry and digital photography analysis for the estimation of leaf area index in a greenhouse tomato culture. in ISPRS-Commission V Mid-Term Symposium on Close Range Image Measurement Technique. 2010. Newcastle upon Tyne, ENGLAND.\u003c/li\u003e\n \u003cli\u003eH. Zhu, J. Dorner, D.L. Rowland, R.C. Derksen, and H.E. Ozkan, Spray penetration into peanut canopies with hydraulic nozzle tips, Biosystems Engineering. 87 (2004) 275-283.\u003c/li\u003e\n \u003cli\u003eY.K. Rui, Y.F. Peng, Z.R. Wang, and J.B. Shen, Stem perimeter, height and biomass of maize (\u003cem\u003eZea mays\u003c/em\u003e L.) grown under different N fertilization regimes in Beijing, China, International Journal of Plant Production. 3 (2009) 85-90.\u003c/li\u003e\n \u003cli\u003eS.Q. Fang, X.L. Yan, and H. Liao, 3D reconstruction and dynamic modeling of root architecture in situ and its application to crop phosphorus research, Plant Journal. 60 (2009) 1096-1108.\u003c/li\u003e\n \u003cli\u003eL.D. Lopez, D. Shantharaj, L. Liu, H. Bais, and J. Yu, Robust image-based 3D modeling of root architecture. 1 (2011) 1-10.\u003c/li\u003e\n \u003cli\u003eJ. Vanderborght, V. Couvreur, F. Meunier, A. Schnepf, H. Vereecken, M. Bouda, and M. Javaux, From hydraulic root architecture models to macroscopic representations of root hydraulics in soil water flow and land surface models, Hydrology and Earth System Sciences. 25 (2021) 4835-4860.\u003c/li\u003e\n \u003cli\u003eA. Junker, M.M. Muraya, K. Weigelt-Fischer, F. Arana-Ceballos, C. Klukas, A.E. Melchinger, R.C. Meyer, D. Riewe, and T. Altmann, Optimizing experimental procedures for quantitative evaluation of crop plant performance in high throughput phenotyping systems, Frontiers in Plant Science. 5 (2015) 1-21.\u003c/li\u003e\n \u003cli\u003eW. Gu, W. Wen, S. Wu, C. Zheng, X. Lu, W. Chang, P. Xiao, and X. Guo, 3D reconstruction of wheat plants by integrating point cloud data and virtual design optimization, Agriculture. 14 (2024) 1-20.\u003c/li\u003e\n \u003cli\u003eJ. Martinez-Guanter, \u0026Aacute;. Ribeiro, G.G. Peteinatos, M. P\u0026eacute;rez-Ruiz, R. Gerhards, J.M. Bengochea-Guevara, J. Machleb, and D. And\u0026uacute;jar, Low-cost three-dimensional modeling of crop plants, Sensors. 19 (2019) 1-14.\u003c/li\u003e\n \u003cli\u003eS. Paulus, J. Behmann, A.K. Mahlein, L. Plumer, and H. Kuhlmann, Low-cost 3D systems: Suitable tools for plant phenotyping, Sensors. 14 (2014) 3001-3018.\u003c/li\u003e\n \u003cli\u003eJ. Mach, L. Krauz, P. P\u0026aacute;ta, and M. Halecky, Development of low-cost multifunctional robotic apparatus for high-throughput plant phenotyping, Smart Agricultural Technology. 9 (2024) 1-17.\u003c/li\u003e\n \u003cli\u003eMetashape Python Reference Release 2.2.0. 2024; Available from: https://www.agisoft.com/pdf/metashape_python_api_2_2_0.pdf.\u003c/li\u003e\n \u003cli\u003eHow to improve the modeling of thin objects in Metashape. 2024; Available from: https://agisoft.freshdesk.com/support/solutions/articles/31000163220-how-to-improve-the-modeling-of-thin-objects-in-metashape.\u003c/li\u003e\n \u003cli\u003eX. Lu, E. Ono, S. Lu, Y. Zhang, P. Teng, M. Aono, Y. Shimizu, F. Hosoi, and K. Omasa, Reconstruction method and optimum range of camera-shooting angle for 3D plant modeling using a multi-camera photography system, Plant Methods. 16 (2020) 1-16.\u003c/li\u003e\n \u003cli\u003eX. Teng, G. Zhou, Y. Wu, C. Huang, W. Dong, and S. Xu, Three-dimensional reconstruction method of rapeseed plants in the whole growth period using an RGB-D camera, Sensors. 21 (2021) 1-20.\u003c/li\u003e\n \u003cli\u003eK. Bobkowska, P. Burdziakowski, J. Szulwic, and K.M. Zielinska-Dabkowska, Seven different lighting conditions in photogrammetric studies of a 3D urban mock-up, Energies. 14 (2021) 1-27.\u003c/li\u003e\n \u003cli\u003eJ. Kolecki, P. Kuras, E. Pastucha, K. Pyka, and M. Sierka, Calibration of industrial cameras for aerial photogrammetric mapping, Remote Sensing. 12 (2020) 1-21.\u003c/li\u003e\n \u003cli\u003eD. Li, L. Xu, X.-s. Tang, S. Sun, X. Cai, and P. Zhang, 3D imaging of greenhouse plants with an inexpensive binocular stereo vision system, Remote Sensing. 9 (2017) 1-27.\u003c/li\u003e\n \u003cli\u003eT.T. Nguyen, D.C. Slaughter, N. Max, J.N. Maloof, and N. Sinha, Structured light-based 3D reconstruction system for plants, Sensors. 15 (2015) 18587-18612.\u003c/li\u003e\n \u003cli\u003eM. Pepe and D. Costantino, Techniques, tools, platforms, and algorithms in close range photogrammetry in building 3D model and 2D representation of objects and complex architectures, Computer-Aided Design and Applications. 18 (2020) 42-65.\u003c/li\u003e\n \u003cli\u003eN. An, S.M. Welch, R.J.C. Markelz, R.L. Baker, C.M. Palmer, J. Ta, J.N. Maloof, and C. Weinig, Quantifying time-series of leaf morphology using 2D and 3D photogrammetry methods for high-throughput plant phenotyping, Computers and Electronics in Agriculture. 135 (2017) 222-232.\u003c/li\u003e\n \u003cli\u003eS. Dandrifosse, A. Bouvry, V. Leemans, B. Dumont, and B. Mercatoris, Imaging wheat canopy through stereo vision: Overcoming the challenges of the laboratory to field transition for morphological features extraction, Frontiers in Plant Science. 11 (2020) 1-15.\u003c/li\u003e\n \u003cli\u003eH. Surmen, Photogrammetry for 3D reconstruction of objects: Effects of geometry, texture and photographing, Image Analysis \u0026amp; Stereology. 42 (2023) 51-63.\u003c/li\u003e\n \u003cli\u003eA. Conn, U.V. Pedmale, J. Chory, and S. Navlakha, High-resolution laser scanning reveals plant architectures that reflect universal network design principles, Cell Systems. 5 (2017) 53-65.\u003c/li\u003e\n \u003cli\u003eA. Bisson-Larriv\u0026eacute;e and J.-B. LeMoine, Photogrammetry and the impact of camera placement and angular intervals between images on model reconstruction, Digital Applications in Archaeology and Cultural Heritage. 26 (2022) 1-12.\u003c/li\u003e\n \u003cli\u003eR. Rossi, S. Costafreda-Aumedes, L. Leolini, C. Leolini, M. Bindi, and M. Moriondo, Implementation of an algorithm for automated phenotyping through plant 3D-modeling: A practical application on the early detection of water stress, Computers and Electronics in Agriculture. 197 (2022) 1-15.\u003c/li\u003e\n \u003cli\u003eD. Rangelov, S. Waanders, K. Waanders, M.v. Keulen, and R. Miltchev. 3D reconstruction techniques and the impact of lighting conditions on reconstruction quality: A comprehensive review. in Preprints. 2025. Preprints.\u003c/li\u003e\n \u003cli\u003eK. Wu, J. Tan, and C. Liu, A novel approach to obtain optimal exposure for 3D shape reconstruction of high dynamic range objects, Measurement Science and Technology. 32 (2021) 1-16.\u003c/li\u003e\n \u003cli\u003eB. Biskup, H. Scharr, U. Schurr, and U. Rascher, A stereo imaging system for measuring structural parameters of plant canopies, Plant Cell and Environment. 30 (2007) 1299-1308.\u003c/li\u003e\n \u003cli\u003eM. Tancik, V. Casser, X. Yan, S. Pradhan, B. Mildenhall, P. Srinivasan, J. Barron, and H. Kretzschmar, Block-nerf: Scalable large scene neural view synthesis. 2022. 8238-8248.\u003c/li\u003e\n \u003cli\u003eC.-S. Liu, J.-J. Lin, and B.-R. Chen, A novel 3D scanning technique for reflective metal surface based on HDR-like image from pseudo exposure image fusion method, Optics and Lasers in Engineering. 168 (2023) 1-10.\u003c/li\u003e\n \u003cli\u003eJ. Franke, T. Kouteck\u0026yacute;, and D. Koutn\u0026yacute;, Comparison of sublimation 3D scanning sprays in terms of their effect on the resulting 3d scan, thickness, and sublimation time, Materials. 16 (2023) 1-15.\u003c/li\u003e\n \u003cli\u003eR. Mendřick\u0026yacute;, Impact of applied anti-reflective material on accuracy of optical 3D digitisation, Materials Science Forum. 919 (2018) 335-344.\u003c/li\u003e\n \u003cli\u003eL.T. De Paolis, V. De Luca, C. Gatto, G. D\u0026rsquo;Errico, and G.I. Paladini. Photogrammetric 3D reconstruction of small objects for a real-time fruition. in Augmented Reality, Virtual Reality, and Computer Graphics. 2020. Cham: Springer International Publishing.\u003c/li\u003e\n \u003cli\u003eT.-N. Doan and C.V. Nguyen, A low-cost digital 3D insect scanner, Information Processing in Agriculture. 11 (2024) 337-355.\u003c/li\u003e\n \u003cli\u003eB. Str\u0026ouml;bel, S. Schmelzle, N. Bl\u0026uuml;thgen, and M. Heethoff, An automated device for the digitization and 3D modelling of insects, combining extended-depth-of-field and all-side multi-view imaging, Zookeys. (2018) 1-27.\u003c/li\u003e\n \u003cli\u003eT.R. Roshan, M. Jafari, M. Golami, and M. Kazemi, Evaluating geometric measurement accuracy based on 3D model reconstruction of nursery tomato plants by Agisoft photoscan software, Computers and Electronics in Agriculture. 221 (2024) 1-14.\u003c/li\u003e\n \u003cli\u003eG. Bernotas, L.C.T. Scorza, M.F. Hansen, I.J. Hales, K.J. Halliday, L.N. Smith, M.L. Smith, and A.J. McCormick, A photometric stereo-based 3D imaging system using computer vision and deep learning for tracking plant growth, Gigascience. 8 (2019) 1-15.\u003c/li\u003e\n \u003cli\u003eJ.Z. Wang, Y. Zhang, and R.R. Gu, Research status and prospects on plant canopy structure measurement using visual sensors based on three-dimensional reconstruction, Agriculture-Basel. 10 (2020) 1-27.\u003c/li\u003e\n \u003cli\u003eS. Wu, W. Wen, W. Gou, X. Lu, W. Zhang, C. Zheng, Z. Xiang, L. Chen, and X. Guo, A miniaturized phenotyping platform for individual plants using multi-view stereo 3D reconstruction, Frontiers in Plant Science. Volume 13 - 2022 (2022) 1-19.\u003c/li\u003e\n \u003cli\u003eJ.A. Gibbs, M.P. Pound, A.P. French, D.M. Wells, E.H. Murchie, and T.P. Pridmore, Active vision and surface reconstruction for 3D plant shoot modelling, IEEE/ACM Transactions on Computational Biology and Bioinformatics. 17 (2020) 1907-1917.\u003c/li\u003e\n \u003cli\u003eD. Li, L. Xu, C. Tan, E.D. Goodman, D. Fu, and L. Xin, Digitization and visualization of greenhouse tomato plants in indoor environments, Sensors. 15 (2015) 4019-4051.\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"plant-methods","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"plme","sideBox":"Learn more about [Plant Methods](http://plantmethods.biomedcentral.com/)","snPcode":"13007","submissionUrl":"https://submission.nature.com/new-submission/13007/3","title":"Plant Methods","twitterHandle":"@PlantMethods","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"BMC/SO AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Close-range photogrammetry, SfM-MVS-based data processing, 3D reconstruction, plant phenotyping, morphological traits, precision agriculture","lastPublishedDoi":"10.21203/rs.3.rs-7178236/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7178236/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"In recent years, non-destructive and non-invasive methods for 3D plant reconstruction have gained importance in plant phenotyping. Morphological traits reflect a plant’s physiological status and serve as key indicators for precision agriculture, crop protection, and food quality assessment. Accurate and efficient 3D modelling enables objective, repeatable monitoring of plant development and health, supporting data-driven decision-making in agricultural and food research. This study presents a cost-effective and flexible photogrammetric methodology for analysing plant morphological traits under controlled laboratory conditions. The system includes an industrial RGB camera mounted on a robotic arm, a rotating platform with an adjustable plant holder, and stable illumination. The key steps involved camera calibration, exposure optimisation, fine-tuning of evaluation algorithm parameters (tweaks), setting the optimal camera-to-object distance, and reducing computational load for 3D model evaluation. Comparative testing revealed that the most effective calibration strategy integrated simultaneous calibration, pre-calibrated parameters, and adaptive fitting, ensuring high reconstruction accuracy and consistent model quality. The optimal acquisition parameters were a 50 milliseconds exposure time, a tweak value of 0.9, and a 16 cm camera-to-object distance. Using more camera positions with fewer frames per position proved more efficient than the reverse. The optimal configuration consisted of three height levels with 40 frames each. Automation and data reduction led to a 75% decrease in processing time, reducing the scan time from 8 minutes to 2.7 minutes per plant. The developed method proved to be a reliable, reproducible, and affordable tool for routine 3D analysis of plant morphology via close-range photogrammetry.","manuscriptTitle":"Implementation of an SfM-MVS-based photogrammetry approach for detailed 3D reconstruction of plants","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-08-05 06:04:16","doi":"10.21203/rs.3.rs-7178236/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-08-20T02:40:46+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-08-19T09:05:28+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-08-19T08:57:54+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-08-19T03:39:42+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"152973463027364388054272558723995373289","date":"2025-08-15T10:05:49+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"325680356854303621489154780902621606484","date":"2025-08-14T04:49:03+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"315984536212593560344842901868668092992","date":"2025-08-11T03:16:12+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"335226648695619758486784143879493299452","date":"2025-08-10T08:14:24+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-07-30T02:37:22+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-07-24T21:33:33+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-07-23T14:52:51+00:00","index":"","fulltext":""},{"type":"submitted","content":"Plant Methods","date":"2025-07-23T07:59:51+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"plant-methods","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"plme","sideBox":"Learn more about [Plant Methods](http://plantmethods.biomedcentral.com/)","snPcode":"13007","submissionUrl":"https://submission.nature.com/new-submission/13007/3","title":"Plant Methods","twitterHandle":"@PlantMethods","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"BMC/SO AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"b89d409b-c510-41df-8c66-f2c5b377b2ce","owner":[],"postedDate":"August 5th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[],"tags":[],"updatedAt":"2025-10-13T16:00:11+00:00","versionOfRecord":{"articleIdentity":"rs-7178236","link":"https://doi.org/10.1186/s13007-025-01445-x","journal":{"identity":"plant-methods","isVorOnly":false,"title":"Plant Methods"},"publishedOn":"2025-10-09 15:57:22","publishedOnDateReadable":"October 9th, 2025"},"versionCreatedAt":"2025-08-05 06:04:16","video":"","vorDoi":"10.1186/s13007-025-01445-x","vorDoiUrl":"https://doi.org/10.1186/s13007-025-01445-x","workflowStages":[]},"version":"v1","identity":"rs-7178236","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7178236","identity":"rs-7178236","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}