[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"item-2942":3,"related-2942":105},{"id":4,"title":5,"url":6,"summary":7,"summary_zh":8,"content":9,"source_name":10,"source_url":6,"published_at":11,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":15,"score_detail":16,"sources":24,"tags":26,"primary_topic":32,"secondary_topics":38,"classification":44,"search_phrases":76,"slug":79,"view_count":22,"doi":80,"paper":81,"created_at":104},2942,"Deep learning-driven multisource remote sensing image fusion: Advances, challenges, and future directions","https:\u002F\u002Fdoi.org\u002F10.1016\u002Fj.engappai.2026.116319","Multisource remote sensing image fusion has become an important solution to a long-standing limitation in Earth observation: individual sensors rarely provide high spatial detail, rich spectral information, reliable structural sensitivity, and frequent temporal coverage at the same time. This review examines how deep learning and artificial intelligence are being used to integrate multispectral, hyperspectral, panchromatic, optical, and synthetic aperture radar imagery for more reliable interpretation of complex ground scenes. It provides a technical synthesis of convolutional neural networks, autoencoders, generative adversarial networks, transformer architectures, diffusion models, and hybrid model driven approaches, with attention to their fusion mechanisms, reconstruction behavior, computational demand, and suitability for operational use. Applications include land cover mapping, precision agriculture, environmental monitoring, urban analysis, disaster assessment, and defense related interpretation. Rather than treating each fusion task separately, this review connects sensor heterogeneity, spatial and spectral resolution trade offs, radiometric correction, geometric correction, registration, noise reduction, and fusion level design within a single framework. The analysis indicates that convolutional models remain effective for stable local detail recovery, adversarial learning can improve visual sharpness but may introduce spectral distortion, transformer models better capture long range spatial and spectral relationships, and diffusion models offer refined reconstruction at greater computational cost. The review further identifies open challenges involving misregistration, spectral bias, limited labeled data, weak generalization across sensors, high memory requirements, and limited interpretability. Future progress should prioritize sensor aware learning, self supervised training, uncertainty aware evaluation, lightweight deployment, and application oriented benchmarks to improve reliability in operational Earth observation.","多源遥感图像融合已成为解决地球观测领域一个长期局限的重要方案：单一传感器很少能够同时提供高空间细节、丰富光谱信息、可靠的结构敏感性以及频繁的时间覆盖。本文综述了如何利用深度学习和人工智能整合多光谱、高光谱、全色、光学和合成孔径雷达（synthetic aperture radar, SAR）影像，以更可靠地解译复杂地表场景。文章对卷积神经网络、自编码器、生成对抗网络、Transformer架构、扩散模型以及混合模型驱动方法进行了技术综合，重点关注其融合机制、重建行为、计算需求以及业务化适用性。应用领域包括土地覆盖制图、精准农业、环境监测、城市分析、灾害评估和国防相关解译。本文并非将每种融合任务分开处理，而是在一个统一框架内将传感器异质性、空间与光谱分辨率权衡、辐射校正、几何校正、配准、降噪和融合层级设计联系起来。分析表明，卷积模型在稳定的局部细节恢复方面仍然有效，对抗学习可以提升视觉锐度但可能引入光谱失真，Transformer模型能更好地捕捉长程空间与光谱关系，而扩散模型以更高的计算成本提供精细重建。本文进一步指出了涉及配准误差、光谱偏差、标注数据有限、跨传感器泛化能力弱、高内存需求以及可解释性有限等开放挑战。未来的进展应优先关注传感器感知学习、自监督训练、不确定性感知评估、轻量化部署以及面向应用的基准测试，以提高业务化地球观测的可靠性。",null,"Engineering Applications of Artificial Intelligence","2026-09-18T00:00:00Z","论文",10,false,82,{"impact":17,"substance":18,"depth":19,"authority":20,"freshness":21,"relevant":22,"comment":23},18,22,19,14,9,1,"发表于核心期刊的综述，系统梳理深度学习多源遥感融合的方法、应用与挑战，对农业遥感与精准农业有直接参考价值，时效性强，值得进入每日精选。",[25],{"name":10,"url":6},[27,28,29,30,31],"农业人工智能","深度学习","精准农业","遥感","多源数据融合",{"code":33,"slug":34,"name":35,"path":36,"parent_code":37},"I2","agricultural-ai","农业AI与大模型","\u002Ftopic\u002Fdigital-agriculture\u002Fagricultural-ai","I",[39],{"code":40,"slug":41,"name":42,"path":43,"parent_code":37},"I3","remote-sensing","遥感与农情监测","\u002Ftopic\u002Fdigital-agriculture\u002Fremote-sensing",{"status":45,"locked":14,"version":46,"method":47,"evidence":48},"classified","2026-10-22.2","rules",{"candidates":49,"agricultural_evidence":51},[50,57,64,71],{"code":33,"score":19,"supported":51,"title":52,"abstract":54,"tags":56},true,[53],"deep learning",[28,53,55],"artificial intelligence",[27,28],{"code":40,"score":17,"supported":51,"title":58,"abstract":60,"tags":63},[59],"remote sensing",[30,59,61,62],"multispectral","hyperspectral",[30],{"code":65,"score":66,"supported":51,"title":67,"abstract":68,"tags":70},"I1",7,[],[29,69],"precision agriculture",[29],{"code":72,"score":22,"supported":14,"title":73,"abstract":74,"tags":75},"I5",[],[],[31],[77,78],"多源遥感 图像融合 深度学习","农业人工智能 多源数据融合 深度学习 精准农业","多源遥感图像融合深度学习-2942","10.1016\u002Fj.engappai.2026.116319",{"doi":80,"openalex_id":82,"authors":83,"venue":10,"cited_by_count":96,"oa_url":6,"card":97,"direction":101,"ingested_from":103},"W7213544281",[84,87,90,93],{"name":85,"orcid":86},"Shahid Karim","https:\u002F\u002Forcid.org\u002F0000-0001-9986-5052",{"name":88,"orcid":89},"Akeel Qadir","https:\u002F\u002Forcid.org\u002F0000-0003-0358-6505",{"name":91,"orcid":92},"Asif Ali Laghari","https:\u002F\u002Forcid.org\u002F0000-0001-5831-5943",{"name":94,"orcid":95},"Irfana Bibi","https:\u002F\u002Forcid.org\u002F0000-0003-2794-504X",0,{"tldr":98,"method":99,"finding":100,"direction":101,"opportunity":102},"综述深度学习多源遥感图像融合方法、挑战与未来方向。","综述CNN、GAN、Transformer、扩散模型等融合机制与重建行为。","CNN擅局部细节，GAN易谱失真，Transformer长程关系强，扩散模型精度高但算力大。","农业遥感与作物表型","面向农业的轻量、自监督、不确定性感知融合与基准数据集构建。","openalex","2026-09-19T23:30:32.767384Z",{"total":106,"page":22,"page_size":106,"items":107},6,[108,171,269,322,375,445],{"id":109,"title":110,"url":111,"summary":112,"summary_zh":113,"content":9,"source_name":114,"source_url":111,"published_at":115,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":116,"score_detail":117,"sources":122,"tags":124,"primary_topic":127,"secondary_topics":128,"classification":130,"search_phrases":153,"slug":156,"view_count":120,"doi":157,"paper":158,"created_at":170},3618,"Multi-Target Remote Sensing Segmentation with Cross-Scale Attention for Urban-Ecological Intelligence","https:\u002F\u002Fdoi.org\u002F10.56201\u002Fijgem.vol.12.no1.2026.pg131.154","Remote sensing image segmentation is essential to extract valuable information from satellite and aerial images to achieve significant applications such as urban planning and ecological monitoring. Yet, it is hard to accurately segment diverse and complicated features because of the constraints of conventional approaches and the requirement to process variable object sizes and complicated boundaries. This research aims to address the above difficulties through the use and assessment of advanced deep learning models specifically UNet, SegNet, and TransU-Net in multi target semantic segmentation. These networks have been applied in single-object segmentation tasks with a focus on structures and roads, while the UNet architecture was additionally utilized for multi-object segmentation comprising buildings, woods, grasslands, water bodies, and cultivated land. Using appropriate remote sensing datasets, the accuracy of the models was thoroughly evaluated using common evaluation metrics, including pixel accuracy, Intersection over Union (IoU), and F1-score. The experimental outcomes illustrate the capabilities of these deep learning methods to provide accurate identification of essential features. Notably, the maximum attainable accuracy for single-object building segmentation was 96.85%, whereas the overall accuracy for multi-object segmentation with the UNet was 84.2%. The experimental results show these deep learning algorithms can effectively separate different targets from remote sensing imagery, thus making them fundamental tools for geospatial analysis and related fields.","遥感图像分割对于从卫星和航空图像中提取有价值信息，以实现城市规划和生态监测等重要应用至关重要。然而，由于传统方法的局限以及处理可变目标尺寸和复杂边界的需要，准确分割多样且复杂的地物十分困难。本研究旨在通过使用和评估先进的深度学习模型，特别是UNet、SegNet和TransU-Net在多目标语义分割中的应用，来解决上述难题。这些网络已被应用于单目标分割任务，重点关注建筑物和道路，而UNet架构还被额外用于多目标分割，包括建筑物、林地、草地、水体和耕地。利用适当的遥感数据集，通过常用评价指标对模型的精度进行了全面评估，包括像素精度、交并比（IoU）和F1分数。实验结果表明了这些深度学习方法在准确识别关键地物方面的能力。值得注意的是，单目标建筑物分割的最高可达精度为96.85%，而使用UNet进行多目标分割的总体精度为84.2%。实验结果表明，这些深度学习算法能够有效地从遥感图像中分离出不同目标，从而使其成为地理空间分析及相关领域的基础工具。","IIARD INTERNATIONAL JOURNAL OF GEOGRAPHY AND ENVIRONMENTAL MANAGEMENT","2026-09-25T00:00:00Z",63,{"impact":118,"substance":17,"depth":119,"authority":13,"freshness":120,"relevant":22,"comment":121},12,15,8,"将UNet、SegNet、TransU-Net用于多目标遥感语义分割，建筑单目标精度96.85%、多目标84.2%，方法常规但结论对农业与生态遥感监测有参考价值。",[123],{"name":114,"url":111},[27,28,125,30,126],"语义分割","生态监测",{"code":40,"slug":41,"name":42,"path":43,"parent_code":37},[129],{"code":33,"slug":34,"name":35,"path":36,"parent_code":37},{"status":45,"locked":14,"version":46,"method":47,"evidence":131},{"candidates":132,"agricultural_evidence":51},[133,137,141,148],{"code":40,"score":119,"supported":51,"title":134,"abstract":135,"tags":136},[59],[30,59],[30],{"code":33,"score":21,"supported":51,"title":138,"abstract":139,"tags":140},[],[28,53],[27,28,125],{"code":142,"score":143,"supported":51,"title":144,"abstract":145,"tags":147},"I4",3,[],[146],"空间分析",[],{"code":149,"score":22,"supported":14,"title":150,"abstract":151,"tags":152},"H6",[],[],[126],[154,155],"UNet 遥感 多目标分割","TransU-Net 城市生态 遥感","UNet遥感多目标分割-3618","10.56201\u002Fijgem.vol.12.no1.2026.pg131.154",{"doi":157,"openalex_id":159,"authors":160,"venue":114,"cited_by_count":96,"oa_url":163,"card":164,"direction":169,"ingested_from":103},"W7214298981",[161],{"name":162,"orcid":9},"Chibueze Favour Aririguzo","https:\u002F\u002Fiiardjournals.org\u002Fget\u002FIJGEM\u002FVOL. 12 NO. 1 2026\u002FMulti-Target Remote Sensing Segmentation 131-154.pdf",{"tldr":165,"method":166,"finding":167,"direction":101,"opportunity":168},"评估UNet、SegNet和TransU-Net在多目标遥感语义分割中的性能，用于城市与生态监测。","使用UNet、SegNet、TransU-Net模型，在遥感数据集上以像素精度、","单目标建筑分割最高精度96.85%，UNet多目标分割总体精度84.2%。","可探索跨尺度注意力机制提升多目标分割精度，并迁移至耕地与作物精细分类。","智慧农业 \u002F 农业物联网","2026-09-27T23:30:28.366758Z",{"id":172,"title":173,"url":174,"summary":175,"summary_zh":176,"content":9,"source_name":177,"source_url":174,"published_at":11,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":178,"score_detail":179,"sources":183,"tags":185,"primary_topic":188,"secondary_topics":194,"classification":197,"search_phrases":232,"slug":235,"view_count":106,"doi":236,"paper":237,"created_at":268},2947,"AI-enabled UAV-based Soil Organic Carbon Mapping in Arid Environments: A Pilot Study Protocol","https:\u002F\u002Fdoi.org\u002F10.2174\u002F0118743315495282260915110324","Introduction Soil organic carbon (SOC) is an important indicator of soil health, agricultural productivity, and carbon sequestration potential. However, accurate and scalable SOC mapping in arid environments is constrained by high spatial heterogeneity and the limitations of conventional soil sampling. This study aims to develop a standardized UAV-enabled framework for high-resolution SOC mapping in arid agricultural environments. Methods A pilot-study protocol integrating UAV-based hyperspectral remote sensing with artificial intelligence and machine learning was developed. The workflow encompasses study-site selection, ground-reference sampling, UAV hyperspectral data acquisition, radiometric and geometric preprocessing, spectral feature extraction and selection, machine-learning model development, validation, uncertainty assessment, and performance evaluation using R 2 , RMSE, and MAE. The protocol also incorporates assessment of environmental confounders, including soil moisture, surface roughness, and crop residues. Results The resulting framework provides a systematic and reproducible workflow for UAV-based SOC estimation, integrating field observations, hyperspectral features, predictive modelling, and uncertainty assessment. It establishes defined procedures for evaluating model robustness and transferability across varying field conditions. Discussion The framework addresses an important methodological gap in UAV-enabled SOC mapping by integrating remote sensing and AI within a standardized pilot-study design. Its emphasis on environmental confounders and uncertainty assessment can improve the reliability and comparability of SOC mapping studies. However, field validation across diverse arid environments remains necessary. Conclusion The proposed protocol provides a practical foundation for reproducible SOC mapping and subsequent field validation, supporting precision agriculture, sustainable soil management, and carbon monitoring, reporting, and verification (MRV) in arid regions.","引言 土壤有机碳（SOC）是衡量土壤健康、农业生产力及碳固存潜力的重要指标。然而，干旱环境中高空间异质性和传统土壤采样的局限性制约了准确且可扩展的SOC制图。本研究旨在开发一个标准化的无人机（UAV）框架，用于干旱农业环境中的高分辨率SOC制图。方法 开发了一套整合无人机高光谱遥感与人工智能及机器学习的试点研究方案。该工作流程涵盖研究地点选择、地面参考采样、无人机高光谱数据采集、辐射与几何预处理、光谱特征提取与选择、机器学习模型开发、验证、不确定性评估，以及使用R²、RMSE和MAE进行的性能评价。该方案还包括对环境混杂因素的评估，包括土壤水分、地表粗糙度和作物残茬。结果 所构建的框架为基于无人机的SOC估算提供了系统且可重复的工作流程，整合了野外观测、高光谱特征、预测建模和不确定性评估。它建立了明确的程序，用于评估模型在不同田间条件下的稳健性和可迁移性。讨论 该框架通过将遥感与人工智能整合于标准化的试点研究设计中，填补了无人机SOC制图领域的重要方法学空白。其对环境混杂因素和不确定性评估的重视，可提高SOC制图研究的可靠性和可比性。然而，仍需在不同干旱环境中进行田间验证。结论 所提出的方案为可重复的SOC制图及后续田间验证提供了实用基础，支持干旱地区的精准农业、可持续土壤管理以及碳监测、报告与核查（MRV）。","The Open Agriculture Journal",67,{"impact":118,"substance":17,"depth":180,"authority":181,"freshness":120,"relevant":22,"comment":182},16,13,"提出无人机高光谱结合AI的干旱区土壤有机碳制图标准化方案，方法框架清晰但尚属试点协议、缺乏实地验证，具备一定参考价值。",[184],{"name":177,"url":174},[186,27,29,30,187],"智慧农业","土壤碳汇",{"code":189,"slug":190,"name":191,"path":192,"parent_code":193},"H2","soil","土壤健康与改良","\u002Ftopic\u002Fenvironment\u002Fsoil","H",[195,196],{"code":33,"slug":34,"name":35,"path":36,"parent_code":37},{"code":40,"slug":41,"name":42,"path":43,"parent_code":37},{"status":45,"locked":14,"version":46,"method":47,"evidence":198},{"candidates":199,"agricultural_evidence":51},[200,205,211,215,219,226],{"code":189,"score":20,"supported":51,"title":201,"abstract":202,"tags":204},[190],[203,190],"土壤",[],{"code":33,"score":13,"supported":51,"title":206,"abstract":207,"tags":210},[],[208,209,55],"机器学习","machine learning",[27],{"code":40,"score":13,"supported":51,"title":212,"abstract":213,"tags":214},[],[30,59,62],[30],{"code":65,"score":120,"supported":51,"title":216,"abstract":217,"tags":218},[],[29,69],[186,29],{"code":220,"score":221,"supported":51,"title":222,"abstract":223,"tags":225},"H4",4,[],[224],"carbon sequestration",[187],{"code":227,"score":143,"supported":51,"title":228,"abstract":229,"tags":231},"H7",[],[230],"干旱",[],[233,234],"无人机 土壤有机碳 制图","AI 高光谱 干旱农业","无人机土壤有机碳制图-2947","10.2174\u002F0118743315495282260915110324",{"doi":236,"openalex_id":238,"authors":239,"venue":177,"cited_by_count":96,"oa_url":174,"card":263,"direction":101,"ingested_from":103},"W7213561504",[240,243,246,249,252,255,257,259,261],{"name":241,"orcid":242},"Moath Awawdeh","https:\u002F\u002Forcid.org\u002F0000-0003-1404-6782",{"name":244,"orcid":245},"Irfan Ahmed","https:\u002F\u002Forcid.org\u002F0000-0002-2172-4177",{"name":247,"orcid":248},"Anees Bashir","https:\u002F\u002Forcid.org\u002F0000-0002-4668-6592",{"name":250,"orcid":251},"Tarig Faisal","https:\u002F\u002Forcid.org\u002F0000-0001-6451-7576",{"name":253,"orcid":254},"Nicky Rahmana Putra","https:\u002F\u002Forcid.org\u002F0000-0003-4886-496X",{"name":256,"orcid":9},"Almaha Jamal",{"name":258,"orcid":9},"Afra Rashed",{"name":260,"orcid":9},"Hamda Yousif",{"name":262,"orcid":9},"Sarah Sadeq",{"tldr":264,"method":265,"finding":266,"direction":101,"opportunity":267},"提出一套无人机高光谱结合AI的干旱区土壤有机碳制图标准化试点方案。","无人机高光谱遥感、地面采样、光谱特征选择与机器学习建模，用R²、RMSE、MAE","构建了可复现的SOC估算流程，并纳入环境混杂因素与不确定性评估。","可在多干旱区开展跨区域验证，探索模型迁移性与不确定性量化方法。","2026-09-19T23:30:33.273156Z",{"id":270,"title":271,"url":272,"summary":273,"summary_zh":274,"content":9,"source_name":275,"source_url":272,"published_at":276,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":277,"score_detail":278,"sources":281,"tags":283,"primary_topic":285,"secondary_topics":286,"classification":288,"search_phrases":304,"slug":307,"view_count":279,"doi":308,"paper":309,"created_at":321},1769,"Land Cover Classification using Hyperspectral Imagery: A Deep Learning Approach","https:\u002F\u002Fdoi.org\u002F10.17485\u002Fijst\u002Fv19i31.487","Objectives: This research aimed to perform proper land cover classification of the Krishnagiri and Dharmapuri districts using hyperspectral imagery. This research also attempted to determine the usefulness of deep learning models in detecting large land cover classes, such as quarry, barren, forest, built-up, and agricultural land, using hyperspectral data. Method: The analysis of the land cover patterns in the study area used hyperspectral images (HSI) taken between 2015 and 2022, which had 230 spectral bands. A total of 3000 hyperspectral images were processed, 2000 of which were utilized for training and 1000 for testing. The images were then preprocessed, segmented into superpixels using superpixel segmentation, and classified. Three deep learning models, a Convolutional Neural Network (CNN), Residual Neural Network 18 (ResNet-18), and Visual Geometry Group-16 (VGG-16), were used as classification models. Standard measures of accuracy, precision, recall, and F-score were used to evaluate the performance and effectiveness of the models. Findings: The results indicate that ResNet-18 outperformed the other models in determining land cover classes using hyperspectral images. The model has shown high classification levels among various types of land, with built-up land having an accuracy level of approximately 95.09, indicating a better ability to extract and classify features than CNN and VGG-16. Novelty: The novelty of this study lies in the combination of hyperspectral image processing with deep learning frameworks to classify regions based on land cover. The relative analysis of CNN, ResNet-18, and VGG-16 demonstrated the usefulness of residual learning networks in enhancing the classification accuracy in hyperspectral remote sensing scenarios. Keywords: Hyperspectral Image (HSI), Land Cover Classification, Deep Learning, Convolutional Neural Network (CNN), ResNet-18, VGG-16, Remote Sensing, Spectral–Spatial Analysis","目标：本研究旨在利用高光谱影像对克里希纳吉里区和达尔马普里区进行准确的土地覆盖分类，并尝试确定深度学习模型在利用高光谱数据检测大型土地覆盖类别（如采石场、荒地、森林、建成区和农业用地）方面的有效性。方法：研究区域的土地覆盖模式分析使用了2015年至2022年间获取的高光谱图像（HSI），共包含230个光谱波段。共处理了3000幅高光谱图像，其中2000幅用于训练，1000幅用于测试。图像经过预处理后，采用超像素分割方法将其分割为超像素，并进行分类。使用三种深度学习模型作为分类模型，即卷积神经网络（CNN）、残差神经网络18（ResNet-18）和视觉几何组-16（VGG-16）。采用准确率、精确率、召回率和F值等标准指标来评估模型的性能和有效性。结果：结果表明，ResNet-18在利用高光谱图像确定土地覆盖类别方面优于其他模型。该模型在各种土地类型中表现出较高的分类水平，其中建成区的准确率约为95.09，表明其在特征提取和分类能力上优于CNN和VGG-16。新颖性：本研究的创新之处在于将高光谱图像处理与深度学习框架相结合，基于土地覆盖对区域进行分类。CNN、ResNet-18和VGG-16的相对分析证明了残差学习网络在提高高光谱遥感场景分类准确性方面的有效性。关键词：高光谱图像（HSI）、土地覆盖分类、深度学习、卷积神经网络（CNN）、ResNet-18、VGG-16、遥感、光谱-空间分析","Indian Journal of Science and Technology","2026-09-05T00:00:00Z",58,{"impact":118,"substance":17,"depth":180,"authority":13,"freshness":279,"relevant":22,"comment":280},2,"研究比较了三种深度学习模型在高光谱影像土地分类中的表现，方法有参考价值，但时效性差，影响范围有限。",[282],{"name":275,"url":272},[27,28,30,284],"土地分类",{"code":40,"slug":41,"name":42,"path":43,"parent_code":37},[287],{"code":33,"slug":34,"name":35,"path":36,"parent_code":37},{"status":45,"locked":14,"version":46,"method":47,"evidence":289},{"candidates":290,"agricultural_evidence":51},[291,295,299],{"code":40,"score":17,"supported":51,"title":292,"abstract":293,"tags":294},[62],[30,59,62],[30],{"code":33,"score":180,"supported":51,"title":296,"abstract":297,"tags":298},[53],[28,53],[27,28],{"code":142,"score":106,"supported":51,"title":300,"abstract":301,"tags":303},[],[146,302],"spatial analysis",[],[305,306],"农业人工智能 土地分类 深度学习 遥感","农业人工智能 土地分类","农业人工智能土地分类深度学习遥感-1769","10.17485\u002Fijst\u002Fv19i31.487",{"doi":308,"openalex_id":310,"authors":311,"venue":275,"cited_by_count":96,"oa_url":272,"card":316,"direction":101,"ingested_from":103},"W7208825957",[312,314],{"name":313,"orcid":9},"P Nithya",{"name":315,"orcid":9},"P Sudhakar",{"tldr":317,"method":318,"finding":319,"direction":101,"opportunity":320},"利用高光谱影像和深度学习模型对印度两地区进行土地覆盖分类，比较CNN、ResNet-18和VGG-1","使用2015-2022年230波段高光谱影像，超像素分割后，用CNN、ResNe","ResNet-18表现最佳，建成区分类准确率约95.09%，优于CNN和VGG-16。","可探索将ResNet-18等深度学习模型应用于高光谱影像的作物精细分类，或结合多时相数据提升农业监测精度。","2026-09-06T23:30:27.341932Z",{"id":323,"title":324,"url":325,"summary":326,"summary_zh":327,"content":9,"source_name":328,"source_url":325,"published_at":329,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":178,"score_detail":330,"sources":334,"tags":336,"primary_topic":339,"secondary_topics":340,"classification":342,"search_phrases":359,"slug":362,"view_count":96,"doi":363,"paper":364,"created_at":374},5570,"Evaluation of Counterfactual Explanation Robustness for Computer Vision Agriculture Yield-Forecast Models Under Satellite Cloud-Mask Gaps: An Interpretability Reliability Bound for Cooperative Storage Decisions","https:\u002F\u002Fdoi.org\u002F10.5281\u002Fzenodo.23545079","Counterfactual explanations are increasingly proposed as the interface through which cooperative storage managers interrogate computer-vision yield-forecast models, yet their reliability under the satellite cloud-mask gaps that routinely afflict agricultural remote sensing has not been bounded. The bound is derived from a Lipschitz continuity assumption on the forecast model and a masking-operator perturbation analysis, and it is expressed as a function of the cloud-mask gap fraction, the model's local Lipschitz constant, and the counterfactual search tolerance. A replication-based evaluation of the before-after-control-impact estimator was conducted to characterise the operating characteristics of the design under a declared variance structure (between-cluster SD = 6.0, residual SD = 4.5, intraclass correlation = 0.64), with power estimated across 2000 design cells. The reliability bound degrades non-linearly as the gap fraction rises, and that the intraclass correlation of 0.64 implies that cooperative-level aggregation of counterfactual recourse is necessary to preserve explanation stability. The contribution is a decision-relevant reliability bound that storage cooperatives can use to set a maximum tolerable cloud-mask gap before counterfactual recourse is suspended.","反事实解释日益被提出作为合作式仓储管理者质询计算机视觉产量预测模型的接口，然而在农业遥感中经常出现的卫星云掩膜缺口条件下，其可靠性尚未得到界定。该界基于预测模型上的Lipschitz连续性假设和掩膜算子扰动分析推导得出，并表示为云掩膜缺口比例、模型局部Lipschitz常数和反事实搜索容差的函数。对前后对照影响估计量进行了基于重复的评估，以表征该设计在给定方差结构（组间标准差=6.0，残差标准差=4.5，组内相关系数=0.64）下的运行特征，并在2000个设计单元中估计了检验效能。可靠性界随缺口比例上升呈非线性退化，且组内相关系数0.64意味着需要对反事实救济进行合作社层面的聚合，以保持解释的稳定性。本文的贡献在于提供了一个与决策相关的可靠性界，仓储合作社可据此设定在暂停反事实救济之前可容忍的最大云掩膜缺口。","Zenodo (CERN European Organization for Nuclear Research)","2026-10-21T00:00:00Z",{"impact":120,"substance":331,"depth":332,"authority":118,"freshness":13,"relevant":22,"comment":333},20,17,"该论文提出反事实解释在卫星云掩膜缺口下的可靠性边界，方法新颖且结论可用于合作社决策，但属细分方法学研究，产业影响有限。",[335],{"name":328,"url":325},[186,27,337,30,338],"产量预测","可解释性",{"code":33,"slug":34,"name":35,"path":36,"parent_code":37},[341],{"code":40,"slug":41,"name":42,"path":43,"parent_code":37},{"status":45,"locked":14,"version":46,"method":47,"evidence":343},{"candidates":344,"agricultural_evidence":51},[345,351,355],{"code":33,"score":118,"supported":51,"title":346,"abstract":348,"tags":350},[347],"computer vision",[349],"计算机视觉",[27],{"code":40,"score":120,"supported":51,"title":352,"abstract":353,"tags":354},[],[30,59],[337,30],{"code":65,"score":22,"supported":14,"title":356,"abstract":357,"tags":358},[],[],[186],[360,361],"农业人工智能 产量预测 可解释性 智慧农业","农业人工智能 产量预测","农业人工智能产量预测可解释性智慧农业-5570","10.5281\u002Fzenodo.23545079",{"doi":363,"openalex_id":365,"authors":366,"venue":328,"cited_by_count":96,"oa_url":325,"card":369,"direction":101,"ingested_from":103},"W7223114002",[367],{"name":368,"orcid":9},"Kadiatou Barry",{"tldr":370,"method":371,"finding":372,"direction":101,"opportunity":373},"推导卫星云掩膜缺口下反事实解释可靠性的理论界，并评估其对合作社储粮决策的影响。","基于Lipschitz连续性假设与掩膜扰动分析推导可靠性界，并用2000个设计单","缺口比例上升使可靠性非线性下降；ICC=0.64表明需合作社层级聚合以保持解释稳定。","可研究不同作物区云掩膜统计特性下的自适应解释暂停阈值与聚合策略。","2026-10-22T11:20:39.240791Z",{"id":376,"title":377,"url":378,"summary":379,"summary_zh":380,"content":9,"source_name":381,"source_url":378,"published_at":382,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":383,"score_detail":384,"sources":387,"tags":389,"primary_topic":392,"secondary_topics":393,"classification":396,"search_phrases":419,"slug":422,"view_count":96,"doi":423,"paper":424,"created_at":444},5006,"Mapping Soil Salinity in Arid Agricultural Regions Using Multi-Source Remote Sensing and a Long-Range Dependency-Enhanced Deep Learning Framework","https:\u002F\u002Fdoi.org\u002F10.3390\u002Fland15101966","Soil salinity estimation across arid agricultural regions is challenging due to pronounced environmental heterogeneity and complex dependencies among multi-source variables. This study proposes a Long-Range Dependency-Enhanced Network (LRD-Net) for regional-scale soil salinity estimation in multiple agricultural oasis regions using multi-source remote sensing and environmental covariates. A dataset comprising 180 field samples from three agricultural oasis regions in Xinjiang, China, was constructed with 25 covariates derived from spectral, topographic, climatic, and soil property data. LRD-Net consists of three stages: a multi-scale feature fusion backbone based on depthwise separable convolutions and cross-scale interaction; a Long-Range Dependency Enhancement Module (LDEM) with an asymmetric dual-branch architecture that captures forward and reverse-aware dependencies through reverse alignment and cross-directional fusion; and multi-scale attention refinement with a regression head for continuous salinity estimation. Under the region-stratified random-split evaluation, LRD-Net achieved an R2 of 0.905, an RMSE of 1.207 g·kg−1, and an MAE of 0.868 g·kg−1. Performance decreased under more spatially independent validation, with an R2 of 0.874 under spatial cross-validation and LORO test R2 values ranging from 0.739 to 0.801, indicating that prediction across spatially and geographically independent samples is more challenging than region-stratified random-split evaluation. The proposed framework effectively integrates multi-source information and captures complementary long-range dependencies, providing a potential approach for regional-scale soil salinity estimation within the studied arid agricultural oasis environments and supporting soil salinization monitoring and sustainable land management in the study region.","干旱农业区域土壤盐分估算面临显著的环境异质性和多源变量间复杂依赖关系的挑战。本研究提出了一种长程依赖增强网络（LRD-Net），利用多源遥感与环境协变量对多个农业绿洲区域进行区域尺度土壤盐分估算。研究构建了包含中国新疆三个农业绿洲区域180个野外样本的数据集，并衍生出25个协变量，涵盖光谱、地形、气候和土壤属性数据。LRD-Net由三个阶段组成：基于深度可分离卷积和跨尺度交互的多尺度特征融合骨干网络；具有非对称双分支架构的长程依赖增强模块（LDEM），通过反向对齐和跨方向融合捕获前向与反向感知依赖关系；以及多尺度注意力细化与回归头，用于连续盐分估算。在区域分层随机划分评估下，LRD-Net取得了R²为0.905、RMSE为1.207 g·kg⁻¹、MAE为0.868 g·kg⁻¹的性能。在空间独立性更强的验证下性能有所下降，空间交叉验证R²为0.874，留一区域测试R²范围为0.739至0.801，表明跨空间和地理独立样本的预测比区域分层随机划分评估更具挑战性。所提出的框架有效整合了多源信息并捕获了互补的长程依赖关系，为研究区干旱农业绿洲环境下的区域尺度土壤盐分估算提供了潜在方法，并支持研究区土壤盐渍化监测与可持续土地管理。","Land","2026-10-14T00:00:00Z",72,{"impact":118,"substance":385,"depth":332,"authority":181,"freshness":21,"relevant":22,"comment":386},21,"方法新颖、数据扎实的遥感土壤盐渍化估算研究，对干旱区盐碱地监测与治理有参考价值，但属细分领域学术进展，公共影响有限。",[388],{"name":381,"url":378},[186,27,30,390,391],"土壤盐渍化","新疆农业",{"code":40,"slug":41,"name":42,"path":43,"parent_code":37},[394,395],{"code":189,"slug":190,"name":191,"path":192,"parent_code":193},{"code":33,"slug":34,"name":35,"path":36,"parent_code":37},{"status":45,"locked":14,"version":46,"method":47,"evidence":397},{"candidates":398,"agricultural_evidence":51},[399,403,407,411,415],{"code":40,"score":119,"supported":51,"title":400,"abstract":401,"tags":402},[59],[30,59],[30],{"code":189,"score":20,"supported":51,"title":404,"abstract":405,"tags":406},[190],[203,190],[],{"code":33,"score":21,"supported":51,"title":408,"abstract":409,"tags":410},[53],[],[27],{"code":227,"score":143,"supported":51,"title":412,"abstract":413,"tags":414},[],[230],[],{"code":65,"score":22,"supported":14,"title":416,"abstract":417,"tags":418},[],[],[186],[420,421],"新疆 农业绿洲 土壤盐渍化 遥感","LRD-Net 土壤盐分 估算","新疆农业绿洲土壤盐渍化遥感-5006","10.3390\u002Fland15101966",{"doi":423,"openalex_id":425,"authors":426,"venue":381,"cited_by_count":96,"oa_url":378,"card":439,"direction":101,"ingested_from":103},"W7221816183",[427,429,432,434,436],{"name":428,"orcid":9},"Mingjie Qian",{"name":430,"orcid":431},"Liangjun Wang","https:\u002F\u002Forcid.org\u002F0009-0006-8718-8326",{"name":433,"orcid":9},"Peng Wang",{"name":435,"orcid":9},"Hangyuan Liu",{"name":437,"orcid":438},"Weitao Chen","https:\u002F\u002Forcid.org\u002F0000-0002-6272-1618",{"tldr":440,"method":441,"finding":442,"direction":101,"opportunity":443},"提出LRD-Net深度学习框架，融合多源遥感与环境协变量，估算新疆干旱绿洲区土壤盐分。","构建180个样点、25个协变量的数据集，设计含长程依赖增强模块的多尺度网络。","区域分层随机划分下R²达0.905，但空间独立验证R²降至0.739-0.874，跨区域预测更难。","可探索跨区域迁移学习与不确定性量化，提升空间独立场景下盐分制图泛化能力。","2026-10-15T23:30:29.667342Z",{"id":446,"title":447,"url":448,"summary":449,"summary_zh":450,"content":9,"source_name":451,"source_url":448,"published_at":382,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":452,"score_detail":453,"sources":455,"tags":457,"primary_topic":459,"secondary_topics":460,"classification":468,"search_phrases":513,"slug":516,"view_count":22,"doi":517,"paper":518,"created_at":540},4988,"Hierarchical cross-modal spatiotemporal attention network for crop yield prediction in complex Himalayan agroecosystems","https:\u002F\u002Fdoi.org\u002F10.3389\u002Ffpls.2026.1933033","Introduction Accurate crop yield prediction is essential for precision agriculture, climate-resilient farm management, and timely decision-making. However, yield prediction remains challenging because of complex, nonlinear, and multi-scale interactions among crop physiology, soil conditions, weather variability, and management practices. Conventional statistical, machine learning, and deep learning approaches, including convolutional neural networks (CNNs), recurrent neural networks (RNNs), and long short-term memory (LSTM) networks, often have limitations in capturing long-range temporal dependencies, hierarchical spatial relationships, heterogeneous data modalities, and predictive uncertainty. Methods This study proposes a Hierarchical Cross-Modal Spatio-Temporal Attention Network (HCM-STAN) for crop yield prediction using a unified attention-based framework for integrating heterogeneous agro-environmental data. The architecture incorporates four principal components: (i) hierarchical spatio-temporal attention operating across tree, block, orchard, and regional scales; (ii) a cross-modal attention module for adaptive fusion of IoT-based soil sensor observations, weather data, plant physiological measurements, and drone-based multispectral imagery; (iii) a soil-aware attention head that explicitly models soil memory, hysteresis effects, and root-zone dynamics; and (iv) an uncertainty-aware prediction module combining Monte Carlo dropout and quantile regression to generate calibrated prediction intervals. The framework was evaluated using a multi-source dataset collected from Himalayan apple orchards during 2023–2026, comprising 1,247 monitored trees across three commercial orchards. Results The proposed HCM-STAN achieved an R 2 of 0.89 for four-week-ahead crop yield prediction, representing a 15.6% improvement over the CNN-LSTM baseline. Furthermore, zero-shot transfer to mango orchards achieved an R 2 of 0.67, corresponding to approximately 75% of the performance obtained under full training. Discussion The results demonstrate that integrating hierarchical spatio-temporal representation learning, cross-modal attention, soil-aware modelling, and uncertainty estimation can effectively address the heterogeneous and dynamic nature of crop yield prediction in complex mountain agroecosystems. The observed zero-shot transfer performance further indicates the potential of HCM-STAN to generalize across crop systems without crop-specific retraining. Overall, the proposed framework provides a promising foundation for reliable, uncertainty-aware yield forecasting and can support precision agriculture, climate-smart farm management, and data-driven decision-making in heterogeneous agroecosystems.","引言 准确的作物产量预测对于精准农业、气候适应性农场管理以及及时决策至关重要。然而，由于作物生理、土壤条件、天气变异和管理措施之间复杂、非线性且多尺度的相互作用，产量预测仍然具有挑战性。传统的统计方法、机器学习以及深度学习方法，包括卷积神经网络（CNN）、循环神经网络（RNN）和长短期记忆（LSTM）网络，在捕捉长程时间依赖关系、层次化空间关系、异构数据模态以及预测不确定性方面往往存在局限。方法 本研究提出了一种用于作物产量预测的层次化跨模态时空注意力网络（HCM-STAN），采用统一的基于注意力的框架来整合异构农业环境数据。该架构包含四个主要组成部分：（i）在树、地块、果园和区域尺度上运行的层次化时空注意力；（ii）用于自适应融合基于物联网的土壤传感器观测、天气数据、植物生理测量和无人机多光谱影像的跨模态注意力模块；（iii）显式建模土壤记忆、滞后效应和根区动态的土壤感知注意力头；（iv）结合蒙特卡洛 dropout 和分位数回归以生成校准预测区间的不确定性感知预测模块。该框架使用2023—2026年在喜马拉雅苹果园采集的多源数据集进行了评估，该数据集包含三个商业果园中1，247棵监测树木。结果 所提出的 HCM-STAN 在提前四周作物产量预测中达到了0.89的 R²，相较于 CNN-LSTM 基线提升了15.6%。此外，零样本迁移至芒果园达到了0.67的 R²，相当于完整训练下所获性能的约75%。讨论 结果表明，整合层次化时空表征学习、跨模态注意力、土壤感知建模和不确定性估计，能够有效应对复杂山地农业生态系统中作物产量预测的异构性和动态性。观察到的零样本迁移性能进一步表明，HCM-STAN 具有在无需作物特定再训练的情况下跨作物系统泛化的潜力。总体而言，该 p","Frontiers in Plant Science",81,{"impact":17,"substance":18,"depth":17,"authority":20,"freshness":21,"relevant":22,"comment":454},"提出分层跨模态时空注意力网络，融合物联网土壤传感、气象、生理与无人机多光谱数据，苹果园四周产量预测R²达0.89并实现芒果园零样本迁移，方法新颖、数据扎实，对智慧农业产量预报有参考价值。",[456],{"name":451,"url":448},[186,27,337,458,30],"多模态融合",{"code":33,"slug":34,"name":35,"path":36,"parent_code":37},[461,467],{"code":462,"slug":463,"name":464,"path":465,"parent_code":466},"B2","fruit","果树","\u002Ftopic\u002Fhorticulture\u002Ffruit","B",{"code":40,"slug":41,"name":42,"path":43,"parent_code":37},{"status":45,"locked":14,"version":46,"method":47,"evidence":469},{"candidates":470,"agricultural_evidence":51},[471,476,484,489,493,497,503,508],{"code":33,"score":472,"supported":51,"title":473,"abstract":474,"tags":475},11,[],[28,208,209,53],[27,458],{"code":462,"score":21,"supported":51,"title":477,"abstract":478,"tags":483},[],[479,480,481,482],"苹果","芒果","orchard","apple",[],{"code":40,"score":120,"supported":51,"title":485,"abstract":486,"tags":488},[],[487,61],"无人机多光谱",[337,30],{"code":65,"score":66,"supported":51,"title":490,"abstract":491,"tags":492},[],[29,69],[186],{"code":189,"score":106,"supported":51,"title":494,"abstract":495,"tags":496},[],[203,190],[],{"code":72,"score":106,"supported":51,"title":498,"abstract":499,"tags":502},[],[500,501],"物联网","IoT",[],{"code":149,"score":143,"supported":51,"title":504,"abstract":505,"tags":507},[],[506],"农业生态",[],{"code":227,"score":143,"supported":51,"title":509,"abstract":510,"tags":512},[],[511],"气候适应",[],[514,515],"喜马拉雅 苹果园 产量预测","HCM-STAN 作物产量预测","喜马拉雅苹果园产量预测-4988","10.3389\u002Ffpls.2026.1933033",{"doi":517,"openalex_id":519,"authors":520,"venue":451,"cited_by_count":96,"oa_url":448,"card":534,"direction":169,"ingested_from":103},"W7221795850",[521,524,526,528,530,532],{"name":522,"orcid":523},"Shabir Ahmad Mir","https:\u002F\u002Forcid.org\u002F0000-0003-2311-218X",{"name":525,"orcid":9},"Umarfarooque Momin",{"name":527,"orcid":9},"Sudeep Kumar E",{"name":529,"orcid":9},"Itfaq Ahmad Mir",{"name":531,"orcid":9},"Ab Rouf Khan",{"name":533,"orcid":9},"Ghulam Mohammad Mir",{"tldr":535,"method":536,"finding":537,"direction":538,"opportunity":539},"提出分层跨模态时空注意力网络HCM-STAN，融合多源农业数据预测喜马拉雅山区苹果产量。","用IoT土壤、气象、生理和无人机多光谱数据，结合分层时空与跨模态注意力及不确定性","四周前产量预测R²达0.89，比CNN-LSTM提升15.6%，零样本迁移芒果R²为0.67。","农业人工智能与决策模型","可探索跨作物零样本迁移的领域自适应机制，并验证模型在更多山地农业生态系统的泛化能力。","2026-10-15T23:30:08.750140Z"]