[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"item-4518":3,"related-4518":79},{"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":18,"tags":20,"primary_topic":26,"secondary_topics":32,"classification":33,"search_phrases":54,"slug":57,"view_count":15,"doi":58,"paper":59,"created_at":78},4518,"Knowledge Distillation for Hybrid CNN–Vision Transformer Architectures: A Systematic Review","https:\u002F\u002Fdoi.org\u002F10.5281\u002Fzenodo.23208859","Hybrid architectures that combine convolutional neural networks (CNNs) with Vision Transformers (ViTs) have become a dominant design pattern in computer vision, pairing the local inductive bias of convolution with the global relational reasoning of self-attention. This dual advantage comes at a computational cost that frequently excludes hybrid models from edge and mobile deployment, motivating a growing body of work that applies knowledge distillation (KD) either to compress hybrid CNN-ViT models directly or to transfer their heterogeneous knowledge into lightweight students of a different architecture family. Distilling from or into a hybrid model is fundamentally harder than distilling between architecturally homogeneous networks, because it requires reconciling two qualitatively different internal representations: spatially local convolutional feature maps and globally relational attention weights. This review synthesizes more than seventy studies published between 2015 and 2026 that address this problem. The literature is organized along three axes: the direction of cross-architecture knowledge transfer (CNN-to-hybrid, hybrid-to-CNN, hybrid-to-hybrid, and ViT-to-ViT within hybrid families), the type of knowledge transferred (response-based, feature-based, attention-based, and relation-based), and the application domain (general vision, medical imaging, remote sensing, object detection and autonomous driving, agriculture, and action recognition or robotics). A comparative synthesis table of representative methods is provided, and six persistent open challenges are identified: the absence of a unified theory of heterogeneous feature alignment, unexamined hyperparameter sensitivity in multi-component losses, domain-fragmented and incomparable evaluation protocols, scarce validation on physical edge hardware, shallow integration with complementary compression techniques, and underreporting of negative or boundary results. A concrete research agenda is proposed to address these gaps. The review is intended to orient researchers designing distillation pipelines for hybrid vision architectures and to consolidate a literature that currently spans disconnected, domain-specific silos.","将卷积神经网络（CNN）与视觉Transformer（ViT）相结合的混合架构已成为计算机视觉中的主流设计范式，它将卷积的局部归纳偏置与自注意力的全局关系推理能力结合在一起。这种双重优势伴随着计算成本，常常使混合模型难以部署于边缘和移动设备，从而推动了越来越多的工作，或直接压缩混合CNN-ViT模型，或将其异构知识迁移到不同架构家族的轻量级学生模型中。从混合模型中进行蒸馏或将其蒸馏到混合模型中，从根本上比在架构同质的网络之间进行蒸馏更为困难，因为这需要协调两种性质不同的内部表示：空间局部的卷积特征图和全局关系的注意力权重。本综述综合了2015年至2026年间发表的七十余项针对该问题的研究。文献沿三个维度进行组织：跨架构知识迁移的方向（CNN到混合、混合到CNN、混合到混合，以及混合家族内的ViT到ViT）、所迁移知识的类型（基于响应、基于特征、基于注意力和基于关系），以及应用领域（通用视觉、医学影像、遥感、目标检测与自动驾驶、农业，以及动作识别或机器人学）。文中提供了代表性方法的比较综合表，并指出了六个持续存在的开放挑战：缺乏异构特征对齐的统一理论、多组件损失中超参数敏感性未经检验、领域碎片化且不可比较的评估协议、在物理边缘硬件上验证稀缺、与互补压缩技术的整合较浅，以及对负面或边界结果报告不足。针对这些空白，提出了具体的研究议程。本综述旨在为设计混合视觉架构蒸馏流程的研究人员提供方向，并整合目前分散于互不关联的领域特定孤岛中的文献。",null,"International Comprehensive Technology and Science Journal","2026-10-07T00:00:00Z","论文",10,false,0,{"impact":15,"substance":15,"depth":15,"authority":15,"freshness":15,"relevant":15,"comment":17},"属计算机视觉通用方法综述，仅在遥感、农业等应用域被顺带提及，与三农及农业信息化主题无实质关联，不建议进入每日精选。",[19],{"name":10,"url":6},[21,22,23,24,25],"农业人工智能","计算机视觉","遥感","知识蒸馏","模型轻量化",{"code":27,"slug":28,"name":29,"path":30,"parent_code":31},"I2","agricultural-ai","农业AI与大模型","\u002Ftopic\u002Fdigital-agriculture\u002Fagricultural-ai","I",[],{"status":34,"locked":14,"version":35,"method":36,"evidence":37},"classified","2026-10-22.2","rules",{"candidates":38,"agricultural_evidence":41},[39,47],{"code":27,"score":40,"supported":41,"title":42,"abstract":43,"tags":46},12,true,[],[44,22,45],"目标检测","computer vision",[21,25,22],{"code":48,"score":49,"supported":41,"title":50,"abstract":51,"tags":53},"I3",7,[],[23,52],"remote sensing",[23],[55,56],"CNN Vision Transformer 知识蒸馏","混合视觉架构 轻量化 边缘部署","CNNVisionTransformer知识蒸馏-4518","10.5281\u002Fzenodo.23208859",{"doi":58,"openalex_id":60,"authors":61,"venue":10,"cited_by_count":15,"oa_url":6,"card":70,"direction":76,"ingested_from":77},"W7220993043",[62,64,66,68],{"name":63,"orcid":9},"Muhammad Adam",{"name":65,"orcid":9},"Abdullahi Ishaq",{"name":67,"orcid":9},"Suleiman Ibrahim",{"name":69,"orcid":9},"Ali Shuaibu Babaa",{"tldr":71,"method":72,"finding":73,"direction":74,"opportunity":75},"系统综述2015-2026年混合CNN-ViT架构的知识蒸馏方法，按迁移方向、知识类型和应用领域分类","综述70余篇文献，按迁移方向、知识类型、应用领域三轴组织并对比。","识别出异构特征对齐理论缺失、评估协议碎片化等六大开放挑战并提出研究议程。","农业人工智能与决策模型","农业领域混合模型蒸馏研究稀缺，可探索面向边缘设备的轻量化作物识别与表型分析。","农业遥感与作物表型","openalex","2026-10-09T23:30:28.968251Z",{"total":80,"page":81,"page_size":80,"items":82},6,1,[83,146,209,262,314,363],{"id":84,"title":85,"url":86,"summary":87,"summary_zh":88,"content":9,"source_name":89,"source_url":86,"published_at":90,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":91,"score_detail":92,"sources":98,"tags":100,"primary_topic":103,"secondary_topics":104,"classification":105,"search_phrases":124,"slug":127,"view_count":15,"doi":128,"paper":129,"created_at":145},4875,"YOLO-Based Leaf Disease Detection in Smart Agriculture: Evolution, Task Adaptation, and Deployment Challenges","https:\u002F\u002Fdoi.org\u002F10.3390\u002Felectronics15204699","Computer vision-based leaf disease detection has become an important technical approach in smart agriculture and agricultural inspection services. Recent advances in artificial intelligence (AI) and deep learning have provided effective solutions for automated leaf disease monitoring, and You Only Look Once (YOLO)-based detectors have attracted increasing attention for balancing detection accuracy and inference speed. However, leaf disease detection remains challenging because lesions often occupy small regions, exhibit subtle visual features, have blurred boundaries, and appear against complex backgrounds, with considerable variation under field conditions. To address these challenges, numerous studies have improved lightweight backbones, feature fusion mechanisms, attention modules, detection heads, loss functions, and training strategies. This critical narrative review examines YOLO-based leaf disease detection, organizing reported adaptations by task characteristics and interpreting their outcomes in relation to datasets, annotations, and evaluation conditions. Representative studies report gains in detection performance, model compactness, or measured inference speed, together with trade-offs that depend on the task and configuration. Nevertheless, several challenges remain, including limited dataset diversity, insufficient cross-crop and cross-environment generalization, inadequate consideration of annotation uncertainty, biased evaluation of lightweight models based only on parameter reduction, and the gap between benchmark performance and real-world deployment. Future research should focus on task-oriented optimization, robust visual perception, improved model generalization, and efficient deployment toward intelligent agricultural monitoring systems.","基于计算机视觉的叶片病害检测已成为智慧农业和农业巡检服务中的一项重要技术途径。近年来，人工智能（AI）与深度学习的进展为自动化叶片病害监测提供了有效解决方案，其中基于You Only Look Once（YOLO）的检测器因在检测精度与推理速度之间取得平衡而受到越来越多的关注。然而，叶片病害检测仍面临诸多挑战，因为病斑通常占据较小区域、具有细微的视觉特征、边界模糊，并且出现在复杂背景中，在田间条件下还存在相当大的变化。为应对这些挑战，大量研究对轻量级骨干网络、特征融合机制、注意力模块、检测头、损失函数和训练策略进行了改进。本批判性叙述综述考察了基于YOLO的叶片病害检测，按任务特征对已报道的改进进行组织，并结合数据集、标注和评估条件解释其结果。代表性研究报告了检测性能、模型紧凑性或实测推理速度方面的提升，同时也揭示了取决于任务和配置的权衡。尽管如此，仍存在若干挑战，包括数据集多样性有限、跨作物和跨环境泛化能力不足、对标注不确定性考虑不充分、仅基于参数减少对轻量级模型进行有偏评估，以及基准性能与实际部署之间的差距。未来研究应聚焦于面向任务的优化、鲁棒的视觉感知、改进的模型泛化能力以及面向智能农业监测系统的高效部署。","Electronics","2026-10-13T00:00:00Z",67,{"impact":40,"substance":93,"depth":94,"authority":95,"freshness":96,"relevant":81,"comment":97},18,16,13,8,"系统梳理YOLO在叶片病害检测中的任务适配与落地瓶颈，对农业AI研发有参考价值，但属综述类论文，产业影响有限。",[99],{"name":89,"url":86},[101,21,22,102,25],"智慧农业","病虫害检测",{"code":27,"slug":28,"name":29,"path":30,"parent_code":31},[],{"status":34,"locked":14,"version":35,"method":36,"evidence":106},{"candidates":107,"agricultural_evidence":41},[108,117],{"code":27,"score":109,"supported":41,"title":110,"abstract":112,"tags":116},20,[111],"YOLO",[113,22,114,115,45,111],"深度学习","deep learning","artificial intelligence",[21,25,102,22],{"code":118,"score":119,"supported":41,"title":120,"abstract":122,"tags":123},"I1",15,[121],"smart agriculture",[101,121],[101],[125,126],"YOLO 叶片病害检测","智慧农业 病害识别 模型部署","YOLO叶片病害检测-4875","10.3390\u002Felectronics15204699",{"doi":128,"openalex_id":130,"authors":131,"venue":89,"cited_by_count":15,"oa_url":86,"card":139,"direction":144,"ingested_from":77},"W7221646013",[132,135,137],{"name":133,"orcid":134},"Xinyi Sun","https:\u002F\u002Forcid.org\u002F0009-0007-5992-1516",{"name":136,"orcid":9},"Liyang Liu",{"name":138,"orcid":9},"Baoquan Yin",{"tldr":140,"method":141,"finding":142,"direction":74,"opportunity":143},"综述了基于YOLO的叶片病害检测的演进、任务适配与部署挑战。","关键方法：轻量骨干、特征融合、注意力、检测头、损失函数与训练策略改进。","YOLO检测性能提升但存在数据集单一、泛化不足、标注不确定与部署差距。","可研究跨作物跨环境泛化、标注不确定性建模及面向真实部署的轻量模型评估。","智慧农业 \u002F 农业物联网","2026-10-14T23:30:08.538550Z",{"id":147,"title":148,"url":149,"summary":150,"summary_zh":151,"content":9,"source_name":152,"source_url":149,"published_at":153,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":154,"score_detail":155,"sources":158,"tags":160,"primary_topic":163,"secondary_topics":164,"classification":165,"search_phrases":193,"slug":196,"view_count":15,"doi":197,"paper":198,"created_at":208},5588,"Certified Temporal Fusion of Artificial Intelligence Climate Analytics for Crop Failure Early Warning Under Sparse Agricultural Sensors: Error Bounds and Famine Relief Targeting","https:\u002F\u002Fdoi.org\u002F10.5281\u002Fzenodo.23545489","Sparse agricultural sensor networks across sub-Saharan Africa leave crop-failure early-warning systems with too little ground observation to support reliable famine relief targeting. The framework treats sensor sparsity as a missing-data problem with a known observation operator and propagates climate-analytics uncertainty through a temporal Kalman-Buys filter whose gain is constrained by a certified risk envelope. Using NASA POWER monthly reanalysis for Accra, Ghana (2000-2023, 312 observations), the analysis establishes a mean temperature of 26.521 C and a warming trend of 0.1622 C per decade, which calibrates the climate-analytics prior. The derived bounds show that relief-targeting error scales with the square root of sensor density and with the Lipschitz constant of the crop-response function, yielding an explicit sensor-placement rule for a target misclassification rate. The contribution is a certified, assumption-explicit bridge between climate analytics and famine relief targeting that states the conditions under which early warning is trustworthy.","撒哈拉以南非洲稀疏的农业传感器网络使得作物歉收预警系统缺乏足够的地面观测数据，难以支撑可靠的饥荒救援目标定位。该框架将传感器稀疏性视为具有已知观测算子的缺失数据问题，并通过时间Kalman-Buys滤波器传播气候分析不确定性，该滤波器的增益受经认证的风险包络约束。利用NASA POWER对加纳阿克拉的月度再分析数据（2000—2023年，312个观测值），分析确定了平均气温为26.521°C、每十年0.1622°C的增暖趋势，据此校准了气候分析先验。推导出的界限表明，救援目标定位误差随传感器密度的平方根以及作物响应函数的Lipschitz常数缩放，从而为目标误分类率给出了明确的传感器布设规则。该研究的贡献在于构建了一个经认证的、假设明确的气候分析与饥荒救援目标定位之间的桥梁，阐明了早期预警可信赖的条件。","Zenodo (CERN European Organization for Nuclear Research)","2026-10-21T00:00:00Z",74,{"impact":40,"substance":156,"depth":93,"authority":40,"freshness":13,"relevant":81,"comment":157},22,"该研究提出一种带误差界认证的时序融合方法，用于稀疏传感器条件下的作物歉收预警与饥荒救援瞄准，方法新颖、结论明确，但案例局限于加纳阿克拉，产业影响有限，适合作为技术前沿条目收录。",[159],{"name":152,"url":149},[101,21,161,23,162],"粮食安全","气候分析",{"code":27,"slug":28,"name":29,"path":30,"parent_code":31},[],{"status":34,"locked":14,"version":35,"method":36,"evidence":166},{"candidates":167,"agricultural_evidence":41},[168,173,180,185,189],{"code":27,"score":169,"supported":41,"title":170,"abstract":171,"tags":172},9,[115],[],[21],{"code":174,"score":175,"supported":41,"title":176,"abstract":177,"tags":179},"I5",3,[],[178],"农业传感",[],{"code":181,"score":81,"supported":14,"title":182,"abstract":183,"tags":184},"A8",[],[],[161],{"code":118,"score":81,"supported":14,"title":186,"abstract":187,"tags":188},[],[],[101],{"code":48,"score":81,"supported":14,"title":190,"abstract":191,"tags":192},[],[],[23],[194,195],"NASA POWER 加纳 气候分析","作物歉收 早期预警 传感器","NASAPOWER加纳气候分析-5588","10.5281\u002Fzenodo.23545489",{"doi":197,"openalex_id":199,"authors":200,"venue":152,"cited_by_count":15,"oa_url":149,"card":203,"direction":74,"ingested_from":77},"W7223111766",[201],{"name":202,"orcid":9},"Fatima-Zahra El Khattabi",{"tldr":204,"method":205,"finding":206,"direction":76,"opportunity":207},"提出带误差界的时间融合框架，在稀疏传感器下用气候分析预警作物歉收。","卡尔曼-布西滤波+NASA POWER再分析数据，风险包络约束增益。","救济瞄准误差随传感器密度平方根和作物响应Lipschitz常数缩放，给出布点规则。","可将该认证误差界框架扩展到多作物多区域，结合卫星遥感弥补地面传感器稀疏。","2026-10-22T11:21:02.682497Z",{"id":210,"title":211,"url":212,"summary":213,"summary_zh":214,"content":9,"source_name":152,"source_url":212,"published_at":153,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":215,"score_detail":216,"sources":219,"tags":221,"primary_topic":224,"secondary_topics":225,"classification":226,"search_phrases":242,"slug":245,"view_count":15,"doi":246,"paper":247,"created_at":261},5587,"Calibration-Shift Inference for Computer Vision Agriculture Yield Models Under Retrofit Sensor Replacement: An Artificial Intelligence Post-Deployment Recalibration Guarantee","https:\u002F\u002Fdoi.org\u002F10.5281\u002Fzenodo.23544195","Computer vision yield models deployed in agricultural fields are routinely paired with retrofit sensor packages that replace or supplement the original imaging hardware, and the resulting calibration shift can silently invalidate the model's predictive guarantees. The analysis characterises the operating characteristics of the design under a declared variance structure with between-cluster standard deviation 6.0, residual standard deviation 4.5 and intraclass correlation 0.64, using 2000 replications per design cell across cluster counts of 8, 16, 24 and 32 and effect sizes of 3, 5 and 8 percentage points. The design attains 80% power at a 5 percentage point effect with 54 clusters. The paper states a formal guarantee: under stated assumptions on the stability of the calibration map, the before-after-control-impact estimator yields a consistent and asymptotically normal estimate of the calibration shift, and the associated confidence interval covers the true shift at the nominal rate. The contribution is a design-theoretic account of how retrofit-induced calibration shift can be detected and bounded without observational data, together with the sample-size requirements that follow.","部署在农田中的计算机视觉产量模型通常与改装传感器组件配合使用，这些组件用于替换或补充原始成像硬件，由此产生的校准偏移可能悄然使模型的预测保证失效。该分析在声明的方差结构下刻画了设计的工作特性，其中组间标准差为6.0、残差标准差为4.5、组内相关系数为0.64，每个设计单元采用2000次重复，聚类数分别为8、16、24和32，效应量为3、5和8个百分点。该设计在54个聚类、5个百分点效应下达到80%的检验力。论文给出了一项形式化保证：在关于校准映射稳定性的既定假设下，前后对照影响（before-after-control-impact）估计量给出校准偏移的一致且渐近正态的估计，且相关置信区间以名义覆盖率覆盖真实偏移。其贡献在于从设计理论角度说明了如何在没有观测数据的情况下检测并界定改装引起的校准偏移，以及随之而来的样本量要求。",65,{"impact":96,"substance":93,"depth":217,"authority":40,"freshness":13,"relevant":81,"comment":218},17,"方法学扎实、统计设计严谨的预印本论文，但偏理论设计层面，缺乏田间实证数据，产业影响有限，适合作为技术参考而非每日精选头条。",[220],{"name":152,"url":212},[101,21,22,222,223],"产量预测","传感器校准",{"code":27,"slug":28,"name":29,"path":30,"parent_code":31},[],{"status":34,"locked":14,"version":35,"method":36,"evidence":227},{"candidates":228,"agricultural_evidence":41},[229,234,238],{"code":27,"score":230,"supported":41,"title":231,"abstract":232,"tags":233},24,[115,45],[22,45],[21,22],{"code":118,"score":81,"supported":14,"title":235,"abstract":236,"tags":237},[],[],[101],{"code":48,"score":81,"supported":14,"title":239,"abstract":240,"tags":241},[],[],[222],[243,244],"计算机视觉 产量预测 传感器校准","农业人工智能 传感器校准 计算机视觉 产量预测","计算机视觉产量预测传感器校准-5587","10.5281\u002Fzenodo.23544195",{"doi":246,"openalex_id":248,"authors":249,"venue":152,"cited_by_count":15,"oa_url":212,"card":256,"direction":74,"ingested_from":77},"W7223104380",[250,252,254],{"name":251,"orcid":9},"Tigist Alemu Gebremariam",{"name":253,"orcid":9},"Hawi Tesfaye Desta",{"name":255,"orcid":9},"Mulugeta Bekele Wolde",{"tldr":257,"method":258,"finding":259,"direction":74,"opportunity":260},"提出改造传感器替换后视觉产量模型校准漂移的检测与置信区间保证方法。","用前后对照影响估计量，基于方差结构做2000次重复的模拟设计分析。","5个百分点效应下54个集群可达80%检验力，估计量一致且渐近正态。","可将该设计理论扩展到多传感器异构替换与在线校准监测，填补真实田间验证空白。","2026-10-22T11:21:01.554995Z",{"id":263,"title":264,"url":265,"summary":266,"summary_zh":267,"content":9,"source_name":152,"source_url":265,"published_at":153,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":268,"score_detail":269,"sources":271,"tags":273,"primary_topic":276,"secondary_topics":277,"classification":278,"search_phrases":294,"slug":297,"view_count":15,"doi":298,"paper":299,"created_at":313},5584,"Distributionally-Robust Evaluation of Computer Vision Agriculture Disease Classifiers Under Climate-Driven Pathogen Range Shifts: An Artificial Intelligence Worst-Case Generalization Bound","https:\u002F\u002Fdoi.org\u002F10.5281\u002Fzenodo.23544150","Computer vision classifiers for crop disease are evaluated almost exclusively on held-out samples drawn from the same distribution as their training data, yet climate-driven pathogen range shifts mean that deployment environments are systematically out of distribution. The framework treats the deployment distribution as an adversarial perturbation of the training distribution within a Wasserstein ball whose radius is calibrated to observed climate displacement. Using NASA POWER monthly reanalysis for Harare, Zimbabwe (2000-2023, 312 monthly observations), the analysis estimates a mean air temperature of 18.914 C, a standard deviation of 3.107 C, and a warming trend of 0.2001 C per decade. These estimates anchor the radius of the ambiguity set and the resulting bound. The bound decomposes worst-case risk into an empirical risk term, a Wasserstein radius term scaled by the Lipschitz constant of the classifier, and a concentration term. Under plausible Lipschitz constants for convolutional backbones, the radius term dominates the empirical risk term once the climate displacement exceeds a threshold that the Harare record already approaches. The contribution is a formal evaluation criterion that replaces the i. i. d. assumption with a climate-calibrated ambiguity set, together with the analytical machinery needed to compute it.","作物病害的计算机视觉分类器几乎完全在与训练数据同分布的留出样本上进行评估，然而气候驱动的病原体分布范围变化意味着部署环境系统性地处于分布之外。该框架将部署分布视为训练分布在Wasserstein球内的对抗性扰动，球的半径根据观测到的气候位移进行校准。利用NASA POWER对津巴布韦哈拉雷的月度再分析数据（2000—2023年，312个月度观测值），分析估计平均气温为18.914°C，标准差为3.107°C，每十年升温趋势为0.2001°C。这些估计值锚定了模糊集的半径及由此得到的界。该界将最坏情况风险分解为经验风险项、由分类器Lipschitz常数缩放后的Wasserstein半径项以及集中项。在卷积骨干网络合理的Lipschitz常数下，一旦气候位移超过哈拉雷记录已接近的阈值，半径项便主导经验风险项。本文的贡献在于提出了一种形式化评估准则，以气候校准的模糊集替代独立同分布（i.i.d.）假设，并提供了计算该准则所需的分析工具。",75,{"impact":40,"substance":156,"depth":93,"authority":95,"freshness":13,"relevant":81,"comment":270},"方法新颖、数据具体，但属方法论预印论文，产业影响有限，适合作为专业精选而非头条。",[272],{"name":152,"url":265},[101,21,22,274,275],"作物病害","气候变化",{"code":27,"slug":28,"name":29,"path":30,"parent_code":31},[],{"status":34,"locked":14,"version":35,"method":36,"evidence":279},{"candidates":280,"agricultural_evidence":41},[281,285,290],{"code":27,"score":230,"supported":41,"title":282,"abstract":283,"tags":284},[115,45],[22,45],[21,22],{"code":286,"score":81,"supported":14,"title":287,"abstract":288,"tags":289},"H7",[],[],[275],{"code":118,"score":81,"supported":14,"title":291,"abstract":292,"tags":293},[],[],[101],[295,296],"NASA POWER 哈拉雷 气温","Wasserstein 分布鲁棒 病害分类","NASAPOWER哈拉雷气温-5584","10.5281\u002Fzenodo.23544150",{"doi":298,"openalex_id":300,"authors":301,"venue":152,"cited_by_count":15,"oa_url":265,"card":308,"direction":74,"ingested_from":77},"W7223087892",[302,304,306],{"name":303,"orcid":9},"Kossi Afeli",{"name":305,"orcid":9},"Akouvi Amegashie",{"name":307,"orcid":9},"Komlan Sélom Dossou",{"tldr":309,"method":310,"finding":311,"direction":74,"opportunity":312},"提出气候校准的分布鲁棒评估准则，替代i.i.d.假设评价作物病害视觉分类器。","用Wasserstein球建模部署分布偏移，以NASA POWER哈拉雷气温数据","气候位移超阈值后，半径项主导经验风险，现有评估会高估模型泛化能力。","可将气候校准的鲁棒评估扩展到多作物多区域，并研究适应分布偏移的鲁棒训练方法。","2026-10-22T11:21:00.036937Z",{"id":315,"title":316,"url":317,"summary":318,"summary_zh":319,"content":9,"source_name":152,"source_url":317,"published_at":153,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":91,"score_detail":320,"sources":322,"tags":324,"primary_topic":326,"secondary_topics":327,"classification":332,"search_phrases":347,"slug":350,"view_count":15,"doi":351,"paper":352,"created_at":362},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。",{"impact":96,"substance":109,"depth":217,"authority":40,"freshness":13,"relevant":81,"comment":321},"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。",[323],{"name":152,"url":317},[101,21,222,23,325],"可解释性",{"code":27,"slug":28,"name":29,"path":30,"parent_code":31},[328],{"code":48,"slug":329,"name":330,"path":331,"parent_code":31},"remote-sensing","遥感与农情监测","\u002Ftopic\u002Fdigital-agriculture\u002Fremote-sensing",{"status":34,"locked":14,"version":35,"method":36,"evidence":333},{"candidates":334,"agricultural_evidence":41},[335,339,343],{"code":27,"score":40,"supported":41,"title":336,"abstract":337,"tags":338},[45],[22],[21],{"code":48,"score":96,"supported":41,"title":340,"abstract":341,"tags":342},[],[23,52],[222,23],{"code":118,"score":81,"supported":14,"title":344,"abstract":345,"tags":346},[],[],[101],[348,349],"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":351,"openalex_id":353,"authors":354,"venue":152,"cited_by_count":15,"oa_url":317,"card":357,"direction":76,"ingested_from":77},"W7223114002",[355],{"name":356,"orcid":9},"Kadiatou Barry",{"tldr":358,"method":359,"finding":360,"direction":76,"opportunity":361},"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":364,"title":365,"url":366,"summary":367,"summary_zh":368,"content":9,"source_name":369,"source_url":366,"published_at":153,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":370,"score_detail":371,"sources":373,"tags":375,"primary_topic":378,"secondary_topics":384,"classification":390,"search_phrases":448,"slug":451,"view_count":15,"doi":452,"paper":453,"created_at":476},5564,"Dynamic thermal imaging for crop water stress detection across phenological stages: Enhancing irrigation scheduling precision","https:\u002F\u002Fdoi.org\u002F10.14719\u002Fpst.17728","Thermal remote sensing and derived indices such as the crop water stress index (CWSI) can assess crop water status and support precision irrigation. However, most operational and artificial intelligence (AI)-based approaches rely on fixed, season-long thresholds to interpret canopy–air temperature differences and often overlook the influence of crop phenology. This review synthesises evidence from grapevine, wheat, maize, potato and several other crops on how developmental stage affects the non-water-stressed baseline (NWSB), CWSI thresholds and irrigation-scheduling algorithms. It covers CWSI theory and estimation practices; crop-specific, phenology-dependent behaviour of baselines and thresholds; a comparison of threshold-based, machine learning, Internet of Things (IoT), digital twin and Food and Agriculture Organization (FAO)-56 approaches with respect to crop phenology; and a conceptual framework for phenology-aware scheduling. Across the reviewed crops, baseline behaviour generally varied systematically across the growing season, the relationship between CWSI and crop water status generally strengthened after canopy closure and CWSI measured during reproductive stages was generally most closely related to yield. A quantitative audit of 56 AI-, thermal-imaging-, or CWSI-based studies found that 78.6 % either did not explicitly incorporate crop phenology or relied on day of year as a temporal proxy, whereas only 5.4 % explicitly encoded phenological stage. The proposed framework considers phenology explicitly in baseline estimation, threshold setting and data fusion or machine learning for irrigation scheduling; the framework will require multi-year, multi-site field validation before operational use. Future research should prioritise standardised phenology definitions and long-term, multi-location studies to establish and validate stage-specific parameters.","热红外遥感及其衍生的作物水分胁迫指数（crop water stress index, CWSI）等指标可评估作物水分状况并支持精准灌溉。然而，大多数业务化方法和基于人工智能（artificial intelligence, AI）的方法依赖固定的全生育期阈值来解释冠层—空气温差，且往往忽视作物物候的影响。本文综述了葡萄、小麦、玉米、马铃薯及其他若干作物在发育阶段如何影响非水分胁迫基线（non-water-stressed baseline, NWSB）、CWSI阈值和灌溉调度算法的相关证据。综述内容涵盖CWSI理论与估算实践；基线和阈值的作物特异性、物候依赖性行为；基于阈值、机器学习、物联网（Internet of Things, IoT）、数字孪生及联合国粮农组织（Food and Agriculture Organization, FAO）-56等方法在作物物候方面的比较；以及物候感知调度的概念框架。在所综述的作物中，基线行为在整个生长季通常呈系统性变化，CWSI与作物水分状况之间的关系在冠层封垄后总体增强，且在生殖阶段测得的CWSI与产量的关系总体最为密切。对56项基于AI、热成像或CWSI的研究进行的定量审查发现，78.6%的研究要么未明确纳入作物物候，要么依赖年积日作为时间代理变量，而仅5.4%的研究明确编码了物候阶段。所提出的框架在基线估算、阈值设定以及用于灌溉调度的数据融合或机器学习中明确考虑物候；该框架在投入业务应用前尚需多年、多站点的田间验证。未来研究应优先建立标准化的物候定义，并开展长期、多地点研究以建立和验证阶段特异性参数。","Plant Science Today",79,{"impact":93,"substance":156,"depth":93,"authority":40,"freshness":169,"relevant":81,"comment":372},"系统综述热成像与CWSI在物候期感知灌溉中的应用，含56项研究量化审计，方法学价值突出，但属综述类论文，影响偏学术层面。",[374],{"name":369,"url":366},[101,21,23,376,377],"作物水分胁迫","精准灌溉",{"code":379,"slug":380,"name":381,"path":382,"parent_code":383},"H3","irrigation","水资源与灌溉","\u002Ftopic\u002Fenvironment\u002Firrigation","H",[385,386],{"code":27,"slug":28,"name":29,"path":30,"parent_code":31},{"code":174,"slug":387,"name":388,"path":389,"parent_code":31},"agricultural-iot","物联网传感与边缘计算","\u002Ftopic\u002Fdigital-agriculture\u002Fagricultural-iot",{"status":34,"locked":14,"version":35,"method":36,"evidence":391},{"candidates":392,"agricultural_evidence":41},[393,398,404,411,417,421,428,435,442],{"code":379,"score":119,"supported":41,"title":394,"abstract":395,"tags":397},[380],[396,380],"灌溉",[377],{"code":27,"score":13,"supported":41,"title":399,"abstract":400,"tags":403},[],[401,402,115],"机器学习","machine learning",[21],{"code":174,"score":169,"supported":41,"title":405,"abstract":406,"tags":410},[],[407,408,409],"物联网","internet of things","IoT",[],{"code":118,"score":49,"supported":41,"title":412,"abstract":413,"tags":416},[],[414,415],"数字孪生","digital twin",[101],{"code":48,"score":49,"supported":41,"title":418,"abstract":419,"tags":420},[],[23,52],[23],{"code":422,"score":80,"supported":41,"title":423,"abstract":424,"tags":427},"A2",[],[425,426],"小麦","wheat",[],{"code":429,"score":80,"supported":41,"title":430,"abstract":431,"tags":434},"A3",[],[432,433],"玉米","maize",[],{"code":436,"score":80,"supported":41,"title":437,"abstract":438,"tags":441},"A5",[],[439,440],"马铃薯","potato",[],{"code":443,"score":175,"supported":41,"title":444,"abstract":445,"tags":447},"B2",[],[446],"葡萄",[],[449,450],"热成像 作物水分胁迫","CWSI 灌溉调度","热成像作物水分胁迫-5564","10.14719\u002Fpst.17728",{"doi":452,"openalex_id":454,"authors":455,"venue":369,"cited_by_count":15,"oa_url":366,"card":471,"direction":144,"ingested_from":77},"W7223088921",[456,459,462,465,468],{"name":457,"orcid":458},"R M Jeya","https:\u002F\u002Forcid.org\u002F0009-0001-3621-8533",{"name":460,"orcid":461},"P. C. Prabu","https:\u002F\u002Forcid.org\u002F0000-0002-7237-9481",{"name":463,"orcid":464},"K P Ragunath","https:\u002F\u002Forcid.org\u002F0000-0002-7851-4437",{"name":466,"orcid":467},"P. Kannan","https:\u002F\u002Forcid.org\u002F0000-0001-7003-3535",{"name":469,"orcid":470},"P. M. Shanmugam","https:\u002F\u002Forcid.org\u002F0000-0002-6706-8419",{"tldr":472,"method":473,"finding":474,"direction":76,"opportunity":475},"综述物候期对热成像CWSI阈值的影响，提出物候感知灌溉调度框架。","综述葡萄、小麦、玉米等作物的热遥感CWSI研究，并审计56项AI\u002F热成像研究。","78.6%研究未显式纳入物候，仅5.4%编码物候期；CWSI与产量在生殖期关联最强。","可开展多站点多年物候期特异性CWSI阈值标定与标准化物候定义研究。","2026-10-22T11:20:24.526683Z"]