International journal of applied earth observation and geoinformation : ITC journal最新文献

{"title":"The approach to UAV image acquisition and processing for very shallow water mapping","authors":"Paulina Kujawa ,&nbsp;Jaroslaw Wajs ,&nbsp;Krzysztof Pleśniak","doi":"10.1016/j.jag.2025.104604","DOIUrl":"10.1016/j.jag.2025.104604","url":null,"abstract":"<div><div>Shallow water areas need to be protected and continuously monitored as a habitat for diverse flora and fauna. These environments are subject to changes caused by both local phenomena, such as tides, and global phenomena, such as global warming. Efficient measurement techniques are needed to optimize the cost and time of data collection and processing. Equally important is to ensure that data processing achieves the highest possible accuracy, especially for depth measurements affected by refraction. The aim of this paper is to present several approaches to data processing, based on the availability of measurement instruments and programming skills, each offering different levels of accuracy. In this study, RGB images were collected from an unmanned aerial vehicle over a Polish lake, together with reference data from a single-beam echo-sounder and GNSS measurements of shallow water profiles. Several processing paths were proposed, including sun glint masking, photogrammetric processing, refraction correction, and the creation of three output models: a point cloud, DEM, and orthomosaic. The expected accuracies are discussed, along with recommendations for the best method, taking into account the strengths and limitations of each approach.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"141 ","pages":"Article 104604"},"PeriodicalIF":7.6,"publicationDate":"2025-06-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144204493","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Detection and mapping of Antarctic lichen using drones, multispectral cameras, and supervised deep learning","authors":"Narmilan Amarasingam ,&nbsp;Juan Sandino ,&nbsp;Ashray Doshi ,&nbsp;Diana King ,&nbsp;Elka Blackman ,&nbsp;Johan Barthelemy ,&nbsp;Barbara Bollard ,&nbsp;Sharon A. Robinson ,&nbsp;Felipe Gonzalez","doi":"10.1016/j.jag.2025.104577","DOIUrl":"10.1016/j.jag.2025.104577","url":null,"abstract":"<div><div>The difficulty of accurately detecting lichens in Antarctic landscapes, due to their fine-scale spatial patterns and low spectral contrast, drives the need for high-resolution drone-based remote sensing imagery to develop and validate robust mapping methods. Few studies have explored the use of remote sensing and deep learning (DL) techniques for mapping and monitoring lichen density in Antarctic regions. This study aims to fill this gap by using multispectral (MS) cameras onboard uncrewed aerial vehicles (UAVs) and DL to detect and map Antarctic lichen through a workflow that enhances detection using a semi-automatic labelling technique based on vegetation indices (VIs). This methodology was validated through a data collection campaign at Robinson Ridge, Windmill Islands, Antarctica in January 2023. Two DL methods were evaluated to classify and map <em>Usnea</em> spp., <em>Umbilicaria</em> and <em>Pseudephebe</em> species (black lichen), moss and non-vegetation: method (1) standalone DL model fitting, namely fully convolutional network (FCN), U-Net, and Deeplabv3+, with semi-automatic labelling thresholding using VIs; and method (2) ensemble stacking by using eXtreme gradient boosting (XGBoost) as the input model, whose predictions are used as features for training a U-Net model. In Method 1, U-Net exhibited the best performance over the other models. Specifically, for <em>Usnea</em> spp., the results demonstrate an intersection over union (IoU) of 84%. Also, the black lichen class obtained an IoU of 86%. In contrast, Method 2, which employed the ensemble stacking technique, demonstrates an IoU of 71% for <em>Usnea</em> spp. and IoU of 75% for black lichen. This study provides promising evidence that using MS cameras on UAVs combined with DL models is an effective approach for detecting and mapping lichen density in Antarctica, though further exploration across diverse regions is recommended to validate its scalability and adaptability.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"141 ","pages":"Article 104577"},"PeriodicalIF":7.6,"publicationDate":"2025-06-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144189724","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Shoreline time series analysis through the Isoradiometric method: Bridging landscape evolution and coastal management","authors":"F. Caldareri ,&nbsp;N. Parrino ,&nbsp;L. Balsamo ,&nbsp;G. Dardanelli ,&nbsp;S. Todaro ,&nbsp;A. Sulli ,&nbsp;A. Maltese","doi":"10.1016/j.jag.2025.104618","DOIUrl":"10.1016/j.jag.2025.104618","url":null,"abstract":"<div><div>The increasing availability of remotely sensed data has enhanced our ability to monitor coastal evolution, yet extracting reliable time series for long-term analysis remains a challenge. This study evaluates the effectiveness of the Isoradiometric shoreline extraction Method in producing consistent time series data across different spatial and temporal scales. We applied the method to about 150 multispectral satellite images spanning 40 years, covering two sandy beaches along Sicily’s coast in the central Mediterranean Sea. Our validation approach focused on assessing method consistency across datasets with different spatial resolutions and revisit times. By comparing Landsat and PlanetScope data, we demonstrated that while high-resolution products capture greater variability in shoreline position, lower-resolution but longer time-span observations effectively identify underlying evolutionary trends. The analysis revealed that manual digitization captures instantaneous swash positions, while the Isoradiometric Method consistently identifies stable morphological features between the low tide terrace and berm, providing more reliable indicators of actual coastal change. This multi-resolution approach proved effective in distinguishing between method-related outliers and paroxysmal events, with the latter typically detected across multiple datasets at corresponding timeframes. The systematic application of the Isoradiometric Method successfully characterized both natural variability patterns and anthropic impacts, providing quantitative baselines for interpreting Quaternary coastal processes while offering practical insights for shoreline monitoring and coastal management strategies. Moreover, we calculated the shifts’ gradient to quantify the rate of change in shoreline position. These results demonstrate: i) the necessity of creating shoreline time series as a tool for geological interpretation through the principle of actualism and as a framework for rationalizing contemporary shoreline monitoring approaches; ii) the Isoradiometric Method enables accurate Earth Observation image processing for this purpose.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"141 ","pages":"Article 104618"},"PeriodicalIF":7.6,"publicationDate":"2025-06-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144195756","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Validated visual features of Multi-Perspective imagery with Explainable Machine learning for detecting rural vacant courtyards in North China","authors":"Zehao Qiao ,&nbsp;Maojun Wang ,&nbsp;Xuexia Zhang ,&nbsp;Juanjuan Zhao ,&nbsp;Xiaojie Zhang ,&nbsp;Tao Liu ,&nbsp;Guangzhong Cao","doi":"10.1016/j.jag.2025.104644","DOIUrl":"10.1016/j.jag.2025.104644","url":null,"abstract":"<div><div>As China’s rural population continues to decline and urbanisation accelerates, the number of vacant courtyards (VCs) has steadily increased. Existing methods for detecting VCs in China face significant challenges regarding data availability, spatial scale, resolution, and reliability, which hinder accurate assessment. This study establishes an innovative approach that evaluates courtyard utilisation status by integrating visual features from horizontal and overhead imagery, distinguishing it from traditional methods that rely solely on remote sensing textures or geometric information. We constructed a systematic set of visual features and employed an interpretable machine learning (XGBoost model) to detect courtyard utilisation status. We identified an optimal model comprising four primary features: from an overhead perspective, plants condition, enclosing wall edge cleanliness, and presence of solar water heaters; and from a horizontal perspective, door and window condition. The research further reveals the asymmetric relationship between these features and courtyard utilisation status, and the spatial heterogeneity underlying this relationship. The model significantly outperforms existing research, achieving an F1 score of 0.834 on the test set while maintaining high accuracy on the external validation dataset. This demonstrates the considerable advantages and potential of this relatively low-cost approach for rapidly detecting VC, providing theoretical support and reliable technical backing for implementing courtyard utilisation monitoring based on image data and sustainable development goals across a broad range of rural areas in the future.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"141 ","pages":"Article 104644"},"PeriodicalIF":7.6,"publicationDate":"2025-06-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144195157","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Estimation of carbon sequestration capacity of urban green infrastructure by fusing multi-source remote sensing data","authors":"Jiahui Chang ,&nbsp;Zhenfeng Shao ,&nbsp;Jinyang Wang ,&nbsp;Zhu Mao ,&nbsp;Tao Cheng ,&nbsp;Xiaodi Xu ,&nbsp;Qingwei Zhuang","doi":"10.1016/j.jag.2025.104643","DOIUrl":"10.1016/j.jag.2025.104643","url":null,"abstract":"<div><div>Urban green infrastructure significantly contributes to the carbon storage functions of urban ecosystems. Accurate selection and efficiently integrating remote sensing data are paramount for evaluating carbon storage at the small-scale of urban green infrastructure. In this study, evaluating the precision of carbon storage estimation by integrating UAV-acquired multi-view spectral images and LiDAR data, complemented by ground-truth validation data. The allometric equations and carbon ratios specific to the tree species. The accuracy evaluation reveals that the R² value and RMSE for the extracted individual tree height variables were 0.75 and 1.76 m. For the estimated carbon storage, the R² reached 0.86, with an RMSE of 28.88 kg C. Additionally, the spatial arrangement and structure of tree species within green infrastructure notably affected carbon storage heterogeneity. This study demonstrates the effectiveness of integrating multi-view spectral imagery and LiDAR in accurately estimating carbon storage in urban green infrastructure. Furthermore, incorporating the small-scale spatial patterns of tree species can enhance the precision of carbon storage estimation.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"141 ","pages":"Article 104643"},"PeriodicalIF":7.6,"publicationDate":"2025-06-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144195156","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"An improved height sampling approach used for global urban building height mapping","authors":"Tingting He ,&nbsp;Yihua Hu ,&nbsp;Fashuai Li ,&nbsp;Yuwei Chen ,&nbsp;Maoxin Zhang ,&nbsp;Qiming Zheng ,&nbsp;Baiyu Dong ,&nbsp;He Ren","doi":"10.1016/j.jag.2025.104633","DOIUrl":"10.1016/j.jag.2025.104633","url":null,"abstract":"<div><div>Building height serves as fundamental information for characterizing urban landscapes and morphology, as influencing various aspects of the urban environment. While traditional methods of obtaining building height are often limited by spatial coverage and proprietary constraints, remote sensing data provides an alternative for indirect estimation. Several height products developed across different spatial scales, yet challenges remain due to the spatial and temporal incompleteness of publicly available building height samples, which introduce inherent uncertainties in global height mapping. This study proposed an improved approach for building height sampling that combines the ALOS AW3D30 and Global Ecosystem Dynamics Investigation (GEDI) data. Both datasets are open-access, providing a more comprehensive and representative sample base for model construction. To address temporal discrepancies between these two data, continuous change detection and classification (CCDC) algorithm was employed to remove invalid height samples. Subsequently, we trained random forest (RF) models using a combination of multi-source remote sensing data, including radar data, optical data, nighttime light data, terrain data, and footprint data, to generate a global urban building height map for the year of 2020. Reference samples from Europe, the United States, and China were employed to validate the model, indicating a high degree of consistency between the references and estimated results (R² = 0.55–0.75, RMSE = 4.71–10.07 m). Moreover, our findings indicated that over 20 % of regions globally experienced rapid urbanization, with average building heights exceeding 10 m, particularly in southern China. The approach proposed in this study provides effective support for building height estimation, particularly in address the limitations of lack of incomplete and representative height samples in global mapping.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"141 ","pages":"Article 104633"},"PeriodicalIF":7.6,"publicationDate":"2025-05-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144178562","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Automated vicarious radiometric validation of spaceborne thermal infrared sensors at non-dedicated validation sites using deep learning-based cloud filtering","authors":"Soushi Kato,&nbsp;Toru Kouyama","doi":"10.1016/j.jag.2025.104617","DOIUrl":"10.1016/j.jag.2025.104617","url":null,"abstract":"<div><div>Vicarious calibration/validation (Cal/Val) is essential for ensuring the radiometric accuracy of spaceborne thermal infrared (TIR) sensors. Anticipating an increase in the number of TIR sensors in the near future, this study developed an automated vicarious radiometric validation method based on <em>in situ</em> water temperature measurements from 14 sites in lakes and bays across Japan, which are not specifically dedicated to Cal/Val of satellite sensors. In addition, we applied a contextual image classification model based on the Swin Transformer architecture to create a fully automated filtering procedure to remove cloudy data. The proposed methods were developed and evaluated using the data acquired by well-calibrated satellite sensors, namely Advanced Spaceborne Thermal Emission and Reflection Radiometer Thermal Infrared Radiometer (ASTER TIR), Landsat 8 Thermal Infrared Sensor (TIRS), and Landsat 9 TIRS-2, as reference targets. To develop a contextual image classification model, we fine-tuned a pre-trained Swin Transformer model with our own training data comprising 921,967 chip images created from ASTER TIR band 13 and Landsat 8 TIRS band 10 images selected from global regions. The developed image classification model achieved an overall accuracy of 96.20 %. However, when applied only to the study areas, the accuracy decreased to 94.85 %, because all the target sites were exclusively located in lakes and bays. The classification model was localized to our study areas by adjusting the probability threshold. Combining contextual image classification with quantitative thresholds, the model successfully classified 90 % of the cloud-free daytime ASTER data and Landsat 8 data. The accuracy for cloud-free classification was 81 % and 86 % for nighttime ASTER data and Landsat 9 data, respectively. Consequently, 169, 58, 500, and 130 matchups were automatically identified for daytime ASTER, nighttime ASTER, Landsat 8, and Landsat 9, respectively. The <em>in situ</em> water temperature for each matchup was converted to top-of-atmosphere brightness temperature (TOA BT) through radiative transfer calculations. <em>In situ</em>-based and satellite-based TOA BT agreed very well within the residual bias error of less than ±0.4 K except for nighttime ASTER data that seemed to be affected by insufficient skin temperature correction. The correlation between <em>in situ</em>-based and satellite-based TOA BT was strong, with <em>R</em>² values ranging from 0.97 to 0.99, for daytime and nighttime ASTER, Landsat 8, and Landsat 9. The statistically estimated offsets between the <em>in situ</em>-based and satellite-based TOA BT were nearly equivalent to previously reported Cal/Val result and within the acceptable range of the sensors’ requirements, indicating that our method and data are suitable for the radiometric validation of spaceborne TIR sensors.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"141 ","pages":"Article 104617"},"PeriodicalIF":7.6,"publicationDate":"2025-05-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144178457","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"UAV-StrawFire: A visible and infrared dataset for real-time straw-fire monitoring with deep learning and image fusion","authors":"Xikun Hu,&nbsp;Ya Jiang,&nbsp;Xiaoyan Xia,&nbsp;Chen Chen,&nbsp;Wenlin Liu,&nbsp;Pengcheng Wan,&nbsp;Kangcheng Bin,&nbsp;Ping Zhong","doi":"10.1016/j.jag.2025.104586","DOIUrl":"10.1016/j.jag.2025.104586","url":null,"abstract":"<div><div>Straw burning poses significant threats to local air quality and nearby public health by emitting harmful pollutants during specific seasons. Traditional satellite-based remote sensing techniques encounter difficulties in monitoring small-scale straw-burning events due to long revisit intervals and low spatial resolution. To address this challenge, unmanned aerial vehicles (UAVs) equipped with imaging sensors have emerged as a rapid and cost-effective solution for monitoring and detecting straw fires. This paper presents the UAV-StrawFire dataset, which comprises RGB images, thermal infrared images, and videos captured during controlled straw residue burning experiments in southern China using drones. The dataset is annotated and labeled to support the application of detection, segmentation, and tracking algorithms. This study addresses three key machine learning tasks using the dataset: (1) flame detection, achieved through a feature-based multi-modal image fusion model (FF-YOLOv5n) reaching a mAP50-95 of 0.5764; (2) flame segmentation, which delineates fire boundaries using the real-time lightweight BiSeNetV2 model, achieving a high mean Intersection over Union (mIoU) score exceeding 0.88; and (3) flame tracking, which monitors the real-time progression of straw burning with a precision of 0.9065 and a success rate of 0.6593 using the Aba-ViTrack algorithm, suitable for on-board processing on UAVs at 50 frames per second (FPS). These experiments provide efficient baseline models for UAV-based straw-burning monitoring with edge computing capabilities. The UAV-StrawFire dataset enables the detection and monitoring of flame regions with varying sizes, textures, and opacities, thereby supporting potential straw-burning control efforts. The dataset is publicly available on IEEE Dataport, offering a valuable resource for researchers in the remote sensing and machine learning communities to advance the development of effective straw-burning monitoring systems.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"141 ","pages":"Article 104586"},"PeriodicalIF":7.6,"publicationDate":"2025-05-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144168952","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Multi-sensor modelling of woody vegetation and canopy cover across natural and modified ecosystems","authors":"Stephen B. Stewart ,&nbsp;Melissa Fedrigo ,&nbsp;Shaun R. Levick ,&nbsp;Anthony P. O’Grady ,&nbsp;Daniel S. Mendham","doi":"10.1016/j.jag.2025.104635","DOIUrl":"10.1016/j.jag.2025.104635","url":null,"abstract":"<div><div>Remote sensing is an essential tool for monitoring the extent and biophysical attributes of vegetation. Multi-sensor approaches, that can reduce the costs of developing high-quality datasets and improve predictive performance, are increasingly common. Despite this trend, the advantages of these data-fusion techniques are rarely reported beyond statistical performance. We use airborne lidar-derived metrics to develop models of canopy cover (CC, %) and woody vegetation (WV, presence/absence) using dry-season imagery from the Sentinel-1 (S1 C-band, 5.5 cm wavelength, Synthetic Aperture Radar) and Sentinel-2 (S2, multispectral optical) satellite constellations across natural and modified agricultural ecosystems in Tasmania, southeast Australia. Validation statistics at 18,876 sample locations demonstrated strong performance for both CC (R² = 0.83, RMSE = 0.13) and WV (OA = 0.94, Kappa = 0.87) when using both S1 and S2 variables for prediction. The small improvement in statistical performance provided by SAR variables (typically 1–2 % for CC and WV) understated the benefits of S1 for discriminating woody vegetation and quantifying canopy cover in non-woody ecosystems (e.g., alpine vegetation, heathlands, wetlands, coastal scrub), demonstrating the complementary benefits of multi-sensor prediction. The emergence and growth of natural capital accounting and frameworks such as the Nature Positive Initiative, mean that high-quality, cost-effective spatial datasets will continue to be in demand. Our study demonstrates the potential of non-commercial, publicly accessible remote sensing imagery to improve fine-scale analyses that may otherwise be cost-prohibitive to apply at scale.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"141 ","pages":"Article 104635"},"PeriodicalIF":7.6,"publicationDate":"2025-05-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144168953","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Multi-scale feature learning for 3D semantic mapping of agricultural fields using UAV point clouds","authors":"Hao Wang ,&nbsp;Yongchao Shan ,&nbsp;Liping Chen ,&nbsp;Mengnan Liu ,&nbsp;Lin Wang ,&nbsp;Zhijun Meng","doi":"10.1016/j.jag.2025.104626","DOIUrl":"10.1016/j.jag.2025.104626","url":null,"abstract":"<div><div>Accurate spatial distribution information of field features is critical for enabling autonomous agricultural machinery navigation. However, current perception systems exhibit limited segmentation performance in complex farm environments due to illumination variations and mutual occlusion among various regions. This paper proposes a low-cost UAV photogrammetry framework for centimeter-level 3D semantic maps of agricultural fields to support autonomous agricultural machinery path planning. The methodology combines UAV-captured images with RTK positioning to reconstruct high-precision 3D point clouds, followed by a novel Local-Global Feature Aggregation Network (LoGA-Net) integrating multi-scale attention mechanisms and geometric constraints. The framework achieves 78.6% mIoU in classifying eight critical agricultural categories: paddy field, dry field, building, vegetation, farm track, paved ground, infrastructure and other static obstacles. Experimental validation demonstrates a 5.9% accuracy improvement over RandLA-Net on the Semantic3D benchmark. This advancement significantly enhances perception accuracy in complex agricultural environments, particularly for field boundary delineation and occluded feature recognition, which directly facilitates robust path planning for unmanned agricultural machinery. The framework provides a scalable technical and data-driven foundation for achieving fully autonomous farm operations, ensuring both operational efficiency and environmental sustainability.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"141 ","pages":"Article 104626"},"PeriodicalIF":7.6,"publicationDate":"2025-05-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144168948","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}