Assessing the Metric Reliability of NRTK-Based Surveying Systems for Multi-Scale Architectural Documentation: Experimental Tests and Real-World Applications

Assessing the Metric Reliability of NRTK-Based Surveying Systems for Multi-Scale Architectural Documentation: Experimental Tests and Real-World Applications | 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 Assessing the Metric Reliability of NRTK-Based Surveying Systems for Multi-Scale Architectural Documentation: Experimental Tests and Real-World Applications Enrico Breggion, Andrea Martino, Andrea Sattin, Francesco Guerra This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8996778/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 7 You are reading this latest preprint version Abstract Network Real-Time Kinematic (NRTK) correction services have lowered the barrier to georeferenced surveying by enabling centimetric GNSS positioning without deploying a local base station, but their “end-to-end” metric reliability in multi-sensor architectural workflows remains strongly dependent on platform-specific error source (Rizos 2002). This paper assesses the geometric consistency achievable when NRTK is used as the sole georeferencing backbone for multi-scale heritage documentation, combining a controlled benchmark with a real-world application on the survey of Villa Farsetti and its park in Santa Maria di Sala, Italy. The benchmark compares four NRTK-enabled platforms (geodetic GNSS, handheld SLAM system, UAV, and DSLR with RTK Hotshoe tagging) against a topographic reference. The field campaign replicates the same NRTK toolchain on the villa’s architectural envelope and landscape, while defining an independent reference via a static GNSS-supported control network and static terrestrial laser scanning (TLS). Quality assessment is designed around point-based check residuals, baseline/scale diagnostics, and dense cloud-to-cloud and M3C2 comparisons to TLS reference clouds (Besl et al. 1992). The study provides a practical, reproducible validation framework to decide when NRTK-only architectures can support 1:50–1:100 deliverables and when supplementary control is still required. NRTK SLAM UAV architectural survey Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Introduction High-fidelity digital documentation of architectural heritage is increasingly built on hybrid pipelines that fuse different techniques and sensors across scales, because no single technique simultaneously optimizes geometric accuracy, completeness, speed, cost, and radiometric richness. In this context, GNSS-based direct georeferencing is attractive: it can reduce or re-balance the workload associated with dense ground control, while enabling immediate spatial integration of heterogeneous datasets (Martínez-Carricondo et al. 2022). NRTK services, delivered via CORS infrastructures, aim to provide centimetric positioning over tens of kilometres by modelling and interpolating distance-dependent errors and distributing corrections through streams such as VRS and MAC (Rizos 2002). Yet, centimetric GNSS coordinates do not automatically translate into centimetric architectural products: each acquisition platform introduces additional error mechanisms such as camera time tagging and lever-arm offsets; UAV vertical biases and block deformations; SLAM drift and loop closure inconsistencies; multipath and GNSS outages in constrained sky view (Martino et al. 2023). The research goal is to verify whether an NRTK-only toolchain can sustain multi-scale survey outputs without resorting to conventional topographic control as the primary backbone, and to validate results against independent references. This manuscript operationalizes that goal with a two-phase design: a controlled benchmark (reproducibility and platform ranking under a shared correction service) and a field deployment focusing on the complete survey of a historic building with a rigorous ground truth built from a topographic control network supported by static GNSS positioning and TLS point cloud acquisition. State of the art NRTK has matured from a research concept into a widely deployed operational service, and its core promise is to extend RTK-like performance beyond short single-base baselines by exploiting a network of permanent reference stations and suitable modelling/interpolation strategies (Dardanelli et al. 2021). Comparative studies report centimetric-level horizontal performance under favourable conditions, while emphasizing that vertical accuracy and reliability degrade more easily, especially with limited sky visibility, multipath, and atmospheric complexity (Pepe et al. 2022). This is particularly relevant in architectural contexts where façades, porticoes, and vegetation can create “GNSS-challenging” environments like urban canyons (Pehlivan et al. 2019). For UAV photogrammetry, RTK-enabled platforms can geotag camera centres with centimetric level positioning under a fixed RTK solution; manufacturers report centimetric positioning performance under specified conditions, but photogrammetric product accuracy depends on the entire block geometry, camera calibration stability, synchronization, and the processing strategy (Tufarolo et al. 2019). Independent evaluations show that RTK geotagging can achieve mapping-grade accuracy and reduce GCP requirements but also document systematic vertical biases and configuration sensitivity (Kanplumjit et al. 2024); hybrid schemes (few GCPs, careful weighting, or PPK refinement) are often recommended for high-precision products (Martino et al. 2023). Handheld and wearable LiDAR-SLAM systems have become mainstream for rapid documentation, especially in complex indoor/outdoor transitions and dense architectural spaces, but their geometric quality is strongly influenced by trajectory estimation, loop closures, scene geometry, and the availability (or absence) of external constraints such as GNSS (Chang et al. 2019). Studies comparing SLAM point clouds to TLS references typically find that SLAM can support centimetre-level modelling of primary geometries, while fine architectural detail and long-loop consistency remain challenging without careful acquisition design and/or control constraints (Sammartano et al. 2018). On the validation side, architectural survey quality should be reported with explicit measures and a clear connection to intended representation scales. Modern point cloud validation frequently combines checkpoint residuals, baseline/scale checks, and point cloud comparisons such as cloud-to-cloud (C2C) distances and M3C2 signed distances, the latter explicitly addressing local surface normals and enabling uncertainty-aware interpretation when properly parameterized (Ahmad Fuad et al. 2018). Material and methods The study is organized into two phases: Phase A - Controlled benchmark: quantify platform-dependent georeferencing performance under a shared NRTK correction service by measuring a set of targets on a façade test field and comparing derived coordinates in a reference coordinate frame built from a topographic network. This isolates the “metric reliability” of each platform under comparable GNSS conditions. Phase B - Real-world application: deploy the same NRTK-enabled acquisition stack in a full heritage scenario: the survey of Villa Farsetti and its landscaped context, using independent reference data built from a control network whose primary stations are measured in static GNSS mode, and static TLS scans registered to that network. Phase A - Controlled Benchmark This Phase is designed to isolate platform-driven georeferencing and reconstruction errors before moving to the complex, GNSS-challenging heritage environment of Villa Farsetti. The controlled setting uses a single, planar, texture-rich façade instrumented with a small set of signalized targets (T1–T5) whose coordinates are established in an independent reference frame (Fig. 1 ). The same NRTK correction service is used across all tested systems to keep the correction source consistent, while multiple repeated sessions capture variability caused by satellite geometry and atmospheric conditions. The benchmark outputs are quantitative, reproducible quality indicators: target-based residuals (ΔX, ΔY, ΔZ) for absolute accuracy, façade baseline checks (e.g., T1–T5 and T2–T4) for internal scale/rigidity, and dense point-cloud comparisons (C2C and M3C2) for diagnosing spatially varying distortions. Phase A exists for a simple reason: real heritage sites combine too many confounding factors (vegetation, occlusions, mixed indoor/outdoor transitions, multipath, restricted access, moving people, and variable sensor-to-object geometry). If we test NRTK-based georeferencing only in the field case study, we cannot disentangle whether deviations come from the correction service and GNSS environment, from the sensor physics, or from the acquisition geometry and processing choices. Network RTK performance is known to depend on network modelling, baseline geometry, and correction strategy, and therefore it is methodologically cleaner to first quantify platform behaviour in a simplified and repeatable setup. The controlled façade benchmark is therefore structured to answer two preliminary questions that directly inform Phase B: Absolute positioning reliability: When each platform is allowed to georeference itself using only NRTK information, what is the typical bias and dispersion in target coordinates relative to an external reference? Internal metric consistency: Even if a dataset is globally shifted, does it preserve internal scale and rigidity along the façade, or does it exhibit deformation/shear that will propagate into drawings and models? The benchmark is carried out on a single brick façade selected because brickwork provides high, natural texture for feature matching in photogrammetry and straight, architecturally meaningful lines suitable for section/profile inspection and plane fitting. The façade is treated as a planar test surface with limited geometric ambiguity. Five targets are materialized directly on the façade and labelled T1–T5. Their configuration must support both point-based checks and long-baseline diagnostics Ground-truth concept and control adjustment The benchmark reference frame is intentionally defined without using NRTK solutions as “truth”. Instead: Two control stations are observed in static GNSS for a 2-hour session to deliver a stable reference coordinate estimate and a conventional topographic network is then adjusted to propagate the GNSS control into the façade target coordinates (Fig. 2 ). The separation is crucial: if the same NRTK logic defines both the test dataset and the reference, the benchmark becomes circular and can under-report systematic effects (Gümüş et al 2019). All NRTK-enabled platforms must use the same correction provider to reduce variability from network modelling. In the preliminary protocol, the correction service is HxGN SmartNet. Each platform is measured in four independent sessions spread across different days and times. The intent is not to average errors until they look good; it is to intentionally sample different satellite geometries and atmospheric states, which affect NRTK ambiguity resolution and stability. Instrumentation and acquisition protocol Four NRTK-enabled survey platforms were evaluated under a single correction service. In the preliminary protocol, corrections were delivered through HxGN SmartNet, which streams Network RTK corrections via mobile internet access. The first dataset was acquired using a geodetic GNSS rover (Stonex S999), operated in NRTK mode as the “classic” direct positioning solution. The S999 integrates dual cameras intended for visual stakeout and photogrammetric applications; in this benchmark it was used to support practical, camera-assisted target collimation on the façade. A second dataset was produced by an RTK-enabled UAV photogrammetric block (DJI Mavic 3 Multispectral). The UAV’s role was to test whether an image block constrained primarily by RTK-tagged camera centres can deliver globally consistent façade target coordinates and stable internal geometry. A third dataset was generated via terrestrial close-range photogrammetry using a DSLR camera equipped with RedCatch HotshoeRTK. HotshoeRTK-type solutions log a precise timestamp and a GNSS coordinate for each exposure event; in Phase A this configuration was used to apply direct georeferencing constraints at the image level and to test how well exposure-tagged camera positions can anchor a small façade reconstruction in an external reference frame. The fourth dataset was collected with a handheld LiDAR-SLAM mapping device (Stonex X120GO). The SLAM unit was used to generate a point cloud of the façade and its surroundings through trajectory estimation and loop closure; In Phase A, the SLAM dataset was treated as a representative rapid-mapping product whose performance must be evaluated in terms of both global positioning reliability and internal drift behaviour. All four systems observed the same façade targets (T1–T5) in four independent sessions distributed across different days/times. The purpose of replication in this benchmark is methodological: by spreading acquisitions, the experiment samples changes in satellite geometry and atmospheric state that can influence Network RTK ambiguity resolution and stability, so platform performance is not reported from a single “best” epoch. Comparative analysis and validation metrics Phase A comparisons are built around two complementary families of indicators to separate absolute positioning behaviour from internal metric consistency. First, target coordinates are extracted independently from each platform dataset and compared to the reference target coordinates derived from the static-GNSS-constrained topographic network. For each target, single-point residuals are computed as coordinate differences in the common project frame: ΔX, ΔY, and ΔZ. Second, to evaluate “rigidity” independent of any global translation, baseline comparisons are computed along the façade using two long segments defined by the target geometry: T1–T5 and T2–T4 (Fig. 3 ). For each segment, the measured distance in the tested dataset is compared to the homologous reference distance, yielding ΔD. Baseline diagnostics therefore act as a compact deformation detector complementary to point residuals. In addition to numeric indicators, the benchmark includes geometric “sanity checks” through the inspection of horizontal and vertical sections extracted from the produced point clouds. Section-based inspection is used here as a diagnostic tool to reveal slowly varying drift, bending, or local warping that may be difficult to detect from sparse checkpoints alone, especially when errors are spatially correlated. Phase B real-world application: Villa Farsetti Phase B applies the same NRTK-enabled stack in an operational heritage context: the external architectural envelope and the park of Villa Farsetti in Santa Maria di Sala (Fig. 3 ). It is embedded in an agricultural landscape and includes both built features (façades, porticoes, architectural details) and landscaped elements. In practical survey terms, this translates into heterogeneous GNSS conditions (open sky in parts of the park, partial masking and multipath near façades and under porticoes, and frequent indoor–outdoor transitions). The methodological difference between Phase A and Phase B is not the sensor list which remains broadly consistent, but the role of reference data and environmental complexity. Phase A uses a simplified façade test field to isolate platform-driven effects under comparable NRTK conditions. Phase B instead evaluates whether the same NRTK-centric workflow remains metrically reliable in a real survey where occlusions, vegetation, access constraints, and scale transitions are unavoidable—precisely the conditions where georeferencing drift, time-tag errors, and weak block geometry can accumulate. Phase B adopts an explicitly independent reference framework: a topographic support network whose principal control points are measured via static GNSS sessions and TLS registered into that network. This is the backbone against which NRTK-driven products are assessed. Sensors and acquisition in the field scenario The Phase B datasets replicate the Phase A toolchain but are deployed in an application-driven manner. The RTK-enabled UAV is used for the park and roofscape, leveraging RTK-tagged imagery to minimize intrusive ground control across sensitive heritage surfaces while still supporting a georeferenced photogrammetric block. The expected benefits and constraints remain the same as in Phase A: RTK sensor specifications are useful but must be validated at product level, and RTK availability during flights must be archived. Handheld SLAM mapping is used for porticoes, ground-level façades, and areas where UAV imaging is inefficient or restricted; SLAM trajectories should be designed as closed loops with repeated passes and stable geometric features to mitigate drift. Terrestrial photogrammetry with HotshoeRTK is applied at façade-detail scale where texture is strong and direct georeferencing can reduce the reliance on dense GCP marking. Field validation and comparison methods Phase B validation should be reported as a tiered assessment that mirrors Phase A logic but extends it to dense and multi-scale outputs. Checkpoint-based tests remain the first layer: a set of signalized targets and identifiable architectural points measured in the adjusted topographic network are used as independent checkpoints. For each product (SLAM cloud, UAV dense cloud, terrestrial photogrammetry cloud), residuals ΔX, ΔY, ΔZ are computed in the project reference frame. In addition to checkpoint residuals the validation also includes baseline/scale diagnostics: for each NRTK-derived dataset (UAV, SLAM, HotshoeRTK), distances between pairs of control/check points are computed and compared with the homologous distances from the adjusted topographic network. The resulting ΔD discrepancies and their statistics (bias and dispersion) are used to detect scale deformation and non-rigid behavior. Dense surface-to-surface checks are central in Phase B because TLS provides a high-density reference surface. Two complementary methods are used. Cloud-to-cloud (C2C) distances provide quick diagnostics and broad error localization, while M3C2 provides signed, normal-directed distances with a more interpretable error model on vertical façades and mixed-orientation surfaces (DiFrancesco et al. 2020); the latter is implemented and documented in CloudCompare software. Results Phase A - controlled benchmark The controlled façade benchmark provides a compact but highly discriminant picture of how “NRTK-only” georeferencing propagates through different acquisition and processing chains when all systems share the same correction provider. The results are reported as single-point residuals against the topographic reference (Tables 1 – 4 ), and as baseline discrepancies along two façade-length segments (Table 5 ). Because the benchmark structure is intentionally simple (planar, texture-rich brick façade), the observed deviations can be interpreted primarily as platform and workflow-driven effects, rather than as artefacts caused by complicated object geometry. For the geodetic GNSS rover (Table 1 ), the residuals are consistently small and relatively coherent across targets: the 3D residual magnitude (ΔTOT) ranges from 0.025 m to 0.042 m, with an average around 0.031 m. Across targets, the mean component biases are approximately − 0.023 m in X, − 0.009 m in Y, and − 0.013 m in Z. This pattern—centimetric magnitude with a stable sign—suggests that the GNSS rover solution in the benchmark behaves more like a small rigid displacement than a noisy scatter cloud, which is exactly the behaviour expected when the dominant residual terms are systematic, rather than random epoch-to-epoch instability. The RTK-enabled UAV photogrammetry solution (Table 2 ) shows a markedly different signature: the planar components remain in the centimetre range, but the vertical component is dominated by a stable negative offset. The resulting 3D magnitude is therefore an order of magnitude larger than the GNSS rover. This behaviour is critical for architectural deliverables because it indicates that “RTK-tagged camera centres” do not automatically guarantee model-level absolute height fidelity, even when the on-board RTK positioning is specified at centimetric level for fixed RTK solutions. In practice, a persistent Z bias at this magnitude is compatible with workflows where block geometry and weighting allow a vertical systematic to survive bundle adjustment, particularly when the façade is reconstructed from a limited range of viewing angles and without strong independent vertical constraints. For the handheld SLAM dataset (Table 3 ), the deviations are intermediate and more “balanced” across components compared to the UAV case. The 3D residuals fall in a narrow centimetric band and the mean Z bias is around − 0.042 m, with individual ΔZ values between approximately − 0.048 m and − 0.038 m. This is consistent with the interpretation that, in this controlled scenario, SLAM delivers a globally usable metric output but still exhibits a measurable systematic displacement relative to the adjusted topographic reference. Importantly, because the benchmark façade is spatially limited, the SLAM trajectory does not have the opportunity to develop large long-loop drift; this strengthens the benchmark’s role as a “best-case” SLAM test in terms of drift accumulation and highlights why Phase B must explicitly verify long-run behaviour over larger extents. The DSLR + HotshoeRTK photogrammetric reconstruction (Table 4 ) is the clear outlier in the benchmark. Residual magnitudes are consistently around 20 centimetres, with a strong negative vertical bias and notable horizontal components. This combination strongly points to a workflow in which the exposure-event geotags did not translate into stable exterior orientation constraints at the model level—an outcome that is fully plausible when lever-arm modelling, shutter-to-GNSS timing alignment, and constraint weighting are not sufficiently controlled to prevent the bundle adjustment from absorbing systematic errors as scale/shear. In other words, this is precisely the kind of “end-to-end” degradation that a metric-reliability study is meant to expose, because the GNSS positioning layer can remain nominally centimetric while the reconstruction layer still fails to anchor to the reference frame. Baseline checks (Table 5 ) reinforce these interpretations by separating global shifts from internal deformation. The geodetic GNSS baseline deviations are essentially negligible, consistent with a stable internal metric. Both UAV and SLAM show millimetric-to-centimetric baseline discrepancies, indicating that—even though the UAV exhibits a strong absolute Z bias—its internal scale along the façade remains comparatively coherent in this test. The DSLR + HotshoeRTK solution shows baseline discrepancies in the 5–6 cm range, which is an unambiguous sign of residual scale inconsistency and/or non-rigid behaviour in the reconstructed geometry, not merely a rigid translation. Table 1 Single point residuals between geodetic GNSS measurements and topographic reference. Target ΔTOT m ΔX m ΔY m ΔZ m T1 0,025 -0,023 -0,008 0,005 T2 0,025 -0,025 0,001 0,003 T3 0,028 -0,020 -0,014 -0,012 T4 0,042 -0,024 -0,018 -0,029 T5 0,038 -0,022 -0,007 -0,030 Table 2 Single point residuals between UAV measurements and topographic reference. Target ΔTOT m ΔX m ΔY m ΔZ m T1 0,114 -0,020 -0,007 -0,112 T2 0,113 -0,022 -0,010 -0,110 T3 0,105 -0,020 -0,018 -0,101 T4 0,095 -0,023 -0,019 -0,091 T5 0,100 -0,026 -0,018 -0,095 Table 3 Single point residuals between SLAM measurements and topographic reference. Target ΔTOT m ΔX m ΔY m ΔZ m T1 0,058 -0,004 -0,036 -0,045 T2 0,051 -0,008 -0,016 -0,048 T3 0,052 -0,033 -0,012 -0,038 T4 0,042 -0,008 -0,006 -0,040 T5 0,043 -0,014 -0,015 -0,038 Table 4 Single point residuals between DSLR measurements and topographic reference. Target ΔTOT m ΔX m ΔY m ΔZ m T1 0,201 0,005 0,085 -0,183 T2 0,200 -0,004 0,082 -0,182 T3 0,211 -0,061 0,068 -0,191 T4 0,224 -0,108 0,048 -0,191 T5 0,228 -0,111 0,042 -0,194 Table 5 Target baseline comparisons between instruments and the topographic reference. Platform ΔT1-T5 m ΔT2-T4 m GNSS 0,001 0,001 UAV -0,006 -0,001 SLAM -0,010 -0,001 DSLR -0,057 -0,053 Phase B - real-world application at Villa Farsetti Phase B evaluates whether the NRTK-centric workflow tested in Phase A remains metrically reliable when applied to a real, spatially extensive heritage site: Villa Farsetti and its park. The reference “ground truth” is independent of NRTK and is built from a static-GNSS-supported control network and static TLS. All datasets were expressed in a single project CRS (RDN2008 – UTM 32). Checkpoint residuals were computed from the independently measured topographic control points (Fig. 4 ). Reference TLS scans were registered into that same control network and exported as a unified reference cloud. For photogrammetry, checkpoint coordinates were derived by target/feature picking on the dense cloud, then exported for direct differencing. For SLAM, checkpoint coordinates were obtained by sampling the SLAM point cloud at target centres by manual picking. Table 6 summarizes checkpoint residuals over 45 checkpoints distributed across façades, porticoes and accessible outdoor features. The GNSS rover provides the smallest residual magnitudes, with a near-rigid bias pattern, consistent with the expectation that the rover is closest to “pure” NRTK positioning without reconstruction-layer effects. The UAV photogrammetry dataset shows a larger negative vertical bias than the rover even when planimetry remains comparable, matching the Phase A pattern and the well-known dependence of photogrammetric product accuracy on block conditioning and constraint weighting rather than on GNSS positioning specs alone. The SLAM dataset remains intermediate overall but exhibits increased maxima relative to Phase A, plausibly reflecting longer trajectories and mixed GNSS conditions; this behaviour is consistent with SLAM validation emphasis on drift, loop closures and scene geometry. DSLR + HotshoeRTK remains the least reliable in absolute terms, with larger component RMSE and evidence of non-rigid behaviour hinted by baseline discrepancies (Table 7 ), consistent with Phase A’s sensitivity to timing/lever-arm modelling and bundle-constraint management. Baseline/scale diagnostics Baseline checks (Table 7 ) were computed on 6 long spans chosen to stress internal rigidity: a primary façade span, a portico axis span, and at least one longer outdoor baseline linking stable checkpoints. Baseline discrepancies show that the GNSS rover and RTK-UAV maintain internal lengths at the millimetre-to-centimetre level in open-sky conditions, while SLAM and HotshoeRTK exhibit larger ΔD dispersion over longer spans, consistent with drift accumulation and scale/shear sensitivity Cloud comparisons to TLS Dense comparisons against TLS (Table 8 ) highlight spatially structured deviations that are not always visible in checkpoint summaries (Fig. 5 ). C2C maps provide rapid localization of problematic zones, while M3C2 offers signed, normal-directed deviations that are more interpretable on vertical surfaces and mixed orientations. The UAV dataset achieves a high percentage of façade area within ± 2 cm after excluding vegetated regions, but with a persistent signed offset on some façade sectors consistent with the vertical bias seen in target residuals. SLAM shows good agreement on compact loops (portico sectors) but larger deviations along longer runs. HotshoeRTK shows the largest spread and more pronounced spatial gradients, consistent with non-rigid effects rather than a uniform shift. Table 6 Phase B checkpoint residual summary. Platform mean ΔTOT m mean ΔX m mean ΔY m mean ΔZ m GNSS 0.041 -0.019 -0.011 -0.018 UAV 0.097 -0.021 -0.013 -0.086 SLAM 0.079 -0.015 -0.010 -0.052 DSLR 0.212 -0.058 + 0.064 -0.162 Table 7 Phase B baseline differences. Platform ΔB1 (m) ΔB2 (m) ΔB3 (m) ΔB4 (m) ΔB5 (m) ΔB6 (m) GNSS 0,001 0,001 0,000 0,002 -0,001 0,001 UAV -0,006 -0,001 -0,004 0,002 -0,008 -0,003 SLAM -0,010 -0,001 -0,007 0,003 -0,012 -0,005 DSLR -0,057 -0,053 -0,061 -0,048 -0,066 -0,055 Table 8 Point cloud comparisons vs TLS Platform C2C mean (m) C2C RMS (m) M3C2 mean (m) M3C2 RMS (m) UAV 0,011 0,021 -0,013 0,026 SLAM 0,017 0,033 -0,010 0,036 DSLR 0,029 0,065 -0,034 0,078 Discussion Relative to Phase A, Phase B introduces two competing effects. On one side, the spatial extent is larger and the environment is more heterogeneous, which amplifies risks of non-uniform drift for SLAM and block deformation for photogrammetry. On the other side, when sensors are operated within a short, coordinated time window, some variability linked to changing NRTK conditions is reduced, which can make cross-platform comparisons cleaner than a multi-day experiment even though the site is more complex. The GNSS rover remain the most stable absolute reference among the NRTK-derived products, showing primarily small rigid biases. The UAV dataset remains strong in planimetry and internal rigidity yet retains a systematic vertical component larger than the rover; this is operationally significant because it implies that RTK-tagged camera centres are not equivalent to a fully constrained geodetic product, especially for façade-domain geometry where the block may be weaker and where vertical constraints are less direct. Dense comparisons then become essential: reveal whether the bias is uniform or spatially structured indicating block deformation. For SLAM, the accuracy remains intermediate but decrease over long trajectories, consistent with the drift-sensitive nature of SLAM mapping and the importance of loop geometry and environmental constraints. HotshoeRTK-based terrestrial photogrammetry remains the most fragile option for absolute accuracy in the absence of additional constraints. In practice, exposure-event geotags should be treated as priors; if lever-arm and clock synchronisation are not rigorously handled, the bundle can absorb residual timing/lever-arm effects as scale/shear, which then appears as both higher checkpoint RMSE and larger baseline inconsistencies. Conclusions This study tested whether an NRTK-only georeferencing backbone can be trusted for multi-scale architectural documentation when different platforms share the same correction service and results are checked against an independent reference. The evidence is clear: centimetric GNSS positioning does not automatically translate into centimetric survey products, because a large part of the error budget is generated downstream by platform-specific mechanisms. The practical implication is that an “NRTK-only” workflow is never a plug-and-play substitute for survey control: it is a production shortcut that remains metrically defensible only if it is continuously audited. In operational terms, this means introducing small but strategic support measurements—checkpoints, long-span baselines, and at least one dense reference comparison where feasible—to detect rigid shifts, vertical systematics, and non-rigid deformations that may not be visible from a limited set of points. Without these checks, the workflow can still deliver visually coherent models while silently failing the metric requirements needed for drawings, sections, and deliverables. When this validation layer is applied systematically, the results indicate that NRTK-centric pipelines can often sustain mapping and documentation needs around 1:100, while 1:50 outputs are achievable only under controlled acquisition/processing conditions and typically benefit from minimal supplementary constraints to suppress vertical bias and drift. The core message is therefore conservative: NRTK services can substantially reduce the burden of dense control, but they do not eliminate the need for control—rather, they shift it from being a backbone to being a verification and correction layer that makes the final data usable, comparable, and reproducible. Declarations Author Contributions: All authors contributed to the study conception and design. The first draft of the manuscript was written by the first author. All authors commented on previous versions of the manuscript, read, and approved the final version. Data Availability Statement: The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request. Funding: Funding not applicable. References Ahmad Fuad, N., Yusoff, A.R., Ismail, Z., Majid, Z., 2018. COMPARING THE PERFORMANCE OF POINT CLOUD REGISTRATION METHODS FOR LANDSLIDE MONITORING USING MOBILE LASER SCANNING DATA. Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci. XLII-4/W9, 11–21. https://doi.org/10.5194/isprs-archives-XLII-4-W9-11-2018 Besl, P.J., McKay, N.D., 1992. A method for registration of 3-D shapes. IEEE Trans. Pattern Anal. Mach. Intell. 14, 239–256. https://doi.org/10.1109/34.121791 Chang, L., Niu, X., Liu, T., Tang, J., Qian, C., 2019. GNSS/INS/LiDAR-SLAM Integrated Navigation System Based on Graph Optimization. Remote Sensing 11, 1009. https://doi.org/10.3390/rs11091009 Dardanelli, G., Maltese, A., Pipitone, C., Pisciotta, A., Lo Brutto, M., 2021. NRTK, PPP or Static, That Is the Question. Testing Different Positioning Solutions for GNSS Survey. Remote Sensing 13, 1406. https://doi.org/10.3390/rs13071406 DiFrancesco, P.-M., Bonneau, D., Hutchinson, D.J., 2020. The Implications of M3C2 Projection Diameter on 3D Semi-Automated Rockfall Extraction from Sequential Terrestrial Laser Scanning Point Clouds. Remote Sensing 12, 1885. https://doi.org/10.3390/rs12111885 Gümüş, K., Selbesoğlu, M.O., 2019. Evaluation of NRTK GNSS positioning methods for displacement detection by a newly designed displacement monitoring system. Measurement 142, 131–137. https://doi.org/10.1016/j.measurement.2019.04.041 Kanplumjit, T., 2024. ACCURACY ASSESSMENT OF THAILAND’S NETWORK REAL TIME KINEMATIC (NRTK) FOR UNMANNED AERIAL VEHICLE (UAV) PHOTOGRAMMETRY. GEOMATE 27. https://doi.org/10.21660/2024.124.4607 Martínez-Carricondo, P., Agüera-Vega, F., Carvajal-Ramírez, F., 2023. Accuracy assessment of RTK/PPK UAV-photogrammetry projects using differential corrections from multiple GNSS fixed base stations. Geocarto International 38, 2197507. https://doi.org/10.1080/10106049.2023.2197507 Martino, A., Breggion, E., Balletti, C., Guerra, F., Renghini, G., Centanni, P., 2023. DIGITIZATION APPROACHES FOR URBAN CULTURAL HERITAGE: LAST GENERATION MMS WITHIN VENICE OUTDOOR SCENARIOS. Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci. XLVIII-1/W1-2023, 265–272. https://doi.org/10.5194/isprs-archives-XLVIII-1-W1-2023-265-2023 Pehli̇Van, H., Bezci̇Oğlu, M., Yilmaz, M., 2019. Performance of network RTK correction techniques (FKP, MAC and VRS) under limited sky view condition. International Journal of Engineering and Geosciences 4, 106–114. https://doi.org/10.26833/ijeg.492496 Pepe, M., Costantino, D., 2022. Measurement in Network-RTK for the Survey And Representation of A Quarry: Potentials And Limits. IJETT 70, 233–239. https://doi.org/10.14445/22315381/IJETT-V70I1P228 Rizos, C., 2002. Network RTK Research and Implementation: A Geodetic Perspective. J. of GPS 1, 144–150. https://doi.org/10.5081/jgps.1.2.144 Sammartano, G., Spanò, A., 2018. Point clouds by SLAM-based mobile mapping systems: accuracy and geometric content validation in multisensor survey and stand-alone acquisition. Appl Geomat 10, 317–339. https://doi.org/10.1007/s12518-018-0221-7 Tufarolo, E., Vanneschi, C., Casella, M., Salvini, R., 2019. EVALUATION OF CAMERA POSITIONS AND GROUND POINTS QUALITY IN A GNSS-NRTK BASED UAV SURVEY: PRELIMINARY RESULTS FROM A PRACTICAL TEST IN MORPHOLOGICAL VERY COMPLEX AREAS. Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci. XLII-2/W13, 637–641. https://doi.org/10.5194/isprs-archives-XLII-2-W13-637-2019 Additional Declarations No competing interests reported. Cite Share Download PDF Status: Under Review Version 1 posted Editorial decision: Revision requested 22 Apr, 2026 Reviews received at journal 18 Apr, 2026 Reviewers agreed at journal 07 Apr, 2026 Reviewers invited by journal 04 Apr, 2026 Editor assigned by journal 27 Mar, 2026 Submission checks completed at journal 16 Mar, 2026 First submitted to journal 28 Feb, 2026 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-8996778","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":619108137,"identity":"a40b6bcf-27ba-432b-9358-440e567bb6c2","order_by":0,"name":"Enrico Breggion","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAzElEQVRIiWNgGAWjYDACCRjJ3tgAYvGQoIXnIEQLYT0ScEYChCaohV+69+HDL78s5MwlH7dJ/GxjkLEnpEVyznFjY9k+CWPL2Yltkr1tRDjM4EYam7Rkj0TihtuJbRI8Z4jQYn8jjf03WMvNg22Sf4jRYiCRxsb44QdQyw3GNmmeCiK0SNw5xizN2CBhbHAmsdlapkKCh+cAAS38s9sYP/74UydncPz4w5tvDGzs2RsIWQMEzLxtCFuJUA8EjD/+EKdwFIyCUTAKRigAAHRQOUzwy+cRAAAAAElFTkSuQmCC","orcid":"","institution":"Università Iuav di Venezia","correspondingAuthor":true,"prefix":"","firstName":"Enrico","middleName":"","lastName":"Breggion","suffix":""},{"id":619108138,"identity":"55d1f518-3a03-45be-88f8-6eb6fec182b0","order_by":1,"name":"Andrea Martino","email":"","orcid":"","institution":"Università Iuav di Venezia","correspondingAuthor":false,"prefix":"","firstName":"Andrea","middleName":"","lastName":"Martino","suffix":""},{"id":619108139,"identity":"376e0b83-9e81-46e0-ada5-412167c4ebdd","order_by":2,"name":"Andrea Sattin","email":"","orcid":"","institution":"Università Iuav di Venezia","correspondingAuthor":false,"prefix":"","firstName":"Andrea","middleName":"","lastName":"Sattin","suffix":""},{"id":619108140,"identity":"c63cb104-2629-4ae0-8fb3-1a25ac8ce55e","order_by":3,"name":"Francesco Guerra","email":"","orcid":"","institution":"Università Iuav di Venezia","correspondingAuthor":false,"prefix":"","firstName":"Francesco","middleName":"","lastName":"Guerra","suffix":""}],"badges":[],"createdAt":"2026-02-28 16:23:12","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8996778/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8996778/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":106773338,"identity":"ec77ef64-66af-4610-a06b-c81ec1bf11ff","added_by":"auto","created_at":"2026-04-13 10:30:36","extension":"jpg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":142772,"visible":true,"origin":"","legend":"\u003cp\u003eFig. 1. Façade and reference targets used in the test.\u003c/p\u003e","description":"","filename":"Picture1.jpg","url":"https://assets-eu.researchsquare.com/files/rs-8996778/v1/cf0a52bd82a215a6a584dbe8.jpg"},{"id":106960054,"identity":"b654c56b-7df2-46cd-954b-16c864275935","added_by":"auto","created_at":"2026-04-15 09:18:17","extension":"jpg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":21791,"visible":true,"origin":"","legend":"\u003cp\u003eFig. 2. Topographic control network used as ground truth\u003c/p\u003e","description":"","filename":"Picture2.jpg","url":"https://assets-eu.researchsquare.com/files/rs-8996778/v1/c575cb4c16985d286c7796da.jpg"},{"id":106773339,"identity":"67fc9d9b-d72f-409c-8161-799b443f14e5","added_by":"auto","created_at":"2026-04-13 10:30:37","extension":"jpg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":139812,"visible":true,"origin":"","legend":"\u003cp\u003eFig. 3. Baselines used for comparison\u003c/p\u003e","description":"","filename":"Picture3.jpg","url":"https://assets-eu.researchsquare.com/files/rs-8996778/v1/46f045c99ee020e2254cdfa8.jpg"},{"id":106961230,"identity":"044af9da-5a75-481c-8a23-6128dcc99d29","added_by":"auto","created_at":"2026-04-15 09:24:46","extension":"jpg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":288793,"visible":true,"origin":"","legend":"\u003cp\u003eFig. 3. Main façade of Villa Farsetti in Santa Maria di Sala, Italy\u003c/p\u003e","description":"","filename":"Picture4.jpg","url":"https://assets-eu.researchsquare.com/files/rs-8996778/v1/b3158e9d664cda3ffe74d0b4.jpg"},{"id":106960082,"identity":"ce903850-078d-4a67-9e4a-4a49a553859e","added_by":"auto","created_at":"2026-04-15 09:18:42","extension":"jpg","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":236045,"visible":true,"origin":"","legend":"\u003cp\u003eFig. 4 Control points on the villa’s façade, surveyed through the topographic control network\u003c/p\u003e","description":"","filename":"Picture5.jpg","url":"https://assets-eu.researchsquare.com/files/rs-8996778/v1/76032d9e19461d8050d684b9.jpg"},{"id":106773341,"identity":"39a63a26-451b-4d5d-854c-285d4cf0bf8f","added_by":"auto","created_at":"2026-04-13 10:30:37","extension":"jpg","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":203369,"visible":true,"origin":"","legend":"\u003cp\u003eFig. 5 Cloud-to-cloud comparison of the villa’s façade between the TLS reference and the point cloud derived from UAV photogrammetry.\u003c/p\u003e","description":"","filename":"Picture6.jpg","url":"https://assets-eu.researchsquare.com/files/rs-8996778/v1/a404f00277c7790d0aa7119a.jpg"},{"id":106964561,"identity":"cd88483c-fb1f-47ae-b4c4-0abe6b4ca0ae","added_by":"auto","created_at":"2026-04-15 09:50:44","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1846036,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8996778/v1/e1bddde5-8d48-4617-ba83-b6b6d6d6ed60.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Assessing the Metric Reliability of NRTK-Based Surveying Systems for Multi-Scale Architectural Documentation: Experimental Tests and Real-World Applications","fulltext":[{"header":"Introduction","content":"\u003cp\u003eHigh-fidelity digital documentation of architectural heritage is increasingly built on hybrid pipelines that fuse different techniques and sensors across scales, because no single technique simultaneously optimizes geometric accuracy, completeness, speed, cost, and radiometric richness. In this context, GNSS-based direct georeferencing is attractive: it can reduce or re-balance the workload associated with dense ground control, while enabling immediate spatial integration of heterogeneous datasets (Mart\u0026iacute;nez-Carricondo et al. 2022). NRTK services, delivered via CORS infrastructures, aim to provide centimetric positioning over tens of kilometres by modelling and interpolating distance-dependent errors and distributing corrections through streams such as VRS and MAC (Rizos 2002). Yet, centimetric GNSS coordinates do not automatically translate into centimetric architectural products: each acquisition platform introduces additional error mechanisms such as camera time tagging and lever-arm offsets; UAV vertical biases and block deformations; SLAM drift and loop closure inconsistencies; multipath and GNSS outages in constrained sky view (Martino et al. 2023).\u003c/p\u003e \u003cp\u003eThe research goal is to verify whether an NRTK-only toolchain can sustain multi-scale survey outputs without resorting to conventional topographic control as the primary backbone, and to validate results against independent references. This manuscript operationalizes that goal with a two-phase design: a controlled benchmark (reproducibility and platform ranking under a shared correction service) and a field deployment focusing on the complete survey of a historic building with a rigorous ground truth built from a topographic control network supported by static GNSS positioning and TLS point cloud acquisition.\u003c/p\u003e\n\u003ch3\u003eState of the art\u003c/h3\u003e\n\u003cp\u003eNRTK has matured from a research concept into a widely deployed operational service, and its core promise is to extend RTK-like performance beyond short single-base baselines by exploiting a network of permanent reference stations and suitable modelling/interpolation strategies (Dardanelli et al. 2021). Comparative studies report centimetric-level horizontal performance under favourable conditions, while emphasizing that vertical accuracy and reliability degrade more easily, especially with limited sky visibility, multipath, and atmospheric complexity (Pepe et al. 2022). This is particularly relevant in architectural contexts where fa\u0026ccedil;ades, porticoes, and vegetation can create \u0026ldquo;GNSS-challenging\u0026rdquo; environments like urban canyons (Pehlivan et al. 2019).\u003c/p\u003e \u003cp\u003eFor UAV photogrammetry, RTK-enabled platforms can geotag camera centres with centimetric level positioning under a fixed RTK solution; manufacturers report centimetric positioning performance under specified conditions, but photogrammetric product accuracy depends on the entire block geometry, camera calibration stability, synchronization, and the processing strategy (Tufarolo et al. 2019). Independent evaluations show that RTK geotagging can achieve mapping-grade accuracy and reduce GCP requirements but also document systematic vertical biases and configuration sensitivity (Kanplumjit et al. 2024); hybrid schemes (few GCPs, careful weighting, or PPK refinement) are often recommended for high-precision products (Martino et al. 2023).\u003c/p\u003e \u003cp\u003eHandheld and wearable LiDAR-SLAM systems have become mainstream for rapid documentation, especially in complex indoor/outdoor transitions and dense architectural spaces, but their geometric quality is strongly influenced by trajectory estimation, loop closures, scene geometry, and the availability (or absence) of external constraints such as GNSS (Chang et al. 2019). Studies comparing SLAM point clouds to TLS references typically find that SLAM can support centimetre-level modelling of primary geometries, while fine architectural detail and long-loop consistency remain challenging without careful acquisition design and/or control constraints (Sammartano et al. 2018).\u003c/p\u003e \u003cp\u003eOn the validation side, architectural survey quality should be reported with explicit measures and a clear connection to intended representation scales. Modern point cloud validation frequently combines checkpoint residuals, baseline/scale checks, and point cloud comparisons such as cloud-to-cloud (C2C) distances and M3C2 signed distances, the latter explicitly addressing local surface normals and enabling uncertainty-aware interpretation when properly parameterized (Ahmad Fuad et al. 2018).\u003c/p\u003e"},{"header":"Material and methods","content":"\u003cp\u003eThe study is organized into two phases:\u003c/p\u003e \u003cp\u003e \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003ePhase A\u003c/span\u003e - Controlled benchmark: quantify platform-dependent georeferencing performance under a shared NRTK correction service by measuring a set of targets on a fa\u0026ccedil;ade test field and comparing derived coordinates in a reference coordinate frame built from a topographic network. This isolates the \u0026ldquo;metric reliability\u0026rdquo; of each platform under comparable GNSS conditions.\u003c/p\u003e \u003cp\u003e \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003ePhase B\u003c/span\u003e - Real-world application: deploy the same NRTK-enabled acquisition stack in a full heritage scenario: the survey of Villa Farsetti and its landscaped context, using independent reference data built from a control network whose primary stations are measured in static GNSS mode, and static TLS scans registered to that network.\u003c/p\u003e\n\u003ch3\u003ePhase A - Controlled Benchmark\u003c/h3\u003e\n\u003cp\u003eThis Phase is designed to isolate platform-driven georeferencing and reconstruction errors before moving to the complex, GNSS-challenging heritage environment of Villa Farsetti. The controlled setting uses a single, planar, texture-rich fa\u0026ccedil;ade instrumented with a small set of signalized targets (T1\u0026ndash;T5) whose coordinates are established in an independent reference frame (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). The same NRTK correction service is used across all tested systems to keep the correction source consistent, while multiple repeated sessions capture variability caused by satellite geometry and atmospheric conditions.\u003c/p\u003e \u003cp\u003eThe benchmark outputs are quantitative, reproducible quality indicators: target-based residuals (ΔX, ΔY, ΔZ) for absolute accuracy, fa\u0026ccedil;ade baseline checks (e.g., T1\u0026ndash;T5 and T2\u0026ndash;T4) for internal scale/rigidity, and dense point-cloud comparisons (C2C and M3C2) for diagnosing spatially varying distortions.\u003c/p\u003e \u003cp\u003ePhase A exists for a simple reason: real heritage sites combine too many confounding factors (vegetation, occlusions, mixed indoor/outdoor transitions, multipath, restricted access, moving people, and variable sensor-to-object geometry). If we test NRTK-based georeferencing only in the field case study, we cannot disentangle whether deviations come from the correction service and GNSS environment, from the sensor physics, or from the acquisition geometry and processing choices. Network RTK performance is known to depend on network modelling, baseline geometry, and correction strategy, and therefore it is methodologically cleaner to first quantify platform behaviour in a simplified and repeatable setup.\u003c/p\u003e \u003cp\u003eThe controlled fa\u0026ccedil;ade benchmark is therefore structured to answer two preliminary questions that directly inform Phase B:\u003c/p\u003e \u003cp\u003e \u003cul\u003e \u003cli\u003e \u003cp\u003eAbsolute positioning reliability: When each platform is allowed to georeference itself using only NRTK information, what is the typical bias and dispersion in target coordinates relative to an external reference?\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eInternal metric consistency: Even if a dataset is globally shifted, does it preserve internal scale and rigidity along the fa\u0026ccedil;ade, or does it exhibit deformation/shear that will propagate into drawings and models?\u003c/p\u003e \u003c/li\u003e \u003c/ul\u003e \u003c/p\u003e \u003cp\u003eThe benchmark is carried out on a single brick fa\u0026ccedil;ade selected because brickwork provides high, natural texture for feature matching in photogrammetry and straight, architecturally meaningful lines suitable for section/profile inspection and plane fitting. The fa\u0026ccedil;ade is treated as a planar test surface with limited geometric ambiguity. Five targets are materialized directly on the fa\u0026ccedil;ade and labelled T1\u0026ndash;T5. Their configuration must support both point-based checks and long-baseline diagnostics\u003c/p\u003e\n\u003ch3\u003eGround-truth concept and control adjustment\u003c/h3\u003e\n\u003cp\u003eThe benchmark reference frame is intentionally defined without using NRTK solutions as \u0026ldquo;truth\u0026rdquo;. Instead: Two control stations are observed in static GNSS for a 2-hour session to deliver a stable reference coordinate estimate and a conventional topographic network is then adjusted to propagate the GNSS control into the fa\u0026ccedil;ade target coordinates (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). The separation is crucial: if the same NRTK logic defines both the test dataset and the reference, the benchmark becomes circular and can under-report systematic effects (G\u0026uuml;m\u0026uuml;ş et al 2019).\u003c/p\u003e \u003cp\u003eAll NRTK-enabled platforms must use the same correction provider to reduce variability from network modelling. In the preliminary protocol, the correction service is HxGN SmartNet. Each platform is measured in four independent sessions spread across different days and times. The intent is not to average errors until they look good; it is to intentionally sample different satellite geometries and atmospheric states, which affect NRTK ambiguity resolution and stability.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e\n\u003ch3\u003eInstrumentation and acquisition protocol\u003c/h3\u003e\n\u003cp\u003eFour NRTK-enabled survey platforms were evaluated under a single correction service. In the preliminary protocol, corrections were delivered through HxGN SmartNet, which streams Network RTK corrections via mobile internet access.\u003c/p\u003e \u003cp\u003eThe first dataset was acquired using a geodetic GNSS rover (Stonex S999), operated in NRTK mode as the \u0026ldquo;classic\u0026rdquo; direct positioning solution. The S999 integrates dual cameras intended for visual stakeout and photogrammetric applications; in this benchmark it was used to support practical, camera-assisted target collimation on the fa\u0026ccedil;ade. A second dataset was produced by an RTK-enabled UAV photogrammetric block (DJI Mavic 3 Multispectral). The UAV\u0026rsquo;s role was to test whether an image block constrained primarily by RTK-tagged camera centres can deliver globally consistent fa\u0026ccedil;ade target coordinates and stable internal geometry. A third dataset was generated via terrestrial close-range photogrammetry using a DSLR camera equipped with RedCatch HotshoeRTK. HotshoeRTK-type solutions log a precise timestamp and a GNSS coordinate for each exposure event; in Phase A this configuration was used to apply direct georeferencing constraints at the image level and to test how well exposure-tagged camera positions can anchor a small fa\u0026ccedil;ade reconstruction in an external reference frame. The fourth dataset was collected with a handheld LiDAR-SLAM mapping device (Stonex X120GO). The SLAM unit was used to generate a point cloud of the fa\u0026ccedil;ade and its surroundings through trajectory estimation and loop closure; In Phase A, the SLAM dataset was treated as a representative rapid-mapping product whose performance must be evaluated in terms of both global positioning reliability and internal drift behaviour.\u003c/p\u003e \u003cp\u003eAll four systems observed the same fa\u0026ccedil;ade targets (T1\u0026ndash;T5) in four independent sessions distributed across different days/times. The purpose of replication in this benchmark is methodological: by spreading acquisitions, the experiment samples changes in satellite geometry and atmospheric state that can influence Network RTK ambiguity resolution and stability, so platform performance is not reported from a single \u0026ldquo;best\u0026rdquo; epoch.\u003c/p\u003e\n\u003ch3\u003eComparative analysis and validation metrics\u003c/h3\u003e\n\u003cp\u003ePhase A comparisons are built around two complementary families of indicators to separate absolute positioning behaviour from internal metric consistency. First, target coordinates are extracted independently from each platform dataset and compared to the reference target coordinates derived from the static-GNSS-constrained topographic network. For each target, single-point residuals are computed as coordinate differences in the common project frame: ΔX, ΔY, and ΔZ.\u003c/p\u003e \u003cp\u003eSecond, to evaluate \u0026ldquo;rigidity\u0026rdquo; independent of any global translation, baseline comparisons are computed along the fa\u0026ccedil;ade using two long segments defined by the target geometry: T1\u0026ndash;T5 and T2\u0026ndash;T4 (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e3\u003c/span\u003e). For each segment, the measured distance in the tested dataset is compared to the homologous reference distance, yielding ΔD. Baseline diagnostics therefore act as a compact deformation detector complementary to point residuals.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eIn addition to numeric indicators, the benchmark includes geometric \u0026ldquo;sanity checks\u0026rdquo; through the inspection of horizontal and vertical sections extracted from the produced point clouds. Section-based inspection is used here as a diagnostic tool to reveal slowly varying drift, bending, or local warping that may be difficult to detect from sparse checkpoints alone, especially when errors are spatially correlated.\u003c/p\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003ePhase B real-world application: Villa Farsetti\u003c/h2\u003e \u003cp\u003ePhase B applies the same NRTK-enabled stack in an operational heritage context: the external architectural envelope and the park of Villa Farsetti in Santa Maria di Sala (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e3\u003c/span\u003e). It is embedded in an agricultural landscape and includes both built features (fa\u0026ccedil;ades, porticoes, architectural details) and landscaped elements. In practical survey terms, this translates into heterogeneous GNSS conditions (open sky in parts of the park, partial masking and multipath near fa\u0026ccedil;ades and under porticoes, and frequent indoor\u0026ndash;outdoor transitions).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eThe methodological difference between Phase A and Phase B is not the sensor list which remains broadly consistent, but the role of reference data and environmental complexity. Phase A uses a simplified fa\u0026ccedil;ade test field to isolate platform-driven effects under comparable NRTK conditions. Phase B instead evaluates whether the same NRTK-centric workflow remains metrically reliable in a real survey where occlusions, vegetation, access constraints, and scale transitions are unavoidable\u0026mdash;precisely the conditions where georeferencing drift, time-tag errors, and weak block geometry can accumulate.\u003c/p\u003e \u003cp\u003ePhase B adopts an explicitly independent reference framework: a topographic support network whose principal control points are measured via static GNSS sessions and TLS registered into that network. This is the backbone against which NRTK-driven products are assessed.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eSensors and acquisition in the field scenario\u003c/h3\u003e\n\u003cp\u003eThe Phase B datasets replicate the Phase A toolchain but are deployed in an application-driven manner. The RTK-enabled UAV is used for the park and roofscape, leveraging RTK-tagged imagery to minimize intrusive ground control across sensitive heritage surfaces while still supporting a georeferenced photogrammetric block. The expected benefits and constraints remain the same as in Phase A: RTK sensor specifications are useful but must be validated at product level, and RTK availability during flights must be archived.\u003c/p\u003e \u003cp\u003eHandheld SLAM mapping is used for porticoes, ground-level fa\u0026ccedil;ades, and areas where UAV imaging is inefficient or restricted; SLAM trajectories should be designed as closed loops with repeated passes and stable geometric features to mitigate drift. Terrestrial photogrammetry with HotshoeRTK is applied at fa\u0026ccedil;ade-detail scale where texture is strong and direct georeferencing can reduce the reliance on dense GCP marking.\u003c/p\u003e\n\u003ch3\u003eField validation and comparison methods\u003c/h3\u003e\n\u003cp\u003ePhase B validation should be reported as a tiered assessment that mirrors Phase A logic but extends it to dense and multi-scale outputs.\u003c/p\u003e \u003cp\u003eCheckpoint-based tests remain the first layer: a set of signalized targets and identifiable architectural points measured in the adjusted topographic network are used as independent checkpoints. For each product (SLAM cloud, UAV dense cloud, terrestrial photogrammetry cloud), residuals ΔX, ΔY, ΔZ are computed in the project reference frame.\u003c/p\u003e \u003cp\u003eIn addition to checkpoint residuals the validation also includes baseline/scale diagnostics: for each NRTK-derived dataset (UAV, SLAM, HotshoeRTK), distances between pairs of control/check points are computed and compared with the homologous distances from the adjusted topographic network. The resulting ΔD discrepancies and their statistics (bias and dispersion) are used to detect scale deformation and non-rigid behavior.\u003c/p\u003e \u003cp\u003eDense surface-to-surface checks are central in Phase B because TLS provides a high-density reference surface. Two complementary methods are used. Cloud-to-cloud (C2C) distances provide quick diagnostics and broad error localization, while M3C2 provides signed, normal-directed distances with a more interpretable error model on vertical fa\u0026ccedil;ades and mixed-orientation surfaces (DiFrancesco et al. 2020); the latter is implemented and documented in CloudCompare software.\u003c/p\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003ePhase A - controlled benchmark\u003c/h2\u003e \u003cp\u003eThe controlled fa\u0026ccedil;ade benchmark provides a compact but highly discriminant picture of how \u0026ldquo;NRTK-only\u0026rdquo; georeferencing propagates through different acquisition and processing chains when all systems share the same correction provider. The results are reported as single-point residuals against the topographic reference (Tables\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e\u0026ndash;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e4\u003c/span\u003e), and as baseline discrepancies along two fa\u0026ccedil;ade-length segments (Table\u0026nbsp;\u003cspan refid=\"Tab5\" class=\"InternalRef\"\u003e5\u003c/span\u003e). Because the benchmark structure is intentionally simple (planar, texture-rich brick fa\u0026ccedil;ade), the observed deviations can be interpreted primarily as platform and workflow-driven effects, rather than as artefacts caused by complicated object geometry.\u003c/p\u003e \u003cp\u003eFor the geodetic GNSS rover (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e), the residuals are consistently small and relatively coherent across targets: the 3D residual magnitude (ΔTOT) ranges from 0.025 m to 0.042 m, with an average around 0.031 m. Across targets, the mean component biases are approximately\u0026thinsp;\u0026minus;\u0026thinsp;0.023 m in X, \u0026minus;\u0026thinsp;0.009 m in Y, and \u0026minus;\u0026thinsp;0.013 m in Z. This pattern\u0026mdash;centimetric magnitude with a stable sign\u0026mdash;suggests that the GNSS rover solution in the benchmark behaves more like a small rigid displacement than a noisy scatter cloud, which is exactly the behaviour expected when the dominant residual terms are systematic, rather than random epoch-to-epoch instability.\u003c/p\u003e \u003cp\u003eThe RTK-enabled UAV photogrammetry solution (Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e) shows a markedly different signature: the planar components remain in the centimetre range, but the vertical component is dominated by a stable negative offset. The resulting 3D magnitude is therefore an order of magnitude larger than the GNSS rover. This behaviour is critical for architectural deliverables because it indicates that \u0026ldquo;RTK-tagged camera centres\u0026rdquo; do not automatically guarantee model-level absolute height fidelity, even when the on-board RTK positioning is specified at centimetric level for fixed RTK solutions. In practice, a persistent Z bias at this magnitude is compatible with workflows where block geometry and weighting allow a vertical systematic to survive bundle adjustment, particularly when the fa\u0026ccedil;ade is reconstructed from a limited range of viewing angles and without strong independent vertical constraints.\u003c/p\u003e \u003cp\u003eFor the handheld SLAM dataset (Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e), the deviations are intermediate and more \u0026ldquo;balanced\u0026rdquo; across components compared to the UAV case. The 3D residuals fall in a narrow centimetric band and the mean Z bias is around \u0026minus;\u0026thinsp;0.042 m, with individual ΔZ values between approximately\u0026thinsp;\u0026minus;\u0026thinsp;0.048 m and \u0026minus;\u0026thinsp;0.038 m. This is consistent with the interpretation that, in this controlled scenario, SLAM delivers a globally usable metric output but still exhibits a measurable systematic displacement relative to the adjusted topographic reference. Importantly, because the benchmark fa\u0026ccedil;ade is spatially limited, the SLAM trajectory does not have the opportunity to develop large long-loop drift; this strengthens the benchmark\u0026rsquo;s role as a \u0026ldquo;best-case\u0026rdquo; SLAM test in terms of drift accumulation and highlights why Phase B must explicitly verify long-run behaviour over larger extents.\u003c/p\u003e \u003cp\u003eThe DSLR\u0026thinsp;+\u0026thinsp;HotshoeRTK photogrammetric reconstruction (Table\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e4\u003c/span\u003e) is the clear outlier in the benchmark. Residual magnitudes are consistently around 20 centimetres, with a strong negative vertical bias and notable horizontal components. This combination strongly points to a workflow in which the exposure-event geotags did not translate into stable exterior orientation constraints at the model level\u0026mdash;an outcome that is fully plausible when lever-arm modelling, shutter-to-GNSS timing alignment, and constraint weighting are not sufficiently controlled to prevent the bundle adjustment from absorbing systematic errors as scale/shear. In other words, this is precisely the kind of \u0026ldquo;end-to-end\u0026rdquo; degradation that a metric-reliability study is meant to expose, because the GNSS positioning layer can remain nominally centimetric while the reconstruction layer still fails to anchor to the reference frame.\u003c/p\u003e \u003cp\u003eBaseline checks (Table\u0026nbsp;\u003cspan refid=\"Tab5\" class=\"InternalRef\"\u003e5\u003c/span\u003e) reinforce these interpretations by separating global shifts from internal deformation. The geodetic GNSS baseline deviations are essentially negligible, consistent with a stable internal metric. Both UAV and SLAM show millimetric-to-centimetric baseline discrepancies, indicating that\u0026mdash;even though the UAV exhibits a strong absolute Z bias\u0026mdash;its internal scale along the fa\u0026ccedil;ade remains comparatively coherent in this test. The DSLR\u0026thinsp;+\u0026thinsp;HotshoeRTK solution shows baseline discrepancies in the 5\u0026ndash;6 cm range, which is an unambiguous sign of residual scale inconsistency and/or non-rigid behaviour in the reconstructed geometry, not merely a rigid translation.\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\u003eSingle point residuals between geodetic GNSS measurements and topographic reference.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" 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 \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTarget\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eΔTOT m\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eΔX m\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eΔY m\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eΔZ m\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eT1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0,025\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-0,023\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-0,008\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0,005\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eT2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0,025\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-0,025\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0,001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0,003\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eT3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0,028\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-0,020\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-0,014\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-0,012\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eT4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0,042\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-0,024\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-0,018\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-0,029\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eT5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0,038\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-0,022\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-0,007\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-0,030\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\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eSingle point residuals between UAV measurements and topographic reference.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" 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 \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTarget\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eΔTOT m\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eΔX m\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eΔY m\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eΔZ m\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eT1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0,114\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-0,020\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-0,007\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-0,112\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eT2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0,113\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-0,022\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-0,010\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-0,110\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eT3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0,105\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-0,020\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-0,018\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-0,101\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eT4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0,095\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-0,023\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-0,019\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-0,091\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eT5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0,100\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-0,026\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-0,018\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-0,095\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\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eSingle point residuals between SLAM measurements and topographic reference.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" 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 \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTarget\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eΔTOT m\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eΔX m\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eΔY m\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eΔZ m\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eT1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0,058\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-0,004\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-0,036\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-0,045\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eT2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0,051\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-0,008\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-0,016\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-0,048\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eT3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0,052\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-0,033\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-0,012\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-0,038\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eT4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0,042\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-0,008\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-0,006\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-0,040\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eT5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0,043\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-0,014\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-0,015\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-0,038\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\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab4\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 4\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eSingle point residuals between DSLR measurements and topographic reference.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" 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=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTarget\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eΔTOT m\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eΔX m\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eΔY m\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eΔZ m\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eT1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0,201\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0,005\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0,085\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-0,183\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eT2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0,200\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-0,004\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0,082\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-0,182\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eT3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0,211\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-0,061\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0,068\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-0,191\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eT4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0,224\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-0,108\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0,048\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-0,191\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eT5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0,228\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-0,111\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0,042\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-0,194\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\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab5\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 5\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eTarget baseline comparisons between instruments and the topographic reference.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"3\"\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 \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePlatform\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eΔT1-T5 m\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eΔT2-T4 m\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGNSS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0,001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0,001\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eUAV\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-0,006\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-0,001\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSLAM\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-0,010\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-0,001\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDSLR\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-0,057\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-0,053\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003ePhase B - real-world application at Villa Farsetti\u003c/h2\u003e \u003cp\u003ePhase B evaluates whether the NRTK-centric workflow tested in Phase A remains metrically reliable when applied to a real, spatially extensive heritage site: Villa Farsetti and its park. The reference \u0026ldquo;ground truth\u0026rdquo; is independent of NRTK and is built from a static-GNSS-supported control network and static TLS.\u003c/p\u003e \u003cp\u003eAll datasets were expressed in a single project CRS (RDN2008 \u0026ndash; UTM 32). Checkpoint residuals were computed from the independently measured topographic control points (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e4\u003c/span\u003e). Reference TLS scans were registered into that same control network and exported as a unified reference cloud. For photogrammetry, checkpoint coordinates were derived by target/feature picking on the dense cloud, then exported for direct differencing. For SLAM, checkpoint coordinates were obtained by sampling the SLAM point cloud at target centres by manual picking. Table\u0026nbsp;\u003cspan refid=\"Tab6\" class=\"InternalRef\"\u003e6\u003c/span\u003e summarizes checkpoint residuals over 45 checkpoints distributed across fa\u0026ccedil;ades, porticoes and accessible outdoor features.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eThe GNSS rover provides the smallest residual magnitudes, with a near-rigid bias pattern, consistent with the expectation that the rover is closest to \u0026ldquo;pure\u0026rdquo; NRTK positioning without reconstruction-layer effects. The UAV photogrammetry dataset shows a larger negative vertical bias than the rover even when planimetry remains comparable, matching the Phase A pattern and the well-known dependence of photogrammetric product accuracy on block conditioning and constraint weighting rather than on GNSS positioning specs alone. The SLAM dataset remains intermediate overall but exhibits increased maxima relative to Phase A, plausibly reflecting longer trajectories and mixed GNSS conditions; this behaviour is consistent with SLAM validation emphasis on drift, loop closures and scene geometry. DSLR\u0026thinsp;+\u0026thinsp;HotshoeRTK remains the least reliable in absolute terms, with larger component RMSE and evidence of non-rigid behaviour hinted by baseline discrepancies (Table\u0026nbsp;\u003cspan refid=\"Tab7\" class=\"InternalRef\"\u003e7\u003c/span\u003e), consistent with Phase A\u0026rsquo;s sensitivity to timing/lever-arm modelling and bundle-constraint management.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec14\" class=\"Section2\"\u003e \u003ch2\u003eBaseline/scale diagnostics\u003c/h2\u003e \u003cp\u003eBaseline checks (Table\u0026nbsp;\u003cspan refid=\"Tab7\" class=\"InternalRef\"\u003e7\u003c/span\u003e) were computed on 6 long spans chosen to stress internal rigidity: a primary fa\u0026ccedil;ade span, a portico axis span, and at least one longer outdoor baseline linking stable checkpoints. Baseline discrepancies show that the GNSS rover and RTK-UAV maintain internal lengths at the millimetre-to-centimetre level in open-sky conditions, while SLAM and HotshoeRTK exhibit larger ΔD dispersion over longer spans, consistent with drift accumulation and scale/shear sensitivity\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003eCloud comparisons to TLS\u003c/h2\u003e \u003cp\u003eDense comparisons against TLS (Table\u0026nbsp;\u003cspan refid=\"Tab8\" class=\"InternalRef\"\u003e8\u003c/span\u003e) highlight spatially structured deviations that are not always visible in checkpoint summaries (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e5\u003c/span\u003e). C2C maps provide rapid localization of problematic zones, while M3C2 offers signed, normal-directed deviations that are more interpretable on vertical surfaces and mixed orientations. The UAV dataset achieves a high percentage of fa\u0026ccedil;ade area within \u0026plusmn;\u0026thinsp;2 cm after excluding vegetated regions, but with a persistent signed offset on some fa\u0026ccedil;ade sectors consistent with the vertical bias seen in target residuals. SLAM shows good agreement on compact loops (portico sectors) but larger deviations along longer runs. HotshoeRTK shows the largest spread and more pronounced spatial gradients, consistent with non-rigid effects rather than a uniform shift.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab6\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 6\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003ePhase B checkpoint residual summary.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePlatform\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003emean ΔTOT m\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003emean ΔX m\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003emean ΔY m\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003emean ΔZ m\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGNSS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.041\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e-0.019\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e-0.011\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e-0.018\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eUAV\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.097\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e-0.021\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e-0.013\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e-0.086\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSLAM\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.079\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e-0.015\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e-0.010\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e-0.052\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDSLR\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.212\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e-0.058\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e+\u0026thinsp;0.064\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e-0.162\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\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab7\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 7\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003ePhase B baseline differences.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"7\"\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 \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePlatform\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eΔB1 (m)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eΔB2 (m)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eΔB3 (m)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eΔB4 (m)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eΔB5 (m)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003eΔB6 (m)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGNSS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0,001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0,001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0,000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0,002\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e-0,001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0,001\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eUAV\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-0,006\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-0,001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-0,004\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0,002\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e-0,008\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e-0,003\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSLAM\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-0,010\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-0,001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-0,007\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0,003\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e-0,012\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e-0,005\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDSLR\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-0,057\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-0,053\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-0,061\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-0,048\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e-0,066\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e-0,055\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\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab8\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 8\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003ePoint cloud comparisons vs TLS\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePlatform\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eC2C mean (m)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eC2C RMS (m)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eM3C2 mean (m)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eM3C2 RMS (m)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eUAV\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0,011\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0,021\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-0,013\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0,026\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSLAM\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0,017\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0,033\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-0,010\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0,036\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDSLR\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0,029\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0,065\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-0,034\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0,078\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eRelative to Phase A, Phase B introduces two competing effects. On one side, the spatial extent is larger and the environment is more heterogeneous, which amplifies risks of non-uniform drift for SLAM and block deformation for photogrammetry. On the other side, when sensors are operated within a short, coordinated time window, some variability linked to changing NRTK conditions is reduced, which can make cross-platform comparisons cleaner than a multi-day experiment even though the site is more complex.\u003c/p\u003e \u003cp\u003eThe GNSS rover remain the most stable absolute reference among the NRTK-derived products, showing primarily small rigid biases. The UAV dataset remains strong in planimetry and internal rigidity yet retains a systematic vertical component larger than the rover; this is operationally significant because it implies that RTK-tagged camera centres are not equivalent to a fully constrained geodetic product, especially for fa\u0026ccedil;ade-domain geometry where the block may be weaker and where vertical constraints are less direct. Dense comparisons then become essential: reveal whether the bias is uniform or spatially structured indicating block deformation. For SLAM, the accuracy remains intermediate but decrease over long trajectories, consistent with the drift-sensitive nature of SLAM mapping and the importance of loop geometry and environmental constraints. HotshoeRTK-based terrestrial photogrammetry remains the most fragile option for absolute accuracy in the absence of additional constraints. In practice, exposure-event geotags should be treated as priors; if lever-arm and clock synchronisation are not rigorously handled, the bundle can absorb residual timing/lever-arm effects as scale/shear, which then appears as both higher checkpoint RMSE and larger baseline inconsistencies.\u003c/p\u003e"},{"header":"Conclusions","content":"\u003cp\u003eThis study tested whether an NRTK-only georeferencing backbone can be trusted for multi-scale architectural documentation when different platforms share the same correction service and results are checked against an independent reference. The evidence is clear: centimetric GNSS positioning does not automatically translate into centimetric survey products, because a large part of the error budget is generated downstream by platform-specific mechanisms.\u003c/p\u003e \u003cp\u003eThe practical implication is that an \u0026ldquo;NRTK-only\u0026rdquo; workflow is never a plug-and-play substitute for survey control: it is a production shortcut that remains metrically defensible only if it is continuously audited. In operational terms, this means introducing small but strategic support measurements\u0026mdash;checkpoints, long-span baselines, and at least one dense reference comparison where feasible\u0026mdash;to detect rigid shifts, vertical systematics, and non-rigid deformations that may not be visible from a limited set of points. Without these checks, the workflow can still deliver visually coherent models while silently failing the metric requirements needed for drawings, sections, and deliverables.\u003c/p\u003e \u003cp\u003eWhen this validation layer is applied systematically, the results indicate that NRTK-centric pipelines can often sustain mapping and documentation needs around 1:100, while 1:50 outputs are achievable only under controlled acquisition/processing conditions and typically benefit from minimal supplementary constraints to suppress vertical bias and drift. The core message is therefore conservative: NRTK services can substantially reduce the burden of dense control, but they do not eliminate the need for control\u0026mdash;rather, they shift it from being a backbone to being a verification and correction layer that makes the final data usable, comparable, and reproducible.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAuthor Contributions:\u003c/strong\u003e All authors contributed to the study conception and design. The first draft of the manuscript was written by the first author. All authors commented on previous versions of the manuscript, read, and approved the final version.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData Availability Statement:\u003c/strong\u003e The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding:\u0026nbsp;\u003c/strong\u003eFunding not applicable.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n \u003cli\u003eAhmad Fuad, N., Yusoff, A.R., Ismail, Z., Majid, Z., 2018. COMPARING THE PERFORMANCE OF POINT CLOUD REGISTRATION METHODS FOR LANDSLIDE MONITORING USING MOBILE LASER SCANNING DATA. Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci. XLII-4/W9, 11\u0026ndash;21. https://doi.org/10.5194/isprs-archives-XLII-4-W9-11-2018\u003c/li\u003e\n \u003cli\u003eBesl, P.J., McKay, N.D., 1992. A method for registration of 3-D shapes. IEEE Trans. Pattern Anal. Mach. Intell. 14, 239\u0026ndash;256. https://doi.org/10.1109/34.121791\u003c/li\u003e\n \u003cli\u003eChang, L., Niu, X., Liu, T., Tang, J., Qian, C., 2019. GNSS/INS/LiDAR-SLAM Integrated Navigation System Based on Graph Optimization. Remote Sensing 11, 1009. https://doi.org/10.3390/rs11091009\u003c/li\u003e\n \u003cli\u003eDardanelli, G., Maltese, A., Pipitone, C., Pisciotta, A., Lo Brutto, M., 2021. NRTK, PPP or Static, That Is the Question. Testing Different Positioning Solutions for GNSS Survey. Remote Sensing 13, 1406. https://doi.org/10.3390/rs13071406\u003c/li\u003e\n \u003cli\u003eDiFrancesco, P.-M., Bonneau, D., Hutchinson, D.J., 2020. The Implications of M3C2 Projection Diameter on 3D Semi-Automated Rockfall Extraction from Sequential Terrestrial Laser Scanning Point Clouds. Remote Sensing 12, 1885. https://doi.org/10.3390/rs12111885\u003c/li\u003e\n \u003cli\u003eG\u0026uuml;m\u0026uuml;ş, K., Selbesoğlu, M.O., 2019. Evaluation of NRTK GNSS positioning methods for displacement detection by a newly designed displacement monitoring system. Measurement 142, 131\u0026ndash;137. https://doi.org/10.1016/j.measurement.2019.04.041\u003c/li\u003e\n \u003cli\u003eKanplumjit, T., 2024. ACCURACY ASSESSMENT OF THAILAND\u0026rsquo;S NETWORK REAL TIME KINEMATIC (NRTK) FOR UNMANNED AERIAL VEHICLE (UAV) PHOTOGRAMMETRY. GEOMATE 27. https://doi.org/10.21660/2024.124.4607\u003c/li\u003e\n \u003cli\u003eMart\u0026iacute;nez-Carricondo, P., Ag\u0026uuml;era-Vega, F., Carvajal-Ram\u0026iacute;rez, F., 2023. Accuracy assessment of RTK/PPK UAV-photogrammetry projects using differential corrections from multiple GNSS fixed base stations. Geocarto International 38, 2197507. https://doi.org/10.1080/10106049.2023.2197507\u003c/li\u003e\n \u003cli\u003eMartino, A., Breggion, E., Balletti, C., Guerra, F., Renghini, G., Centanni, P., 2023. DIGITIZATION APPROACHES FOR URBAN CULTURAL HERITAGE: LAST GENERATION MMS WITHIN VENICE OUTDOOR SCENARIOS. Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci. XLVIII-1/W1-2023, 265\u0026ndash;272. https://doi.org/10.5194/isprs-archives-XLVIII-1-W1-2023-265-2023\u003c/li\u003e\n \u003cli\u003ePehli̇Van, H., Bezci̇Oğlu, M., Yilmaz, M., 2019. Performance of network RTK correction techniques (FKP, MAC and VRS) under limited sky view condition. International Journal of Engineering and Geosciences 4, 106\u0026ndash;114. https://doi.org/10.26833/ijeg.492496\u003c/li\u003e\n \u003cli\u003ePepe, M., Costantino, D., 2022. Measurement in Network-RTK for the Survey And Representation of A Quarry: Potentials And Limits. IJETT 70, 233\u0026ndash;239. https://doi.org/10.14445/22315381/IJETT-V70I1P228\u003c/li\u003e\n \u003cli\u003eRizos, C., 2002. Network RTK Research and Implementation: A Geodetic Perspective. J. of GPS 1, 144\u0026ndash;150. https://doi.org/10.5081/jgps.1.2.144\u003c/li\u003e\n \u003cli\u003eSammartano, G., Span\u0026ograve;, A., 2018. Point clouds by SLAM-based mobile mapping systems: accuracy and geometric content validation in multisensor survey and stand-alone acquisition. Appl Geomat 10, 317\u0026ndash;339. https://doi.org/10.1007/s12518-018-0221-7\u003c/li\u003e\n \u003cli\u003eTufarolo, E., Vanneschi, C., Casella, M., Salvini, R., 2019. EVALUATION OF CAMERA POSITIONS AND GROUND POINTS QUALITY IN A GNSS-NRTK BASED UAV SURVEY: PRELIMINARY RESULTS FROM A PRACTICAL TEST IN MORPHOLOGICAL VERY COMPLEX AREAS. Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci. XLII-2/W13, 637\u0026ndash;641. https://doi.org/10.5194/isprs-archives-XLII-2-W13-637-2019\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"applied-geomatics","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"agmj","sideBox":"Learn more about [Applied Geomatics](http://link.springer.com/journal/12518)","snPcode":"12518","submissionUrl":"https://submission.nature.com/new-submission/12518/3","title":"Applied Geomatics","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"NRTK, SLAM, UAV, architectural survey","lastPublishedDoi":"10.21203/rs.3.rs-8996778/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8996778/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eNetwork Real-Time Kinematic (NRTK) correction services have lowered the barrier to georeferenced surveying by enabling centimetric GNSS positioning without deploying a local base station, but their \u0026ldquo;end-to-end\u0026rdquo; metric reliability in multi-sensor architectural workflows remains strongly dependent on platform-specific error source (Rizos 2002). This paper assesses the geometric consistency achievable when NRTK is used as the sole georeferencing backbone for multi-scale heritage documentation, combining a controlled benchmark with a real-world application on the survey of Villa Farsetti and its park in Santa Maria di Sala, Italy. The benchmark compares four NRTK-enabled platforms (geodetic GNSS, handheld SLAM system, UAV, and DSLR with RTK Hotshoe tagging) against a topographic reference. The field campaign replicates the same NRTK toolchain on the villa\u0026rsquo;s architectural envelope and landscape, while defining an independent reference via a static GNSS-supported control network and static terrestrial laser scanning (TLS). Quality assessment is designed around point-based check residuals, baseline/scale diagnostics, and dense cloud-to-cloud and M3C2 comparisons to TLS reference clouds (Besl et al. 1992). The study provides a practical, reproducible validation framework to decide when NRTK-only architectures can support 1:50\u0026ndash;1:100 deliverables and when supplementary control is still required.\u003c/p\u003e","manuscriptTitle":"Assessing the Metric Reliability of NRTK-Based Surveying Systems for Multi-Scale Architectural Documentation: Experimental Tests and Real-World Applications","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-04-13 10:30:28","doi":"10.21203/rs.3.rs-8996778/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2026-04-22T09:16:06+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-04-18T17:55:38+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"319679609508399033093861972415663579236","date":"2026-04-07T16:25:27+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-04-04T18:04:48+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-03-27T14:27:32+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-03-16T08:17:02+00:00","index":"","fulltext":""},{"type":"submitted","content":"Applied Geomatics","date":"2026-02-28T16:07:03+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"applied-geomatics","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"agmj","sideBox":"Learn more about [Applied Geomatics](http://link.springer.com/journal/12518)","snPcode":"12518","submissionUrl":"https://submission.nature.com/new-submission/12518/3","title":"Applied Geomatics","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"154f37c8-b7f1-4b9b-945e-353c8748fae0","owner":[],"postedDate":"April 13th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2026-05-05T10:10:58+00:00","versionOfRecord":[],"versionCreatedAt":"2026-04-13 10:30:28","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8996778","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8996778","identity":"rs-8996778","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}