Science of Remote Sensing最新文献

{"title":"Advancing vegetation segmentation from ALS point clouds: From benchmarking to GreenSegNet-A","authors":"Aditya Aditya ,&nbsp;Bharat Lohani ,&nbsp;Jagannath Aryal ,&nbsp;Stephan Winter","doi":"10.1016/j.srs.2026.100382","DOIUrl":"10.1016/j.srs.2026.100382","url":null,"abstract":"<div><div>Accurate large-scale vegetation segmentation is essential to maintain vegetation inventories, which are vital for informed ecological planning, landscape management, and long-term sustainability and liveability of the environment. Advancements in deep learning (DL) coupled with the increasing availability of airborne laser scanning (ALS) point clouds hold significant potential for detailed and large-scale vegetation segmentation. Yet, ALS-based vegetation segmentation has received limited attention, leading to ambiguity in model selection. To address this research gap, we present a comprehensive benchmarking of point-based DL models for vegetation segmentation. Seven representative DL models, KPConv, RandLANet, SCFNet, PointNeXt, SPoTr, PointMetaBase, and GreenSegNet, have been implemented on three different datasets, Eclair, Dales, and WHU-Urban3D. Evaluated through a ten-fold cross-validation strategy, the results reveal strong but inconsistent performances. KPConv records the highest mean intersection over union (mIoU) on the Eclair dataset with 96.24% while GreenSegNet dominates on Dales dataset, reaching 93.91%. GreenSegNet also outperforms other models on the WHU-Urban3D dataset, achieving a mIoU of 79.27%. These findings highlight both the promise and the limitations of existing models, including the vegetation-specific GreenSegNet, which also exhibited inconsistent behavior on ALS data due to sparsity, nadir-view perspective, and canopy occlusions. Building on these insights, we propose GreenSegNet-A, a DL architecture explicitly tailored for ALS vegetation segmentation. Incorporated with a novel ALS-adaptive module, GreenSegNet-A achieves mIoU scores of 96.56% (Eclair), 94.29% (Dales), and 80.87% (WHU-Urban3D). Statistical tests confirm its efficacy, while ablation studies validate the design choices. Although the model has a slightly higher parameter count than GreenSegNet, it remains lighter compared to other models. Overall, GreenSegNet-A establishes a strong performance baseline for ALS vegetation segmentation within the scope of our evaluation. The source code is available at <span><span>this URL</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":101147,"journal":{"name":"Science of Remote Sensing","volume":"13 ","pages":"Article 100382"},"PeriodicalIF":5.2,"publicationDate":"2026-01-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146077417","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":"Assessing the impact of polarization on soil moisture retrieval using C-band SAR data across diverse crop structures","authors":"Vaibhav Gupta ,&nbsp;Dharmendra Kumar Pandey ,&nbsp;Nicolas Baghdadi ,&nbsp;Mehrez Zribi ,&nbsp;Sekhar Muddu","doi":"10.1016/j.srs.2026.100377","DOIUrl":"10.1016/j.srs.2026.100377","url":null,"abstract":"<div><div>SAR operating at shorter wavelengths, exhibits shallower penetration depth and interacts primarily with the upper canopy layer. Consequently, its scattering mechanisms differ from those observed at longer wavelengths. Using this as a basis, this study presents a comprehensive evaluation of Soil Moisture retrieval using C-band data from two satellites missions, EOS-04 and Sentinel-1A through the application of Water Cloud Model. The in-situ datasets were acquired over southern India between June 2022 and January 2024. A total of 43 Sentinel-1A and 32 EOS-04 images were analysed alongside in situ measurements of SM and LAI collected over four crops, as well as bare soil. The novelty of this work lies in the comparative assessment of polarization configurations operating at the shorter wavelength, providing new insights into their relative sensitivity, retrieval performance and development of scattering mechanism across diverse canopy structures. The WCM was calibrated using LAI as the vegetation descriptor, the resulting SM estimates achieved RMSE ranging from 6.28 % to 10.15 %. HH polarization exhibited greater sensitivity under dense canopies, such as turmeric, whereas VV yielded slightly higher overall retrieval accuracy for most crop structures. Analysis of the scattering behaviour revealed that vegetation influence becomes dominant in VV at relatively lower biomass at this wavelength. Results also indicated that at higher SM levels, sensitivity and retrieval accuracy decline due to saturation effects. Overall, this study provides insights into polarization-dependent scattering mechanism in shorter wavelength and highlighting the importance of accounting for crop structure for SM retrieval.</div></div>","PeriodicalId":101147,"journal":{"name":"Science of Remote Sensing","volume":"13 ","pages":"Article 100377"},"PeriodicalIF":5.2,"publicationDate":"2026-01-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146077416","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":"Satellite remote sensing of hydro-biogeochemical responses to near-coastal water dynamics in global river mouth areas","authors":"Youngwook Kim ,&nbsp;Ji-Hyung Park ,&nbsp;Jinyang Du","doi":"10.1016/j.srs.2026.100379","DOIUrl":"10.1016/j.srs.2026.100379","url":null,"abstract":"<div><div>Land-margin ecosystems surrounding river mouths are hydro-biogeochemical hotspots where water, carbon, and nutrients are exchanged between land and ocean. The land-margin ecosystems have recently experienced significant variations in surface water extent (Fw) due to increasing intensity of climate and environmental changes. The variations of the Fw at the river mouth areas are closely linked with the changes in biogeochemical cycles, including greenhouse gas emissions, ocean chlorophyll production and nutrient exports from land-margin ecosystems. Multi-source environmental remote sensing data records were used to investigate how changes in Fw affect hydroclimates, including precipitation, surface soil moisture, and root-zone soil moisture, and biogeochemical fluxes associated with heterotrophic respiration, atmospheric CH₄, and terrigenous dissolved organic matter (tDOM). The study focused on 253 major river mouth sites, identified within the boundary of each river mouth using the MERIT-Hydro map derived from a digital elevation model. The long-term (2003–2022) satellite-derived Fw data showed a strong increasing trend in the mean annual Fw over global land areas and major river mouths. However, the Fw trends varied across aridity zones in response to climate and environmental changes —likely due to the changes in surface dryness and permafrost melting dynamics —with 46 % of river mouths showing a decreasing Fw trend, indicating lower surface wetness conditions. Fw generally showed positive correlations with heterotrophic respiration in the area surrounding river mouths. Its relationship with atmospheric CH₄ concentration was also positive in river mouth areas located in semi-arid and sub-humid zones. Particularly, in arid regions, the increasing Fw led to enhance heterotrophic respiration, but significantly reduced atmosphere CH₄ concentrations. The deceased flux of tDOM exported from land to water may be linked to the reduced runoffs from river mouth areas as indicated by the Fw decreases. The decreased Fw lowered tDOM exports to coastal waters in 61 % of the studied river mouth areas. The results highlight that long-term satellite-derived Fw observations, alongside multi-source remote sensing data, are critical for monitoring surface wetness in land-margin ecosystems and assessing its impact on hydro-biogeochemical fluxes in near-coastal environments.</div></div>","PeriodicalId":101147,"journal":{"name":"Science of Remote Sensing","volume":"13 ","pages":"Article 100379"},"PeriodicalIF":5.2,"publicationDate":"2026-01-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146078057","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":"HCV-CVAE: A hierarchical convolutional variational transformer for thin cloud removal in remote sensing imagery","authors":"Yan Zhang ,&nbsp;Feng Han ,&nbsp;Juwei Xiang ,&nbsp;Jiwu Guan ,&nbsp;Song Wang","doi":"10.1016/j.srs.2026.100380","DOIUrl":"10.1016/j.srs.2026.100380","url":null,"abstract":"<div><div>To address the challenge faced by existing thin-cloud removal methods in balancing global structure reconstruction and local texture restoration under complex cloud conditions, this paper proposes a remote sensing image de-clouding approach based on a Hierarchical Convolutional Variational Vision Transformer (HCV-CVAE). Built upon the conventional CVAE framework, the proposed model introduces an HCV-ViT encoder that integrates the strengths of convolutional networks and Transformers to enhance local texture representation while capturing global semantic dependencies. Furthermore, strategies such as KL-divergence annealing, cross-dimensional weighted mutual information loss, and test-time augmentation are incorporated to improve the stability of the latent space and the robustness of the generation process. The proposed approach exhibits superior performance over existing algorithms on the RICE2 and T-Cloud datasets, with the highest PSNR and SSIM reaching 40.93 dB and 0.9872, respectively. The HCV-CVAE effectively restores fine details and spectral characteristics beneath clouds while maintaining global structural consistency, exhibiting significant advantages in both visual quality and quantitative metrics. All implementation code and pretrained models are publicly available at: <span><span>https://github.com/Kyperio/HCV-CVAE</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":101147,"journal":{"name":"Science of Remote Sensing","volume":"13 ","pages":"Article 100380"},"PeriodicalIF":5.2,"publicationDate":"2026-01-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146077415","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":"Observing irrigation using SWOT SAR Ka-band data from daily calibration and validation acquisitions","authors":"Henri Bazzi ,&nbsp;Nicolas Baghdadi ,&nbsp;Cecile Cazals ,&nbsp;Sami Najem ,&nbsp;Damien Desroches ,&nbsp;Frédéric Frappart ,&nbsp;Mehrez Zribi ,&nbsp;François Charron","doi":"10.1016/j.srs.2026.100378","DOIUrl":"10.1016/j.srs.2026.100378","url":null,"abstract":"<div><div>While primarily designed for ocean and inland water monitoring through Interferometric SAR (InSAR) technology, the Surface Water and Ocean Topography (SWOT) Ka-band SAR sensor also presents a novel potential for agricultural applications. This study explores the sensitivity of SWOT's Ka-band backscatter to soil moisture variations, focusing on detecting irrigation events using daily observations collected during the calibration/validation (Cal/Val) phase. Daily backscatter variations from the SWOT Level 1B High-Rate Single-look Complex product were examined over an experimental irrigated grassland site, in response to irrigation events and rainfall. The analysis included first evaluating the stability of SWOT Ka-band backscatter signal, the temporal responses to both irrigation and rainfall, and the influence of vegetation density on Ka-band SAR signal penetration. Main findings showed that the Ka-band SAR data was sensitive to soil moisture variation due to irrigation, inducing an increased backscattering by an average of 4.3 dB on the same day of irrigation. For some cases of flooded vegetation persisting after irrigation, specular reflection and/or double-bounce scattering mechanisms were observed, causing an extreme increase in the Ka-band backscattering. Following complete infiltration, irrigation events induced an average increase of about 2 dB one day after irrigation which dropped back to previous levels two days later due to natural soil drying. Despite the Ka-band's short wavelength, typically limiting canopy penetration, SWOT's near-vertical incidence angle appears to enhance its ability to penetrate dense vegetation cover reaching the soil surface and detecting soil moisture dynamics. These findings open new perspectives for leveraging the daily CAL/VAL SWOT acquisitions to map irrigated areas and support agricultural water management.</div></div>","PeriodicalId":101147,"journal":{"name":"Science of Remote Sensing","volume":"13 ","pages":"Article 100378"},"PeriodicalIF":5.2,"publicationDate":"2026-01-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146037867","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":"A novel two-stage adversarial joint learning model for reconstructing InSAR phase in decorrelated areas","authors":"Mahmoud Abdallah ,&nbsp;Songbo Wu ,&nbsp;Xiaoli Ding","doi":"10.1016/j.srs.2026.100373","DOIUrl":"10.1016/j.srs.2026.100373","url":null,"abstract":"<div><div>Interferograms are basic observables of any Interferometric Synthetic Aperture Radar (InSAR) measurements. Interferometric decorrelation, however, often reduces the quality of interferograms, sometimes to an extent where no interferometric measurements can be properly carried out. Techniques such as applying a filter can help in reducing the impact of noise in interferograms but often cannot overcome the problem of decorrelation satisfactorily. This paper presents an approach based on a novel two-stage generative adversarial network (GAN) tailored for reconstructing interferometric phase values in decorrelated areas. The approach comprises an edge mapping stage (EMS) and a phase predicting stage (PPS). During the edge mapping stage, a pre-trained convolutional neural network (CNN) identifies fringe lines, while a GAN reconnects the discontinuous fringes. In the phase predicting stage, a second GAN uses the reconnected fringes as a guide to reconstruct the phase information. The model was trained on simulated datasets, achieving an overall accuracy (OA) of 84 % in fringe reconnection and a structural similarity index (SSIM) of 96 %. We validated the proposed model with real-world case studies, successfully reconstructing the phases of co-seismic deformation interferograms for the Tonopah, Nevada earthquake (M 6.5, May 15, 2020) and the Western Xizang earthquake (M 6.3, July 22, 2020). We also evaluated the adaptability of the proposed model using topographic mapping datasets. The experimental results achieved a cross-correlation range of 0.72–0.87 when reconstructing phase information over the Greater Bay Area (GBA) with fine-tuning, indicating potential applicability of the approach to a broader range of InSAR applications.</div></div>","PeriodicalId":101147,"journal":{"name":"Science of Remote Sensing","volume":"13 ","pages":"Article 100373"},"PeriodicalIF":5.2,"publicationDate":"2026-01-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146037866","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":"Discriminating winter wheat yellow rust and Fusarium head blight using Sentinel-2 imagery at a regional scale","authors":"Zhiqin Gui ,&nbsp;Huiqin Ma ,&nbsp;Jingcheng Zhang ,&nbsp;Wenjiang Huang ,&nbsp;Lin Yuan ,&nbsp;Kehui Ren","doi":"10.1016/j.srs.2026.100371","DOIUrl":"10.1016/j.srs.2026.100371","url":null,"abstract":"&lt;div&gt;&lt;div&gt;Yellow rust (&lt;em&gt;Puccinia striiformis&lt;/em&gt; f. sp. &lt;em&gt;Tritici&lt;/em&gt;, YR) and Fusarium head blight (&lt;em&gt;Fusarium graminearum&lt;/em&gt;, FHB) are two major wheat diseases. These two diseases frequently pose concurrent risks to grain security, particularly in high-yielding wheat regions of eastern China. Accurate regional-scale discrimination of wheat YR and FHB is essential for developing effective green and intelligent disease management strategies. While satellite remote sensing shows potential for regional crop disease monitoring, conventional machine learning modeling approaches widely employed often fail to exploit the spectral-spatial information inherent in imagery. Meanwhile, the scarcity of ground-based disease survey samples limits the application of emerging sample-driven deep learning methods. This study evaluated the effectiveness of 27 sample-feature-algorithm combinatorial modeling strategies for discriminating regional-scale wheat YR and FHB using Sentinel-2 imagery. We augmented disease samples using a stepwise approach that combines marking diseased field vector boundaries with sliding window segmentation (SWS), horizontal-vertical flipping (HVF), and multi-angle rotation (MAR). Recursive feature elimination with cross-validation (RFECV) was employed to optimize spectral and textural features, yielding in two distinct feature sets: disease-sensitive spectral features (SFs) and spectral-textural combined features (STCFs). The original spectral bands (OSBs) served as a third feature set. These sample sets and feature sets were input into several fundamentally distinct algorithms to construct wheat YR and FHB discrimination models. These include three commonly used machine learning (ML) methods, namely, support vector machine (SVM), random forest (RF), and extreme gradient boosting (XGBoost). Additionally, include two deep learning methods, namely, the two-dimensional convolutional neural network (2D-CNN) and the spectral-spatial attention network (SSAN). The results indicated that three ML algorithms exhibited stable performance across all three feature sets under SWS-based sample augmentation. SVM yielded the best overall accuracy, but texture features provided only limited improvement over the SVM model compared with RF and XGBoost. The OSBs outperformed SFs and STCFs in 2D-CNN and SSAN modeling, achieving an overall accuracy (OA) comparable to that of SVM under SWS + HVF + MAR-based sample augmentation. Specifically, the SWS + HVF + MAR-OSBs-SSAN model demonstrated superior performance metrics. This model achieved an average accuracy of 81.8 %, a Kappa coefficient of 0.704, a G-means of 0.892, and an F1-score of 81.1 %. These accuracy results surpassed those of the SWS-STCFS-SVM model, even though the latter achieved the highest OA of 82.8 %. Sample augmentation yielded limited gains in modeling for the 2D-CNN but demonstrated more significant gains for the SSAN. Overall, the STCFs-based SVM modeling strategy remains preferab","PeriodicalId":101147,"journal":{"name":"Science of Remote Sensing","volume":"13 ","pages":"Article 100371"},"PeriodicalIF":5.2,"publicationDate":"2026-01-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146037869","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":"A cost-effective method for mapping land cover at national scale","authors":"John R. Dymond ,&nbsp;James D. Shepherd ,&nbsp;Richard Law ,&nbsp;Brent Martin ,&nbsp;Jan Schindler ,&nbsp;Stella Belliss","doi":"10.1016/j.srs.2026.100376","DOIUrl":"10.1016/j.srs.2026.100376","url":null,"abstract":"<div><div>Timely and accurate national-scale land-cover mapping is essential for resource management. However, achieving this with limited resources is a challenge, particularly in mountainous and ecologically diverse regions with frequent cloud cover like New Zealand. We present a cost-effective, scalable methodology for land-cover classification that integrates Sentinel-2 imagery, spectral decision rules, temporal NDVI analysis, and deep learning (U-Net) within a unified, reproducible workflow. Our approach generates land-cover maps at a spatial resolution of 10 m. National classification was generated in less than 12 h of computing time. Validation against 4500 samples stratified by map class yielded an overall classification accuracy of 96 %, outperforming leading global products. This method balances automation with expert-informed logic, enabling accurate differentiation of challenging classes such as exotic forest, indigenous forest, and croplands. Although developed for New Zealand, the workflow should be adaptable to other countries seeking low-cost, high-frequency land-cover mapping. These land-cover maps can support a range of environmental applications, including carbon accounting, biodiversity assessment, erosion modelling, and detection of land-use change.</div></div>","PeriodicalId":101147,"journal":{"name":"Science of Remote Sensing","volume":"13 ","pages":"Article 100376"},"PeriodicalIF":5.2,"publicationDate":"2026-01-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146037871","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":"Utilising mobile laser scanning point clouds to assess harvesting quality in thinning stands","authors":"Anwar Sagar ,&nbsp;Johannes Pohjala ,&nbsp;Jesse Muhojoki ,&nbsp;Anubhav Dhital ,&nbsp;Harri Kaartinen ,&nbsp;Kalle Kärhä ,&nbsp;Kalervo Järvelin ,&nbsp;Reza Ghabcheloo ,&nbsp;Juha Hyyppä ,&nbsp;Ville Kankare","doi":"10.1016/j.srs.2026.100374","DOIUrl":"10.1016/j.srs.2026.100374","url":null,"abstract":"<div><div>Forestry is entering a new era where precision and innovation converge through advanced mobile laser scanning (MLS) technologies. Traditional methods of assessing harvesting quality, often manual, time-consuming, and prone to human error, are being replaced by objective, data-driven approaches. In this study, we conducted high-resolution point cloud scanning across four forest stands (11 ha) in Central Finland using the handheld GeoSLAM ZEB Horizon LiDAR system. We aimed to evaluate the capacity of MLS to measure harvesting attributes related to stand density, tree dimensions, and strip road characteristics, to assess the impact of the Ponsse Plc Thinning Density Assistant (TDA), and to detect defective tree stems. Within a 5-ha subset, 11 potentially anomalous trees were identified. A spatially precise tree map was created using QGIS and a separate map application, enabling comparison between manual field measurements and digital measurements. The findings indicate a strong concordance between automated and traditional assessments. With few exceptions, the results were consistent with established Best Practices for Sustainable Forest Management. Preliminary tests of a novel algorithm for curved stem detection further suggest the potential of MLS for automated defect recognition. A strip road width model was also developed to estimate the average strip road width within the forest stand. These findings underscore MLS as a powerful tool for enhancing accuracy, efficiency, and objectivity in modern forest management.</div></div>","PeriodicalId":101147,"journal":{"name":"Science of Remote Sensing","volume":"13 ","pages":"Article 100374"},"PeriodicalIF":5.2,"publicationDate":"2026-01-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146037870","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":"Mapping hidden heritage: Self-supervised pre-training on high-resolution LiDAR DEM derivatives for archaeological stone wall detection","authors":"Zexian Huang ,&nbsp;Mashnoon Islam ,&nbsp;Brian Armstrong ,&nbsp;Billy Bell ,&nbsp;Kourosh Khoshelham ,&nbsp;Martin Tomko","doi":"10.1016/j.srs.2026.100372","DOIUrl":"10.1016/j.srs.2026.100372","url":null,"abstract":"<div><div>Historic dry-stone walls hold significant cultural and environmental importance, serving as historical markers and contributing to ecosystem preservation and wildfire management during dry seasons in Australia. However, many of these stone structures in remote or vegetated landscapes remain undocumented due to limited accessibility and the high cost of manual mapping. Deep learning–based segmentation offers a scalable approach for automated mapping of such features, but challenges remain: 1. the visual occlusion of low-lying dry-stone walls by dense vegetation and 2. the scarcity of labeled training data. This study presents <strong>DINO-CV</strong>, a self-supervised cross-view pre-training framework based on knowledge distillation, designed for accurate and data-efficient mapping of dry-stone walls using <strong>Digital Elevation Models (DEMs)</strong> derived from high-resolution airborne LiDAR. By learning invariant geometric and geomorphic features across DEM-derived views, (i.e., Multi-directional Hillshade and Visualization for Archaeological Topography), DINO-CV addresses the occlusion by vegetation and data scarcity challenges. Applied to the <strong>Budj Bim Cultural Landscape</strong> at Victoria, Australia, a UNESCO World Heritage site, the approach achieves a mean Intersection over Union (<em>mIoU</em>) of <em>68.6%</em> on test areas and maintains <em>63.8%</em> mIoU when fine-tuned with only 10% labeled data. These results demonstrate the potential of self-supervised learning on high-resolution DEM derivatives for large-scale, automated mapping of cultural heritage features in complex and vegetated environments. Beyond archaeology, this approach offers a scalable solution for environmental monitoring and heritage preservation across inaccessible or environmentally sensitive regions.</div></div>","PeriodicalId":101147,"journal":{"name":"Science of Remote Sensing","volume":"13 ","pages":"Article 100372"},"PeriodicalIF":5.2,"publicationDate":"2026-01-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145977578","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}