Optimizing Soybean Breeding: High-Throughput Phenotyping for Stink Bug Resistance and High Yields

Optimizing Soybean Breeding: High-Throughput Phenotyping for Stink Bug Resistance and High Yields | 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 Optimizing Soybean Breeding: High-Throughput Phenotyping for Stink Bug Resistance and High Yields Maiara Oliveira, Alexandre Hild Aono, Patricia Braga, Adriano Abreu Moreira, and 6 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7736736/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract The stink bug complex is one of the most damaging pests of soybean, reducing yield and seed quality. Genetic resistance remains the most sustainable and effective management strategy, but its quantitative inheritance and labor-intensive field phenotyping make its implementation in breeding programs challenging. This study explored high-throughput phenotyping (HTP) using unmanned aerial vehicles (UAVs) equipped with RGB cameras to evaluate a soybean population and the potential of phenotyping to stink bug resistance by correlating image-derived features and machine learning (ML) models. A population of 304 soybean lines was evaluated in alpha-lattice design trials across two seasons under natural infestations. Five resistance-related traits, grain yield (GY), hundred-seed weight (HSW), number of days to maturity (NDM), tolerance (TOL), and leaf retention (LR), were manually scored and linked to UAV-derived vegetation indices (VIs) and texture indices (TIs). Three ML models (AdaBoost, SVM, MLP) were tested to predict these traits from aerial features. Results showed that VIs, particularly Visible Atmospherically Resistant Index at the first percentile (VARI_P25), were consistently associated with resistance-related traits, while decision tree analysis highlighted TIs at 45° and 135° as complementary sources of structural information. Prediction ability was highest for GY, HSW, and NDM, especially in flights near flowering and maturity, but remained low for TOL and LR. Integrating multiple flights modestly improved accuracy, whereas cross-season predictions were unreliable. Nonetheless, indices such as VARI_P25 provided useful cross-season correlations for HSW and TOL, enabling early screening of less promising lines. This pioneering study demonstrates that UAV–ML pipelines can capture genetic signals of stink bug resistance in soybean, despite environmental complexity. These findings open new avenues for resistance phenotyping, supporting more efficient breeding strategies and accelerating genetic gains in soybean improvement. Crop protection Glycine max Machine learning models Unmanned aerial vehicles (UAVs) Vegetation index Texture indices Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Figure 9 Figure 10 Highlights VARI_P25 consistently emerged as the most robust index, with high heritability, strong trait correlations, and key importance in ML models. UAV-based RGB imagery revealed strong correlations between image-derived indices and resistance-related traits, often more informative than ML predictions. Texture indices (TIs) provided complementary structural information in decision tree–based models, though further studies are needed to validate their role. Cross-season predictions remain challenging, but UAV-derived indices proved valuable for early screening and resource optimization in breeding pipelines. Introduction Soybean ( Glycine max ) is one of the most widely cultivated oilseed crops in the world due to its versatility, being used in the production of oil, meal, biodiesel, and proteins for animal and human nutrition (Hartman et al., 2011 ). In the 2023/24 season, global soybean production totaled 396.9 million tons (USDA, 2024 ). Brazil is currently the world’s leading soybean producer, with an estimated output of 154.6 million tons and an average yield of 3.5 t/ha (CONAB, 2024). Nowadays, soybean is cultivated in almost all regions of Brazil, favoring the emergence of various insect pests and pathogens that cause significant losses in grain yield (GY) and quality (Boerma and Walker, 2005 ). Among these pests, stink bugs (Hemiptera: Pentatomidae) are particularly damaging, with potential yield losses of up to 125 kg/ha per insect per linear meter (Guedes et al., 2012 ). These insects feed directly on soybean pods, affecting the grains and causing losses not only in yield but also in the physiological and phytosanitary quality of the seeds (Bortolotto et al., 2015 ). In addition, stink bug’s feeding process causes soybean delayed maturity, which hinders the determination of the harvest time and mechanical harvesting due to leaf retention and green stems (Panizzi and Hirose, 1995 ; Rossetto et al., 1995 ). Genetic resistance remains one of the most efficient and sustainable strategies for stink bug control and is often integrated with chemical approaches. However, resistance to stink bugs is a complex, polygenic, and quantitatively inherited trait, which makes its selection in breeding programs particularly challenging (Godoi and Pinheiro, 2009 ). Plant breeding programs targeting quantitative traits require a large and diverse population, which needs to be evaluated in a feasible and precise way. In this sense, phenotyping is one of the main bottlenecks of breeding, as it typically relies on manual and visual measurements, which are time-consuming and imprecise. This limitation restricts the number of lines that can be tested (Bhat and Yu, 2021 ). Developing high throughput phenotyping (HTP) methodologies has enhanced the phenotypic evaluation of genotypes, optimizing diverse breeding programs (Li et al., 2014 ). HTP methodologies, particularly utilizing autonomous platforms like unmanned aerial vehicles (UAVs) coupled with cameras, have emerged as valuable tools in assisting breeders in selecting and developing commercial cultivars. Furthermore, the use of images facilitates non-destructive evaluation of numerous traits across a large number of lines at various growth stages and locations simultaneously (Furbank and Tester, 2011 ; Reynolds et al., 2019 ). The incorporation of image-based phenotyping in soybean breeding has been prominent in predicting many important traits, including maturity (Trevisan et al., 2020 ; Yu et al., 2016 ), biomass (Moreira et al., 2021 ; Sakurai et al., 2022 ), drought resistance (Zhou et al., 2020 ), and yield estimation (Maimaitijiang et al., 2020 ; Zhou et al., 2022 ). However, the use of HTP for quantifying and identifying insect injury in soybeans is still in its early development. Marston et al. ( 2020 ) demonstrated the effectiveness of using drone-based multispectral imagery and machine learning (ML) models to identify stress caused by soybean aphids, noting a reduction in near-infrared reflectance as aphid populations increased. Conversely, Iost-Filho et al. ( 2022 ) found that stink bug infestation did not significantly alter spectral reflectance in soybean under controlled conditions, even when using hyperspectral sensors. Thus, applying these technologies in field conditions for the selection of resistant genotypes still requires further research to increase the efficiency and accuracy of the selection process (Goggin et al., 2015 ). This study aimed to evaluate the potential of UAV-based RGB imagery and image-derived features to support phenotyping for resistance to the stink bug complex in soybean under natural field infestations. A population of 290 soybean lines was assessed across two seasons using an alpha-lattice design. By contrasting different image-derived descriptors and their association with conventional phenotyping measurements, this study highlights effective strategies for soybean image-based phenotyping in resistance evaluation. Furthermore, we assessed the predictive performance of ML algorithms in predicting key agronomic traits from image features, providing an end-to-end workflow with strong potential to support soybean breeding. Materials and Methods Plant Material and Field Experiment For the conduct of this study, we utilized a breeding population consisting of 290 soybean lines, part of the germplasm bank of the Laboratory of Genetic Diversity and Breeding at Luiz de Queiroz College of Agriculture/University of São Paulo (ESALQ/USP). These lines resulted from crosses among the commercial cultivars BRS-133, CD-215, Conquista, Dowling, IAC-100, and Pintado, which were recombined with the plant introduction (PI) genotypes: PI 200487 (Kinoshita), PI 471904 (Orba), PI 200526 (Shiranui), and PI 459025 (Bing Nan). Progenies from these crosses were subsequently recombined with nine populations that exhibited genetic variability for stink bug resistance. These genotypes underwent two selection cycles focusing on resistance to the soybean stink bug complex, in addition to one selection cycle for high GY. Phenotypic evaluations were performed during the 2019/2020 and 2020/2021 growing seasons in the municipality of Piracicaba/SP (Brazil), at the Anhumas Experimental Genetics Station (22°50'00' S, 48°05'00' W, altitude of 460 m). In both seasons, the 290 lines, along with 14 commercial checks recommended for the region and/or that exhibited resistance to stink bugs, were evaluated using a 16x19 alpha-lattice experimental design with three replicates. Experimental plots consisted of two rows, each 4 meters in length, spaced 0.5 meters apart, and with a seeding density of 15 seeds per meter. The experiments were conducted following the recommended agronomic management for soybean cultivation, except that chemical control of insects was not performed, aiming to ensure the presence of stink bugs in the experimental area, enabling the evaluation of the lines' resistance to this insect. To monitor the population and verify the presence of stink bugs in the experimental area, beat cloths were employed from flowering to full maturation. Beat cloths were used weekly throughout the entire reproductive cycle of the plant, covering the period of stink bug occurrence and attack. Evaluations involved the use of a beat cloth along a 1-meter crop row, with random beating across the entire experiment. Adults, nymphs (third, fourth, and fifth instars), and the total population (adults and nymphs) of soybean stink bugs were counted. Manually Measured Agronomic Traits The evaluation of soybean genotypes involved the manual measurement of five key agronomic traits. This encompassed the assessment of the number of days to maturity (NDM), which represented the duration from planting until 95% of pods achieved maturity, thereby indicating the genotype's cycle. Leaf retention (LR) was appraised at maturity using a visual scale ranging from 1 to 5, where a score of 1 indicated plants undergoing normal senescence, while a score of 5 indicated plants retaining multiple stems and green leaves, rendering harvest impractical. GY was determined post-harvest, representing the total mass of seeds produced in the plot, corrected for 13% grain moisture, and converted to kilograms per hectare (kg.ha − 1 ). Healthy seed weight (HSW) was determined as the weight of seeds devoid of stink bug damage, evaluated after harvest and grain processing (kg.ha − 1 ). This measure was obtained during seed processing, where they undergo separation through a spiral mechanism, where shriveled, green, and malformed grains are separated by the action of gravity and centrifugation (da Rocha et al., 2014 ). Tolerance (TOL) was calculated to assess the genotype's resilience to stink bug injury, quantified through an index obtained by the following formula: $$\:TOL=\:\left(1-\left(\frac{GY-HSW}{GY}\right)\right)\:\times\:100$$ 1 UAV Data Collection Aerial images were collected by a DJI Phantom 4 Pro V2.0 drone (DJI, Shenzhen, China) equipped with a 20-megapixel RGB camera capturing images in the visible spectrum. Flight planning was conducted using the FieldAgent digital platform (Sentera, Minnesota, USA) (Fig. 1 a). The camera captured sequential images of the area through progressive scanning. During the crop season of 2019/2020, six flights were conducted, while in the 2020/2021 season, this number was expanded to ten different dates. Detailed information on the days after planting (DAP) when each flight was conducted can be found in Fig. 2 . All flights were executed at the height of 20 meters, with a ground sampling distance (GSD) of 1.8 centimeters. An 80% level of both frontal and lateral overlap was maintained between the acquired images. Additionally, flights were conducted on sunny, cloud-free days, within the time frame of 10 a.m. to 2 p.m., aiming to optimize lighting conditions and minimize environmental variabilities. Image Processing After acquiring the aerial images, individual orthomosaics were created for each field (Fig. 1 b) using Agisoft MetaShape software (Agisoft LLC, St. Petersburg, Russia), following the standard protocol provided by the software for orthomosaic generation. Subsequently, the orthomosaics were imported into QGIS version 3.02 software (QGIS Development Team, 2018 ), where shapefiles were delineated, defining the boundaries of each experimental plot (Fig. 1 c.1). To mitigate potential interferences resulting from leaf overlap between adjacent plots, a negative buffer of 0.10 m was applied. Each plot was identified with an ID based on the experimental design. The FieldImageR 0.6.0 package (Matias et al., 2020 ) implemented in R was used to crop each experimental area simultaneously, employing the shapefile as the plot boundary (Fig. 1 c.2). After individualizing and cropping each plot, background removal was performed, including shadows, soil, and vegetative residues, using the methodology proposed by Yuan et al. ( 2019 ), which employs a plant segmentation method using vegetation indices to maximize segmentation accuracy. All the steps were implemented using Python v3 (Van Rossum and Drake, 2009 ) and the OpenCV library (Pisarevsky, 2008 ). First, all the plot images were rotated to a 90-degree orientation to calculate the vegetation indices of modified excess green (MExG) and the color index of vegetation extraction (CIVE). By subtracting these indices (MExG - CIVE), we scaled the resulting image into a 0-255 range and used this matrix to create a segmentation mask with Otsu's thresholding technique (Otsu, 1979 ) (Fig. 1 c.4). Image-Derived Traits For each segmented image corresponding to an orthomosaic at different flight dates, we calculated different vegetation indices and features describing their physical properties, including color, gradient, and texture. We employed the Python v3 programming language (Van Rossum and Drake, 2009 ) with the libraries OpenCV (Pisarevsky, 2008 ) and scikit-image (Van Der Walt et al., 2014 ). Initially, the Canopy Cover Index (CC) was computed, representing the percentage of remaining pixels after image segmentation. Additionally, six VIs were calculated (Table 1 ), including the triangular greenness index (TGI), the green leaf index (GLI), the normalized green-red difference index (NGRDI), the red-green-blue index (RGBVI), and the visible atmospherically resistant index (VARI). Based on the images resulting from each index calculation, we summarized the value distribution into different descriptive statistics: mean, standard deviation (std), skewness (skew), kurtosis, and 25, 50, and 75 percentiles, denoted as P25, P50, and P75, respectively. Table 1 Summary of vegetation indices (VI) used in this study. Each VI was calculated using an RGB image, where the image channels are red (R), green (G), and blue (B). Vegetation Index Name Formula References TGI Triangular greenness index \(\:G-0.39\times\:R-0.61\times\:B\) (2) (Hunt et al., 2005 ) GLI Green leaf index \(\:\frac{(2\times\:G-R-B)}{(2\times\:G+R+B)}\) (3) (Bassine et al., 2019 ) NGRDI Normalized green-red difference index \(\:\frac{G-R}{G+R}\) (4) (Hunt et al., 2005 ) RGBVI Red-green-blue index \(\:\frac{{G}^{2}-R*B}{{G}^{2}+R*B}\) (5) (Bendig et al., 2015 ) VARI Visible atmospherically resistant index \(\:\frac{G-R}{G+R-B}\) (6) (Gitelson et al., 2002 ) To reflect different wavelength values, we created a gray-level histogram with 64 bins, considering each bin quantification as a different attribute. Texture measures were calculated using gray level co-occurrence matrices (GLCMs) to describe repetitive patterns in the image, varying in size and providing information about roughness and regularity. For each image, we constructed four different GLCMs corresponding to angles of 0, 45, 90, and 135 degrees. These matrices were summarized into six different features based in Texture Indices (TIs): energy, correlation, angular second moment (ASM), dissimilarity, and contrast (Haralick et al., 1973 ). Therefore, the complete characterization of each image was based on 124 features: 1 feature measuring CC, 7 descriptive statistics for each of the 5 vegetation indices (35 features), 64 features corresponding to color histogram quantifications (Hist_1 to Hist_64), and 6 features for each of the 4 GLCMs (24 features). Associations between these features and the manually measured traits were performed with Pearson correlations in software R version 4.3.2 (R Core Team, 2023 ). Statistical Analysis of Phenotypic Data The phenotypic data from all trials were analyzed in R (R Core Team, 2023 ) using the ASReml-R package version 4.3.2 (Butler et al., 2018 ). Variance components and genetic parameters were estimated using the restricted maximum likelihood (REML) method (Patterson and Thompson, 1971 ), while genetic values were predicted using the best linear unbiased prediction (BLUP) method (Henderson, 1982 ). The significance of random effects was assessed through likelihood ratio tests (LRTs). For the five manually measured traits in the study and each of the six indices obtained from images, statistical analyses were conducted for each year, considering each flight individually. The analyses followed the model: $$\:{y}_{ijk}=\mu\:+{g}_{i}+{r}_{k}{\:+\:b}_{j\left(k\right)}+{\epsilon\:}_{ijk}$$ 7 where, \(\:{y}_{ijk}\) represents the phenotype observed for the i -th genotype, within the j -th block, across the k -th replicate; \(\:\mu\:\) the overall mean; \(\:{g}_{i}\) the random effect associated with the i -th genotype ( \(\:{g}_{i}\:\sim\:N(0,\:{\sigma\:}_{g}^{2})\) with \(\:{\sigma\:}_{g}^{2}\) representing the genetic variance); \(\:\:{r}_{k}\) the fixed effect of the k -th replicate; \(\:{b}_{j\left(k\right)}\) the random effect of the j -th block within the k -th repetition ( \(\:{b}_{j}\:\sim\:N(0,\:{\sigma\:}_{j}^{2})\) with \(\:{\sigma\:}_{j}^{2}\) representing the block variance); and \(\:{\epsilon\:}_{ijk}\) the model residual, assuming that the residuals are independent and normally distributed, i.e., \(\:{\epsilon\:}_{ijk}\:\sim\:N(0,\:{\sigma\:}^{2})\) , where \(\:{\sigma\:}^{2}\) is the residual variance. Heritability in the broad sense ( \(\:{h}^{2}\) ) was estimated using the method proposed by Piepho and Mohring (2007): $$\:\left({h}^{2}\right)=\frac{{\sigma\:}_{g}^{2}}{{\sigma\:}_{g}^{2}+\frac{{\sigma\:}^{2}}{\:r\:}}$$ 8 where \(\:{\sigma\:}_{g}^{2}\) and \(\:{\sigma\:}^{2}\) are the genotypic and error variances, respectively, and r is the number of replicates. The correlation between the true and predicted random effects was employed as a metric to evaluate the accuracy: $$\:acc=\sqrt{1-\frac{PEV\left({\widehat{a}}_{i}\right)}{{\widehat{\sigma\:}}_{a}^{2}}}$$ 9 where \(\:PEV\left({\widehat{a}}_{i}\right)\) is the prediction error variance of the predicted random effect and \(\:{\widehat{\sigma\:}}_{a}^{2}\) is the estimated variance of the random effect. Using the BLUPs of each genotype, Pearson's correlations were performed between the traits. Machine Learning Methods The scikit-learn library (Van Der Walt et al., 2014 ), developed in Python v3 programming language (Van Rossum and Drake, 2009 ), was employed to implement the ML-based prediction models. We utilized the features calculated from images collected via UAV as independent variables and the five visual characteristics as dependent variables. Each visual characteristic was associated with the UAV-collected indices according to the flight date. For the GY, HSW, TOL and NDM characteristics, regression models were applied, while for LR, a classification model was adopted. Subsequently, six distinct forecasting scenarios were tested. Two scenarios considered all flights conducted in each season (2019/2020 and 2020/2021), while two others focused exclusively on the last three flights of each season. Additionally, two scenarios were designed to test cross-year prediction: the first utilized data from 2019/2020 to predict the phenological characteristics of 2020/2021 based solely on flights near the maturation period, and the second incorporated all flights. Three machine learning methods were explored in this study: support vector machine for regression (SVR) (Vapnik, 2000 ), multilayer perceptron (MLP) neural network (Pal and Mitra, 1992 ), and adaptive boosting (AdaBoost) (Freund and Schapire, 1997 ). Additionally, we applied a decision tree to investigate which characteristics obtained through images are more important for predicting the manually measured agronomic traits, estimating the feature importance of each image feature using the Gini index. The performance of the models was evaluated through a 10-fold cross-validation scenario. For the regression models, the Pearson correlation coefficient ( R ) and mean squared error ( MSE ) were used as evaluation criteria. For the classification models, accuracy and precision metrics were employed to assess the model's ability to classify the test data correctly. These metrics were calculated by comparing the predicted and actual values in the training and test sets. Results Agronomic Trait Evaluations To develop cultivars resistant to stink bugs, the first step is to identify promising genotypes under natural field pressure. Because infestation intensity varies between seasons, monitoring is essential to characterize the selection environment. In this study, stink bug incidence was recorded weekly with beat cloth sampling throughout the reproductive period. Infestation levels in both seasons remained within the recommended thresholds for effective pest management, and UAV flights were scheduled to coincide with infestation peaks (Fig. 3 ). The 2019/2020 season exhibited a significantly higher infestation, reaching an average of 18 stink bugs per beat cloth at the peak of infestation, 100 days after planting (DAP). In contrast, the subsequent 2020/2021 season had a lower infestation, with a peak of only 3 stink bugs per beat cloth at 120 DAP, along with other smaller peaks recorded at 80 and 100 DAP, averaging 2.5 stink bugs. These contrasting infestation dynamics provided two distinct selection environments for evaluating soybean resistance to the stink bug complex. To evaluate the response of soybean genotypes under natural stink bug infestation, variance component analysis was performed for five agronomic traits (LR, NDM, GY, TOL, and HSW) across two growing seasons. This approach allowed partitioning the total phenotypic variation into genetic and block effects, as well as estimating trait h² and prediction acc . Both variance components were significant according to LRT tests (Fig. 4 ). Genetic variance (σG) explained a large proportion of trait variation in both years. In 2019, σG accounted for 34% of the variation in GY and up to 44% for HSW, TOL, and LR. In 2020, the pattern was similar, although the contribution of σG was reduced for GY and HSW, while remaining high for NDM and TOL. h² were consistently high, ranging from 0.72 (NDM, GY) to 0.77 (HSW) in 2019, and from 0.65 to 0.90 in 2020. acc followed the same trend, with values between 0.83 and 0.87 in 2019 and slightly higher in 2020, with a range from 0.79 to 0.94, particularly for GY and TOL. These results highlight the strong genetic control of the evaluated traits and the reliability of phenotypic predictions across seasons. Figure 5 shows the genetic correlations among the manually measured traits across the two seasons. All correlations were significant, though their magnitude and direction varied between years. Notably, there is a strong correlation between GY and HSW, the latter often used in the selection of more productive and stink bug-resistant genotypes. This correlation was higher in the 2020 season, reaching a value of 0.87, compared to 0.53 in the 2019 season. Image-Derived Traits After image pre-processing, a total of 124 indices were extracted per flight, including color, texture, and gradient histogram features. To reduce redundancy, indices were filtered based on their correlation with manually measured traits, using a threshold of r > 0.6, resulting in a filtered set of 38 features. Among these, 19 were VIs derived from different descriptive statistics. Given the high correlation among these indices (Supplementary Fig. 1), we prioritized those with the highest correlations across most flights. This resulted in the selection of four key VIs at the 25th percentile (VARI_P25, RGBVI_P25, NGRDI_P25, and GLI_P25). For TIs, all those with a correlation greater than 0.6 in any flight were subjected to Principal Component Analysis (PCA), using the first principal component for subsequent analyses (Supplementary Fig. 2). In total, 18 TIs were selected and included in the PCA: ASM, Dissimilarity, Energy, and Homogeneity at angles of 0°, 45°, 90°, and 135°, as well as Correlation at angles of 45° and 135°. These TIs are complementary, as each captures different aspects of the structural and textural variation of the vegetation. Therefore, TI were summarized into the first principal component, which captured the main structural variation. Additionally, the CC index was included in the selection due to its high correlation with the evaluated traits. In total, six indices (VARI_P25, RGBVI_P25, NGRDI_P25, GLI_P25, CC, and PCA) were carried forward for correlation analyses with agronomic traits. Color histogram quantifications did not exceed the correlation threshold and were excluded. After selecting the indices, we integrated them into mixed models to understand how the comprehension of different characteristics contributes to overall plant performance and facilitates the identification of genotypes with desirable traits. Figure 6 shows the correlation structure among indices across flight dates in both seasons. As expected, correlations were stronger between flights conducted at closer intervals and declined as the temporal distance increased. This is evident in the clusters, where closer flight times show significant correlations, whereas more distant flights present correlations close to zero. Given this pattern, each flight time was analyzed individually to understand how different growth stages influence the evaluated traits. By considering each flight time as a distinct dependent variable, we obtain a more precise and detailed analysis, which allows us to identify the specific moment in the crop stages that is most suitable for performing flights to assess the traits of interest. The decomposition of variance components and genetic parameters for the six image-derived traits across flight dates are shown in Supplementary Figs. 3 and 4. It is observed that, across all flight dates, the traits CC and PCA, showed the lowest h² estimates for both seasons. In 2019, flights at 90 and 97 DAP (grain filling period) presented the lowest h² estimates for all traits. Additionally, the VIs demonstrated similarity in h², with the Vari_P25 at 106 DAP standing out with a h² of 0.76. Overall, the acc estimates ranged from moderate to high, though some flight periods showed lower values. In 2020, flights at 57, 65, and 122 DAP were associated with lower heritability, while VARI_P25 again showed the highest estimate (h² = 0.77) at 107 DAP. Acc followed a similar pattern to the previous season. Likelihood ratio tests confirmed that genetic variance significantly explained trait variation in all cases, except for PCA at 97 DAP in 2019. Building upon these findings, we next examined the relationships between image-derived indices and manually measured traits. Figure 7 shows the genetic correlations between UAV-based indices and field measurements across flight dates. Strong correlations were observed for all manually measured traits in both seasons, indicating that specific flight periods provided more informative associations with key traits relevant to soybean resistance to stink bugs. The strongest correlations were observed for LR and NDM in the 2019 season, during flights close to harvest and maturity, reaching 0.76. This same pattern was observed in the subsequent season, but with slightly lower correlation magnitudes. For GY and HSW, strong correlations were also found in the 2020 season during the early flights (57 to 78 DAP), while mid-season flights (85–107 DAP) showed weaker associations. Moreover, GY and HSW strongly correlated with multiple indices in the later flights. For TOL, although the correlations were not as strong as with the other traits, in the 2019 season, where stink bug infestation was more pronounced, significant correlations above − 0.5 and 0.5 were observed during flights close to maturity with the CC and PCA indices. The VIs exhibit similar patterns, reflecting their shared capacity to capture related aspects of canopy reflectance. Likewise, CC, derived from soil–leaf segmentation, shows patterns similar to those of the VI based on RGB bands. In the early flights, VIs and the CC show positive correlations with GY, HSW, and TOL, which turn negative in later flights. In contrast, the PCA index displayed distinct and often inverse correlation patterns, suggesting that texture-based metrics captured complementary structural information not reflected by VIs. In summary, these analyses revealed consistent and informative patterns across seasons, with VIs generally outperforming TIs in terms of correlation and h². Among them, VARI_P25 was particularly informative, showing strong and consistent associations with multiple traits across flights and seasons, together with high h² estimates. In 2020/2021, early flights (57–78 DAP) were most informative for GY and HSW, whereas later flights (107–122 DAP) captured stronger associations for TOL, NDM, and LR. In 2019/2020, when no flights were conducted near flowering, the most informative periods occurred closer to maturity. Machine Learning-Based Prediction We evaluated the predictive ability (PA) of three machine learning models (AdaBoost, MLP, and SVM) for five manually measured traits using UAV-derived indices. Predictions were made for each flight date to identify the most informative periods for each trait (Supplementary Tables 1–2). Across both seasons, AdaBoost consistently achieved the highest PA, particularly for grain yield (GY). In 2020/2021, early flights (65–78 DAP) produced the strongest results, with AdaBoost correlations exceeding 0.80, while predictive ability declined in later flights. In 2019/2020, the best predictions occurred later, at 120 DAP. Similar patterns were observed for HSW, with AdaBoost again outperforming the other models. In 2020/2021, HSW prediction was strongest in early flights (65–78 DAP), whereas in 2019/2020 the best results occurred at 120–127 DAP. For NDM and TOL, the MLP model showed consistently lower performance. AdaBoost achieved the highest PA for NDM in flights near physiological maturity. A similar trend was observed for TOL, where late-season flights improved model accuracy and both AdaBoost and SVM produced comparable performance. These patterns were consistent when evaluated with MSE, reinforcing the superiority of AdaBoost during late reproductive stages. For LR, classification models were used, with accuracy and precision as evaluation metrics. In general, the three models showed similar accuracy, with AdaBoost and MLP demonstrating superior results, especially in flights close to maturity. However, the PAs for this trait were relatively low, not exceeding 0.4 in accuracy. In the 2019/2020 season, the highest accuracy was found in the flight at 127 DAP, with 0.37 using the MLP model, while in 2020/2021, AdaBoost reached the highest accuracy at 122 DAP (0.39). Precision values showed similar trends, with AdaBoost slightly outperforming the other models. Overall, predictive ability was higher in 2020/2021 than in 2019/2020 for GY and HSW, likely due to the greater number of flights conducted near flowering in that season. In contrast, NDM showed relatively stable predictive performance across years, with its best results in 2019/2020. TOL and LR exhibited low predictive ability in both years. Figure 8 summarizes the flights that yielded the highest PA for each trait, based on Pearson correlation for regression models (GY, HSW, TOL, NDM) and accuracy for classification models (LR). We further evaluated six prediction scenarios by varying the combination of flights and the target season: (i) all flights from 2019; (ii) all flights from 2020; (iii) the last three flights of 2019; (iv) the last three flights of 2020; (v) training with 2019 data to predict 2020; and (vi) the reverse. Supplementary Figs. 5A–B present the predictive ability (PA) of the three models for each trait under these scenarios. Figure 9 shows that, within a season, aggregating all flights generally improved PA compared to using single flights, particularly with the AdaBoost model. The gains were most evident for GY and HSW in 2020/2021, where AdaBoost achieved the highest overall correlations. Similar trends were observed in 2019/2020, confirming that multi-flight integration captures temporal variation more effectively. By contrast, cross-season predictions performed poorly across all models and traits, highlighting the difficulty of transferring models between years despite large training sets. These results emphasize both the value of integrating multiple flights within a season and the current limitations of cross-season forecasting. Finally, we evaluated cross-season correlations between image-derived indices and manually measured traits (Fig. 10 ). Unlike the ML models, which performed poorly when trained in 2019/2020 and tested in 2020/2021, several cross-season correlations were significant. Notably, HSW and TOL in 2019/2020 showed strong correlations with VARI_P25 in 2020/2021, particularly at 107 and 114 DAP, in some cases exceeding within-season correlations. In contrast, GY displayed consistently weaker cross-season correlations, reinforcing its complex and environment-dependent nature. We also conducted a decision tree analysis to understand the importance of the predictor variables. As expected, some features demonstrated significant relevance, while others exhibited low importance. Furthermore, the importance of a variable varied depending on the flight period and the year. However, certain VIs and TIs stood out, as evidenced in Supplementary Figs. 6 to 10, which present the importance graph of each predictor variable in the flights that demonstrated a predictive capacity greater than 0.6 for the traits NDM, GY, and HSW. For LR e TOL, due to the low PA presented for these traits, only the most important flights were included in the graph. For GY, VARI_P25 was the most important feature in late-season flights of 2019, while texture indices (TIs) such as ASM (0° and 45°) and Homogeneity (45°) contributed strongly during early flights (57–78 DAP). HSW showed a similar trend, but TIs were more influential than VIs, particularly at 45° and 135°, with VARI_P25 again emerging in late flights (122 DAP). For NDM, VARI_P25 remained prominent, alongside Energy (135°) and Dissimilarity (45°) in 2019, with some gradient histograms showing specific importance for this trait. By contrast, no single variable dominated for LR and TOL; however, Contrast (45°) contributed to LR in the final 2020 flight, while VARI_P25 was again the main predictor for TOL. Discussion UAV-based HTP represents an effective tool for evaluating traits of interest in plant breeding programs. This methodology provides breeders with the ability to assess various stages of crop development and analyze a large number of genotypes across different locations, in a shorter time and with greater efficiency (Reynolds et al., 2019 ).. In this study, we leveraged UAV-RGB imagery to evaluate soybean performance under natural stink bug pressure, relating image-derived indices to conventional field phenotypes and assessing their ability to predict those phenotypes. Additionally, we expanded such an evaluation by incorporating ML models and their capability to predict these traits using images captured by UAV. HTP protocols represent a strategy with potential to overcome the limitations associated with traditional large-scale phenotyping, allowing for a more detailed evaluation of traits of interest. Considering that resistance to the stink bug complex is a quantitative trait, the breeding process becomes complex, requiring extensive field evaluations (Furbank and Tester, 2011 ) Stink Bug Resistance Phenotyping in Soybean Breeding In soybean breeding, phenotyping for resistance is well described for specific traits that condition resistance, such as pod damage, LR, grain filling period, crop cycle, grain size, and the HSW. HSW, in particular, has been reported as strongly associated with resistance and yield (da Rocha et al., 2014 ). In our study, correlations among resistance-related traits, productivity, and phenology support their use as joint selection criteria. Differences in correlation magnitudes between seasons likely reflect contrasting infestation levels: under very low pressure, trait expression is muted, whereas very high pressure can saturate injury across genotypes, both scenarios reducing discrimination among lines (Panizzi and Corrêa-Ferreira, 1997 ). Heritability estimates for the evaluated traits were high, suggesting that these traits can be employed in selecting more resistant and productive genotypes. Previous studies under natural stink bug infestation reported comparable patterns: in an F2:3 population, Santos et al. ( 2018 ) observed h² ≈ 0.80 (NDM), 0.70 (GY, HSW) e 0.20 (LR);in a RILs population, da Rocha et al. ( 2015 ) reported h² ≈ 0.75 (NDM), 0.35 (HSW), 0.24 (GY) e 0.67 (LR)., both populations originating from the cross between IAC 100 (resistant) and CD 215 (susceptible). Variation in h² across studies is expected for quantitative traits such as GY and HSW, given strong environmental modulation (Han et al., 2012 ). For LR, its value heavily depends on the evaluator's assessment, making it challenging to consistently achieve high heritability estimates (Godoi and Pinheiro, 2009 ; Pinheiro et al., 2005 ). We observed that \(\:{\sigma\:}_{g}^{2}\) explained a substantial proportion of the total phenotypic variance, indicating the presence of variability in the evaluated population and enabling the selection of genotypes. This variability is influenced by the parental lines involved in the crosses, which are quite diverse, including IAC 100, a source reported to carry multiple mechanisms of stink bug resistance (McPherson et al., 2007 ; Veiga et al., 1999 ). Thoughtful parental choice to maximize useful diversity remains central to breeding gains (Falconer et al., 1996 ). Despite their relevance, visual assessment becomes impractical and inaccurate on a large scale, a common challenge encountered in plant breeding programs, especially when evaluating insect resistance (Tang and He, 2021 ). Although UAV-based approaches have been explored in soybean for various applications, their use for evaluating resistance to stink bugs remains limited (Iost-Filho et al., 2022 ; Marston et al., 2020 , 2022 ). UAV-based imaging enables quantitative estimation of canopy reflectance-derived VIs, which relate to physiological processes, while TIs capture spatial/structural canopy information (Araus and Cairns, 2014 ; Li et al., 2014 ). Image-based Phenotyping for Stink Bug Resistance Given the utility of aerial images, their application extends to evaluating a wide range of traits in plant breeding programs. VIs are measures related to photosynthetic activity and the physiological state of plants, especially concerning their response to stress and overall health (Krause et al., 2020 ). These indices can capture subtle differences in plant composition and photosynthetic activity, which may be influenced by different environmental conditions and biotic stresses (Kefauver et al., 2015 ). Additionally, VIs are used in many studies for predicting agricultural productivity due to their stable and superior performance (Ballester et al., 2017 ; Zhou et al., 2017 ). Therefore, variations in VIs among different tested genotypes can provide insights into how plants respond and adapt to different stresses. In our study, VIs demonstrated high heritability and prediction accuracy, with strong correlations to traits associated with stink bug resistance, particularly in flights close to maturity and under higher stink bug pressure. Comparable findings have been reported in other crops. Li et al. ( 2023 ) showed that UAV-based RGB VIs effectively identified delayed senescence in wheat, while Resende et al. ( 2024 ) found that early flights (V5–V8) in maize yielded the strongest associations between GLI and GY. However, they also emphasized that these correlations were highly variable depending on environmental conditions and crop development stage. A methodological distinction of our work was the use of the first percentile of pixel values instead of plot means. One likely explanation is that stink bug infestations are typically patchy and spatially heterogeneous across the field and even within individual plots (Fernandes et al., 2019). Thus, considering only the first percentile of the pixels ensures that we are evaluating those plants that were attacked by stink bugs. This strategy allows for a more accurate assessment of resistance to stink bug attacks and can be an interesting strategy for breeding programs focused on insect resistance. In soybean, Bai et al. ( 2022 ) demonstrated that UAV-RGB imagery improved yield prediction under lodging conditions, particularly around flowering, underscoring the importance of capturing critical phenological stages. These authors also noted the contribution of texture information to yield estimation. Similarly, Maimaitijiang et al. ( 2020 ) reported that incorporating canopy texture features improved yield prediction accuracy, with R² ranging from 0.65 to 0.72. TIs represent spatial variations in color intensity in images and can capture repetitive patterns that are useful for assessing aspects such as canopy uniformity, which are associated with plant vigor and, consequently, productivity (Ma et al., 2022 ). Consistent with this, our results highlighted VARI_P25 as the most informative index, with superior heritability and consistent correlations across traits. Nonetheless, texture indices (TIs) also proved valuable in specific contexts, as evidenced by the PCA-based TI score correlating with tolerance (TOL) in 2019/2020 (r = 0.58), surpassing other VIs. Thus, while VIs are generally more effective, TIs can provide complementary information in certain environments or traits. Overall, these findings demonstrate that UAV-derived indices are strongly associated with visual characteristics and can be used individually for selection. However, exploiting their combined predictive power requires integrative approaches, such as ML, to fully capture the complexity of image-based datasets. Machine Learning Prediction of Stink Bug Resistance Several studies have combined UAV imagery with ML to address challenges in soybean improvement. Hassanijalilian et al. ( 2020 ) evaluated decision tree–based models to classify iron deficiency chlorosis in 40 soybean cultivars using RGB images and found that AdaBoost outperformed random forest (RF) and decision trees, achieving an average F1-score of 0.75. Similarly, RF and artificial neural networks effectively distinguished dicamba-tolerant from susceptible genotypes, with classification accuracies ranging from 0.69 to 0.75 (Vieira et al., 2022 ). Notably, LR resembles IDC and dicamba injury in that it is also expressed through visible chlorophyll loss in the leaves. In our study, AdaBoost consistently delivered the strongest results for most traits, both when using individual flights and when aggregating multiple flights, confirming the effectiveness of decision tree–based models for these prediction tasks. Nonetheless, PAs were generally lower than those reported in the studies above. A likely explanation lies in the contrasting stink bug pressures across seasons: in 2019/2020, severe infestations produced uniformly high scores, while in 2020/2021, symptoms were minimal and most scores clustered at the lower end this restricted phenotypic variation reduced model discrimination. These findings reinforce the importance of maintaining intermediate pest pressure levels to maximize trait expression and enable effective resistance screening (Panizzi and Corrêa-Ferreira, 1997 ). For GY and HSW, the highest PAs were observed in flights near flowering, although informative predictions were also obtained at later growth stages. Zhou et al. ( 2020 ) reported a similar pattern using a convolutional neural network to estimate soybean yield under water stress, with images collected at both early and late reproductive stages providing strong predictions. By incorporating all flights to predict GY, the results indicate that a greater number of flights contributes to capturing a higher degree of phenotypic and environmental variability. Flights conducted close to flowering can capture stand-related variation, such as incomplete plant establishment or uneven development, which may confound predictions and not fully reflect the genetic basis of resistance. In contrast, flights near maturity captured stink bug injury more effectively during the critical reproductive stages (R5–R7), when feeding causes the greatest yield losses (da Rocha et al., 2014 ). Thus, late flights more accurately represent genetic differences in resistance and yield potential, whereas early flights are less informative for resistance screening since stink bug damage is not yet phenotypically expressed. Although UAVs are widely used in plant breeding, their application for stink bug phenotyping in the field has not yet been reported. Some studies, have explored UAV-based approaches for detecting soybean aphid stress. Marston et al. ( 2022 ) achieved up to 89.4% classification accuracy using hyperspectral reflectance with support vector machines, while Marston et al. ( 2020 ) demonstrated that multispectral imagery coupled with linear regression could also identify aphid-induced stress. However, these studies were conducted with controlled infestation in cages and using few or only one genotype, conditions that differ from those found in breeding programs. Despite these limitations, they highlight the potential of advanced sensors, such as multispectral and hyperspectral cameras, to improve resistance screening under field conditions. Overall, the best PAs in our study were achieved in the 2020/2021 season, particularly for HSW and GY. Yet caution is needed when selecting based on HSW under low stink bug pressure, as this may favor highly productive lines rather than resistant ones (da Rocha et al., 2014 ). Conversely, the high selection pressure in 2019/2020 hindered the identification of resistant genotypes by masking phenotypic differences and reducing model discrimination, since tolerance mechanisms can buffer yield losses under moderate feeding but are overcome under extreme pest pressure (Smith, 2005 ). These contrasting scenarios underscore the importance of maintaining intermediate pest pressure, as overly low or high infestations limit discrimination among genotypes (Corrêa-Ferreira & Panizzi, 1999 ). Incorporating a greater number of flights improved PAs across traits, although the increment was modest, suggesting that repeated temporal sampling helps capture both genetic and environmental variability (Rutkoski et al., 2016 ; Sun et al., 2017 ). Crop prediction in plant breeding programs using data from previous years can be a valuable tool, but it presents considerable challenges, mainly due to environmental variations between seasons (Crossa et al., 2017 ; Yang et al., 2017 ). This is particularly evident when UAV-acquired images are used, as differences in climatic and management conditions across flights directly affect prediction accuracy. Still, VARI_P25 maintained moderate cross-season correlations with traits such as HSW and TOL, in some cases exceeding within-season associations. Although modest, these correlations can be useful for an initial genotype screening, allowing the exclusion of less promising materials and consequently optimizing resources and evaluation time in subsequent cycles (Zhou et al., 2022 ). By contrast, correlations for GY were considerably lower between seasons, indicating that this trait is more strongly modulated by environmental variation. We did not identify consistent patterns among the most important indices in the decision tree analysis. Each flight and each characteristic exhibited different patterns of influence from the indices used. Still, VARI_P25 emerged as the most recurrently important index, appearing among the flights with the highest predictive capacity for nearly all traits, with the exception of LR. Beyond its importance in the ML models, VARI_P25 also showed strong direct correlations with manually measured traits. Biologically, this index reflects canopy vigor and stress through greenness (Gitelson et al., 2002 ), and has been previously linked to grain weight under stress (Rodene et al., 2022 ), yield (Silva et al., 2022 ), and leaf area (Hasan et al., 2019 ; Zhang et al., 2019 ). These consistent associations across traits and studies highlight VARI_P25 as a particularly promising index for stink bug resistance phenotyping. In addition to VIs, TIs also contributed substantially to model performance. TIs calculated at 45° and 135° angles were often among the most influential predictors. Unlike VIs, which capture physiological status through reflectance, TIs quantify canopy structure and spatial patterns, traits that may indirectly reflect plant responses to pest damage (Ma et al., 2022 ). Future studies should further investigate the role of TIs in traits such as LR, which is notoriously difficult to assess visually and can easily be confounded with late maturity. Under stink bug attack, leaves and stems remain green while pods mature, a contrast that texture indices may be able to capture more reliably than visual assessment (Pinheiro et al., 2005 ; Godoi and Pinheiro, 2009 ). In summary, ML approaches coupled with UAV imagery proved effective for predicting soybean traits associated with stink bug resistance, with AdaBoost consistently delivering the strongest results. While environmental variation and extremes in pest pressure constrained prediction accuracy, integrating multiple flights and combining VIs with TIs enhanced the detection of phenotypic signals. These findings underscore the potential of UAV-ML pipelines to support resistance phenotyping, provided that experimental designs ensure balanced pest pressure and consistent temporal sampling. Conclusion This study provides the first comprehensive evaluation of stink bug resistance in soybean using UAV-based imagery integrated with ML models. We show that VI, particularly VARI_P25, are consistently associated with key traits, while TIs offer complementary information, underscoring the value of combining both approaches. Prediction ability was strongly influenced by pest pressure and flight timing, highlighting the importance of monitoring at critical reproductive stages and maintaining intermediate infestation levels for reliable resistance screening. Despite challenges such as limited cross-season transferability, our findings establish UAV–ML pipelines as a promising framework for phenotyping complex resistance traits. Future research integrating advanced sensors, multi-season datasets, and optimized experimental designs will be key to fully leveraging these tools for accelerating soybean breeding for stink bug resistance. Declarations Impact We introduce UAV-RGB imaging with ML models that derive actionable indices from field imagery to quantify canopy signals of stink bug resistance under natural infestations, enabling objective, scalable screening of breeding lines. This data-driven approach fits Precision Agriculture by using sensing and analytics to make phenotyping more efficient, less labor-intensive, and faster at advancing superior genotypes. CRediT Author Contribution Statement MO: Conceptualization, Methodology, Software, Validation, Formal analysis, Investigation, Data Curation, Writing - Original Draft, Writing - Review & Editing, Visualization. AA: Methodology, Software, Validation, Formal analysis, Investigation, Data Curation, Writing - Review & Editing, Visualization. PB: Conceptualization, Methodology, Investigation, Data Curation. AM: Conceptualization, Methodology, Investigation, Data Curation. FC: Data Curation. FK: Data Curation. FF: Data Curation. JPazini: Data Curation, Writing - Review & Editing. PY: Validation, Data Curation, Resources, Supervision. JPinheiro: Conceptualization, Validation, Investigation, Data Curation, Writing - Review & Editing, Resources, Supervision, Project administration, Funding acquisition Competing Interests The authors have no relevant financial or nonfinancial interests to disclose. Acknowledgement This work was supported by grants from the Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq), Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP), and the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES – Financial Code 001). AA received a PhD fellowship from FAPESP (2019/03232-6). Data Availability Data will be made available on request. References Araus, J. L., and Cairns, J. E. (2014). Field high-throughput phenotyping: the new crop breeding frontier. Trends in Plant Science , 19 (1), 52–61. https://doi.org/10.1016/J.TPLANTS.2013.09.008 Bai, D., Li, D., Zhao, C., Wang, Z., Shao, M., Guo, B., Liu, Y., Wang, Q., Li, J., Guo, S., Wang, R., Li, Y. H., Qiu, L. J., and Jin, X. (2022). Estimation of soybean yield parameters under lodging conditions using RGB information from unmanned aerial vehicles. Frontiers in Plant Science , 13 , 1012293. https://doi.org/10.3389/FPLS.2022.1012293/BIBTEX Ballester, C., Hornbuckle, J., Brinkhoff, J., Smith, J., and Quayle, W. (2017). Assessment of In-Season Cotton Nitrogen Status and Lint Yield Prediction from Unmanned Aerial System Imagery. Remote Sensing 2017, Vol. 9, Page 1149 , 9 (11), 1149. https://doi.org/10.3390/RS9111149 Bassine, F. Z., Errami, A., and Khaldoun, M. (2019). Vegetation Recognition Based on UAV Image Color Index. Proceedings - 2019 IEEE International Conference on Environment and Electrical Engineering and 2019 IEEE Industrial and Commercial Power Systems Europe, EEEIC/I and CPS Europe 2019 . https://doi.org/10.1109/EEEIC.2019.8783830 Bendig, J., Yu, K., Aasen, H., Bolten, A., Bennertz, S., Broscheit, J., Gnyp, M. L., and Bareth, G. (2015). Combining UAV-based plant height from crop surface models, visible, and near infrared vegetation indices for biomass monitoring in barley. International Journal of Applied Earth Observation and Geoinformation , 39 , 79–87. https://doi.org/10.1016/J.JAG.2015.02.012 Bhat, J. A., and Yu, D. (2021). High-throughput NGS-based genotyping and phenotyping: Role in genomics-assisted breeding for soybean improvement. Legume Science , 3 (3), e81. https://doi.org/10.1002/LEG3.81 Boerma, H. R., and Walker, D. R. (2005). Discovery and utilization of QTLs for insect resistance in soybean. In Genetics of Adaptation (pp. 181–189). Springer-Verlag. https://doi.org/10.1007/1-4020-3836-4_16 Bortolotto, O. C., Pomari-Fernandes, A., De, R. C. O., Bueno, F., De, A., Da Kruz, Y. K. S., Queiroz, A. P., Sanzovo, A., and Ferreira, R. B. (2015). The use of soybean integrated pest management in Brazil: a review. Agronomy Science and Biotechnology , 1 (1), 25–32. www.asbjournal.comFreelyavailableonline Butler, D. G., Cullis, B. R., Gilmour, A. R., Gogel, B. J., and Thompson, R. (2018). ASReml-R Reference Manual Version 4 ASReml estimates variance components under a general linear mixed model by residual maximum likelihood (REML) . http://asreml.org CONAB - Companhia Nacional de Abastacimento. 2024. Acompanhamento da safra brasileira: grãos (2023/24), 8° Levantamento. In safra 2023/24, n 8 – oitavo levantamento (Vol. 11, Issue 1). https://www.conab.gov.br/info-agro/safras/graos/boletim-da-safra-de-graos Corrêa-Ferreira, B. S., and Panizzi, A. R. (1999). Percevejos da soja e seu manejo. Embrapa-CNPSo Documentos, 124 , 1–45. Crossa, J., Pérez-Rodríguez, P., Cuevas, J., Montesinos-López, O., Jarquín, D., de los Campos, G., Burgueño, J., González-Camacho, J. M., Pérez-Elizalde, S., Beyene, Y., Dreisigacker, S., Singh, R., Zhang, X., Gowda, M., Roorkiwal, M., Rutkoski, J., and Varshney, R. K. (2017). Genomic Selection in Plant Breeding: Methods, Models, and Perspectives. Trends in Plant Science , 22 (11), 961–975. https://doi.org/10.1016/J.TPLANTS.2017.08.011 da Rocha, F., Bermudez, F., Ferreira, M. C., de Oliveira, K. C., and Pinheiro, J. B. (2014). Critério efetivo de seleção para avaliar a resistência ao complexo de percevejos em soja. Crop Breeding and Applied Biotechnology , 14 (3), 174–179. https://doi.org/10.1590/1984-70332014v14n3a26 da Rocha, F., Vieira, C. C., Ferreira, M. C., de Oliveira, K. C., Moreira, F. F., and Pinheiro, J. B. (2015). Selection of soybean lines exhibiting resistance to stink bug complex in distinct environments. Food and Energy Security , 4 (2), 133–143. https://doi.org/10.1002/FES3.57 Falconer, D. S., Mackay, T. F. C., and Frankham, R. (1996). Introduction to Quantitative Genetics . In Trends in Genetics (Fourth Edition). Trends in Genetics. Freund, Y., and Schapire, R. E. (1997). A Decision-Theoretic Generalization of On-Line Learning and an Application to Boosting. Journal of Computer and System Sciences , 55 (1), 119–139. https://doi.org/10.1006/JCSS.1997.1504 Furbank, R. T., and Tester, M. (2011). Phenomics - technologies to relieve the phenotyping bottleneck. Trends in Plant Science , 16 (12), 635–644. https://doi.org/10.1016/j.tplants.2011.09.005 Gitelson, A. A., Kaufman, Y. J., Stark, R., and Rundquist, D. (2002). Novel algorithms for remote estimation of vegetation fraction. Remote Sensing of Environment , 80 (1), 76–87. https://doi.org/10.1016/S0034-4257(01)00289-9 Godoi, C. R. C., and Pinheiro, J. B. (2009). Genetic parameters and selection strategies for soybean genotypes resistant to the stink bug-complex. Genetics and Molecular Biology , 32 (2), 328–336. https://doi.org/10.1590/S1415-47572009000200020 Goggin, F. L., Lorence, A., and Topp, C. N. (2015). Applying high-throughput phenotyping to plant–insect interactions: picturing more resistant crops. Current Opinion in Insect Science , 9 , 69–76. https://doi.org/10.1016/J.COIS.2015.03.002 Guedes, J. V. C., Arnemann, J. A., Stürmer, G. R., Melo, A. A., Bigolin, M., Perini, C. R., and Sari, B. G. (2012). Percevejos da soja: novos cenários, novo manejo. Revista Plantio Direto , 12 , 28–34. Han, Y., Li, D., Zhu, D., Li, H., Li, X., Teng, W., and Li, W. (2012). QTL analysis of soybean seed weight across multi-genetic backgrounds and environments. Theoretical and Applied Genetics , 125 (4), 671–683. https://doi.org/10.1007/S00122-012-1859-X/FIGURES/2 Haralick, R. M., Dinstein, I., and Shanmugam, K. (1973). Textural Features for Image Classification. IEEE Transactions on Systems, Man and Cybernetics , SMC - 3 (6), 610–621. https://doi.org/10.1109/TSMC.1973.4309314 Hartman, G. L., West, E. D., and Herman, T. K. (2011). Crops that feed the World 2. Soybean-worldwide production, use, and constraints caused by pathogens and pests. Food Security , 3 (1), 5–17. https://doi.org/10.1007/S12571-010-0108-X/METRICS Hasan, U., Sawut, M., and Chen, S. (2019). Estimating the Leaf Area Index of Winter Wheat Based on Unmanned Aerial Vehicle RGB-Image Parameters. Sustainability 2019, 11 (23), 6829. https://doi.org/10.3390/SU11236829 Hassanijalilian, O., Igathinathane, C., Bajwa, S., and Nowatzki, J. (2020). Rating iron deficiency in soybean using image processing and decision-tree based models. Remote Sensing, 12 (24), 4143. https://doi.org/10.3390/rs12244143 Henderson, C. R. (1982). Analysis of Covariance in the Mixed Model: Higher-Level, Nonhomogeneous, and Random Regressions. Biometrics , 38 (3), 623. https://doi.org/10.2307/2530044 Hunt, E. R., Cavigelli, M., Daughtry, C. S. T., McMurtrey, J. E., and Walthall, C. L. (2005). Evaluation of digital photography from model aircraft for remote sensing of crop biomass and nitrogen status. Precision Agriculture , 6 (4), 359–378. https://doi.org/10.1007/S11119-005-2324-5/METRICS Iost-Filho, F. H., Pazini, J. de B., de Medeiros, A. D., Rosalen, D. L., and Yamamoto, P. T. (2022). Assessment of Injury by Four Major Pests in Soybean Plants Using Hyperspectral Proximal Imaging. Agronomy , 12 (7), 1516. https://doi.org/10.3390/AGRONOMY12071516/S1 Kefauver, S. C., El-Haddad, G., Vergara-Diaz, O., and Araus, J. L. (2015). RGB picture vegetation indexes for High-Throughput Phenotyping Platforms (HTPPs). Remote Sensing for Agriculture, Ecosystems, and Hydrology XVII , 9637 , 96370J. https://doi.org/10.1117/12.2195235 Krause, M. R., Mondal, S., Crossa, J., Singh, R. P., Pinto, F., Haghighattalab, A., Shrestha, S., Rutkoski, J., Gore, M. A., Sorrells, M. E., and Poland, J. (2020). Aerial high-throughput phenotyping enables indirect selection for grain yield at the early generation, seed-limited stages in breeding programs. Crop Science , 60 (6), 3096–3114. https://doi.org/10.1002/CSC2.20259 Li, L., Zhang, Q., and Huang, D. (2014). A Review of Imaging Techniques for Plant Phenotyping. Sensors (Basel, Switzerland) , 14 (11), 20078. https://doi.org/10.3390/S141120078 Li, L., Guo, W., Carroll, M. E., Singh, A., Swetnam, T. L., Merchant, N., & Sarkar, S. et al. (2023). UAV-based RGB imagery and ground measurements for high-throughput phenotyping of senescence and QTL mapping in bread wheat. Crop Science, 63 (6), 3292–3309. https://doi.org/10.1002/csc2.21086 Ma, Y., Ma, L., Zhang, Q., Huang, C., Yi, X., Chen, X., Hou, T., Lv, X., and Zhang, Z. (2022). Cotton Yield Estimation Based on Vegetation Indices and Texture Features Derived From RGB Image. Frontiers in Plant Science , 13 , 925986. https://doi.org/10.3389/FPLS.2022.925986/BIBTEX Maimaitijiang, M., Sagan, V., Sidike, P., Hartling, S., Esposito, F., and Fritschi, F. B. (2020). Soybean yield prediction from UAV using multimodal data fusion and deep learning. Remote Sensing of Environment , 237 , 111599. https://doi.org/10.1016/J.RSE.2019.111599 Marston, Z. P. D., Cira, T. M., Hodgson, E. W., Knight, J. F., MacRae, I. V., Koch, R. L., and Rondon, S. (2020). Detection of Stress Induced by Soybean Aphid (Hemiptera: Aphididae) Using Multispectral Imagery from Unmanned Aerial Vehicles. Journal of Economic Entomology , 113 (2), 779–786. https://doi.org/10.1093/JEE/TOZ306 Marston, Z. P. D., Cira, T. M., Knight, J. F., Mulla, D., Alves, T. M., Hodgson, E. W., Ribeiro, A. V., MacRae, I. V., and Koch, R. L. (2022). Linear Support Vector Machine Classification of Plant Stress From Soybean Aphid (Hemiptera: Aphididae) Using Hyperspectral Reflectance. Journal of Economic Entomology , 115 (5), 1557–1563. https://doi.org/10.1093/JEE/TOAC077 Matias, F. I., Caraza-Harter, M. V., and Endelman, J. B. (2020). FIELDimageR: An R package to analyze orthomosaic images from agricultural field trials. The Plant Phenome Journal , 3 (1), e20005. https://doi.org/10.1002/PPJ2.20005 McPherson, R. M., Buss, G. R., and Roberts, P. M. (2007). Assessing Stink Bug Resistance in Soybean Breeding Lines Containing Genes from Germplasm IAC-100. Journal of Economic Entomology , 100 (4), 1456–1463. https://doi.org/10.1093/JEE/100.4.1456 Moreira, F. F., Rojas de Oliveira, H., Lopez, M. A., Abughali, B. J., Gomes, G., Cherkauer, K. A., Brito, L. F., and Rainey, K. M. (2021). High-Throughput Phenotyping and Random Regression Models Reveal Temporal Genetic Control of Soybean Biomass Production. Frontiers in Plant Science , 12 , 715983. https://doi.org/10.3389/FPLS.2021.715983/BIBTEX Otsu, N. (1979). A Threshold Selection Method from Gray-Level Histograms. IEEE Trans Syst Man Cybern , SMC - 9 (1), 62–66. https://doi.org/10.1109/TSMC.1979.4310076 Pal, S. K., and Mitra, S. (1992). Multilayer Perceptron, Fuzzy Sets, and Classification. IEEE Transactions on Neural Networks , 3 (5), 683–697. https://doi.org/10.1109/72.159058 Panizzi, A. R., and Hirose, E. (1995). Survival, reproduction and starvation resistance of adult southern green stink bug (Heteroptera: Pnetatomidae) reared on sesame or soybean. Annals of the Entomological Society of America , 88 (5), 661–665. https://doi.org/10.1093/aesa/88.5.661 Panizzi, A. R., and Corrêa-Ferreira, B. S. (1997). Dynamics in the insect fauna adaptation to soybean in the tropics. Trends in Entomology, 1 , 71–88. Patterson, H. D., and Thompson, R. (1971). Recovery of inter-block information when block sizes are unequal. Biometrika , 58 (3), 545–554. https://doi.org/10.1093/BIOMET/58.3.545 Piepho, H. P., and Möhring, J. (2007). Computing heritability and selection response from unbalanced plant breeding trials. Genetics , 177 (3), 1881–1888. https://doi.org/10.1534/GENETICS.107.074229 Pinheiro, J. B., Vello, N. A., Rossetto, C. J., and Zucchi, M. I. (2005). Potential of soybean genotypes as insect resistance sources. Biotechnology , September 2005 , 294–301. https://doi.org/10.12702/1984-7033.v05n03a06 Pisarevsky, V. (2008). OpenCV, The Open Computer Vision Library . QGIS Development Team. (2018). QGIS Geographic Information System. Open Source Geospatial Foundation Project . http://qgis.osgeo.org R Core Team. (2023). R: A Language and Environment for Statistical Computing_. R Foundation for Statistical Computing (4.3.1; p. https://www.R-project.org/). Resende, E. L., Bruzi, A. T., Cardoso, E. da S., Carneiro, V. Q., Souza, V. A. P., Barros, P. H. F. C., & Pereira, R. R. (2024). High-throughput phenotyping: Application in maize breeding. AgriEngineering, 6 (2), 1078–1092. https://doi.org/10.3390/agriengineering6020062 Reynolds, D., Baret, F., Welcker, C., Bostrom, A., Ball, J., Cellini, F., Lorence, A., Chawade, A., Khafif, M., Noshita, K., Mueller-Linow, M., Zhou, J., and Tardieu, F. (2019). What is cost-efficient phenotyping? Optimizing costs for different scenarios. Plant Science , 282 , 14–22. https://doi.org/10.1016/j.plantsci.2018.06.015 Rodene, E., Xu, G., Palali Delen, S., Zhao, X., Smith, C., Ge, Y., Schnable, J., and Yang, J. (2022). A UAV-based high-throughput phenotyping approach to assess time-series nitrogen responses and identify trait-associated genetic components in maize. The Plant Phenome Journal , 5 (1), e20030. https://doi.org/10.1002/PPJ2.20030 Rossetto, C. J., Gallo, P., Razera, L., Bortoletto, N., Igue, T., Medina, P., Tisselli, O., Aquilera, V., and Pinheiro, J. (1995). Mechanisms of resistance to stink bug complex in the soybean cultivar IAC-100. Anais- Sociedade Entomologica Do Brasil , 24 (3), 517–522. Rutkoski, J., Poland, J., Mondal, S., Autrique, E., Pérez, L. G., Crossa, J., Reynolds, M., and Singh, R. (2016). Canopy temperature and vegetation indices from high-throughput phenotyping improve accuracy of pedigree and genomic selection for grain yield in wheat. G3: Genes, Genomes, Genetics, 6 (9), 2799–2808. https://doi.org/10.1534/g3.116.032888 Sakurai, K., Toda, Y., Kajiya-Kanegae, H., Ohmori, Y., Yamasaki, Y., Takahashi, H., Takanashi, H., Tsuda, M., Tsujimoto, H., Kaga, A., Nakazono, M., Fujiwara, T., and Iwata, H. (2022). Time-series multispectral imaging in soybean for improving biomass and genomic prediction accuracy. The Plant Genome , 15 (4), e20244. https://doi.org/10.1002/TPG2.20244 Santos, M. F., Möller, M., Clough, S. J., and Pinheiro, J. B. (2018). Heritability of agronomic traits correlated with reduced stink bug damage in an F2:3 soybean population derived from IAC-100. Journal of Crop Improvement , 32 (1), 1–18. https://doi.org/10.1080/15427528.2017.1370404 Silva, M. F. E., Maciel, G. M., Gallis, R. B. A., Barbosa, R. L., Carneiro, V. Q., Rezende, W. S., and Siquieroli, A. C. S. (2022). High-throughput phenotyping by RGB and multispectral imaging analysis of genotypes in sweet corn. Horticultura Brasileira , 40 (1), 92–98. https://doi.org/10.1590/S0102-0536-2022012 Smith, C. M. (2005). Plant resistance to arthropods: Molecular and conventional approaches. Springer Dordrecht 426 (1) 101-122. https://doi.org/10.1007/1-4020-3702-3 Sun, J., Rutkoski, J. E., Poland, J. A., Crossa, J., Jannink, J.-L., and Sorrells, M. E. (2017). Multitrait, random regression, or simple repeatability model in high-throughput phenotyping data improve genomic prediction for wheat grain yield. The Plant Genome, 10 (2), 1–12. https://doi.org/10.3835/plantgenome2016.11.0111 Tang, H., and He, Z. (2021). Advances and challenges in quantitative delineation of the genetic architecture of complex traits. Quantitative Biology (Beijing, China) , 9 (2), 168. https://doi.org/10.15302/J-QB-021-0249 Trevisan, R., Pérez, O., Schmitz, N., Diers, B., and Martin, N. (2020). High-Throughput Phenotyping of Soybean Maturity Using Time Series UAV Imagery and Convolutional Neural Networks. Remote Sensing , 12 (21), 3617. https://doi.org/10.3390/rs12213617 USDA, U. S. D. of A. (2024). World Agricultural Production . Foreign Agricultural Service/USDA. https://apps.fas.usda.gov/psdonline/circulars/production.pdf Van Der Walt, S., Schönberger, J. L., Nunez-Iglesias, J., Boulogne, F., Warner, J. D., Yager, N., Gouillart, E., and Yu, T. (2014). Scikit-image: Image processing in python. PeerJ , 2014 (1). https://doi.org/10.7717/PEERJ.453 Van Rossum, G., and Drake, F. L. (2009). Python 3 Reference Manual . CreateSpace. Vapnik, V. (2000). SVM method of estimating density, conditional probability, and conditional density. Proceedings - IEEE International Symposium on Circuits and Systems , 2 . https://doi.org/10.1109/ISCAS.2000.856437 Veiga, R. F. A., Rossetto, C. J., Razera, L. F., Gallo, P. B., Bertoletoo, N., Medina, P. F., Tisselli Filho, O., and Cione, J. (1999). caracterização morfológica e agronômica do cultivar de soja ‘Iac-100.’ IAC - Boletim Técnico 177 - Instituto Agronômico , May 2017 , 23p. Vieira, C. C., Sarkar, S., Tian, F., Zhou, J., Jarquin, D., Nguyen, H. T., Zhou, J., and Chen, P. (2022). Differentiate Soybean Response to Off-Target Dicamba Damage Based on UAV Imagery and Machine Learning. Remote Sensing 2022, Vol. 14, Page 1618 , 14 (7), 1618. https://doi.org/10.3390/RS14071618 Yang, G., Liu, J., Zhao, C., Li, Z., Huang, Y., Yu, H., Xu, B., Yang, X., Zhu, D., Zhang, X., Zhang, R., Feng, H., Zhao, X., Li, Z., Li, H., and Yang, H. (2017). Unmanned Aerial Vehicle Remote Sensing for Field-Based Crop Phenotyping: Current Status and Perspectives. Frontiers in Plant Science , 8 . https://doi.org/10.3389/FPLS.2017.01111 Yu, N., Li, L., Schmitz, N., Tian, L. F., Greenberg, J. A., and Diers, B. W. (2016). Development of methods to improve soybean yield estimation and predict plant maturity with an unmanned aerial vehicle based platform. Remote Sensing of Environment , 187 , 91–101. https://doi.org/10.1016/J.RSE.2016.10.005 Yuan, W., Wijewardane, N. K., Jenkins, S., Bai, G., Ge, Y., and Graef, G. L. (2019). Early Prediction of Soybean Traits through Color and Texture Features of Canopy RGB Imagery. Scientific Reports , 9 (1). https://doi.org/10.1038/s41598-019-50480-x Zhang, J., Virk, S., Porter, W., Kenworthy, K., Sullivan, D., and Schwartz, B. (2019). Applications of unmanned aerial vehicle based imagery in turfgrass field trials. Frontiers in Plant Science , 10 , 429695. https://doi.org/10.3389/FPLS.2019.00279/BIBTEX Zhou, J., Zhou, J., Ye, H., Ali, M. L., Nguyen, H. T., and Chen, P. (2020). Classification of soybean leaf wilting due to drought stress using UAV-based imagery. Computers and Electronics in Agriculture , 175 , 105576. https://doi.org/10.1016/J.COMPAG.2020.105576 Zhou, X., Zheng, H. B., Xu, X. Q., He, J. Y., Ge, X. K., Yao, X., Cheng, T., Zhu, Y., Cao, W. X., and Tian, Y. C. (2017). Predicting grain yield in rice using multi-temporal vegetation indices from UAV-based multispectral and digital imagery. ISPRS Journal of Photogrammetry and Remote Sensing , 130 , 246–255. https://doi.org/10.1016/J.ISPRSJPRS.2017.05.003 Zhou, J., Beche, E., Vieira, C. C., Yungbluth, D., Zhou, J., Scaboo, A., & Chen, P. (2022). Improve soybean variety selection accuracy using UAV-based high-throughput phenotyping technology. Frontiers in Plant Science, 12 , 768742. https://doi.org/10.3389/fpls.2021.768742 Additional Declarations No competing interests reported. Supplementary Files Suplementar1MOv2.docx Cite Share Download PDF Status: Posted Version 1 posted 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-7736736","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":540673703,"identity":"b22776e1-2d7e-4a6b-93b9-0cf9807a766d","order_by":0,"name":"Maiara Oliveira","email":"","orcid":"","institution":"University of São Paulo (USP)","correspondingAuthor":false,"prefix":"","firstName":"Maiara","middleName":"","lastName":"Oliveira","suffix":""},{"id":540673704,"identity":"183baaf4-f411-45ae-9063-82f9f75ee48b","order_by":1,"name":"Alexandre Hild Aono","email":"","orcid":"","institution":"University of Campinas (UNICAMP)","correspondingAuthor":false,"prefix":"","firstName":"Alexandre","middleName":"Hild","lastName":"Aono","suffix":""},{"id":540673705,"identity":"def69daf-ad25-4d7f-9580-0af18900455a","order_by":2,"name":"Patricia Braga","email":"","orcid":"","institution":"University of São Paulo (USP)","correspondingAuthor":false,"prefix":"","firstName":"Patricia","middleName":"","lastName":"Braga","suffix":""},{"id":540673706,"identity":"bc228179-ed74-42de-8b08-614d4d49337f","order_by":3,"name":"Adriano Abreu Moreira","email":"","orcid":"","institution":"University of São Paulo (USP)","correspondingAuthor":false,"prefix":"","firstName":"Adriano","middleName":"Abreu","lastName":"Moreira","suffix":""},{"id":540673707,"identity":"ddb39f22-bdc5-4b5c-a47c-737ea5911173","order_by":4,"name":"Fernanda Smaniotto Campion","email":"","orcid":"","institution":"University of São Paulo (USP)","correspondingAuthor":false,"prefix":"","firstName":"Fernanda","middleName":"Smaniotto","lastName":"Campion","suffix":""},{"id":540673708,"identity":"f0c63eca-9003-43e1-9bbe-969f48d5b8e0","order_by":5,"name":"Felipe Augusto Krause","email":"","orcid":"","institution":"University of São Paulo (USP)","correspondingAuthor":false,"prefix":"","firstName":"Felipe","middleName":"Augusto","lastName":"Krause","suffix":""},{"id":540673709,"identity":"5f98ecee-7bb0-44de-9bc2-a4b19b0aaa7d","order_by":6,"name":"Fernando Henrique Iost Filho","email":"","orcid":"","institution":"University of São Paulo (USP)","correspondingAuthor":false,"prefix":"","firstName":"Fernando","middleName":"Henrique Iost","lastName":"Filho","suffix":""},{"id":540673710,"identity":"0f5d7ee5-914f-4161-a8c9-b85a1468d9ab","order_by":7,"name":"Juliano Bastos Pazini","email":"","orcid":"","institution":"University of São Paulo (USP)","correspondingAuthor":false,"prefix":"","firstName":"Juliano","middleName":"Bastos","lastName":"Pazini","suffix":""},{"id":540673711,"identity":"4c119150-a13a-4f39-9345-84259743ff3d","order_by":8,"name":"Pedro Takao Yamamoto","email":"","orcid":"","institution":"University of São Paulo (USP)","correspondingAuthor":false,"prefix":"","firstName":"Pedro","middleName":"Takao","lastName":"Yamamoto","suffix":""},{"id":540673712,"identity":"b178e3c4-d2ba-4f16-9d1e-1779920e3a1f","order_by":9,"name":"José Baldin Pinheiro","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAs0lEQVRIie3PoQ7CMBCA4VtqGzSKvcItvFBRNYOUN1iC4BVK1ocYHtGlydSWWgRiCcnUBA+AoEWAPUlCf3UVX64HkEr9YJmGzHKEVXwwKoFI1nQCy0gANhWZsPow2lndZOMHy/YXysdMh63Badtcd4LpiUK0QMfRBcKRcUsi8hGJRN+TSfneItCWVGJ6FW5xxSnc4jSFFPXxPM5Ply/80N4ViVSfkQMFAOTfkZNAKpVK/WEviWw+OVV4KZwAAAAASUVORK5CYII=","orcid":"","institution":"University of São Paulo (USP)","correspondingAuthor":true,"prefix":"","firstName":"José","middleName":"Baldin","lastName":"Pinheiro","suffix":""}],"badges":[],"createdAt":"2025-09-29 00:08:13","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7736736/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7736736/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":96239022,"identity":"d9d61854-3378-41cf-89cd-026adc8182c2","added_by":"auto","created_at":"2025-11-19 07:01:26","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":7099114,"visible":true,"origin":"","legend":"","description":"","filename":"Artigo1MOv2finalPrecisionAgriculture.docx","url":"https://assets-eu.researchsquare.com/files/rs-7736736/v1/7a3e94784751a0e34318512b.docx"},{"id":96239137,"identity":"783bda01-854d-4afb-9984-dad16585fe94","added_by":"auto","created_at":"2025-11-19 07:03:09","extension":"json","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":11628,"visible":true,"origin":"","legend":"","description":"","filename":"5332ff3861364ddda40bc1c9942d629f.json","url":"https://assets-eu.researchsquare.com/files/rs-7736736/v1/137508502156795400c7a2bc.json"},{"id":95823627,"identity":"a41268ca-4712-40f8-a9cb-e79bd1fc49e1","added_by":"auto","created_at":"2025-11-13 11:04:42","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":3613050,"visible":true,"origin":"","legend":"","description":"","filename":"Suplementar1MOv2.docx","url":"https://assets-eu.researchsquare.com/files/rs-7736736/v1/0ff968bf303557a2bdefeaac.docx"},{"id":95823616,"identity":"7fe0563e-181c-4110-98a7-d2248ebfd96d","added_by":"auto","created_at":"2025-11-13 11:04:41","extension":"xml","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":210797,"visible":true,"origin":"","legend":"","description":"","filename":"5332ff3861364ddda40bc1c9942d629f1enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-7736736/v1/a7a2efa45c759f1d059ff93e.xml"},{"id":96239004,"identity":"e7e2a996-97c0-4509-88fb-c4d569c80378","added_by":"auto","created_at":"2025-11-19 07:00:35","extension":"png","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":745805,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7736736/v1/a825938d3e4c594cf3e60b8d.png"},{"id":96239006,"identity":"5a8aabf4-81cc-4586-a6d7-0a025f20141d","added_by":"auto","created_at":"2025-11-19 07:00:40","extension":"png","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":174581,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage10.png","url":"https://assets-eu.researchsquare.com/files/rs-7736736/v1/535d4ef159bd4888c6320209.png"},{"id":95823599,"identity":"3deac875-e254-4a19-b030-c05ef68b9cad","added_by":"auto","created_at":"2025-11-13 11:04:41","extension":"png","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":130853,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7736736/v1/f44de82ece3cc20bc6bdd541.png"},{"id":95823607,"identity":"a3c55639-eb6d-44d4-95c5-5e15dd8080a3","added_by":"auto","created_at":"2025-11-13 11:04:41","extension":"png","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":104534,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7736736/v1/9ddf9d6e1d67bc7b16a99a52.png"},{"id":95823604,"identity":"fe1ac15a-7fe2-4443-8657-f7c8c88c8cf3","added_by":"auto","created_at":"2025-11-13 11:04:41","extension":"png","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":39894,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-7736736/v1/a7dcd00e4fc856e4395fd567.png"},{"id":95823621,"identity":"ecd05552-983b-42d2-814b-84d65456e433","added_by":"auto","created_at":"2025-11-13 11:04:41","extension":"png","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":246992,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-7736736/v1/f0514e7221f5d1f06bed97a8.png"},{"id":96238978,"identity":"befa236d-56d3-4393-b401-c32cdacde9ac","added_by":"auto","created_at":"2025-11-19 06:59:21","extension":"png","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":29686,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-7736736/v1/3c2a2f51b2fcddd7f91f4512.png"},{"id":96238983,"identity":"3fbf9cc8-93e2-4409-9412-d9648d201977","added_by":"auto","created_at":"2025-11-19 06:59:33","extension":"png","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":958486,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage7.png","url":"https://assets-eu.researchsquare.com/files/rs-7736736/v1/a0989ec20f6ada3e54cc4ccd.png"},{"id":96238984,"identity":"e895bd56-d0d3-40b3-8868-cde1ff4aa498","added_by":"auto","created_at":"2025-11-19 06:59:34","extension":"png","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":108753,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage8.png","url":"https://assets-eu.researchsquare.com/files/rs-7736736/v1/f08637523cf995747b1889f7.png"},{"id":95823610,"identity":"527f0539-1eda-4004-a15b-d9057130d81f","added_by":"auto","created_at":"2025-11-13 11:04:41","extension":"png","order_by":13,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":30586,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage9.png","url":"https://assets-eu.researchsquare.com/files/rs-7736736/v1/1141d59abcb35adac1b166e9.png"},{"id":95823613,"identity":"bf304871-a674-49ca-b220-7352539d1ca9","added_by":"auto","created_at":"2025-11-13 11:04:41","extension":"png","order_by":14,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":134887,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7736736/v1/f1633b9bbaf090ac8c2287d6.png"},{"id":96239557,"identity":"65f8d57c-d288-43c2-b590-67981b6e55fd","added_by":"auto","created_at":"2025-11-19 07:06:58","extension":"png","order_by":15,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":35730,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage10.png","url":"https://assets-eu.researchsquare.com/files/rs-7736736/v1/f46e534b885df7d8af2b4233.png"},{"id":95823615,"identity":"1b5419d8-eb91-4128-b243-bf2cc0ded4ec","added_by":"auto","created_at":"2025-11-13 11:04:41","extension":"png","order_by":16,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":31246,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7736736/v1/db151fea556053e5e08f19f7.png"},{"id":96239077,"identity":"a45607d6-0ec4-4bc9-b062-fcd19d593ca9","added_by":"auto","created_at":"2025-11-19 07:02:22","extension":"png","order_by":17,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":36110,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7736736/v1/0af5f7bc75a23328aef3948a.png"},{"id":95823626,"identity":"24a7e1f4-11df-4004-890e-26f61b35a855","added_by":"auto","created_at":"2025-11-13 11:04:42","extension":"png","order_by":18,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":8644,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-7736736/v1/acd58c87699fdd978a4bc0d9.png"},{"id":95823623,"identity":"e327e715-d791-4ffc-a8e7-7bbcfb91fc64","added_by":"auto","created_at":"2025-11-13 11:04:41","extension":"png","order_by":19,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":53219,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-7736736/v1/0f2ac6973195a603f3e07b63.png"},{"id":95823619,"identity":"91ed7e11-e9d4-44df-a8ae-b1f917559360","added_by":"auto","created_at":"2025-11-13 11:04:41","extension":"png","order_by":20,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":12730,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-7736736/v1/8de6c595a6c7aef66884f5ad.png"},{"id":95823628,"identity":"9316aa75-c10d-4ca9-bea4-eedc993a8c15","added_by":"auto","created_at":"2025-11-13 11:04:42","extension":"png","order_by":21,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":223646,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage7.png","url":"https://assets-eu.researchsquare.com/files/rs-7736736/v1/3c5e25e1431616287d76e14a.png"},{"id":95823625,"identity":"febb2cc8-1f23-417e-90d0-c29467014151","added_by":"auto","created_at":"2025-11-13 11:04:42","extension":"png","order_by":22,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":36954,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage8.png","url":"https://assets-eu.researchsquare.com/files/rs-7736736/v1/b89486688041fb2dc06cc3ea.png"},{"id":96239275,"identity":"f7c9ebb8-a5dc-4017-8d92-bcb671f18198","added_by":"auto","created_at":"2025-11-19 07:05:58","extension":"png","order_by":23,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":41776,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage9.png","url":"https://assets-eu.researchsquare.com/files/rs-7736736/v1/f1a19e441a2b0c987402170e.png"},{"id":96239365,"identity":"23743caf-a000-4bc3-be4d-76fb2f4bed96","added_by":"auto","created_at":"2025-11-19 07:06:23","extension":"xml","order_by":24,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":205995,"visible":true,"origin":"","legend":"","description":"","filename":"5332ff3861364ddda40bc1c9942d629f1structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-7736736/v1/f1e68485d1aa800ad11a1cfa.xml"},{"id":95823629,"identity":"d0635341-b028-4f8d-8557-d99da7344c1a","added_by":"auto","created_at":"2025-11-13 11:04:42","extension":"html","order_by":25,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":223621,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-7736736/v1/19f1ed253b0766da7669d5d6.html"},{"id":96239110,"identity":"75d82b7a-dbb3-432a-b9bc-8c93c29a24f6","added_by":"auto","created_at":"2025-11-19 07:02:53","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":745805,"visible":true,"origin":"","legend":"\u003cp\u003eFlowchart of procedures performed, from acquisition to data analysis.\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7736736/v1/76f2d4510dc7264669c400a0.png"},{"id":95823596,"identity":"562f05a5-393b-43ec-8ea2-e16ee4ebb0b4","added_by":"auto","created_at":"2025-11-13 11:04:41","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":434597,"visible":true,"origin":"","legend":"\u003cp\u003eRepresentation of soybean development stages, indicating the days after planting (DAP) at which drone flights were conducted in the two growing seasons.\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7736736/v1/9fd882580cb3ade6004f32b5.png"},{"id":95823594,"identity":"3d090538-ed12-4014-b5ac-307492b9f29a","added_by":"auto","created_at":"2025-11-13 11:04:41","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":126752,"visible":true,"origin":"","legend":"\u003cp\u003eAverage occurrence of stink bugs during the reproductive period of soybean in the 2019/2020 (Fig. 3A) and 2020/2021 (Fig. 3B) seasons, obtained through beat cloth sampling.\u003c/p\u003e","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7736736/v1/7775ba98c062d596e5efddd7.png"},{"id":96238957,"identity":"a45e75aa-bd49-418b-8449-c5cfd76800b1","added_by":"auto","created_at":"2025-11-19 06:58:18","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":46363,"visible":true,"origin":"","legend":"\u003cp\u003eExplained Percent of variation by variance components, heritability, and accuracy for grain yield (GY), hundred-seed weight (HSW), tolerance (TOL), number of days to maturity (NDM), and leaf retention (LR) in two growing seasons.\u003c/p\u003e","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-7736736/v1/515e9320650016e923d95caa.png"},{"id":96239247,"identity":"816abd9f-ff31-47de-8653-82a0ebf16d9c","added_by":"auto","created_at":"2025-11-19 07:05:52","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":397496,"visible":true,"origin":"","legend":"\u003cp\u003eGenetic correlation between the visual characteristics evaluated: grain yield (GY), hundred-seed weight (HSW), tolerance (TOL), number of days to maturity (NDM), and leaf retention (LR) in two growing seasons, 2019/2020 (blue) and 2020/2021 (red).\u003c/p\u003e","description":"","filename":"floatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-7736736/v1/eea3151a9ad93e905f029a33.png"},{"id":95823601,"identity":"7ff41f61-2c15-4495-acb8-30b84441b22b","added_by":"auto","created_at":"2025-11-13 11:04:41","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":29686,"visible":true,"origin":"","legend":"\u003cp\u003eHeatmaps showing the correlation patterns between image-based indices (NGRDI, GLI, RGBVI, VARI, CC, and PCA) and flight times at different days after planting (DAP) during two growing seasons (2019/2020 and 2020/2021).\u003c/p\u003e","description":"","filename":"floatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-7736736/v1/7806dd446ad2909e08828068.png"},{"id":96239230,"identity":"16faaf51-620d-42c4-99ff-f0bc010e1ec2","added_by":"auto","created_at":"2025-11-19 07:05:45","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":841619,"visible":true,"origin":"","legend":"\u003cp\u003eGenetic correlations between image-derived indices and manually measured traits across different flights for the 2019/2020 (Fig 7A) and 2020/2021 (Fig. 7B) growing seasons. Correlations are displayed for various indices (NGRDI _P25, GLI_P25, RGBVI_P25, Vari_P25, CC, PCA) and traits (GY, HSW, LR, NDM, TOL) at multiple DAP (days after planting).\u003c/p\u003e","description":"","filename":"floatimage7.png","url":"https://assets-eu.researchsquare.com/files/rs-7736736/v1/ce67b2c7ce1a35bcf85a76e2.png"},{"id":95823602,"identity":"90335f83-cd00-46ec-a065-50617c27dc38","added_by":"auto","created_at":"2025-11-13 11:04:41","extension":"png","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":108753,"visible":true,"origin":"","legend":"\u003cp\u003eThe boxplot presents the predictive ability (PA) values of the best machine learning models for each manually measured agronomic trait at different flight dates during the 2019/2020 and 2020/2021 seasons. The evaluations were performed using 10-fold cross-validation repeated 50 times, with Pearson correlation for regression models (GY, HSW, TOL, and NDM) and accuracy for classification models (LR). The flight dates and models are indicated on the x-axis, while the PA are shown on the y-axis.\u003c/p\u003e","description":"","filename":"floatimage8.png","url":"https://assets-eu.researchsquare.com/files/rs-7736736/v1/c65241a6f191fabd7ca06613.png"},{"id":96239292,"identity":"3b51cab5-94f7-4bae-a766-d6665f957f1a","added_by":"auto","created_at":"2025-11-19 07:06:01","extension":"png","order_by":9,"title":"Figure 9","display":"","copyAsset":false,"role":"figure","size":30586,"visible":true,"origin":"","legend":"\u003cp\u003ePredictive abilities (PA) of three different ML models—AdaBoost, MLP, and SVM—evaluated for the five manually measured traits across different tested scenarios. The evaluations were performed using 10-fold cross-validation repeated 50 times PAs were assessed using Pearson correlation for the regression models on traits GY, NDM, HSW, and TOL. For the classification model applied to the LR trait, accuracy was used as the evaluation metric.\u003c/p\u003e","description":"","filename":"floatimage9.png","url":"https://assets-eu.researchsquare.com/files/rs-7736736/v1/faa3c1030aa899f4be823156.png"},{"id":96239041,"identity":"c7d4e71a-ed3d-463b-b88d-4430c6712577","added_by":"auto","created_at":"2025-11-19 07:01:51","extension":"png","order_by":10,"title":"Figure 10","display":"","copyAsset":false,"role":"figure","size":174581,"visible":true,"origin":"","legend":"\u003cp\u003eCorrelations between the VARI_P25 index and the phenotypic traits GY, HSW, LR, NDM, and TOL in the 2019/2020 and 2020/2021 seasons.\u003c/p\u003e","description":"","filename":"floatimage10.png","url":"https://assets-eu.researchsquare.com/files/rs-7736736/v1/c1c13dde5bdbaf5b2290e87b.png"},{"id":96254497,"identity":"e8f840d9-7cd8-437c-a11d-60059b9078a0","added_by":"auto","created_at":"2025-11-19 07:46:46","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":3677380,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7736736/v1/f83dec3f-a23b-4920-bb21-198438cdb8e0.pdf"},{"id":95823624,"identity":"840827b2-753d-4dfd-ab80-f010ac297964","added_by":"auto","created_at":"2025-11-13 11:04:42","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":3613050,"visible":true,"origin":"","legend":"","description":"","filename":"Suplementar1MOv2.docx","url":"https://assets-eu.researchsquare.com/files/rs-7736736/v1/fa805e94f751d3eb916db10e.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Optimizing Soybean Breeding: High-Throughput Phenotyping for Stink Bug Resistance and High Yields","fulltext":[{"header":"Highlights","content":"\u003cp\u003eVARI_P25 consistently emerged as the most robust index, with high heritability, strong trait correlations, and key importance in ML models.\u003c/p\u003e\u003cp\u003eUAV-based RGB imagery revealed strong correlations between image-derived indices and resistance-related traits, often more informative than ML predictions.\u003c/p\u003e\u003cp\u003eTexture indices (TIs) provided complementary structural information in decision tree\u0026ndash;based models, though further studies are needed to validate their role.\u003c/p\u003e\u003cp\u003eCross-season predictions remain challenging, but UAV-derived indices proved valuable for early screening and resource optimization in breeding pipelines.\u003c/p\u003e"},{"header":"Introduction","content":"\u003cp\u003eSoybean (\u003cem\u003eGlycine max\u003c/em\u003e) is one of the most widely cultivated oilseed crops in the world due to its versatility, being used in the production of oil, meal, biodiesel, and proteins for animal and human nutrition (Hartman et al., \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2011\u003c/span\u003e). In the 2023/24 season, global soybean production totaled 396.9\u0026nbsp;million tons (USDA, \u003cspan citationid=\"CR63\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Brazil is currently the world\u0026rsquo;s leading soybean producer, with an estimated output of 154.6\u0026nbsp;million tons and an average yield of 3.5 t/ha (CONAB, 2024). Nowadays, soybean is cultivated in almost all regions of Brazil, favoring the emergence of various insect pests and pathogens that cause significant losses in grain yield (GY) and quality (Boerma and Walker, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2005\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eAmong these pests, stink bugs (Hemiptera: Pentatomidae) are particularly damaging, with potential yield losses of up to 125 kg/ha per insect per linear meter (Guedes et al., \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2012\u003c/span\u003e). These insects feed directly on soybean pods, affecting the grains and causing losses not only in yield but also in the physiological and phytosanitary quality of the seeds (Bortolotto et al., \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2015\u003c/span\u003e). In addition, stink bug\u0026rsquo;s feeding process causes soybean delayed maturity, which hinders the determination of the harvest time and mechanical harvesting due to leaf retention and green stems (Panizzi and Hirose, \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e1995\u003c/span\u003e; Rossetto et al., \u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e1995\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eGenetic resistance remains one of the most efficient and sustainable strategies for stink bug control and is often integrated with chemical approaches. However, resistance to stink bugs is a complex, polygenic, and quantitatively inherited trait, which makes its selection in breeding programs particularly challenging (Godoi and Pinheiro, \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2009\u003c/span\u003e). Plant breeding programs targeting quantitative traits require a large and diverse population, which needs to be evaluated in a feasible and precise way. In this sense, phenotyping is one of the main bottlenecks of breeding, as it typically relies on manual and visual measurements, which are time-consuming and imprecise. This limitation restricts the number of lines that can be tested (Bhat and Yu, \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2021\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eDeveloping high throughput phenotyping (HTP) methodologies has enhanced the phenotypic evaluation of genotypes, optimizing diverse breeding programs (Li et al., \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2014\u003c/span\u003e). HTP methodologies, particularly utilizing autonomous platforms like unmanned aerial vehicles (UAVs) coupled with cameras, have emerged as valuable tools in assisting breeders in selecting and developing commercial cultivars. Furthermore, the use of images facilitates non-destructive evaluation of numerous traits across a large number of lines at various growth stages and locations simultaneously (Furbank and Tester, \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2011\u003c/span\u003e; Reynolds et al., \u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e2019\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eThe incorporation of image-based phenotyping in soybean breeding has been prominent in predicting many important traits, including maturity (Trevisan et al., \u003cspan citationid=\"CR62\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; Yu et al., \u003cspan citationid=\"CR70\" class=\"CitationRef\"\u003e2016\u003c/span\u003e), biomass (Moreira et al., \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Sakurai et al., \u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e2022\u003c/span\u003e), drought resistance (Zhou et al., \u003cspan citationid=\"CR73\" class=\"CitationRef\"\u003e2020\u003c/span\u003e), and yield estimation (Maimaitijiang et al., \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; Zhou et al., \u003cspan citationid=\"CR75\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). However, the use of HTP for quantifying and identifying insect injury in soybeans is still in its early development.\u003c/p\u003e\u003cp\u003eMarston et al. (\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e2020\u003c/span\u003e) demonstrated the effectiveness of using drone-based multispectral imagery and machine learning (ML) models to identify stress caused by soybean aphids, noting a reduction in near-infrared reflectance as aphid populations increased. Conversely, Iost-Filho et al. (\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2022\u003c/span\u003e) found that stink bug infestation did not significantly alter spectral reflectance in soybean under controlled conditions, even when using hyperspectral sensors. Thus, applying these technologies in field conditions for the selection of resistant genotypes still requires further research to increase the efficiency and accuracy of the selection process (Goggin et al., \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e2015\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eThis study aimed to evaluate the potential of UAV-based RGB imagery and image-derived features to support phenotyping for resistance to the stink bug complex in soybean under natural field infestations. A population of 290 soybean lines was assessed across two seasons using an alpha-lattice design. By contrasting different image-derived descriptors and their association with conventional phenotyping measurements, this study highlights effective strategies for soybean image-based phenotyping in resistance evaluation. Furthermore, we assessed the predictive performance of ML algorithms in predicting key agronomic traits from image features, providing an end-to-end workflow with strong potential to support soybean breeding.\u003c/p\u003e"},{"header":"Materials and Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\u003ch2\u003ePlant Material and Field Experiment\u003c/h2\u003e\u003cp\u003eFor the conduct of this study, we utilized a breeding population consisting of 290 soybean lines, part of the germplasm bank of the Laboratory of Genetic Diversity and Breeding at Luiz de Queiroz College of Agriculture/University of S\u0026atilde;o Paulo (ESALQ/USP). These lines resulted from crosses among the commercial cultivars BRS-133, CD-215, Conquista, Dowling, IAC-100, and Pintado, which were recombined with the plant introduction (PI) genotypes: PI 200487 (Kinoshita), PI 471904 (Orba), PI 200526 (Shiranui), and PI 459025 (Bing Nan). Progenies from these crosses were subsequently recombined with nine populations that exhibited genetic variability for stink bug resistance. These genotypes underwent two selection cycles focusing on resistance to the soybean stink bug complex, in addition to one selection cycle for high GY.\u003c/p\u003e\u003cp\u003ePhenotypic evaluations were performed during the 2019/2020 and 2020/2021 growing seasons in the municipality of Piracicaba/SP (Brazil), at the Anhumas Experimental Genetics Station (22\u0026deg;50'00' S, 48\u0026deg;05'00' W, altitude of 460 m). In both seasons, the 290 lines, along with 14 commercial checks recommended for the region and/or that exhibited resistance to stink bugs, were evaluated using a 16x19 alpha-lattice experimental design with three replicates. Experimental plots consisted of two rows, each 4 meters in length, spaced 0.5 meters apart, and with a seeding density of 15 seeds per meter. The experiments were conducted following the recommended agronomic management for soybean cultivation, except that chemical control of insects was not performed, aiming to ensure the presence of stink bugs in the experimental area, enabling the evaluation of the lines' resistance to this insect.\u003c/p\u003e\u003cp\u003eTo monitor the population and verify the presence of stink bugs in the experimental area, beat cloths were employed from flowering to full maturation. Beat cloths were used weekly throughout the entire reproductive cycle of the plant, covering the period of stink bug occurrence and attack. Evaluations involved the use of a beat cloth along a 1-meter crop row, with random beating across the entire experiment. Adults, nymphs (third, fourth, and fifth instars), and the total population (adults and nymphs) of soybean stink bugs were counted.\u003c/p\u003e\u003c/div\u003e\n\u003ch3\u003eManually Measured Agronomic Traits\u003c/h3\u003e\n\u003cp\u003eThe evaluation of soybean genotypes involved the manual measurement of five key agronomic traits. This encompassed the assessment of the number of days to maturity (NDM), which represented the duration from planting until 95% of pods achieved maturity, thereby indicating the genotype's cycle. Leaf retention (LR) was appraised at maturity using a visual scale ranging from 1 to 5, where a score of 1 indicated plants undergoing normal senescence, while a score of 5 indicated plants retaining multiple stems and green leaves, rendering harvest impractical. GY was determined post-harvest, representing the total mass of seeds produced in the plot, corrected for 13% grain moisture, and converted to kilograms per hectare (kg.ha\u003csup\u003e\u0026minus;\u0026thinsp;1\u003c/sup\u003e). Healthy seed weight (HSW) was determined as the weight of seeds devoid of stink bug damage, evaluated after harvest and grain processing (kg.ha\u003csup\u003e\u0026minus;\u0026thinsp;1\u003c/sup\u003e). This measure was obtained during seed processing, where they undergo separation through a spiral mechanism, where shriveled, green, and malformed grains are separated by the action of gravity and centrifugation (da Rocha et al., \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2014\u003c/span\u003e). Tolerance (TOL) was calculated to assess the genotype's resilience to stink bug injury, quantified through an index obtained by the following formula:\u003cdiv id=\"Equ1\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ1\" name=\"EquationSource\"\u003e\n$$\\:TOL=\\:\\left(1-\\left(\\frac{GY-HSW}{GY}\\right)\\right)\\:\\times\\:100$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e1\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e\n\u003ch3\u003eUAV Data Collection\u003c/h3\u003e\n\u003cp\u003eAerial images were collected by a DJI Phantom 4 Pro V2.0 drone (DJI, Shenzhen, China) equipped with a 20-megapixel RGB camera capturing images in the visible spectrum. Flight planning was conducted using the FieldAgent digital platform (Sentera, Minnesota, USA) (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003ea). The camera captured sequential images of the area through progressive scanning.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eDuring the crop season of 2019/2020, six flights were conducted, while in the 2020/2021 season, this number was expanded to ten different dates. Detailed information on the days after planting (DAP) when each flight was conducted can be found in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e. All flights were executed at the height of 20 meters, with a ground sampling distance (GSD) of 1.8 centimeters. An 80% level of both frontal and lateral overlap was maintained between the acquired images. Additionally, flights were conducted on sunny, cloud-free days, within the time frame of 10 a.m. to 2 p.m., aiming to optimize lighting conditions and minimize environmental variabilities.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\n\u003ch3\u003eImage Processing\u003c/h3\u003e\n\u003cp\u003eAfter acquiring the aerial images, individual orthomosaics were created for each field (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eb) using Agisoft MetaShape software (Agisoft LLC, St. Petersburg, Russia), following the standard protocol provided by the software for orthomosaic generation. Subsequently, the orthomosaics were imported into QGIS version 3.02 software (QGIS Development Team, \u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e2018\u003c/span\u003e), where shapefiles were delineated, defining the boundaries of each experimental plot (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003ec.1). To mitigate potential interferences resulting from leaf overlap between adjacent plots, a negative buffer of 0.10 m was applied. Each plot was identified with an ID based on the experimental design. The FieldImageR 0.6.0 package (Matias et al., \u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e2020\u003c/span\u003e) implemented in R was used to crop each experimental area simultaneously, employing the shapefile as the plot boundary (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003ec.2).\u003c/p\u003e\u003cp\u003eAfter individualizing and cropping each plot, background removal was performed, including shadows, soil, and vegetative residues, using the methodology proposed by Yuan et al. (\u003cspan citationid=\"CR71\" class=\"CitationRef\"\u003e2019\u003c/span\u003e), which employs a plant segmentation method using vegetation indices to maximize segmentation accuracy. All the steps were implemented using Python v3 (Van Rossum and Drake, \u003cspan citationid=\"CR65\" class=\"CitationRef\"\u003e2009\u003c/span\u003e) and the OpenCV library (Pisarevsky, \u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e2008\u003c/span\u003e). First, all the plot images were rotated to a 90-degree orientation to calculate the vegetation indices of modified excess green (MExG) and the color index of vegetation extraction (CIVE). By subtracting these indices (MExG - CIVE), we scaled the resulting image into a 0-255 range and used this matrix to create a segmentation mask with Otsu's thresholding technique (Otsu, \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e1979\u003c/span\u003e) (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003ec.4).\u003c/p\u003e\n\u003ch3\u003eImage-Derived Traits\u003c/h3\u003e\n\u003cp\u003eFor each segmented image corresponding to an orthomosaic at different flight dates, we calculated different vegetation indices and features describing their physical properties, including color, gradient, and texture. We employed the Python v3 programming language (Van Rossum and Drake, \u003cspan citationid=\"CR65\" class=\"CitationRef\"\u003e2009\u003c/span\u003e) with the libraries OpenCV (Pisarevsky, \u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e2008\u003c/span\u003e) and scikit-image (Van Der Walt et al., \u003cspan citationid=\"CR64\" class=\"CitationRef\"\u003e2014\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eInitially, the Canopy Cover Index (CC) was computed, representing the percentage of remaining pixels after image segmentation. Additionally, six VIs were calculated (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e), including the triangular greenness index (TGI), the green leaf index (GLI), the normalized green-red difference index (NGRDI), the red-green-blue index (RGBVI), and the visible atmospherically resistant index (VARI). Based on the images resulting from each index calculation, we summarized the value distribution into different descriptive statistics: mean, standard deviation (std), skewness (skew), kurtosis, and 25, 50, and 75 percentiles, denoted as P25, P50, and P75, respectively.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eSummary of vegetation indices (VI) used in this study. Each VI was calculated using an RGB image, where the image channels are red (R), green (G), and blue (B).\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"4\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eVegetation Index\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eName\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eFormula\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eReferences\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eTGI\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTriangular greenness index\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:G-0.39\\times\\:R-0.61\\times\\:B\\)\u003c/span\u003e\u003c/span\u003e (2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e(Hunt et al., \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2005\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eGLI\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eGreen leaf index\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\frac{(2\\times\\:G-R-B)}{(2\\times\\:G+R+B)}\\)\u003c/span\u003e\u003c/span\u003e (3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e(Bassine et al., \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2019\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNGRDI\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eNormalized green-red difference index\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\frac{G-R}{G+R}\\)\u003c/span\u003e\u003c/span\u003e (4)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e(Hunt et al., \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2005\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRGBVI\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eRed-green-blue index\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\frac{{G}^{2}-R*B}{{G}^{2}+R*B}\\)\u003c/span\u003e\u003c/span\u003e (5)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e(Bendig et al., \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2015\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eVARI\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eVisible atmospherically resistant index\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\frac{G-R}{G+R-B}\\)\u003c/span\u003e\u003c/span\u003e (6)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e(Gitelson et al., \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2002\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003eTo reflect different wavelength values, we created a gray-level histogram with 64 bins, considering each bin quantification as a different attribute. Texture measures were calculated using gray level co-occurrence matrices (GLCMs) to describe repetitive patterns in the image, varying in size and providing information about roughness and regularity. For each image, we constructed four different GLCMs corresponding to angles of 0, 45, 90, and 135 degrees. These matrices were summarized into six different features based in Texture Indices (TIs): energy, correlation, angular second moment (ASM), dissimilarity, and contrast (Haralick et al., \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e1973\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eTherefore, the complete characterization of each image was based on 124 features: 1 feature measuring CC, 7 descriptive statistics for each of the 5 vegetation indices (35 features), 64 features corresponding to color histogram quantifications (Hist_1 to Hist_64), and 6 features for each of the 4 GLCMs (24 features). Associations between these features and the manually measured traits were performed with Pearson correlations in software R version 4.3.2 (R Core Team, \u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e2023\u003c/span\u003e).\u003c/p\u003e\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\u003ch2\u003eStatistical Analysis of Phenotypic Data\u003c/h2\u003e\u003cp\u003eThe phenotypic data from all trials were analyzed in R (R Core Team, \u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e2023\u003c/span\u003e) using the ASReml-R package version 4.3.2 (Butler et al., \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). Variance components and genetic parameters were estimated using the restricted maximum likelihood (REML) method (Patterson and Thompson, \u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e1971\u003c/span\u003e), while genetic values were predicted using the best linear unbiased prediction (BLUP) method (Henderson, \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e1982\u003c/span\u003e). The significance of random effects was assessed through likelihood ratio tests (LRTs).\u003c/p\u003e\u003cp\u003eFor the five manually measured traits in the study and each of the six indices obtained from images, statistical analyses were conducted for each year, considering each flight individually. The analyses followed the model:\u003cdiv id=\"Equ2\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ2\" name=\"EquationSource\"\u003e\n$$\\:{y}_{ijk}=\\mu\\:+{g}_{i}+{r}_{k}{\\:+\\:b}_{j\\left(k\\right)}+{\\epsilon\\:}_{ijk}$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e7\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003ewhere, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{y}_{ijk}\\)\u003c/span\u003e\u003c/span\u003e represents the phenotype observed for the \u003cem\u003ei\u003c/em\u003e-th genotype, within the \u003cem\u003ej\u003c/em\u003e-th block, across the \u003cem\u003ek\u003c/em\u003e-th replicate; \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\mu\\:\\)\u003c/span\u003e\u003c/span\u003e the overall mean; \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{g}_{i}\\)\u003c/span\u003e\u003c/span\u003e the random effect associated with the \u003cem\u003ei\u003c/em\u003e-th genotype (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{g}_{i}\\:\\sim\\:N(0,\\:{\\sigma\\:}_{g}^{2})\\)\u003c/span\u003e\u003c/span\u003e with \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\sigma\\:}_{g}^{2}\\)\u003c/span\u003e\u003c/span\u003e representing the genetic variance); \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\:{r}_{k}\\)\u003c/span\u003e\u003c/span\u003e the fixed effect of the \u003cem\u003ek\u003c/em\u003e-th replicate; \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{b}_{j\\left(k\\right)}\\)\u003c/span\u003e\u003c/span\u003ethe random effect of the \u003cem\u003ej\u003c/em\u003e-th block within the \u003cem\u003ek\u003c/em\u003e-th repetition (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{b}_{j}\\:\\sim\\:N(0,\\:{\\sigma\\:}_{j}^{2})\\)\u003c/span\u003e\u003c/span\u003e with \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\sigma\\:}_{j}^{2}\\)\u003c/span\u003e\u003c/span\u003erepresenting the block variance); and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\epsilon\\:}_{ijk}\\)\u003c/span\u003e\u003c/span\u003e the model residual, assuming that the residuals are independent and normally distributed, i.e., \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\epsilon\\:}_{ijk}\\:\\sim\\:N(0,\\:{\\sigma\\:}^{2})\\)\u003c/span\u003e\u003c/span\u003e, where \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\sigma\\:}^{2}\\)\u003c/span\u003e\u003c/span\u003e is the residual variance.\u003c/p\u003e\u003cp\u003eHeritability in the broad sense (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{h}^{2}\\)\u003c/span\u003e\u003c/span\u003e) was estimated using the method proposed by Piepho and Mohring (2007):\u003cdiv id=\"Equ3\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ3\" name=\"EquationSource\"\u003e\n$$\\:\\left({h}^{2}\\right)=\\frac{{\\sigma\\:}_{g}^{2}}{{\\sigma\\:}_{g}^{2}+\\frac{{\\sigma\\:}^{2}}{\\:r\\:}}$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e8\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003ewhere \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\sigma\\:}_{g}^{2}\\)\u003c/span\u003e\u003c/span\u003e and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\sigma\\:}^{2}\\)\u003c/span\u003e\u003c/span\u003e are the genotypic and error variances, respectively, and \u003cem\u003er\u003c/em\u003e is the number of replicates. The correlation between the true and predicted random effects was employed as a metric to evaluate the accuracy:\u003cdiv id=\"Equ4\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ4\" name=\"EquationSource\"\u003e\n$$\\:acc=\\sqrt{1-\\frac{PEV\\left({\\widehat{a}}_{i}\\right)}{{\\widehat{\\sigma\\:}}_{a}^{2}}}$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e9\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003ewhere \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:PEV\\left({\\widehat{a}}_{i}\\right)\\)\u003c/span\u003e\u003c/span\u003e is the prediction error variance of the predicted random effect and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\widehat{\\sigma\\:}}_{a}^{2}\\)\u003c/span\u003e\u003c/span\u003e is the estimated variance of the random effect. Using the BLUPs of each genotype, Pearson's correlations were performed between the traits.\u003c/p\u003e\u003c/div\u003e\n\u003ch3\u003eMachine Learning Methods\u003c/h3\u003e\n\u003cp\u003eThe scikit-learn library (Van Der Walt et al., \u003cspan citationid=\"CR64\" class=\"CitationRef\"\u003e2014\u003c/span\u003e), developed in Python v3 programming language (Van Rossum and Drake, \u003cspan citationid=\"CR65\" class=\"CitationRef\"\u003e2009\u003c/span\u003e), was employed to implement the ML-based prediction models. We utilized the features calculated from images collected via UAV as independent variables and the five visual characteristics as dependent variables. Each visual characteristic was associated with the UAV-collected indices according to the flight date. For the GY, HSW, TOL and NDM characteristics, regression models were applied, while for LR, a classification model was adopted. Subsequently, six distinct forecasting scenarios were tested. Two scenarios considered all flights conducted in each season (2019/2020 and 2020/2021), while two others focused exclusively on the last three flights of each season. Additionally, two scenarios were designed to test cross-year prediction: the first utilized data from 2019/2020 to predict the phenological characteristics of 2020/2021 based solely on flights near the maturation period, and the second incorporated all flights.\u003c/p\u003e\u003cp\u003eThree machine learning methods were explored in this study: support vector machine for regression (SVR) (Vapnik, \u003cspan citationid=\"CR66\" class=\"CitationRef\"\u003e2000\u003c/span\u003e), multilayer perceptron (MLP) neural network (Pal and Mitra, \u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e1992\u003c/span\u003e), and adaptive boosting (AdaBoost) (Freund and Schapire, \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e1997\u003c/span\u003e). Additionally, we applied a decision tree to investigate which characteristics obtained through images are more important for predicting the manually measured agronomic traits, estimating the feature importance of each image feature using the Gini index.\u003c/p\u003e\u003cp\u003eThe performance of the models was evaluated through a 10-fold cross-validation scenario. For the regression models, the Pearson correlation coefficient (\u003cem\u003eR\u003c/em\u003e) and mean squared error (\u003cem\u003eMSE\u003c/em\u003e) were used as evaluation criteria. For the classification models, accuracy and precision metrics were employed to assess the model's ability to classify the test data correctly. These metrics were calculated by comparing the predicted and actual values in the training and test sets.\u003c/p\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\u003ch2\u003eAgronomic Trait Evaluations\u003c/h2\u003e\u003cp\u003eTo develop cultivars resistant to stink bugs, the first step is to identify promising genotypes under natural field pressure. Because infestation intensity varies between seasons, monitoring is essential to characterize the selection environment. In this study, stink bug incidence was recorded weekly with beat cloth sampling throughout the reproductive period. Infestation levels in both seasons remained within the recommended thresholds for effective pest management, and UAV flights were scheduled to coincide with infestation peaks (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e).\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eThe 2019/2020 season exhibited a significantly higher infestation, reaching an average of 18 stink bugs per beat cloth at the peak of infestation, 100 days after planting (DAP). In contrast, the subsequent 2020/2021 season had a lower infestation, with a peak of only 3 stink bugs per beat cloth at 120 DAP, along with other smaller peaks recorded at 80 and 100 DAP, averaging 2.5 stink bugs. These contrasting infestation dynamics provided two distinct selection environments for evaluating soybean resistance to the stink bug complex.\u003c/p\u003e\u003cp\u003eTo evaluate the response of soybean genotypes under natural stink bug infestation, variance component analysis was performed for five agronomic traits (LR, NDM, GY, TOL, and HSW) across two growing seasons. This approach allowed partitioning the total phenotypic variation into genetic and block effects, as well as estimating trait \u003cem\u003eh\u0026sup2;\u003c/em\u003e and prediction \u003cem\u003eacc\u003c/em\u003e. Both variance components were significant according to LRT tests (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e).\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eGenetic variance (σG) explained a large proportion of trait variation in both years. In 2019, σG accounted for 34% of the variation in GY and up to 44% for HSW, TOL, and LR. In 2020, the pattern was similar, although the contribution of σG was reduced for GY and HSW, while remaining high for NDM and TOL. h\u0026sup2; were consistently high, ranging from 0.72 (NDM, GY) to 0.77 (HSW) in 2019, and from 0.65 to 0.90 in 2020. acc followed the same trend, with values between 0.83 and 0.87 in 2019 and slightly higher in 2020, with a range from 0.79 to 0.94, particularly for GY and TOL. These results highlight the strong genetic control of the evaluated traits and the reliability of phenotypic predictions across seasons.\u003c/p\u003e\u003cp\u003eFigure \u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e shows the genetic correlations among the manually measured traits across the two seasons. All correlations were significant, though their magnitude and direction varied between years. Notably, there is a strong correlation between GY and HSW, the latter often used in the selection of more productive and stink bug-resistant genotypes. This correlation was higher in the 2020 season, reaching a value of 0.87, compared to 0.53 in the 2019 season.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e\u003ch2\u003eImage-Derived Traits\u003c/h2\u003e\u003cp\u003eAfter image pre-processing, a total of 124 indices were extracted per flight, including color, texture, and gradient histogram features. To reduce redundancy, indices were filtered based on their correlation with manually measured traits, using a threshold of r\u0026thinsp;\u0026gt;\u0026thinsp;0.6, resulting in a filtered set of 38 features. Among these, 19 were VIs derived from different descriptive statistics. Given the high correlation among these indices (Supplementary Fig.\u0026nbsp;1), we prioritized those with the highest correlations across most flights. This resulted in the selection of four key VIs at the 25th percentile (VARI_P25, RGBVI_P25, NGRDI_P25, and GLI_P25).\u003c/p\u003e\u003cp\u003eFor TIs, all those with a correlation greater than 0.6 in any flight were subjected to Principal Component Analysis (PCA), using the first principal component for subsequent analyses (Supplementary Fig.\u0026nbsp;2). In total, 18 TIs were selected and included in the PCA: ASM, Dissimilarity, Energy, and Homogeneity at angles of 0\u0026deg;, 45\u0026deg;, 90\u0026deg;, and 135\u0026deg;, as well as Correlation at angles of 45\u0026deg; and 135\u0026deg;. These TIs are complementary, as each captures different aspects of the structural and textural variation of the vegetation. Therefore, TI were summarized into the first principal component, which captured the main structural variation.\u003c/p\u003e\u003cp\u003eAdditionally, the CC index was included in the selection due to its high correlation with the evaluated traits. In total, six indices (VARI_P25, RGBVI_P25, NGRDI_P25, GLI_P25, CC, and PCA) were carried forward for correlation analyses with agronomic traits. Color histogram quantifications did not exceed the correlation threshold and were excluded.\u003c/p\u003e\u003cp\u003eAfter selecting the indices, we integrated them into mixed models to understand how the comprehension of different characteristics contributes to overall plant performance and facilitates the identification of genotypes with desirable traits. Figure\u0026nbsp;6 shows the correlation structure among indices across flight dates in both seasons. As expected, correlations were stronger between flights conducted at closer intervals and declined as the temporal distance increased. This is evident in the clusters, where closer flight times show significant correlations, whereas more distant flights present correlations close to zero. Given this pattern, each flight time was analyzed individually to understand how different growth stages influence the evaluated traits. By considering each flight time as a distinct dependent variable, we obtain a more precise and detailed analysis, which allows us to identify the specific moment in the crop stages that is most suitable for performing flights to assess the traits of interest.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e\u003cp\u003eThe decomposition of variance components and genetic parameters for the six image-derived traits across flight dates are shown in Supplementary Figs.\u0026nbsp;3 and 4. It is observed that, across all flight dates, the traits CC and PCA, showed the lowest h\u0026sup2; estimates for both seasons. In 2019, flights at 90 and 97 DAP (grain filling period) presented the lowest h\u0026sup2; estimates for all traits. Additionally, the VIs demonstrated similarity in h\u0026sup2;, with the Vari_P25 at 106 DAP standing out with a h\u0026sup2; of 0.76. Overall, the acc estimates ranged from moderate to high, though some flight periods showed lower values. In 2020, flights at 57, 65, and 122 DAP were associated with lower heritability, while VARI_P25 again showed the highest estimate (h\u0026sup2; = 0.77) at 107 DAP. Acc followed a similar pattern to the previous season. Likelihood ratio tests confirmed that genetic variance significantly explained trait variation in all cases, except for PCA at 97 DAP in 2019.\u003c/p\u003e\u003cp\u003eBuilding upon these findings, we next examined the relationships between image-derived indices and manually measured traits. Figure\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e7\u003c/span\u003e shows the genetic correlations between UAV-based indices and field measurements across flight dates. Strong correlations were observed for all manually measured traits in both seasons, indicating that specific flight periods provided more informative associations with key traits relevant to soybean resistance to stink bugs.\u003c/p\u003e\u003cp\u003eThe strongest correlations were observed for LR and NDM in the 2019 season, during flights close to harvest and maturity, reaching 0.76. This same pattern was observed in the subsequent season, but with slightly lower correlation magnitudes. For GY and HSW, strong correlations were also found in the 2020 season during the early flights (57 to 78 DAP), while mid-season flights (85\u0026ndash;107 DAP) showed weaker associations. Moreover, GY and HSW strongly correlated with multiple indices in the later flights. For TOL, although the correlations were not as strong as with the other traits, in the 2019 season, where stink bug infestation was more pronounced, significant correlations above \u0026minus;\u0026thinsp;0.5 and 0.5 were observed during flights close to maturity with the CC and PCA indices.\u003c/p\u003e\u003cp\u003eThe VIs exhibit similar patterns, reflecting their shared capacity to capture related aspects of canopy reflectance. Likewise, CC, derived from soil\u0026ndash;leaf segmentation, shows patterns similar to those of the VI based on RGB bands. In the early flights, VIs and the CC show positive correlations with GY, HSW, and TOL, which turn negative in later flights. In contrast, the PCA index displayed distinct and often inverse correlation patterns, suggesting that texture-based metrics captured complementary structural information not reflected by VIs.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eIn summary, these analyses revealed consistent and informative patterns across seasons, with VIs generally outperforming TIs in terms of correlation and h\u0026sup2;. Among them, VARI_P25 was particularly informative, showing strong and consistent associations with multiple traits across flights and seasons, together with high h\u0026sup2; estimates. In 2020/2021, early flights (57\u0026ndash;78 DAP) were most informative for GY and HSW, whereas later flights (107\u0026ndash;122 DAP) captured stronger associations for TOL, NDM, and LR. In 2019/2020, when no flights were conducted near flowering, the most informative periods occurred closer to maturity.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec14\" class=\"Section2\"\u003e\u003ch2\u003eMachine Learning-Based Prediction\u003c/h2\u003e\u003cp\u003eWe evaluated the predictive ability (PA) of three machine learning models (AdaBoost, MLP, and SVM) for five manually measured traits using UAV-derived indices. Predictions were made for each flight date to identify the most informative periods for each trait (Supplementary Tables\u0026nbsp;1\u0026ndash;2).\u003c/p\u003e\u003cp\u003eAcross both seasons, AdaBoost consistently achieved the highest PA, particularly for grain yield (GY). In 2020/2021, early flights (65\u0026ndash;78 DAP) produced the strongest results, with AdaBoost correlations exceeding 0.80, while predictive ability declined in later flights. In 2019/2020, the best predictions occurred later, at 120 DAP. Similar patterns were observed for HSW, with AdaBoost again outperforming the other models. In 2020/2021, HSW prediction was strongest in early flights (65\u0026ndash;78 DAP), whereas in 2019/2020 the best results occurred at 120\u0026ndash;127 DAP.\u003c/p\u003e\u003cp\u003eFor NDM and TOL, the MLP model showed consistently lower performance. AdaBoost achieved the highest PA for NDM in flights near physiological maturity. A similar trend was observed for TOL, where late-season flights improved model accuracy and both AdaBoost and SVM produced comparable performance. These patterns were consistent when evaluated with MSE, reinforcing the superiority of AdaBoost during late reproductive stages.\u003c/p\u003e\u003cp\u003eFor LR, classification models were used, with accuracy and precision as evaluation metrics. In general, the three models showed similar accuracy, with AdaBoost and MLP demonstrating superior results, especially in flights close to maturity. However, the PAs for this trait were relatively low, not exceeding 0.4 in accuracy. In the 2019/2020 season, the highest accuracy was found in the flight at 127 DAP, with 0.37 using the MLP model, while in 2020/2021, AdaBoost reached the highest accuracy at 122 DAP (0.39). Precision values showed similar trends, with AdaBoost slightly outperforming the other models.\u003c/p\u003e\u003cp\u003eOverall, predictive ability was higher in 2020/2021 than in 2019/2020 for GY and HSW, likely due to the greater number of flights conducted near flowering in that season. In contrast, NDM showed relatively stable predictive performance across years, with its best results in 2019/2020. TOL and LR exhibited low predictive ability in both years. Figure\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e8\u003c/span\u003e summarizes the flights that yielded the highest PA for each trait, based on Pearson correlation for regression models (GY, HSW, TOL, NDM) and accuracy for classification models (LR).\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eWe further evaluated six prediction scenarios by varying the combination of flights and the target season: (i) all flights from 2019; (ii) all flights from 2020; (iii) the last three flights of 2019; (iv) the last three flights of 2020; (v) training with 2019 data to predict 2020; and (vi) the reverse. Supplementary Figs.\u0026nbsp;5A\u0026ndash;B present the predictive ability (PA) of the three models for each trait under these scenarios.\u003c/p\u003e\u003cp\u003eFigure \u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e9\u003c/span\u003e shows that, within a season, aggregating all flights generally improved PA compared to using single flights, particularly with the AdaBoost model. The gains were most evident for GY and HSW in 2020/2021, where AdaBoost achieved the highest overall correlations. Similar trends were observed in 2019/2020, confirming that multi-flight integration captures temporal variation more effectively. By contrast, cross-season predictions performed poorly across all models and traits, highlighting the difficulty of transferring models between years despite large training sets. These results emphasize both the value of integrating multiple flights within a season and the current limitations of cross-season forecasting.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eFinally, we evaluated cross-season correlations between image-derived indices and manually measured traits (Fig.\u0026nbsp;\u003cspan refid=\"Fig9\" class=\"InternalRef\"\u003e10\u003c/span\u003e). Unlike the ML models, which performed poorly when trained in 2019/2020 and tested in 2020/2021, several cross-season correlations were significant. Notably, HSW and TOL in 2019/2020 showed strong correlations with VARI_P25 in 2020/2021, particularly at 107 and 114 DAP, in some cases exceeding within-season correlations. In contrast, GY displayed consistently weaker cross-season correlations, reinforcing its complex and environment-dependent nature.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eWe also conducted a decision tree analysis to understand the importance of the predictor variables. As expected, some features demonstrated significant relevance, while others exhibited low importance. Furthermore, the importance of a variable varied depending on the flight period and the year. However, certain VIs and TIs stood out, as evidenced in Supplementary Figs.\u0026nbsp;6 to 10, which present the importance graph of each predictor variable in the flights that demonstrated a predictive capacity greater than 0.6 for the traits NDM, GY, and HSW. For LR e TOL, due to the low PA presented for these traits, only the most important flights were included in the graph.\u003c/p\u003e\u003cp\u003eFor GY, VARI_P25 was the most important feature in late-season flights of 2019, while texture indices (TIs) such as ASM (0\u0026deg; and 45\u0026deg;) and Homogeneity (45\u0026deg;) contributed strongly during early flights (57\u0026ndash;78 DAP). HSW showed a similar trend, but TIs were more influential than VIs, particularly at 45\u0026deg; and 135\u0026deg;, with VARI_P25 again emerging in late flights (122 DAP). For NDM, VARI_P25 remained prominent, alongside Energy (135\u0026deg;) and Dissimilarity (45\u0026deg;) in 2019, with some gradient histograms showing specific importance for this trait. By contrast, no single variable dominated for LR and TOL; however, Contrast (45\u0026deg;) contributed to LR in the final 2020 flight, while VARI_P25 was again the main predictor for TOL.\u003c/p\u003e\u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eUAV-based HTP represents an effective tool for evaluating traits of interest in plant breeding programs. This methodology provides breeders with the ability to assess various stages of crop development and analyze a large number of genotypes across different locations, in a shorter time and with greater efficiency (Reynolds et al., \u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e2019\u003c/span\u003e).. In this study, we leveraged UAV-RGB imagery to evaluate soybean performance under natural stink bug pressure, relating image-derived indices to conventional field phenotypes and assessing their ability to predict those phenotypes. Additionally, we expanded such an evaluation by incorporating ML models and their capability to predict these traits using images captured by UAV. HTP protocols represent a strategy with potential to overcome the limitations associated with traditional large-scale phenotyping, allowing for a more detailed evaluation of traits of interest. Considering that resistance to the stink bug complex is a quantitative trait, the breeding process becomes complex, requiring extensive field evaluations (Furbank and Tester, \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2011\u003c/span\u003e)\u003c/p\u003e\u003cdiv id=\"Sec16\" class=\"Section2\"\u003e\u003ch2\u003eStink Bug Resistance Phenotyping in Soybean Breeding\u003c/h2\u003e\u003cp\u003eIn soybean breeding, phenotyping for resistance is well described for specific traits that condition resistance, such as pod damage, LR, grain filling period, crop cycle, grain size, and the HSW. HSW, in particular, has been reported as strongly associated with resistance and yield (da Rocha et al., \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2014\u003c/span\u003e). In our study, correlations among resistance-related traits, productivity, and phenology support their use as joint selection criteria. Differences in correlation magnitudes between seasons likely reflect contrasting infestation levels: under very low pressure, trait expression is muted, whereas very high pressure can saturate injury across genotypes, both scenarios reducing discrimination among lines (Panizzi and Corr\u0026ecirc;a-Ferreira, \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e1997\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eHeritability estimates for the evaluated traits were high, suggesting that these traits can be employed in selecting more resistant and productive genotypes. Previous studies under natural stink bug infestation reported comparable patterns: in an F2:3 population, Santos et al. (\u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003e2018\u003c/span\u003e) observed h\u0026sup2; \u0026asymp; 0.80 (NDM), 0.70 (GY, HSW) e 0.20 (LR);in a RILs population, da Rocha et al. (\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2015\u003c/span\u003e) reported h\u0026sup2; \u0026asymp; 0.75 (NDM), 0.35 (HSW), 0.24 (GY) e 0.67 (LR)., both populations originating from the cross between IAC 100 (resistant) and CD 215 (susceptible). Variation in h\u0026sup2; across studies is expected for quantitative traits such as GY and HSW, given strong environmental modulation (Han et al., \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2012\u003c/span\u003e). For LR, its value heavily depends on the evaluator's assessment, making it challenging to consistently achieve high heritability estimates (Godoi and Pinheiro, \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2009\u003c/span\u003e; Pinheiro et al., \u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e2005\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eWe observed that \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\sigma\\:}_{g}^{2}\\)\u003c/span\u003e\u003c/span\u003e explained a substantial proportion of the total phenotypic variance, indicating the presence of variability in the evaluated population and enabling the selection of genotypes. This variability is influenced by the parental lines involved in the crosses, which are quite diverse, including IAC 100, a source reported to carry multiple mechanisms of stink bug resistance (McPherson et al., \u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e2007\u003c/span\u003e; Veiga et al., \u003cspan citationid=\"CR67\" class=\"CitationRef\"\u003e1999\u003c/span\u003e). Thoughtful parental choice to maximize useful diversity remains central to breeding gains (Falconer et al., \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e1996\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eDespite their relevance, visual assessment becomes impractical and inaccurate on a large scale, a common challenge encountered in plant breeding programs, especially when evaluating insect resistance (Tang and He, \u003cspan citationid=\"CR61\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). Although UAV-based approaches have been explored in soybean for various applications, their use for evaluating resistance to stink bugs remains limited (Iost-Filho et al., \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Marston et al., \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e2020\u003c/span\u003e, \u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). UAV-based imaging enables quantitative estimation of canopy reflectance-derived VIs, which relate to physiological processes, while TIs capture spatial/structural canopy information (Araus and Cairns, \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2014\u003c/span\u003e; Li et al., \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2014\u003c/span\u003e).\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec17\" class=\"Section2\"\u003e\u003ch2\u003eImage-based Phenotyping for Stink Bug Resistance\u003c/h2\u003e\u003cp\u003eGiven the utility of aerial images, their application extends to evaluating a wide range of traits in plant breeding programs. VIs are measures related to photosynthetic activity and the physiological state of plants, especially concerning their response to stress and overall health (Krause et al., \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). These indices can capture subtle differences in plant composition and photosynthetic activity, which may be influenced by different environmental conditions and biotic stresses (Kefauver et al., \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2015\u003c/span\u003e). Additionally, VIs are used in many studies for predicting agricultural productivity due to their stable and superior performance (Ballester et al., \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2017\u003c/span\u003e; Zhou et al., \u003cspan citationid=\"CR74\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). Therefore, variations in VIs among different tested genotypes can provide insights into how plants respond and adapt to different stresses. In our study, VIs demonstrated high heritability and prediction accuracy, with strong correlations to traits associated with stink bug resistance, particularly in flights close to maturity and under higher stink bug pressure.\u003c/p\u003e\u003cp\u003eComparable findings have been reported in other crops. Li et al. (\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2023\u003c/span\u003e) showed that UAV-based RGB VIs effectively identified delayed senescence in wheat, while Resende et al. (\u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e2024\u003c/span\u003e) found that early flights (V5\u0026ndash;V8) in maize yielded the strongest associations between GLI and GY. However, they also emphasized that these correlations were highly variable depending on environmental conditions and crop development stage. A methodological distinction of our work was the use of the first percentile of pixel values instead of plot means. One likely explanation is that stink bug infestations are typically patchy and spatially heterogeneous across the field and even within individual plots (Fernandes et al., 2019). Thus, considering only the first percentile of the pixels ensures that we are evaluating those plants that were attacked by stink bugs. This strategy allows for a more accurate assessment of resistance to stink bug attacks and can be an interesting strategy for breeding programs focused on insect resistance.\u003c/p\u003e\u003cp\u003eIn soybean, Bai et al. (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2022\u003c/span\u003e) demonstrated that UAV-RGB imagery improved yield prediction under lodging conditions, particularly around flowering, underscoring the importance of capturing critical phenological stages. These authors also noted the contribution of texture information to yield estimation. Similarly, Maimaitijiang et al. (\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2020\u003c/span\u003e) reported that incorporating canopy texture features improved yield prediction accuracy, with R\u0026sup2; ranging from 0.65 to 0.72. TIs represent spatial variations in color intensity in images and can capture repetitive patterns that are useful for assessing aspects such as canopy uniformity, which are associated with plant vigor and, consequently, productivity (Ma et al., \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2022\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eConsistent with this, our results highlighted VARI_P25 as the most informative index, with superior heritability and consistent correlations across traits. Nonetheless, texture indices (TIs) also proved valuable in specific contexts, as evidenced by the PCA-based TI score correlating with tolerance (TOL) in 2019/2020 (r\u0026thinsp;=\u0026thinsp;0.58), surpassing other VIs. Thus, while VIs are generally more effective, TIs can provide complementary information in certain environments or traits. Overall, these findings demonstrate that UAV-derived indices are strongly associated with visual characteristics and can be used individually for selection. However, exploiting their combined predictive power requires integrative approaches, such as ML, to fully capture the complexity of image-based datasets.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec18\" class=\"Section2\"\u003e\u003ch2\u003eMachine Learning Prediction of Stink Bug Resistance\u003c/h2\u003e\u003cp\u003eSeveral studies have combined UAV imagery with ML to address challenges in soybean improvement. Hassanijalilian et al. (\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2020\u003c/span\u003e) evaluated decision tree\u0026ndash;based models to classify iron deficiency chlorosis in 40 soybean cultivars using RGB images and found that AdaBoost outperformed random forest (RF) and decision trees, achieving an average F1-score of 0.75. Similarly, RF and artificial neural networks effectively distinguished dicamba-tolerant from susceptible genotypes, with classification accuracies ranging from 0.69 to 0.75 (Vieira et al., \u003cspan citationid=\"CR68\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Notably, LR resembles IDC and dicamba injury in that it is also expressed through visible chlorophyll loss in the leaves.\u003c/p\u003e\u003cp\u003eIn our study, AdaBoost consistently delivered the strongest results for most traits, both when using individual flights and when aggregating multiple flights, confirming the effectiveness of decision tree\u0026ndash;based models for these prediction tasks. Nonetheless, PAs were generally lower than those reported in the studies above. A likely explanation lies in the contrasting stink bug pressures across seasons: in 2019/2020, severe infestations produced uniformly high scores, while in 2020/2021, symptoms were minimal and most scores clustered at the lower end this restricted phenotypic variation reduced model discrimination. These findings reinforce the importance of maintaining intermediate pest pressure levels to maximize trait expression and enable effective resistance screening (Panizzi and Corr\u0026ecirc;a-Ferreira, \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e1997\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eFor GY and HSW, the highest PAs were observed in flights near flowering, although informative predictions were also obtained at later growth stages. Zhou et al. (\u003cspan citationid=\"CR73\" class=\"CitationRef\"\u003e2020\u003c/span\u003e) reported a similar pattern using a convolutional neural network to estimate soybean yield under water stress, with images collected at both early and late reproductive stages providing strong predictions. By incorporating all flights to predict GY, the results indicate that a greater number of flights contributes to capturing a higher degree of phenotypic and environmental variability. Flights conducted close to flowering can capture stand-related variation, such as incomplete plant establishment or uneven development, which may confound predictions and not fully reflect the genetic basis of resistance. In contrast, flights near maturity captured stink bug injury more effectively during the critical reproductive stages (R5\u0026ndash;R7), when feeding causes the greatest yield losses (da Rocha et al., \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2014\u003c/span\u003e). Thus, late flights more accurately represent genetic differences in resistance and yield potential, whereas early flights are less informative for resistance screening since stink bug damage is not yet phenotypically expressed.\u003c/p\u003e\u003cp\u003eAlthough UAVs are widely used in plant breeding, their application for stink bug phenotyping in the field has not yet been reported. Some studies, have explored UAV-based approaches for detecting soybean aphid stress. Marston et al. (\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e2022\u003c/span\u003e) achieved up to 89.4% classification accuracy using hyperspectral reflectance with support vector machines, while Marston et al. (\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e2020\u003c/span\u003e) demonstrated that multispectral imagery coupled with linear regression could also identify aphid-induced stress. However, these studies were conducted with controlled infestation in cages and using few or only one genotype, conditions that differ from those found in breeding programs. Despite these limitations, they highlight the potential of advanced sensors, such as multispectral and hyperspectral cameras, to improve resistance screening under field conditions.\u003c/p\u003e\u003cp\u003eOverall, the best PAs in our study were achieved in the 2020/2021 season, particularly for HSW and GY. Yet caution is needed when selecting based on HSW under low stink bug pressure, as this may favor highly productive lines rather than resistant ones (da Rocha et al., \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2014\u003c/span\u003e). Conversely, the high selection pressure in 2019/2020 hindered the identification of resistant genotypes by masking phenotypic differences and reducing model discrimination, since tolerance mechanisms can buffer yield losses under moderate feeding but are overcome under extreme pest pressure (Smith, \u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e2005\u003c/span\u003e). These contrasting scenarios underscore the importance of maintaining intermediate pest pressure, as overly low or high infestations limit discrimination among genotypes (Corr\u0026ecirc;a-Ferreira \u0026amp; Panizzi, \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e1999\u003c/span\u003e). Incorporating a greater number of flights improved PAs across traits, although the increment was modest, suggesting that repeated temporal sampling helps capture both genetic and environmental variability (Rutkoski et al., \u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e2016\u003c/span\u003e; Sun et al., \u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e2017\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eCrop prediction in plant breeding programs using data from previous years can be a valuable tool, but it presents considerable challenges, mainly due to environmental variations between seasons (Crossa et al., \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2017\u003c/span\u003e; Yang et al., \u003cspan citationid=\"CR69\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). This is particularly evident when UAV-acquired images are used, as differences in climatic and management conditions across flights directly affect prediction accuracy. Still, VARI_P25 maintained moderate cross-season correlations with traits such as HSW and TOL, in some cases exceeding within-season associations. Although modest, these correlations can be useful for an initial genotype screening, allowing the exclusion of less promising materials and consequently optimizing resources and evaluation time in subsequent cycles (Zhou et al., \u003cspan citationid=\"CR75\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). By contrast, correlations for GY were considerably lower between seasons, indicating that this trait is more strongly modulated by environmental variation.\u003c/p\u003e\u003cp\u003eWe did not identify consistent patterns among the most important indices in the decision tree analysis. Each flight and each characteristic exhibited different patterns of influence from the indices used. Still, VARI_P25 emerged as the most recurrently important index, appearing among the flights with the highest predictive capacity for nearly all traits, with the exception of LR. Beyond its importance in the ML models, VARI_P25 also showed strong direct correlations with manually measured traits. Biologically, this index reflects canopy vigor and stress through greenness (Gitelson et al., \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2002\u003c/span\u003e), and has been previously linked to grain weight under stress (Rodene et al., \u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e2022\u003c/span\u003e), yield (Silva et al., \u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003e2022\u003c/span\u003e), and leaf area (Hasan et al., \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Zhang et al., \u003cspan citationid=\"CR72\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). These consistent associations across traits and studies highlight VARI_P25 as a particularly promising index for stink bug resistance phenotyping.\u003c/p\u003e\u003cp\u003eIn addition to VIs, TIs also contributed substantially to model performance. TIs calculated at 45\u0026deg; and 135\u0026deg; angles were often among the most influential predictors. Unlike VIs, which capture physiological status through reflectance, TIs quantify canopy structure and spatial patterns, traits that may indirectly reflect plant responses to pest damage (Ma et al., \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Future studies should further investigate the role of TIs in traits such as LR, which is notoriously difficult to assess visually and can easily be confounded with late maturity. Under stink bug attack, leaves and stems remain green while pods mature, a contrast that texture indices may be able to capture more reliably than visual assessment (Pinheiro et al., \u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e2005\u003c/span\u003e; Godoi and Pinheiro, \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2009\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eIn summary, ML approaches coupled with UAV imagery proved effective for predicting soybean traits associated with stink bug resistance, with AdaBoost consistently delivering the strongest results. While environmental variation and extremes in pest pressure constrained prediction accuracy, integrating multiple flights and combining VIs with TIs enhanced the detection of phenotypic signals. These findings underscore the potential of UAV-ML pipelines to support resistance phenotyping, provided that experimental designs ensure balanced pest pressure and consistent temporal sampling.\u003c/p\u003e\u003c/div\u003e"},{"header":"Conclusion","content":"\u003cp\u003eThis study provides the first comprehensive evaluation of stink bug resistance in soybean using UAV-based imagery integrated with ML models. We show that VI, particularly VARI_P25, are consistently associated with key traits, while TIs offer complementary information, underscoring the value of combining both approaches. Prediction ability was strongly influenced by pest pressure and flight timing, highlighting the importance of monitoring at critical reproductive stages and maintaining intermediate infestation levels for reliable resistance screening. Despite challenges such as limited cross-season transferability, our findings establish UAV\u0026ndash;ML pipelines as a promising framework for phenotyping complex resistance traits. Future research integrating advanced sensors, multi-season datasets, and optimized experimental designs will be key to fully leveraging these tools for accelerating soybean breeding for stink bug resistance.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003ch2\u003eImpact\u003c/h2\u003e\u003cp\u003eWe introduce UAV-RGB imaging with ML models that derive actionable indices from field imagery to quantify canopy signals of stink bug resistance under natural infestations, enabling objective, scalable screening of breeding lines. This data-driven approach fits \u003cem\u003ePrecision Agriculture\u003c/em\u003e by using sensing and analytics to make phenotyping more efficient, less labor-intensive, and faster at advancing superior genotypes.\u003c/p\u003e\u003cb\u003eCRediT Author Contribution Statement\u003c/b\u003e\u003c/p\u003e\u003cp\u003eMO: Conceptualization, Methodology, Software, Validation, Formal analysis, Investigation, Data Curation, Writing - Original Draft, Writing - Review \u0026amp; Editing, Visualization. AA: Methodology, Software, Validation, Formal analysis, Investigation, Data Curation, Writing - Review \u0026amp; Editing, Visualization. PB: Conceptualization, Methodology, Investigation, Data Curation. AM: Conceptualization, Methodology, Investigation, Data Curation. FC: Data Curation. FK: Data Curation. FF: Data Curation. JPazini: Data Curation, Writing - Review \u0026amp; Editing. PY: Validation, Data Curation, Resources, Supervision. JPinheiro: Conceptualization, Validation, Investigation, Data Curation, Writing - Review \u0026amp; Editing, Resources, Supervision, Project administration, Funding acquisition\u003c/p\u003e\u003cp\u003e\u003ch2\u003eCompeting Interests\u003c/h2\u003e\u003cp\u003eThe authors have no relevant financial or nonfinancial interests to disclose.\u003c/p\u003e\u003c/p\u003e\u003ch2\u003eAcknowledgement\u003c/h2\u003e\u003cp\u003eThis work was supported by grants from the Conselho Nacional de Desenvolvimento Cient\u0026iacute;fico e Tecnol\u0026oacute;gico (CNPq), Funda\u0026ccedil;\u0026atilde;o de Amparo \u0026agrave; Pesquisa do Estado de S\u0026atilde;o Paulo (FAPESP), and the Coordena\u0026ccedil;\u0026atilde;o de Aperfei\u0026ccedil;oamento de Pessoal de N\u0026iacute;vel Superior (CAPES \u0026ndash; Financial Code 001). AA received a PhD fellowship from FAPESP (2019/03232-6).\u003c/p\u003e\u003ch2\u003eData Availability\u003c/h2\u003e\u003cp\u003eData will be made available on request.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eAraus, J. L., and Cairns, J. E. (2014). Field high-throughput phenotyping: the new crop breeding frontier. \u003cem\u003eTrends in Plant Science\u003c/em\u003e, \u003cem\u003e19\u003c/em\u003e(1), 52\u0026ndash;61. https://doi.org/10.1016/J.TPLANTS.2013.09.008\u003c/li\u003e\n\u003cli\u003eBai, D., Li, D., Zhao, C., Wang, Z., Shao, M., Guo, B., Liu, Y., Wang, Q., Li, J., Guo, S., Wang, R., Li, Y. H., Qiu, L. J., and Jin, X. (2022). Estimation of soybean yield parameters under lodging conditions using RGB information from unmanned aerial vehicles. \u003cem\u003eFrontiers in Plant Science\u003c/em\u003e, \u003cem\u003e13\u003c/em\u003e, 1012293. https://doi.org/10.3389/FPLS.2022.1012293/BIBTEX\u003c/li\u003e\n\u003cli\u003eBallester, C., Hornbuckle, J., Brinkhoff, J., Smith, J., and Quayle, W. (2017). Assessment of In-Season Cotton Nitrogen Status and Lint Yield Prediction from Unmanned Aerial System Imagery. \u003cem\u003eRemote Sensing 2017, Vol. 9, Page 1149\u003c/em\u003e, \u003cem\u003e9\u003c/em\u003e(11), 1149. https://doi.org/10.3390/RS9111149\u003c/li\u003e\n\u003cli\u003eBassine, F. Z., Errami, A., and Khaldoun, M. (2019). Vegetation Recognition Based on UAV Image Color Index. \u003cem\u003eProceedings - 2019 IEEE International Conference on Environment and Electrical Engineering and 2019 IEEE Industrial and Commercial Power Systems Europe, EEEIC/I and CPS Europe 2019\u003c/em\u003e. https://doi.org/10.1109/EEEIC.2019.8783830\u003c/li\u003e\n\u003cli\u003eBendig, J., Yu, K., Aasen, H., Bolten, A., Bennertz, S., Broscheit, J., Gnyp, M. L., and Bareth, G. (2015). Combining UAV-based plant height from crop surface models, visible, and near infrared vegetation indices for biomass monitoring in barley. \u003cem\u003eInternational Journal of Applied Earth Observation and Geoinformation\u003c/em\u003e, \u003cem\u003e39\u003c/em\u003e, 79\u0026ndash;87. https://doi.org/10.1016/J.JAG.2015.02.012\u003c/li\u003e\n\u003cli\u003eBhat, J. A., and Yu, D. (2021). High-throughput NGS-based genotyping and phenotyping: Role in genomics-assisted breeding for soybean improvement. \u003cem\u003eLegume Science\u003c/em\u003e, \u003cem\u003e3\u003c/em\u003e(3), e81. https://doi.org/10.1002/LEG3.81\u003c/li\u003e\n\u003cli\u003eBoerma, H. R., and Walker, D. R. (2005). Discovery and utilization of QTLs for insect resistance in soybean. In \u003cem\u003eGenetics of Adaptation\u003c/em\u003e (pp. 181\u0026ndash;189). Springer-Verlag. https://doi.org/10.1007/1-4020-3836-4_16\u003c/li\u003e\n\u003cli\u003eBortolotto, O. C., Pomari-Fernandes, A., De, R. C. O., Bueno, F., De, A., Da Kruz, Y. K. S., Queiroz, A. P., Sanzovo, A., and Ferreira, R. B. (2015). The use of soybean integrated pest management in Brazil: a review. \u003cem\u003eAgronomy Science and Biotechnology\u003c/em\u003e, \u003cem\u003e1\u003c/em\u003e(1), 25\u0026ndash;32. www.asbjournal.comFreelyavailableonline\u003c/li\u003e\n\u003cli\u003eButler, D. G., Cullis, B. R., Gilmour, A. R., Gogel, B. J., and Thompson, R. (2018). \u003cem\u003eASReml-R Reference Manual Version 4 ASReml estimates variance components under a general linear mixed model by residual maximum likelihood (REML)\u003c/em\u003e. http://asreml.org\u003c/li\u003e\n\u003cli\u003eCONAB - Companhia Nacional de Abastacimento. 2024. Acompanhamento da safra brasileira: gr\u0026atilde;os (2023/24), 8\u0026deg; Levantamento. In \u003cem\u003esafra 2023/24, n 8 \u0026ndash; oitavo levantamento\u003c/em\u003e (Vol. 11, Issue 1). https://www.conab.gov.br/info-agro/safras/graos/boletim-da-safra-de-graos\u003c/li\u003e\n\u003cli\u003eCorr\u0026ecirc;a-Ferreira, B. S., and Panizzi, A. R. (1999). Percevejos da soja e seu manejo. \u003cem\u003eEmbrapa-CNPSo Documentos, 124\u003c/em\u003e, 1\u0026ndash;45.\u003c/li\u003e\n\u003cli\u003eCrossa, J., P\u0026eacute;rez-Rodr\u0026iacute;guez, P., Cuevas, J., Montesinos-L\u0026oacute;pez, O., Jarqu\u0026iacute;n, D., de los Campos, G., Burgue\u0026ntilde;o, J., Gonz\u0026aacute;lez-Camacho, J. M., P\u0026eacute;rez-Elizalde, S., Beyene, Y., Dreisigacker, S., Singh, R., Zhang, X., Gowda, M., Roorkiwal, M., Rutkoski, J., and Varshney, R. K. (2017). Genomic Selection in Plant Breeding: Methods, Models, and Perspectives. \u003cem\u003eTrends in Plant Science\u003c/em\u003e, \u003cem\u003e22\u003c/em\u003e(11), 961\u0026ndash;975. https://doi.org/10.1016/J.TPLANTS.2017.08.011\u003c/li\u003e\n\u003cli\u003eda Rocha, F., Bermudez, F., Ferreira, M. C., de Oliveira, K. C., and Pinheiro, J. B. (2014). Crit\u0026eacute;rio efetivo de sele\u0026ccedil;\u0026atilde;o para avaliar a resist\u0026ecirc;ncia ao complexo de percevejos em soja. \u003cem\u003eCrop Breeding and Applied Biotechnology\u003c/em\u003e, \u003cem\u003e14\u003c/em\u003e(3), 174\u0026ndash;179. https://doi.org/10.1590/1984-70332014v14n3a26\u003c/li\u003e\n\u003cli\u003eda Rocha, F., Vieira, C. C., Ferreira, M. C., de Oliveira, K. C., Moreira, F. F., and Pinheiro, J. B. (2015). Selection of soybean lines exhibiting resistance to stink bug complex in distinct environments. \u003cem\u003eFood and Energy Security\u003c/em\u003e, \u003cem\u003e4\u003c/em\u003e(2), 133\u0026ndash;143. https://doi.org/10.1002/FES3.57\u003c/li\u003e\n\u003cli\u003eFalconer, D. S., Mackay, T. F. C., and Frankham, R. (1996). Introduction to Quantitative Genetics . In \u003cem\u003eTrends in Genetics\u003c/em\u003e (Fourth Edition). Trends in Genetics.\u003c/li\u003e\n\u003cli\u003eFreund, Y., and Schapire, R. E. (1997). A Decision-Theoretic Generalization of On-Line Learning and an Application to Boosting. \u003cem\u003eJournal of Computer and System Sciences\u003c/em\u003e, \u003cem\u003e55\u003c/em\u003e(1), 119\u0026ndash;139. https://doi.org/10.1006/JCSS.1997.1504\u003c/li\u003e\n\u003cli\u003eFurbank, R. T., and Tester, M. (2011). Phenomics - technologies to relieve the phenotyping bottleneck. \u003cem\u003eTrends in Plant Science\u003c/em\u003e, \u003cem\u003e16\u003c/em\u003e(12), 635\u0026ndash;644. https://doi.org/10.1016/j.tplants.2011.09.005\u003c/li\u003e\n\u003cli\u003eGitelson, A. A., Kaufman, Y. J., Stark, R., and Rundquist, D. (2002). Novel algorithms for remote estimation of vegetation fraction. \u003cem\u003eRemote Sensing of Environment\u003c/em\u003e, \u003cem\u003e80\u003c/em\u003e(1), 76\u0026ndash;87. https://doi.org/10.1016/S0034-4257(01)00289-9\u003c/li\u003e\n\u003cli\u003eGodoi, C. R. C., and Pinheiro, J. B. (2009). Genetic parameters and selection strategies for soybean genotypes resistant to the stink bug-complex. \u003cem\u003eGenetics and Molecular Biology\u003c/em\u003e, \u003cem\u003e32\u003c/em\u003e(2), 328\u0026ndash;336. https://doi.org/10.1590/S1415-47572009000200020\u003c/li\u003e\n\u003cli\u003eGoggin, F. L., Lorence, A., and Topp, C. N. (2015). Applying high-throughput phenotyping to plant\u0026ndash;insect interactions: picturing more resistant crops. \u003cem\u003eCurrent Opinion in Insect Science\u003c/em\u003e, \u003cem\u003e9\u003c/em\u003e, 69\u0026ndash;76. https://doi.org/10.1016/J.COIS.2015.03.002\u003c/li\u003e\n\u003cli\u003eGuedes, J. V. C., Arnemann, J. A., St\u0026uuml;rmer, G. R., Melo, A. A., Bigolin, M., Perini, C. R., and Sari, B. G. (2012). Percevejos da soja: novos cen\u0026aacute;rios, novo manejo. \u003cem\u003eRevista Plantio Direto\u003c/em\u003e, \u003cem\u003e12\u003c/em\u003e, 28\u0026ndash;34.\u003c/li\u003e\n\u003cli\u003eHan, Y., Li, D., Zhu, D., Li, H., Li, X., Teng, W., and Li, W. (2012). QTL analysis of soybean seed weight across multi-genetic backgrounds and environments. \u003cem\u003eTheoretical and Applied Genetics\u003c/em\u003e, \u003cem\u003e125\u003c/em\u003e(4), 671\u0026ndash;683. https://doi.org/10.1007/S00122-012-1859-X/FIGURES/2\u003c/li\u003e\n\u003cli\u003eHaralick, R. M., Dinstein, I., and Shanmugam, K. (1973). Textural Features for Image Classification. \u003cem\u003eIEEE Transactions on Systems, Man and Cybernetics\u003c/em\u003e, \u003cem\u003eSMC\u003c/em\u003e-\u003cem\u003e3\u003c/em\u003e(6), 610\u0026ndash;621. https://doi.org/10.1109/TSMC.1973.4309314\u003c/li\u003e\n\u003cli\u003eHartman, G. L., West, E. D., and Herman, T. K. (2011). Crops that feed the World 2. Soybean-worldwide production, use, and constraints caused by pathogens and pests. \u003cem\u003eFood Security\u003c/em\u003e, \u003cem\u003e3\u003c/em\u003e(1), 5\u0026ndash;17. https://doi.org/10.1007/S12571-010-0108-X/METRICS\u003c/li\u003e\n\u003cli\u003eHasan, U., Sawut, M., and Chen, S. (2019). Estimating the Leaf Area Index of Winter Wheat Based on Unmanned Aerial Vehicle RGB-Image Parameters. \u003cem\u003eSustainability 2019, 11\u003c/em\u003e(23), 6829. https://doi.org/10.3390/SU11236829\u003c/li\u003e\n\u003cli\u003eHassanijalilian, O., Igathinathane, C., Bajwa, S., and Nowatzki, J. (2020). Rating iron deficiency in soybean using image processing and decision-tree based models. \u003cem\u003eRemote Sensing, 12\u003c/em\u003e(24), 4143. https://doi.org/10.3390/rs12244143\u003c/li\u003e\n\u003cli\u003eHenderson, C. R. (1982). Analysis of Covariance in the Mixed Model: Higher-Level, Nonhomogeneous, and Random Regressions. \u003cem\u003eBiometrics\u003c/em\u003e, \u003cem\u003e38\u003c/em\u003e(3), 623. https://doi.org/10.2307/2530044\u003c/li\u003e\n\u003cli\u003eHunt, E. R., Cavigelli, M., Daughtry, C. S. T., McMurtrey, J. E., and Walthall, C. L. (2005). Evaluation of digital photography from model aircraft for remote sensing of crop biomass and nitrogen status. \u003cem\u003ePrecision Agriculture\u003c/em\u003e, \u003cem\u003e6\u003c/em\u003e(4), 359\u0026ndash;378. https://doi.org/10.1007/S11119-005-2324-5/METRICS\u003c/li\u003e\n\u003cli\u003eIost-Filho, F. H., Pazini, J. de B., de Medeiros, A. D., Rosalen, D. L., and Yamamoto, P. T. (2022). Assessment of Injury by Four Major Pests in Soybean Plants Using Hyperspectral Proximal Imaging. \u003cem\u003eAgronomy\u003c/em\u003e, \u003cem\u003e12\u003c/em\u003e(7), 1516. https://doi.org/10.3390/AGRONOMY12071516/S1\u003c/li\u003e\n\u003cli\u003eKefauver, S. C., El-Haddad, G., Vergara-Diaz, O., and Araus, J. L. (2015). RGB picture vegetation indexes for High-Throughput Phenotyping Platforms (HTPPs). \u003cem\u003eRemote Sensing for Agriculture, Ecosystems, and Hydrology XVII\u003c/em\u003e, \u003cem\u003e9637\u003c/em\u003e, 96370J. https://doi.org/10.1117/12.2195235\u003c/li\u003e\n\u003cli\u003eKrause, M. R., Mondal, S., Crossa, J., Singh, R. P., Pinto, F., Haghighattalab, A., Shrestha, S., Rutkoski, J., Gore, M. A., Sorrells, M. E., and Poland, J. (2020). Aerial high-throughput phenotyping enables indirect selection for grain yield at the early generation, seed-limited stages in breeding programs. \u003cem\u003eCrop Science\u003c/em\u003e, \u003cem\u003e60\u003c/em\u003e(6), 3096\u0026ndash;3114. https://doi.org/10.1002/CSC2.20259\u003c/li\u003e\n\u003cli\u003eLi, L., Zhang, Q., and Huang, D. (2014). A Review of Imaging Techniques for Plant Phenotyping. \u003cem\u003eSensors (Basel, Switzerland)\u003c/em\u003e, \u003cem\u003e14\u003c/em\u003e(11), 20078. https://doi.org/10.3390/S141120078\u003c/li\u003e\n\u003cli\u003eLi, L., Guo, W., Carroll, M. E., Singh, A., Swetnam, T. L., Merchant, N., \u0026amp; Sarkar, S. et al. (2023). UAV-based RGB imagery and ground measurements for high-throughput phenotyping of senescence and QTL mapping in bread wheat. \u003cem\u003eCrop Science, 63\u003c/em\u003e(6), 3292\u0026ndash;3309. https://doi.org/10.1002/csc2.21086\u003c/li\u003e\n\u003cli\u003eMa, Y., Ma, L., Zhang, Q., Huang, C., Yi, X., Chen, X., Hou, T., Lv, X., and Zhang, Z. (2022). Cotton Yield Estimation Based on Vegetation Indices and Texture Features Derived From RGB Image. \u003cem\u003eFrontiers in Plant Science\u003c/em\u003e, \u003cem\u003e13\u003c/em\u003e, 925986. https://doi.org/10.3389/FPLS.2022.925986/BIBTEX\u003c/li\u003e\n\u003cli\u003eMaimaitijiang, M., Sagan, V., Sidike, P., Hartling, S., Esposito, F., and Fritschi, F. B. (2020). Soybean yield prediction from UAV using multimodal data fusion and deep learning. \u003cem\u003eRemote Sensing of Environment\u003c/em\u003e, \u003cem\u003e237\u003c/em\u003e, 111599. https://doi.org/10.1016/J.RSE.2019.111599\u003c/li\u003e\n\u003cli\u003eMarston, Z. P. D., Cira, T. M., Hodgson, E. W., Knight, J. F., MacRae, I. V., Koch, R. L., and Rondon, S. (2020). Detection of Stress Induced by Soybean Aphid (Hemiptera: Aphididae) Using Multispectral Imagery from Unmanned Aerial Vehicles. \u003cem\u003eJournal of Economic Entomology\u003c/em\u003e, \u003cem\u003e113\u003c/em\u003e(2), 779\u0026ndash;786. https://doi.org/10.1093/JEE/TOZ306\u003c/li\u003e\n\u003cli\u003eMarston, Z. P. D., Cira, T. M., Knight, J. F., Mulla, D., Alves, T. M., Hodgson, E. W., Ribeiro, A. V., MacRae, I. V., and Koch, R. L. (2022). Linear Support Vector Machine Classification of Plant Stress From Soybean Aphid (Hemiptera: Aphididae) Using Hyperspectral Reflectance. \u003cem\u003eJournal of Economic Entomology\u003c/em\u003e, \u003cem\u003e115\u003c/em\u003e(5), 1557\u0026ndash;1563. https://doi.org/10.1093/JEE/TOAC077\u003c/li\u003e\n\u003cli\u003eMatias, F. I., Caraza-Harter, M. V., and Endelman, J. B. (2020). FIELDimageR: An R package to analyze orthomosaic images from agricultural field trials. \u003cem\u003eThe Plant Phenome Journal\u003c/em\u003e, \u003cem\u003e3\u003c/em\u003e(1), e20005. https://doi.org/10.1002/PPJ2.20005\u003c/li\u003e\n\u003cli\u003eMcPherson, R. M., Buss, G. R., and Roberts, P. M. (2007). Assessing Stink Bug Resistance in Soybean Breeding Lines Containing Genes from Germplasm IAC-100. \u003cem\u003eJournal of Economic Entomology\u003c/em\u003e, \u003cem\u003e100\u003c/em\u003e(4), 1456\u0026ndash;1463. https://doi.org/10.1093/JEE/100.4.1456\u003c/li\u003e\n\u003cli\u003eMoreira, F. F., Rojas de Oliveira, H., Lopez, M. A., Abughali, B. J., Gomes, G., Cherkauer, K. A., Brito, L. F., and Rainey, K. M. (2021). High-Throughput Phenotyping and Random Regression Models Reveal Temporal Genetic Control of Soybean Biomass Production. \u003cem\u003eFrontiers in Plant Science\u003c/em\u003e, \u003cem\u003e12\u003c/em\u003e, 715983. https://doi.org/10.3389/FPLS.2021.715983/BIBTEX\u003c/li\u003e\n\u003cli\u003eOtsu, N. (1979). A Threshold Selection Method from Gray-Level Histograms. \u003cem\u003eIEEE Trans Syst Man Cybern\u003c/em\u003e, \u003cem\u003eSMC\u003c/em\u003e-\u003cem\u003e9\u003c/em\u003e(1), 62\u0026ndash;66. https://doi.org/10.1109/TSMC.1979.4310076\u003c/li\u003e\n\u003cli\u003ePal, S. K., and Mitra, S. (1992). Multilayer Perceptron, Fuzzy Sets, and Classification. \u003cem\u003eIEEE Transactions on Neural Networks\u003c/em\u003e, \u003cem\u003e3\u003c/em\u003e(5), 683\u0026ndash;697. https://doi.org/10.1109/72.159058\u003c/li\u003e\n\u003cli\u003ePanizzi, A. R., and Hirose, E. (1995). Survival, reproduction and starvation resistance of adult southern green stink bug (Heteroptera: Pnetatomidae) reared on sesame or soybean. \u003cem\u003eAnnals of the Entomological Society of America\u003c/em\u003e, \u003cem\u003e88\u003c/em\u003e(5), 661\u0026ndash;665. https://doi.org/10.1093/aesa/88.5.661\u003c/li\u003e\n\u003cli\u003ePanizzi, A. R., and Corr\u0026ecirc;a-Ferreira, B. S. (1997). Dynamics in the insect fauna adaptation to soybean in the tropics. \u003cem\u003eTrends in Entomology, 1\u003c/em\u003e, 71\u0026ndash;88.\u003c/li\u003e\n\u003cli\u003ePatterson, H. D., and Thompson, R. (1971). Recovery of inter-block information when block sizes are unequal. \u003cem\u003eBiometrika\u003c/em\u003e, \u003cem\u003e58\u003c/em\u003e(3), 545\u0026ndash;554. https://doi.org/10.1093/BIOMET/58.3.545\u003c/li\u003e\n\u003cli\u003ePiepho, H. P., and M\u0026ouml;hring, J. (2007). Computing heritability and selection response from unbalanced plant breeding trials. \u003cem\u003eGenetics\u003c/em\u003e, \u003cem\u003e177\u003c/em\u003e(3), 1881\u0026ndash;1888. https://doi.org/10.1534/GENETICS.107.074229\u003c/li\u003e\n\u003cli\u003ePinheiro, J. B., Vello, N. A., Rossetto, C. J., and Zucchi, M. I. (2005). Potential of soybean genotypes as insect resistance sources. \u003cem\u003eBiotechnology\u003c/em\u003e, \u003cem\u003eSeptember 2005\u003c/em\u003e, 294\u0026ndash;301. https://doi.org/10.12702/1984-7033.v05n03a06\u003c/li\u003e\n\u003cli\u003ePisarevsky, V. (2008). \u003cem\u003eOpenCV, The Open Computer Vision Library\u003c/em\u003e.\u003c/li\u003e\n\u003cli\u003eQGIS Development Team. (2018). \u003cem\u003eQGIS Geographic Information System. Open Source Geospatial Foundation Project\u003c/em\u003e. http://qgis.osgeo.org\u003c/li\u003e\n\u003cli\u003eR Core Team. (2023). \u003cem\u003eR: A Language and Environment for Statistical Computing_. R Foundation for Statistical Computing\u003c/em\u003e (4.3.1; p. https://www.R-project.org/).\u003c/li\u003e\n\u003cli\u003eResende, E. L., Bruzi, A. T., Cardoso, E. da S., Carneiro, V. Q., Souza, V. A. P., Barros, P. H. F. C., \u0026amp; Pereira, R. R. (2024). High-throughput phenotyping: Application in maize breeding. \u003cem\u003eAgriEngineering, 6\u003c/em\u003e(2), 1078\u0026ndash;1092. https://doi.org/10.3390/agriengineering6020062\u003c/li\u003e\n\u003cli\u003eReynolds, D., Baret, F., Welcker, C., Bostrom, A., Ball, J., Cellini, F., Lorence, A., Chawade, A., Khafif, M., Noshita, K., Mueller-Linow, M., Zhou, J., and Tardieu, F. (2019). What is cost-efficient phenotyping? Optimizing costs for different scenarios. \u003cem\u003ePlant Science\u003c/em\u003e, \u003cem\u003e282\u003c/em\u003e, 14\u0026ndash;22. https://doi.org/10.1016/j.plantsci.2018.06.015\u003c/li\u003e\n\u003cli\u003eRodene, E., Xu, G., Palali Delen, S., Zhao, X., Smith, C., Ge, Y., Schnable, J., and Yang, J. (2022). A UAV-based high-throughput phenotyping approach to assess time-series nitrogen responses and identify trait-associated genetic components in maize. \u003cem\u003eThe Plant Phenome Journal\u003c/em\u003e, \u003cem\u003e5\u003c/em\u003e(1), e20030. https://doi.org/10.1002/PPJ2.20030\u003c/li\u003e\n\u003cli\u003eRossetto, C. J., Gallo, P., Razera, L., Bortoletto, N., Igue, T., Medina, P., Tisselli, O., Aquilera, V., and Pinheiro, J. (1995). Mechanisms of resistance to stink bug complex in the soybean cultivar IAC-100. \u003cem\u003eAnais- Sociedade Entomologica Do Brasil\u003c/em\u003e, \u003cem\u003e24\u003c/em\u003e(3), 517\u0026ndash;522.\u003c/li\u003e\n\u003cli\u003eRutkoski, J., Poland, J., Mondal, S., Autrique, E., P\u0026eacute;rez, L. G., Crossa, J., Reynolds, M., and Singh, R. (2016). Canopy temperature and vegetation indices from high-throughput phenotyping improve accuracy of pedigree and genomic selection for grain yield in wheat. \u003cem\u003eG3: Genes, Genomes, Genetics, 6\u003c/em\u003e(9), 2799\u0026ndash;2808.\u003cem\u003e \u003c/em\u003ehttps://doi.org/10.1534/g3.116.032888\u003c/li\u003e\n\u003cli\u003eSakurai, K., Toda, Y., Kajiya-Kanegae, H., Ohmori, Y., Yamasaki, Y., Takahashi, H., Takanashi, H., Tsuda, M., Tsujimoto, H., Kaga, A., Nakazono, M., Fujiwara, T., and Iwata, H. (2022). Time-series multispectral imaging in soybean for improving biomass and genomic prediction accuracy. \u003cem\u003eThe Plant Genome\u003c/em\u003e, \u003cem\u003e15\u003c/em\u003e(4), e20244. https://doi.org/10.1002/TPG2.20244\u003c/li\u003e\n\u003cli\u003eSantos, M. F., M\u0026ouml;ller, M., Clough, S. J., and Pinheiro, J. B. (2018). Heritability of agronomic traits correlated with reduced stink bug damage in an F2:3 soybean population derived from IAC-100. \u003cem\u003eJournal of Crop Improvement\u003c/em\u003e, \u003cem\u003e32\u003c/em\u003e(1), 1\u0026ndash;18. https://doi.org/10.1080/15427528.2017.1370404\u003c/li\u003e\n\u003cli\u003eSilva, M. F. E., Maciel, G. M., Gallis, R. B. A., Barbosa, R. L., Carneiro, V. Q., Rezende, W. S., and Siquieroli, A. C. S. (2022). High-throughput phenotyping by RGB and multispectral imaging analysis of genotypes in sweet corn. \u003cem\u003eHorticultura Brasileira\u003c/em\u003e, \u003cem\u003e40\u003c/em\u003e(1), 92\u0026ndash;98. https://doi.org/10.1590/S0102-0536-2022012\u003c/li\u003e\n\u003cli\u003eSmith, C. M. (2005). Plant resistance to arthropods: Molecular and conventional approaches. \u003cem\u003eSpringer Dordrecht 426\u003c/em\u003e(1) 101-122. https://doi.org/10.1007/1-4020-3702-3\u003c/li\u003e\n\u003cli\u003eSun, J., Rutkoski, J. E., Poland, J. A., Crossa, J., Jannink, J.-L., and Sorrells, M. E. (2017). Multitrait, random regression, or simple repeatability model in high-throughput phenotyping data improve genomic prediction for wheat grain yield. \u003cem\u003eThe Plant Genome, 10\u003c/em\u003e(2), 1\u0026ndash;12. https://doi.org/10.3835/plantgenome2016.11.0111\u003c/li\u003e\n\u003cli\u003eTang, H., and He, Z. (2021). Advances and challenges in quantitative delineation of the genetic architecture of complex traits. \u003cem\u003eQuantitative Biology (Beijing, China)\u003c/em\u003e, \u003cem\u003e9\u003c/em\u003e(2), 168. https://doi.org/10.15302/J-QB-021-0249\u003c/li\u003e\n\u003cli\u003eTrevisan, R., P\u0026eacute;rez, O., Schmitz, N., Diers, B., and Martin, N. (2020). High-Throughput Phenotyping of Soybean Maturity Using Time Series UAV Imagery and Convolutional Neural Networks. \u003cem\u003eRemote Sensing\u003c/em\u003e, \u003cem\u003e12\u003c/em\u003e(21), 3617. https://doi.org/10.3390/rs12213617\u003c/li\u003e\n\u003cli\u003eUSDA, U. S. D. of A. (2024). \u003cem\u003eWorld Agricultural Production\u003c/em\u003e. Foreign Agricultural Service/USDA. https://apps.fas.usda.gov/psdonline/circulars/production.pdf\u003c/li\u003e\n\u003cli\u003eVan Der Walt, S., Sch\u0026ouml;nberger, J. L., Nunez-Iglesias, J., Boulogne, F., Warner, J. D., Yager, N., Gouillart, E., and Yu, T. (2014). Scikit-image: Image processing in python. \u003cem\u003ePeerJ\u003c/em\u003e, \u003cem\u003e2014\u003c/em\u003e(1). https://doi.org/10.7717/PEERJ.453\u003c/li\u003e\n\u003cli\u003eVan Rossum, G., and Drake, F. L. (2009). \u003cem\u003ePython 3 Reference Manual\u003c/em\u003e. CreateSpace.\u003c/li\u003e\n\u003cli\u003eVapnik, V. (2000). SVM method of estimating density, conditional probability, and conditional density. \u003cem\u003eProceedings - IEEE International Symposium on Circuits and Systems\u003c/em\u003e, \u003cem\u003e2\u003c/em\u003e. https://doi.org/10.1109/ISCAS.2000.856437\u003c/li\u003e\n\u003cli\u003eVeiga, R. F. A., Rossetto, C. J., Razera, L. F., Gallo, P. B., Bertoletoo, N., Medina, P. F., Tisselli Filho, O., and Cione, J. (1999). caracteriza\u0026ccedil;\u0026atilde;o morfol\u0026oacute;gica e agron\u0026ocirc;mica do cultivar de soja \u0026lsquo;Iac-100.\u0026rsquo; \u003cem\u003eIAC - Boletim T\u0026eacute;cnico 177 - Instituto Agron\u0026ocirc;mico\u003c/em\u003e, \u003cem\u003eMay 2017\u003c/em\u003e, 23p.\u003c/li\u003e\n\u003cli\u003eVieira, C. C., Sarkar, S., Tian, F., Zhou, J., Jarquin, D., Nguyen, H. T., Zhou, J., and Chen, P. (2022). Differentiate Soybean Response to Off-Target Dicamba Damage Based on UAV Imagery and Machine Learning. \u003cem\u003eRemote Sensing 2022, Vol. 14, Page 1618\u003c/em\u003e, \u003cem\u003e14\u003c/em\u003e(7), 1618. https://doi.org/10.3390/RS14071618\u003c/li\u003e\n\u003cli\u003eYang, G., Liu, J., Zhao, C., Li, Z., Huang, Y., Yu, H., Xu, B., Yang, X., Zhu, D., Zhang, X., Zhang, R., Feng, H., Zhao, X., Li, Z., Li, H., and Yang, H. (2017). Unmanned Aerial Vehicle Remote Sensing for Field-Based Crop Phenotyping: Current Status and Perspectives. \u003cem\u003eFrontiers in Plant Science\u003c/em\u003e, \u003cem\u003e8\u003c/em\u003e. https://doi.org/10.3389/FPLS.2017.01111\u003c/li\u003e\n\u003cli\u003eYu, N., Li, L., Schmitz, N., Tian, L. F., Greenberg, J. A., and Diers, B. W. (2016). Development of methods to improve soybean yield estimation and predict plant maturity with an unmanned aerial vehicle based platform. \u003cem\u003eRemote Sensing of Environment\u003c/em\u003e, \u003cem\u003e187\u003c/em\u003e, 91\u0026ndash;101. https://doi.org/10.1016/J.RSE.2016.10.005\u003c/li\u003e\n\u003cli\u003eYuan, W., Wijewardane, N. K., Jenkins, S., Bai, G., Ge, Y., and Graef, G. L. (2019). Early Prediction of Soybean Traits through Color and Texture Features of Canopy RGB Imagery. \u003cem\u003eScientific Reports\u003c/em\u003e, \u003cem\u003e9\u003c/em\u003e(1). https://doi.org/10.1038/s41598-019-50480-x\u003c/li\u003e\n\u003cli\u003eZhang, J., Virk, S., Porter, W., Kenworthy, K., Sullivan, D., and Schwartz, B. (2019). Applications of unmanned aerial vehicle based imagery in turfgrass field trials. \u003cem\u003eFrontiers in Plant Science\u003c/em\u003e, \u003cem\u003e10\u003c/em\u003e, 429695. https://doi.org/10.3389/FPLS.2019.00279/BIBTEX\u003c/li\u003e\n\u003cli\u003eZhou, J., Zhou, J., Ye, H., Ali, M. L., Nguyen, H. T., and Chen, P. (2020). Classification of soybean leaf wilting due to drought stress using UAV-based imagery. \u003cem\u003eComputers and Electronics in Agriculture\u003c/em\u003e, \u003cem\u003e175\u003c/em\u003e, 105576. https://doi.org/10.1016/J.COMPAG.2020.105576\u003c/li\u003e\n\u003cli\u003eZhou, X., Zheng, H. B., Xu, X. Q., He, J. Y., Ge, X. K., Yao, X., Cheng, T., Zhu, Y., Cao, W. X., and Tian, Y. C. (2017). Predicting grain yield in rice using multi-temporal vegetation indices from UAV-based multispectral and digital imagery. \u003cem\u003eISPRS Journal of Photogrammetry and Remote Sensing\u003c/em\u003e, \u003cem\u003e130\u003c/em\u003e, 246\u0026ndash;255. https://doi.org/10.1016/J.ISPRSJPRS.2017.05.003\u003c/li\u003e\n\u003cli\u003eZhou, J., Beche, E., Vieira, C. C., Yungbluth, D., Zhou, J., Scaboo, A., \u0026amp; Chen, P. (2022). Improve soybean variety selection accuracy using UAV-based high-throughput phenotyping technology. \u003cem\u003eFrontiers in Plant Science, 12\u003c/em\u003e, 768742. https://doi.org/10.3389/fpls.2021.768742\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"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":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Crop protection, Glycine max, Machine learning models, Unmanned aerial vehicles (UAVs), Vegetation index, Texture indices","lastPublishedDoi":"10.21203/rs.3.rs-7736736/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7736736/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eThe stink bug complex is one of the most damaging pests of soybean, reducing yield and seed quality. Genetic resistance remains the most sustainable and effective management strategy, but its quantitative inheritance and labor-intensive field phenotyping make its implementation in breeding programs challenging. This study explored high-throughput phenotyping (HTP) using unmanned aerial vehicles (UAVs) equipped with RGB cameras to evaluate a soybean population and the potential of phenotyping to stink bug resistance by correlating image-derived features and machine learning (ML) models. A population of 304 soybean lines was evaluated in alpha-lattice design trials across two seasons under natural infestations. Five resistance-related traits, grain yield (GY), hundred-seed weight (HSW), number of days to maturity (NDM), tolerance (TOL), and leaf retention (LR), were manually scored and linked to UAV-derived vegetation indices (VIs) and texture indices (TIs). Three ML models (AdaBoost, SVM, MLP) were tested to predict these traits from aerial features. Results showed that VIs, particularly Visible Atmospherically Resistant Index at the first percentile (VARI_P25), were consistently associated with resistance-related traits, while decision tree analysis highlighted TIs at 45° and 135° as complementary sources of structural information. Prediction ability was highest for GY, HSW, and NDM, especially in flights near flowering and maturity, but remained low for TOL and LR. Integrating multiple flights modestly improved accuracy, whereas cross-season predictions were unreliable. Nonetheless, indices such as VARI_P25 provided useful cross-season correlations for HSW and TOL, enabling early screening of less promising lines. This pioneering study demonstrates that UAV–ML pipelines can capture genetic signals of stink bug resistance in soybean, despite environmental complexity. These findings open new avenues for resistance phenotyping, supporting more efficient breeding strategies and accelerating genetic gains in soybean improvement.\u003c/p\u003e","manuscriptTitle":"Optimizing Soybean Breeding: High-Throughput Phenotyping for Stink Bug Resistance and High Yields","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-11-13 11:04:36","doi":"10.21203/rs.3.rs-7736736/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"124929cb-5098-44c4-94f8-a7716b7852c9","owner":[],"postedDate":"November 13th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2026-04-30T09:24:09+00:00","versionOfRecord":[],"versionCreatedAt":"2025-11-13 11:04:36","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7736736","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7736736","identity":"rs-7736736","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}