[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"item-2151":3},{"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,"view_count":32,"doi":33,"paper":34,"created_at":52},2151,"An adaptive deep learning framework for multi-temporal crop and drought stress monitoring in precision agriculture","https:\u002F\u002Fdoi.org\u002F10.1038\u002Fs41598-026-70304-z","Precision agriculture increasingly relies on data-driven methods to address the challenges of crop stress and drought monitoring under changing climatic conditions. Hence, a deep learning (DL) based innovative framework is required for assessing crop stress dynamics, aimed at improving agricultural productivity and ensuring regional food security, because conventional models often struggle to capture subtle spatial and spectral variations, resulting in limited accuracy and generalizability. To overcome these challenges, an innovative and adaptive DL-based approach has been proposed to utilize multi-temporal Sentinel-2 satellite images collected over Haldharmau village, Gonda District, Uttar Pradesh. The model effectively integrates local spectral and spatial details with broader contextual features, enabling crop classification and stress detection to support precision agriculture. The crop stress and drought maps have been generated with the help of a CNN+ViT model (i.e., A hybrid DL-based classification model) based classified images in conjunction with NDVI and NDWI images to evaluate drought-induced crop variability across 2023, 2024, and 2025. The monthly (January to June) stress maps derived from this process provide valuable insights into temporal patterns of crop exposure. The proposed framework offers quantitative insights into temporal and spatial patterns of crop vulnerability and resilience across semi‐arid agricultural landscapes. Therefore, the proposed framework advances precision agriculture modeling by integrating satellite images, deep learning, and phenological analysis, and is readily transferable to other drought-prone regions with analogous crop systems.","精准农业日益依赖数据驱动方法来应对气候变化条件下作物胁迫与干旱监测的挑战。因此，需要一种基于深度学习（DL）的创新框架来评估作物胁迫动态，以提高农业生产力并保障区域粮食安全，因为传统模型往往难以捕捉细微的空间和光谱变化，导致精度和泛化能力有限。为克服这些挑战，提出了一种创新且自适应的基于深度学习的方法，利用在北方邦戈达县哈尔达毛村采集的多时相Sentinel-2卫星影像。该模型有效整合了局部光谱与空间细节以及更广泛的上下文特征，实现了作物分类和胁迫检测，以支持精准农业。作物胁迫和干旱地图借助CNN+ViT模型（即基于混合深度学习的分类模型）生成的分类影像，结合NDVI和NDWI影像，评估了2023年、2024年和2025年干旱引起的作物变异性。通过该过程得到的逐月（1月至6月）胁迫地图为作物暴露的时间格局提供了有价值的见解。所提出的框架为半干旱农业景观中作物脆弱性和恢复力的时空格局提供了定量认识。因此，该框架通过整合卫星影像、深度学习和物候分析，推进了精准农业建模，并可随时推广至具有类似作物系统的其他干旱易发地区。",null,"Scientific Reports","2026-09-09T00:00:00Z","论文",10,false,76,{"impact":17,"substance":18,"depth":19,"authority":20,"freshness":21,"relevant":22,"comment":23},16,21,17,14,8,1,"基于CNN+ViT混合模型与Sentinel-2多时相影像的作物分类与干旱胁迫监测框架，方法新颖、数据跨三年，对旱区精准农业有参考价值。",[25],{"name":10,"url":6},[27,28,29,30,31],"智慧农业","农业人工智能","精准农业","遥感监测","作物干旱胁迫",0,"10.1038\u002Fs41598-026-70304-z",{"doi":33,"openalex_id":35,"authors":36,"venue":10,"cited_by_count":32,"oa_url":43,"card":44,"direction":50,"ingested_from":51},"W7212034680",[37,40],{"name":38,"orcid":39},"Gausiya Yasmeen","https:\u002F\u002Forcid.org\u002F0000-0002-7853-1376",{"name":41,"orcid":42},"Tasneem Ahmed","https:\u002F\u002Forcid.org\u002F0000-0003-2702-3168","https:\u002F\u002Fwww.nature.com\u002Farticles\u002Fs41598-026-70304-z_reference.pdf",{"tldr":45,"method":46,"finding":47,"direction":48,"opportunity":49},"提出CNN+ViT混合深度学习框架，利用多时相Sentinel-2影像监测作物分类与干旱胁迫。","CNN+ViT混合模型结合NDVI、NDWI，基于Sentinel-2多时相影像","框架可量化半干旱区作物时空脆弱性与恢复力，生成月度胁迫图。","农业遥感与作物表型","可探索多源遥感与时序模型融合，提升跨区域干旱胁迫预警的泛化能力。","智慧农业 \u002F 农业物联网","openalex","2026-09-11T23:30:13.461760Z"]