Increasing hydro-climatic variability is challenging the stability of meteorological baselines in tropical rice agro-ecosystems. This study analyzed 3,652 daily ERA5-Land observations from 1 January 2016 to 30 December 2025 for Malaysia’s Barat Laut Selangor (BLS) rice granary using structural-break analysis, Granger predictive association analysis, and machine-learning models, including Random Forest (RF) and Multiple Linear Regression (MLR), to compare the predictability of Layer 1 soil moisture and ERA5-Land low-vegetation Leaf Area Index (LAI) using hydro-climatic predictors. Structural-break analysis indicated a temporal discontinuity in monthly precipitation, although sensitivity analyses showed that its precise timing was method-dependent and uncertain. Three-month accumulated precipitation–potential evaporation balances showed recurrent periods during which ERA5-Land potential evaporation exceeded precipitation, indicating periods of increased atmospheric water demand. Granger predictive analysis identified a significant lagged predictive association between Layer 1 soil moisture and LAI, whereas immediate meteorological variables showed weaker predictive associations with LAI. Under chronological temporal validation, RF showed strong predictive performance for Layer 1 soil moisture (R² = 0.9311), whereas prediction of ERA5-Land low-vegetation LAI remained limited (MLR: R² = 0.2910; RF: R² = 0.2460). Across 49 expanding-window rolling-origin folds, RF maintained comparatively strong soil-moisture performance (R² = 0.7374 ± 0.1688), while LAI prediction remained weak (MLR: R² = 0.2185 ± 0.1182; RF: R² = 0.1983 ± 0.1914). The contrasting predictability of the two targets suggests that the selected hydro-climatic predictors represented the physical soil-water state more effectively than the model-derived vegetation indicator. The absence of improved LAI performance with RF further indicates that greater model complexity alone did not resolve this predictive limitation. Potential influences of crop phenology, irrigation, management practices, spatial-scale differences, and reanalysis-product uncertainty remain plausible but untested because direct agronomic observations were unavailable. Future modeling should integrate longer observational records, independent vegetation measurements, crop calendars, irrigation information, and field-level agronomic data.
Satellite images have recently been made accessible using web services in a web-based Geographic Information System (WebGIS). Web Map Service (WMS) and Web Processing Service (WPS) are the main services used by the information system for viewing and processing remotely sensed data online, respectively. While WMS is widely used to render satellite images for visual interpretation, few image processing functions are implemented using WPS. Formulating the service to classify satellite images available via WMS provides a very accessible, efficient and cost-effective satellite image processing resource for many applications. Artificial Neural Network (ANN) computing is one of the machine learning methods used for satellite image classification. Unlike the conventional per-pixel parametric statistics classifier, an ANN is a distribution-assumption-independent algorithm capable of processing complex nonlinear relationships, which is useful for identifying spectral and spatial patterns of land use types in satellite images. This study focuses on the formulation of WPS to classify satellite images made available as a color composite WMS using ANN. The results of the study show that an accurate satellite image classifier using an ANN can be implemented with WPS, using WMS URL as input. It is further demonstrated that the WPS-based ANN classifier could be easily accessed using WebGIS for easy implementation. This study provides a method to increase the usability of satellite image color composite WMSs, from data exclusively used for visual interpretation, to WPS inputs for a wide array of web-based image-processing applications.
To solve the semantic segmentation in urban street scenes, such as modal information fragmentation, low accuracy of small target segmentation, and insufficient robustness in complex environments, this study proposes a semantic segmentation method that integrates modal memory sharing and computer vision. First, a modal memory sharing module based on color image modality and two-dimensional attention mechanism is designed to efficiently transmit and reuse cross-modal information. Secondly, a computer vision semantic segmentation architecture combined with hierarchical networks is constructed to adapt to the hierarchical feature distribution of urban street scenes and strengthen the extraction and fusion of fine-grained semantic features. The results showed that the proposed method performed well in average intersection over union ratio, pixel accuracy, and F1 on urban scene targets. On the Cityscapes data set, the pixel accuracy of the tree category reached 0.863, and the average intersection over union ratio of the vehicle category was 0.712. On the CamVid data set, the pixel accuracy of the tree category reached 0.836, and the average intersection over union ratio of the vehicle category was 0.653. The proposed method achieves high-precision segmentation of both large-scale features and small-scale targets. The proposed method effectively solves low multi-modal information utilization and insufficient segmentation accuracy in urban street scenes, and provides reliable technical support for downstream applications such as intelligent transportation and urban governance.
Advances in sensor technology have enabled the development of spectral-based remote sensing approaches to detect plastic litter in the environment. This study modelled mixed-pixel conditions for freshwater environments and proposes a new spectral index for detecting plastic litter in these complex environments. Based on the freshwater zone framework, synthetic spectra for mixed-pixel environmental conditions were generated by linearly mixing plastic spectra with those of selected background surfaces. Using the spectral absorption feature of plastics at 1729 nm, several candidate indices were formulated. Index E was selected as the optimal index for plastic detection across all freshwater zones, as it demonstrated the highest separability across all freshwater zones. However, the separability of this optimal index remains constrained to the non-plastic materials assessed in this study. Therefore, extending the separability assessment to a wide range of material substrates in natural freshwater environments is necessary to determine its robustness and improve its transferability to other heterogeneous environments.
Artificial intelligence has become increasingly important for Earth observation and remote sensing image analysis; however, deploying deep learning models under limited labeled data and computational resources remains a significant challenge. This study proposes a lightweight transfer learning framework with dynamic data augmentation for few-shot remote sensing image classification. Built upon the EfficientNet-Lite architecture, the proposed framework integrates dynamic data augmentation, synthetic data generation using text-to-image generative models, transfer learning, and ensemble prediction to improve classification performance while preserving computational efficiency. The framework was evaluated on four benchmark remote sensing datasets (OPS-SAT, UC Merced Land Use, WHU-RS19, and RSSCN7). Experimental results demonstrate a 5.33% absolute increase in classification accuracy and a 43% reduction in the loss score compared with a recent EfficientNet-Lite-based baseline. The proposed framework achieves an effective balance between predictive performance, computational efficiency, and model compactness, making it suitable for remote sensing applications operating under limited computational resources and scarce labeled data. The proposed framework is particularly applicable to Earth observation tasks such as environmental monitoring, land-use and land-cover mapping, disaster response, agricultural observation, and onboard satellite image analysis, where both computational efficiency and reliable classification performance are essential. These findings demonstrate that combining lightweight architectures with transfer learning and advanced data augmentation strategies provides an effective and practical framework for few-shot remote sensing image classification and facilitates the deployment of deep learning models in real-world Earth observation applications.
Real-time monitoring of plain-area hydraulic sluices requires reliable detection and tracking of moving targets including unauthorized personnel, abnormally approaching vessels, and floating debris within the forebay. Conventional visual algorithms suffer from high false alarm rates and tracking failures caused by water ripples, drastic illumination shifts, and target occlusion by sluice piers. To address these issues, this study develops a high-precision intelligent monitoring framework integrating an Improved Gaussian Mixture Model and an anti-occlusion Improved Kernelized Correlation Filter. First, an adaptive update coefficient and optimized matching threshold are embedded into Gaussian Mixture Model to suppress periodic water surface noise unique to sluice scenes. Second, Histogram of Oriented Gradients and grayscale features are fused, while an Average Peak Correlation Energy-based dual-indicator confidence evaluation, together with Kalman motion prediction, is introduced to construct a robust Improved Kernelized Correlation Filter tracker against occlusion-induced drift. Finally, the detection and tracking modules constitute a mutually feedback closed-loop pipeline, combined with electronic fence limits and dwell-time thresholds to realize automatic intrusion identification and early warning. Experimental results show that the proposed method achieves an F1-score of 94.12%, a tracking success rate of 84.32%, a minimum center location error of 10.04 pixels, a detection precision of 94.47%, and a tracking precision of 90.86%. The algorithm runs stably at 112 FPS on ordinary CPUs without GPU acceleration, offering a lightweight and low-cost technical solution for real-time unattended safety monitoring of plain-area hydraulic sluices.
Objective The spatial distribution and dynamic changes of water bodies, vegetation, beaches, and buildings in coastal tourist attractions directly affect regional tourism planning, ecological protection, and disaster early warning. However, in remote sensing images of these areas, the spectral overlap of ground features is severe and the boundaries are blurred. Coupled with the continuous interference of tides and fog, existing methods struggle to efficiently and accurately acquire the spatial information of these key elements. This study aims to construct an information extraction model that can effectively address these problems. Methods Interval type II fuzzy C-means clustering is used as the core segmentation algorithm. Particle swarm optimization and the Scherpennis index are used to adaptively adjust the clustering parameters. Gray-level co-occurrence matrix is introduced to extract texture features, and Fourier descriptors are used to extract shape features. Finally, the two types of features are fused and input into a classifier to complete ground feature identification. Results The accuracy of the core segmentation algorithm reaches 95.34%. Under a noise level of 10, the model achieved an average cross-union ratio of 98.10% for the water body, a beach boundary positioning error of 2.24 pixels, and shape similarity indices of 0.952, 0.904, 0.868, and 0.936 for the water body, vegetation, beach, and buildings, respectively. Conclusion The model proposed in this study outperforms existing comparative models in terms of segmentation accuracy, noise resistance, and geometric feature preservation, providing effective technical support for the dynamic monitoring and intelligent management of coastal tourism resources.
Although several database forensic models and frameworks have been developed for different applications and data types over the years, few studies have attempted to synthesise existing literature thereby limiting our understanding of the extent to which current frameworks provide spatially explicit frameworks for geospatial data. To address this scientific gap, the study performed a systematic review of existing work on auditing geospatial data manipulations within databases as a first step towards development of spatially explicit forensic audit model and framework for geospatial data. Results indicate that existing research has been predominantly generic with a focus on relational databases redundant phases. Additionally, the forensic techniques are mostly biased towards generating evidence admissible for litigation purposes. This consequently left other capabilities such as operational analysis and spatial trend evaluation unexplored. Thus, this study developed an integrated model and multifaceted framework for geospatial data that extends beyond litigation-oriented outcomes. Specifically, the model provides capabilities for spatio-temporal trend analysis and management of spatial data acquisition related projects. These capabilities are not only critical in expanding our understanding of principles and techniques but best practices related to the management of spatial manipulations within databases as well. The results from these study have the potential to simplify troubleshooting, ensure accountability, attribution and enhance integrity and security.
Coral bleaching is a significant global marine environmental issue closely linked to oceanographic and climatic variability. Bleaching events can significantly affect coral reef ecosystems and coastal communities that depend on marine resources. To support local-scale coral bleaching assessment in the Sunda Strait, this study integrated sea surface temperature (SST), sea surface salinity (SSS), ocean acidification (OA), and chlorophyll a (Chl-a) concentration as environmental indicators associated with coral bleaching conditions. Multitemporal Landsat 8/9 Operational Land Imager and Thermal Infrared Sensor (OLI-TIRS) imagery acquired from 2023 to 2024 was used to derive SST, SSS, OA, and Chl-a distributions. The derived environmental parameters were evaluated together with coral bleaching status based on Degree Heating Week (DHW) data provided by the National Oceanic and Atmospheric Administration (NOAA). The results demonstrate that local-scale coral bleaching conditions in the Sunda Strait exhibit substantial spatio-temporal variability compared with broader global-scale observations. Integrated analysis of SST, SSS, OA, and Chl-a revealed varying levels of environmental stress and coral vulnerability across the investigated reef areas. Although species-specific bleaching observations were not available in this study, the results provide important information for understanding local coral bleaching dynamics and supporting future reef monitoring and mitigation strategies in the Sunda Strait.
Small uncrewed aerial systems (sUAS, or drones) have become standard tools across many scientific fields by enabling affordable, on-demand aerial surveys. A single sUAS survey can generate hundreds to thousands of overlapping images that, once processed, produce large raster products (orthomosaics and digital surface models) that are often many gigabytes in size, making adherence to the FAIR data principles difficult. Repositories such as Zenodo and PANGAEA satisfy the findability requirement through appropriate metadata and persistent identifiers, but they do not provide visualization or spatial browsing of geospatial raster imagery. Assessing whether an archived dataset is fit for reuse therefore requires downloading large files and importing them into specialized GIS software before any visual inspection is possible. The friction introduced by these additional steps may limit reuse. Limited reuse is especially wasteful in remote regions such as Greenland, where sUAS surveys are expensive and logistically demanding to conduct and where features of interest from multiple disciplines are concentrated within the same small coastal footprints, giving each survey high cross-disciplinary reuse potential. To illustrate the value of visualization and spatial browsing for published sUAS datasets, we present the Greenland Drone Explorer (GDE), a WebGIS that aggregates dispersed, already-published Greenland sUAS datasets and enables 2D and 2.5D preview of orthomosaics without download or GIS software. As a vendor-neutral alternative to the proprietary platform on which GDE is built, we pair it with an open-source Jupyter notebook that reproduces the same capability by retrieving data from Zenodo, generating cloud-optimized GeoTIFFs, and displaying them in a browser-based map. Thus, GDE augments the baseline findability provided by repositories with spatial, visual discovery that may lower the barrier to reuse of dispersed sUAS raster imagery.
Amidst the global restructuring of urban commerce, China’s catering industry has transitioned from a phase of exogenous shock-response to a phase of structural "involution" (Neijuan). While existing research often relies on static snapshots or focuses on megacities, this study proposes the Morphological Density-Percentile Decoupling (MDPD) framework to track longitudinal survival trajectories across urban hierarchical tiers from 2017 to 2024. Using Hunan Province as a representative regional laboratory, we implement a rigorous multi-stage data audit and sensitivity analysis to mitigate the uncertainties of digital traces. Grounded in Adaptive Cycle Theory, our analysis reveals a fundamental morphological shift: while 2020 mobility shocks are consistent with a "flight to quality" toward primary cores, the 2024 landscape may reflect a Structural Middle-Market Expansion toward Tier-II buffer clusters. We identify a "Resilience Paradox" where hyper-saturation in high-density zones may be associated with increased vulnerability as described by "Agglomeration Diseconomies." By contrasting regional hubs with Tier-1 megacities, this research provides a scale-agnostic resilience framework that captures how commercial systems reorganize in response to observed changes in spatial distribution and survival patterns, which may suggest broader socioeconomic mechanisms such as consumption downgrading and centrifugal spatial diffusion. Ultimately, while these findings provide a robust provincial-scale benchmark for inland China, they serve as a comparative baseline to inform future studies across diverse regional economic contexts.
Delineating contiguous phenological regions (phenoregions) from time-series data requires methods that balance spatial contiguity, temporal homogeneity, and computational scalability, yet guidance on method selection under realistic spatial conditions remains limited. We systematically evaluate six spatially explicit regionalization methods—Automatic Regionalization with Initial Seed Location (Arisel), Regionalization with Dynamically Constrained Agglomerative Clustering and Partitioning (Redcap), Spatial “K”luster Analysis by Tree Edge Removal (Skater), Extended Simple Linear Iterative Clustering (ESLIC), Spatial K-Means (SKM), and Spatially Constrained Spectral Clustering (SCSC)—using (i) a controlled synthetic landscape and (ii) real-world remote-sensing datasets comprising NDVI and EVI time series from different sensors in Mt. Kenya National Park and the Argentine Chaco. Methods are assessed with respect to cluster homogeneity, spatial contiguity, agreement with reference solutions, and runtime, enabling analysis of performance trade-offs across spatial scales. Results show that Arisel, Redcap, Skater, and ESLIC consistently emerged as the strongest methods, although their relative performance varied by landscape context and by whether runtime was considered. Arisel performed best on quality-focused evaluations in the synthetic and Mt. Kenya cases, but its high computational cost reduced its suitability as dataset size increased. Redcap and Skater were robust high-performing alternatives, while ESLIC provided the best overall balance between delineation quality and efficiency as computational cost is given greater weight. Beyond identifying high-performing methods, the framework offers an objective, transparent, and reusable basis for method selection, supporting informed algorithm choice in phenological analysis and in a broader class of geospatial and remote-sensing regionalization applications where contiguity, homogeneity, and scalability must be jointly considered.
Accurate wildfire spread prediction is essential for effective mitigation in Thailand, where fires are frequent, small, and often arise from separate ignition sources. Existing research, largely trained on datasets from other regions, does not capture Thailand's distinct fire behavior, and most approaches rely solely on spatial models. We address this limitation by constructing a Thailand-specific wildfire dataset collected from 2024 to 2025, covering major forest regions across the country, and by developing a preprocessing strategy that removes newly ignited fire clusters using a density-based spatial clustering of applications with noise (DBSCAN) algorithm, ensuring that the model learns true propagation patterns. We apply the spatio-temporal model dubbed Simpler Yet Better Video Prediction (SimVP) for 1–3 day forecasting, enhanced with an Atrous Spatial Pyramid Pooling (ASPP) module to improve multi-scale spatial feature extraction and a cyclical temporal encoding (CTE) to incorporate seasonal cues. Experiments on our dataset show that our model achieves an IoU of 0.685 and an average F1-score of 0.718, outperforming all spatial and spatio-temporal baselines, with performance improvements most evident in 1–3-day-ahead forecasts. These results demonstrate the effectiveness of tailored preprocessing and multi-scale spatio-temporal modeling for wildfire forecasting in Thailand.
The disparity in the distribution of urban green space has grown increasingly noticeable as urbanization has accelerated. Its unevenness causes significant disparities in how inhabitants in various locations enjoy green space resources, which has a major impact on population health and well-being despite being a crucial component in enhancing residents' quality of life. Accurately extracting and analyzing green space data under various conditions is a pressing issue in the research of the equitable allocation of urban green space. An improved fully convolutional network is utilized to fuse multiple sources of remote sensing street view data. By extracting urban green space information, spatial analysis methods are applied to assess its fairness. The experimental results indicated that the green space coverage rates varied greatly among different regions. The green space coverage rates of regions A, B, C, and D were 35%, 12%, 42%, and 8%, respectively. Green space accessibility was better in the city center area, where the average distance for residents to reach the nearest green space was 500m. In the urban fringe areas, the average distance reached 1500 meters. In terms of specific spatial distribution, the area of green space per square kilometer in the city center area was 1.2 hectares, while it was only 0.4 hectares in the city fringe area. In addition, comparing the green space information extracted by the improved full convolutional network model with the actual survey data, the model's accuracy reached 95.6% and the recall reached 93.2%. The model demonstrated high extraction precision. The study's findings can offer a scientific foundation for urban green space planning, which can support the city's sustainable growth and raise citizens' standards of living.
Drone mapping images exhibit significant scale changes, and traditional semantic segmentation methods cannot balance feature fusion and spatial detail preservation, resulting in a decrease in segmentation accuracy. Therefore, an image semantic segmentation model based on multiscale cross-layer interaction and coordinated symmetric cascade is proposed for Unmanned Aerial Vehicle applications. The model uses multi-scale attention to concatenate convolutional networks and Transformers for cross layer interaction and coordination information, and then uses boundary loss functions for semantic segmentation of pixels. The experimental results show that the model achieved a segmentation accuracy of 98.23%, an average pixel accuracy of 97.48%, a boundary F1 score of over 93%, and a small object recall rate of over 88%. These results demonstrate that the proposed model performs well in complex boundary scenes, can handle complex and subtle boundary scenes, has excellent image semantic segmentation performance, ensures efficient image segmentation while achieving effective fine-grained image interpretation, and provides strong technical support for the development of unmanned aerial vehicles.
To analyze the driving factors and mechanisms of water quality changes in the Huangshui River Basin (HRB), this study uses random forest to construct an analysis model of the driving factors of water quality changes in the HRB. This study collected various water quality indicators from 32 springs in the HRB, and used a random forest model to analyze the impact of three major driving factors: climate and hydrology, topography, and human activities on water quality changes in the HRB. The results demonstrated that the R2 of the model reached 0.943 and the RMSE was only 0.589, which could accurately simulate the complex nonlinear relationship between various pollutants and driving factors in the water body. The study found that human activities were the core driving source of water quality deterioration in the HRB. Among them, the amount of chemical fertilizer use and the amount of hydropower development contributed the most to water quality changes, reaching 15.8% and 15.3%. The use of chemical fertilizers caused eutrophication and excessive fluoride in spring areas through surface runoff. Hydropower development reduced the flow rate through water reduction sections, prolonged the water exchange cycle, led to the accumulation of heavy metals and radionuclides, and had an impact on water quality. The proposed random forest model can analyze the driving factors and mechanisms of water quality changes in the HRB, thereby helping relevant managers provide appropriate prevention and control methods.
LiDAR systems estimate distances based on the transmission and reception times of laser beams, generating object representations in the form of point clouds. This research aims to develop a cost-efficient 3D mapping system by integrating LiDAR, GNSS, and camera sensors within a Mobile Mapping Vehicle (MMV) platform, where the sensors are mounted on a moving vehicle to enable continuous data acquisition. The main novelty of this research lies in the development of a low-cost multisensor fusion framework that integrates high-accuracy LiDAR-based geometric mapping with RGB-based 3D camera data. This approach not only produces precise object shape reconstructions but also enables automatic object type identification using the You Only Look Once (YOLO) artificial intelligence method. This approach overcomes the limitations of conventional LiDAR data, which lack adequate visual information to distinguish between object categories. The experimental results demonstrated that the LiDAR measurements yielded a maximum difference of 0.332 m and a minimum difference of 4.9 × 10−5m compared with the ground truth, indicating good geometric accuracy consistent with Level of Detail (LOD) 2 quality. However, LiDAR data alone remain limited in identifying object types owing to the absence of RGB information. To address this limitation, 3D camera model data were integrated, providing enhanced visualization and enabling the effective classification of residential and commercial objects.
This paper examines strategies that can be used to determine the appropriate binarization of predictive land-cover maps to produce categorical land-cover maps, in this study used to separate peatland from non-peatland. Seven different strategies were applied to two predictive peatland maps, and the accuracy of the resultant binary land cover maps was evaluated. The main objective was to find the most effective approach to include as much peatland as possible, while simultaneously keeping the amount of noise (false positives) in the peatland map at a minimum.The best overall results were obtained with metrics related to correlation in the confusion matrix. Cohen’s Kappa and the F1 score (defined as the harmonic mean of precision and recall) both reached their maximum at the same cutoff value, producing a land cover map with relatively high recall and limited amounts of noise in terms of false positive results.Maximizing the F1 score does not necessarily produce the optimal result for all applications. The intended purpose of the map must also be considered when deciding whether it is more important to increase true positive results or minimize false positives. In this study, selecting the cutoff point by maximizing Cohen’s Kappa or the F1 score proved to be the most effective overall strategy for dichotomizing the maps. Other strategies may be more appropriate when the distribution of the predictive scores is more balanced or there is a partisan preference for enhancing either user’s accuracy or producer’s accuracy.
Remote sensing-based estimation of sugar cane yield is constrained by spectral saturation under high-biomass canopies and by the fact that satellite platforms do not directly capture fine-scale edaphic variability or farmer management decisions. Consequently, most existing sugar cane yield models rely primarily on optical indices and neglect the combined effects of radar backscatter, soil and climate gradients, and agronomic management at field scale. This study addresses this gap by evaluating the incremental effect of different information sources on sugar cane yield estimation (t/ha) through progressive incorporation of optical sensor variables (vegetation indices), synthetic aperture radar (SAR), edaphic and meteorological variables, and agronomic management records at the field scale in the Cauca River Valley, Colombia, one of the most productive areas of sugar cane worldwide. Multiple linear regression (ordinary least squares, OLS) and linear mixed-effects models (LMM) were implemented to evaluate how each group of variables improves performance metrics when introduced sequentially. Models based exclusively on SAR information exhibited limited performance (R² < 0.40), whereas combining SAR with optical indices increased predictive capacity (R² < 0.55). Incorporating agronomic management variables further enhanced model accuracy, with the best LMM configuration reaching R² = 0.72 and RMSE = 13.9 t/ha. These results demonstrate that agronomic management decisions increased the explained variance of the model by approximately 19% and reduced prediction error by 4.23 t/ha relative to models without management information. The findings highlight the importance of explicitly representing management in operational and transferable yield estimation models and provide guidance for extending similar multisource approaches to other irrigated sugar cane growing regions.
Orthophoto maps are essential for spatial analysis in interpreting land management and topographic studies. However, there have been challenges achieving accurate spatial accuracy for orthophoto generation due to constraints of having the proper image acquisition device. This study, therefore, investigates orthophoto accuracy in part of Ibaraki Prefecture using old aerial photographs with different image acquisition devices—a surveying scanner, low- and high-resolution digital cameras, and low- and high-resolution general-purpose scanners, with and without geometric correction. The validation accuracy results show that the RMSEXY errors achieved by the surveying scanner, high-resolution digital camera, and low- and high-resolution general-purpose scanners with geometric correction were 1.887m, 1.336m, 2.024m, and 1.649m, respectively. However, orthophotos generated from general-purpose scanners without geometric correction showed higher RMSEXY errors, ranging from 15.589m to 17.032m. The findings indicate that high-resolution general-purpose scanners with geometric correction achieved the comparable positional accuracy to surveying scanners. In contrast, general-purpose scanners without geometric correction tend to degrade positional accuracy. This study provides a practical assessment for the suitability of image acquisition device to generate orthophoto maps and highlights the applicability of low-cost scanning devices for orthophoto generation.
