[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"item-3503":3,"related-3503":58},{"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,"search_phrases":32,"slug":35,"view_count":36,"doi":37,"paper":38,"created_at":57},3503,"Adversarial Patch and Camouflage Attacks on Aerial Object Detection: A Geometry-Aware Review","https:\u002F\u002Fdoi.org\u002F10.3390\u002Fs26196057","Uncrewed aerial vehicles increasingly rely on on-board deep object detectors for applications such as surveillance, delivery, agriculture, and traffic monitoring. The vulnerability of these detectors to adversarial patches has therefore become a practical safety concern. Research on adversarial patch and camouflage attacks against aerial object detection has expanded rapidly, yet existing reviews consider aerial systems only as one application domain among many and do not examine the unique geometric conditions of aerial imagery. This review surveys adversarial patch and camouflage attacks, together with the corresponding defense mechanisms, for aerial, UAV, and remote sensing object detection. The literature is organized using a taxonomy based on two dimensions: the physical medium of the perturbation and the application domain it targets. Existing methods are then compared with respect to physical robustness, evaluation datasets and detectors, threat models, and their treatment of viewpoint variation. A recurring observation across the literature is that viewpoint variation is represented primarily through object scale, while explicit modeling of viewing geometry remains limited. This trend is closely linked to the characteristics of the aerial datasets used for evaluation, which typically provide realistic imagery but limited geometric information. Consequently, questions concerning viewpoint-dependent placement, degradation, and naturalness remain only partially explored. The review concludes by identifying the major research gaps, future directions, and deployment challenges that are likely to shape the next generation of aerial adversarial attack and defense research.","无人驾驶飞行器（UAV）在监视、配送、农业和交通监测等应用中日益依赖机载深度目标检测器。因此，这些检测器对对抗性补丁的脆弱性已成为一个实际的安全问题。针对空中目标检测的对抗性补丁与伪装攻击研究迅速扩展，然而现有综述仅将空中系统视为众多应用领域之一，并未考察航空影像独特的几何条件。本综述考察了针对空中、无人机及遥感目标检测的对抗性补丁与伪装攻击以及相应的防御机制。文献按照基于两个维度的分类体系进行组织：扰动的物理介质及其所针对的应用领域。随后，现有方法在物理鲁棒性、评估数据集与检测器、威胁模型以及其对视角变化的处理方面进行了比较。文献中反复出现的一个观察是，视角变化主要通过目标尺度来表示，而对观察几何的显式建模仍然有限。这一趋势与用于评估的航空数据集的特征密切相关，这些数据集通常提供真实影像但几何信息有限。因此，关于视角依赖的放置、退化及自然性的问题仅得到部分探讨。本综述最后指出了可能塑造下一代空中对抗攻击与防御研究的主要研究空白、未来方向及部署挑战。",null,"Sensors","2026-09-24T00:00:00Z","论文",10,false,67,{"impact":17,"substance":18,"depth":19,"authority":20,"freshness":21,"relevant":22,"comment":23},8,20,17,13,9,1,"系统梳理航空与无人机目标检测的对抗补丁与伪装攻击及防御，指出视角几何建模不足的研究空白，对农业遥感与无人机应用安全有参考价值。",[25],{"name":10,"url":6},[27,28,29,30,31],"无人机","农业人工智能","目标检测","遥感","对抗攻击",[33,34],"无人机 对抗补丁 目标检测","航空图像 对抗攻击","无人机对抗补丁目标检测-3503",0,"10.3390\u002Fs26196057",{"doi":37,"openalex_id":39,"authors":40,"venue":10,"cited_by_count":36,"oa_url":6,"card":50,"direction":54,"ingested_from":56},"W7214217753",[41,44,47],{"name":42,"orcid":43},"Sandesh Shrestha","https:\u002F\u002Forcid.org\u002F0000-0001-5298-4468",{"name":45,"orcid":46},"K. T. Y. Mahima","https:\u002F\u002Forcid.org\u002F0000-0003-4975-9408",{"name":48,"orcid":49},"Asanka G. Perera","https:\u002F\u002Forcid.org\u002F0000-0003-4021-3943",{"tldr":51,"method":52,"finding":53,"direction":54,"opportunity":55},"综述无人机与遥感目标检测中的对抗补丁与伪装攻击及防御，并提出几何感知分类体系。","基于扰动介质与应用领域的分类法，比较物理鲁棒性、数据集、威胁模型与视角处理。","现有研究多以目标尺度代替视角变化，缺乏显式视角几何建模，因数据集几何信息有限。","农业遥感与作物表型","可构建含视角几何标注的农业航拍数据集，研究视角依赖的对抗补丁放置与自然性防御。","openalex","2026-09-25T23:30:31.219576Z",{"total":59,"page":22,"page_size":59,"items":60},6,[61,105,156,209,250,290],{"id":62,"title":63,"url":64,"summary":65,"summary_zh":66,"content":9,"source_name":67,"source_url":64,"published_at":68,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":69,"score_detail":70,"sources":74,"tags":76,"search_phrases":79,"slug":82,"view_count":36,"doi":83,"paper":84,"created_at":104},2798,"Deep Learning-Based Small-Object Detection in UAV Imagery: A Symmetry-Informed Review of Scale Variation and Scenario Constraints","https:\u002F\u002Fdoi.org\u002F10.3390\u002Fsym18091548","As UAV platforms are increasingly employed in military reconnaissance, disaster monitoring, precision agriculture, and other remote-sensing applications, small-object detection in UAV imagery is a critical yet challenging task. Existing reviews primarily address either general small-object detection or general object detection in UAV imagery, whereas reviews specifically devoted to small-object detection in UAV imagery remain scarce. This review examines existing deep learning-based research on small-object detection in UAV imagery from a problem-driven and deployment-oriented perspective. A symmetry-informed perspective was also adopted to analyze how variations in object-scale, viewpoint, imaging quality, and scene conditions affected feature representation and detection stability. Building on five major challenges in UAV small-object detection, a three-tier framework encompassing detection paradigms, key technical strategies, and scenario-specific constraints was established. Representative UAV datasets were further examined to characterize object-scale distributions, while existing methods were reviewed in terms of multiscale representation, feature enhancement, training optimization, and lightweight deployment. The applicability of different technical approaches was then discussed across five typical UAV scenarios. In addition, under relatively consistent experimental conditions, representative detection models based on two-stage, one-stage, and Transformer-based paradigms were analyzed in terms of detection accuracy, computational efficiency, and deployment feasibility. Finally, current limitations and future directions were summarized for scale-robust, scenario-adaptive, and resource-aware small-object detection in UAV imagery.","随着无人机平台在军事侦察、灾害监测、精准农业等遥感应用中的日益普及，无人机图像中的小目标检测成为一项关键但具有挑战性的任务。现有综述主要关注通用小目标检测或无人机图像中的通用目标检测，而专门针对无人机图像中小目标检测的综述仍然匮乏。本综述从问题驱动和部署导向的视角，审视了现有的基于深度学习的无人机图像小目标检测研究。同时采用对称性视角，分析了目标尺度、视角、成像质量和场景条件的变化如何影响特征表示和检测稳定性。基于无人机小目标检测中的五大挑战，构建了一个涵盖检测范式、关键技术策略和场景特定约束的三层框架。进一步考察了代表性无人机数据集以刻画目标尺度分布，同时从多尺度表示、特征增强、训练优化和轻量化部署等方面综述了现有方法。随后讨论了不同技术方法在五种典型无人机场景中的适用性。此外，在相对一致的实验条件下，对基于两阶段、单阶段和Transformer范式的代表性检测模型在检测精度、计算效率和部署可行性方面进行了分析。最后，总结了当前局限性及未来方向，以推动无人机图像中尺度鲁棒、场景自适应和资源感知的小目标检测研究。","Symmetry","2026-09-16T00:00:00Z",76,{"impact":71,"substance":18,"depth":72,"authority":20,"freshness":21,"relevant":22,"comment":73},16,18,"无人机小目标检测的系统性综述，提出对称性视角与三层框架，对农业遥感与精准农业部署有实质参考价值。",[75],{"name":67,"url":64},[77,27,28,78,29],"智慧农业","农业遥感",[80,81],"农业人工智能 农业遥感 智慧农业 目标检测","农业人工智能 农业遥感","农业人工智能农业遥感智慧农业目标检测-2798","10.3390\u002Fsym18091548",{"doi":83,"openalex_id":85,"authors":86,"venue":67,"cited_by_count":36,"oa_url":64,"card":99,"direction":54,"ingested_from":56},"W7213430256",[87,89,92,94,96],{"name":88,"orcid":9},"Ling Wen",{"name":90,"orcid":91},"Anmin Gong","https:\u002F\u002Forcid.org\u002F0000-0002-2715-2296",{"name":93,"orcid":9},"Caihong Ma",{"name":95,"orcid":9},"Shengzhou Ma",{"name":97,"orcid":98},"Yanzhe Zhang","https:\u002F\u002Forcid.org\u002F0009-0008-6529-1247",{"tldr":100,"method":101,"finding":102,"direction":54,"opportunity":103},"综述无人机图像小目标检测的深度学习研究，提出对称性视角与三层框架。","文献综述，分析代表性数据集与两阶段、单阶段、Transformer模型。","尺度变化和场景约束是核心挑战，需尺度鲁棒、场景自适应和资源感知检测。","可探索农业场景下尺度自适应与轻量化部署的平衡，以及跨场景泛化的小目标检测。","2026-09-17T23:30:37.479759Z",{"id":106,"title":107,"url":108,"summary":109,"summary_zh":110,"content":9,"source_name":111,"source_url":108,"published_at":112,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":113,"score_detail":114,"sources":118,"tags":120,"search_phrases":122,"slug":125,"view_count":36,"doi":126,"paper":127,"created_at":155},2500,"Estimation of crop canopy nitrogen content using deep transfer learning with PROSAIL-PRO model and UAV hyperspectral imagery","https:\u002F\u002Fdoi.org\u002F10.1016\u002Fj.compag.2026.112385","Canopy nitrogen content (CNC) is a key indicator for assessing crop nitrogen status. The non-destructive and rapid estimation of CNC can be performed using unmanned aerial vehicle (UAV)-based hyperspectral remote sensing on large-scale; however, it is often constrained by the scarcity of ground truth data relative to model complexity, which can lead to overfitting and poor generalizability. To address these challenges, a novel inversion framework was proposed by integrating the PROSAIL-PRO physical model with deep transfer learning. Generalized physical features were extracted using pre-trained deep neural network on large-scale PROSAIL-PRO simulated spectra. Subsequently, a fine-tuning strategy was applied in which shallow-level parameters were frozen while deep-level parameters were updated, thereby effectively decoupling intrinsic physical features from site-specific variations. Four deep learning architectures including convolutional neural network (CNN), Residual Neural Network (ResNet18), and their respective channel attention-enhanced variants were evaluated and compared with six classical statistical models including a backpropagation neural network (BPNN). The results demonstrated that transfer-learning-optimized deep models significantly outperformed both baseline and traditional methods in terms of accuracy and generalizability. Specifically, the CNN-based transfer model achieved the best performance on the wheat dataset ( R 2 = 0.8597 and RMSECV = 1.7642), whereas the ResNet18-based transfer model achieved the best performance on the maize dataset ( R 2 = 0.6921 and RMSECV = 2.7944). These findings confirmed that PROSAIL-PRO driven deep transfer learning effectively mitigated overfitting in small-sample hyperspectral datasets. By ensuring high-precision inversion and robust generalization, the proposed physics-guided data-driven fusion approach provides a promising solution for crop nitrogen monitoring.","冠层氮含量（Canopy Nitrogen Content, CNC）是评估作物氮素状况的关键指标。基于无人机（UAV）高光谱遥感可以在大尺度上对CNC进行无损快速估测；然而，该任务常受限于地面真值数据相对模型复杂度而言较为稀缺的问题，这会导致过拟合和泛化能力较差。为应对这些挑战，该研究提出了一种将PROSAIL-PRO物理模型与深度迁移学习相结合的新型反演框架。利用在大规模PROSAIL-PRO模拟光谱上预训练的深度神经网络提取广义物理特征。随后，采用冻结浅层参数、更新深层参数的微调策略，从而有效解耦内在物理特征与站点特异性变异。研究评估了四种深度学习架构，包括卷积神经网络（CNN）、残差神经网络（ResNet18）及其各自的通道注意力增强变体，并与包括反向传播神经网络（BPNN）在内的六种经典统计模型进行了比较。结果表明，经迁移学习优化的深度模型在精度和泛化能力方面均显著优于基线和传统方法。具体而言，基于CNN的迁移模型在小麦数据集上取得了最佳性能（R² = 0.8597，RMSECV = 1.7642），而基于ResNet18的迁移模型在玉米数据集上取得了最佳性能（R² = 0.6921，RMSECV = 2.7944）。这些发现证实，PROSAIL-PRO驱动的深度迁移学习有效缓解了小样本高光谱数据集中的过拟合问题。通过确保高精度反演和稳健泛化，所提出的物理引导的数据驱动融合方法为作物氮素监测提供了一种有前景的解决方案。","Computers and Electronics in Agriculture","2026-09-14T00:00:00Z",81,{"impact":72,"substance":115,"depth":72,"authority":116,"freshness":21,"relevant":22,"comment":117},22,14,"将PROSAIL-PRO物理模型与深度迁移学习结合，缓解小样本高光谱反演过拟合，方法新颖且验证扎实，对作物氮素遥感监测有实用价值。",[119],{"name":111,"url":108},[77,27,28,30,121],"作物氮素监测",[123,124],"作物氮素监测 农业人工智能 智慧农业 无人机","作物氮素监测 农业人工智能","作物氮素监测农业人工智能智慧农业无人机-2500","10.1016\u002Fj.compag.2026.112385",{"doi":126,"openalex_id":128,"authors":129,"venue":111,"cited_by_count":36,"oa_url":108,"card":150,"direction":54,"ingested_from":56},"W7212964562",[130,132,134,136,138,141,143,145,148],{"name":131,"orcid":9},"Jing Zhao",{"name":133,"orcid":9},"Hong Li",{"name":135,"orcid":9},"Junping Liu",{"name":137,"orcid":9},"Wei Chen",{"name":139,"orcid":140},"Xin Guo","https:\u002F\u002Forcid.org\u002F0000-0002-1583-5364",{"name":142,"orcid":9},"Yunlong Wu",{"name":144,"orcid":9},"Menglong Zhao",{"name":146,"orcid":147},"Junaid Nawaz Chauhdary","https:\u002F\u002Forcid.org\u002F0000-0001-7398-5646",{"name":149,"orcid":9},"Zhaoxia Yan",{"tldr":151,"method":152,"finding":153,"direction":54,"opportunity":154},"融合PROSAIL-PRO物理模型与深度迁移学习，用无人机高光谱估算作物冠层氮含量。","PROSAIL-PRO模拟光谱预训练CNN\u002FResNet18，冻结浅层微调深层，","迁移学习模型精度与泛化性最优，小麦R²=0.86、玉米R²=0.69，缓解小样本过拟合。","可探索多作物多生育期迁移、物理模型参数不确定性传播及跨传感器泛化能力。","2026-09-15T23:30:01.518007Z",{"id":157,"title":158,"url":159,"summary":160,"summary_zh":161,"content":9,"source_name":162,"source_url":159,"published_at":163,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":164,"score_detail":165,"sources":167,"tags":169,"search_phrases":171,"slug":174,"view_count":36,"doi":175,"paper":176,"created_at":208},2320,"Genotype-Aware Prediction of Soybean Seed Composition from Multimodal UAV Imagery of the Standing Crop","https:\u002F\u002Fdoi.org\u002F10.3390\u002Frs18183121","Geospatial artificial intelligence (GeoAI) integrates multimodal remote sensing with deep learning to model complex agricultural systems at scale. Within this framework, accurate and non-destructive prediction of seed composition from in-season standing crops is essential for breeding and precision agriculture. This study developed an end-to-end convolutional neural network (CNN) framework to estimate eight seed traits (protein, oil, sucrose, fiber, starch, ash, complex and simple carbohydrates) from UAV-based multisensor imagery and associated genotype and phenological metadata. A total of 372 soybean samples were collected over two growing seasons (2020–2021) from two fields in Missouri, with UAV flights capturing multispectral (MSI), thermal (THR), and LiDAR (LDR) data at four time points spanning vegetative to reproductive growth stages. CNN models were trained in single- and multi-date configurations, incorporating genotype (GEN) and days after sowing (DAS) as additional features. The highest accuracy was achieved for sucrose (R2 = 0.80), followed by simple carbohydrate (R2 = 0.70) and starch (R2 = 0.55), with notable gains from GEN and DAS. Multi-date models incorporating earlier acquisitions often matched or outperformed later or all-date combinations. Among modalities, MSI provided the most robust estimates, with limited added value from LDR or THR. Unlike feature-based pipelines prone to multicollinearity, this image-to-trait approach enables automated, scalable prediction of soybean seed composition for in-season, field-level assessment.","地理空间人工智能（GeoAI）将多模态遥感与深度学习相结合，以在大尺度上对复杂农业系统进行建模。在该框架下，对当季未收获作物进行准确且无损的种子成分预测，对于育种和精准农业至关重要。本研究开发了一种端到端的卷积神经网络（CNN）框架，利用基于无人机（UAV）的多传感器影像以及相关的基因型和物候元数据，估算八种种子性状（蛋白质、油分、蔗糖、纤维、淀粉、灰分、复杂碳水化合物和简单碳水化合物）。研究在2020—2021两个生长季内，从密苏里州两块田地共采集372份大豆样本，无人机飞行在从营养生长期到生殖生长期的四个时间点获取了多光谱（MSI）、热红外（THR）和激光雷达（LDR）数据。CNN模型以单日期和多日期配置进行训练，并将基因型（GEN）和播种后天数（DAS）作为附加特征纳入。蔗糖的估算精度最高（R² = 0.80），其次为简单碳水化合物（R² = 0.70）和淀粉（R² = 0.55），且GEN和DAS的加入带来了显著提升。纳入早期采集数据的多日期模型往往与后期或全日期组合的表现相当，甚至更优。在各模态中，MSI提供了最稳健的估算结果，而LDR或THR的附加价值有限。与易受多重共线性影响的特征工程流程不同，这种从图像到性状的方法能够实现大豆种子成分的自动化、可扩展预测，用于当季田块尺度的评估。","Remote Sensing","2026-09-11T00:00:00Z",80,{"impact":72,"substance":115,"depth":72,"authority":116,"freshness":17,"relevant":22,"comment":166},"基于多模态无人机影像与基因型信息的大豆籽粒成分无损预测研究，方法新颖、数据扎实，对智慧育种与精准农业有参考价值。",[168],{"name":162,"url":159},[77,27,28,30,170],"大豆育种",[172,173],"农业人工智能 大豆育种 智慧农业 无人机","农业人工智能 大豆育种","农业人工智能大豆育种智慧农业无人机-2320","10.3390\u002Frs18183121",{"doi":175,"openalex_id":177,"authors":178,"venue":162,"cited_by_count":36,"oa_url":159,"card":203,"direction":54,"ingested_from":56},"W7212223520",[179,182,185,188,191,193,195,198,201],{"name":180,"orcid":181},"Vasit Sagan","https:\u002F\u002Forcid.org\u002F0000-0003-4375-2096",{"name":183,"orcid":184},"Kristen Rhodes","https:\u002F\u002Forcid.org\u002F0000-0002-8469-5043",{"name":186,"orcid":187},"Sourav Bhadra","https:\u002F\u002Forcid.org\u002F0000-0002-5832-4695",{"name":189,"orcid":190},"Haireti Alifu","https:\u002F\u002Forcid.org\u002F0000-0002-7369-6657",{"name":192,"orcid":9},"Aviskar Giri",{"name":194,"orcid":9},"Ashutosh Pawar",{"name":196,"orcid":197},"Bishal Roy","https:\u002F\u002Forcid.org\u002F0000-0001-9912-5505",{"name":199,"orcid":200},"Supria Sarkar","https:\u002F\u002Forcid.org\u002F0000-0001-9467-5786",{"name":202,"orcid":9},"Felix Fritschi",{"tldr":204,"method":205,"finding":206,"direction":54,"opportunity":207},"用无人机多模态影像与基因型数据训练CNN，预测大豆八种种子成分。","372份大豆样本，多光谱\u002F热红外\u002FLiDAR影像加基因型与播后天数，端到端CNN","蔗糖预测最优(R²=0.80)，基因型与播期提升精度，多光谱最稳健。","可探索基因型-表型-环境交互建模，并迁移至其他作物与多时相早期预测。","2026-09-13T23:30:22.776177Z",{"id":210,"title":211,"url":212,"summary":213,"summary_zh":214,"content":9,"source_name":111,"source_url":212,"published_at":163,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":215,"score_detail":216,"sources":219,"tags":221,"search_phrases":223,"slug":226,"view_count":36,"doi":227,"paper":228,"created_at":249},2284,"Lightweight nighttime citrus fruit detection in UAV imagery using an enhanced NightNet","https:\u002F\u002Fdoi.org\u002F10.1016\u002Fj.compag.2026.112378","Lightweight nighttime citrus fruit detection in UAV imagery using an enhanced NightNet。Computers and Electronics in Agriculture","基于增强型NightNet的无人机图像轻量化夜间柑橘果实检测。《计算机与电子农业》",77,{"impact":71,"substance":217,"depth":19,"authority":116,"freshness":21,"relevant":22,"comment":218},21,"核心期刊论文，提出轻量化夜间柑橘检测方法，方法新颖且面向无人机夜间作业场景，具备细分领域参考价值。",[220],{"name":111,"url":212},[77,27,28,30,222],"柑橘",[224,225],"农业人工智能 智慧农业 无人机 柑橘","农业人工智能 智慧农业","农业人工智能智慧农业无人机柑橘-2284","10.1016\u002Fj.compag.2026.112378",{"doi":227,"openalex_id":229,"authors":230,"venue":111,"cited_by_count":36,"oa_url":9,"card":244,"direction":54,"ingested_from":56},"W7212284586",[231,233,236,238,240,242],{"name":232,"orcid":9},"Zhixian Wu",{"name":234,"orcid":235},"Juntao Xiong","https:\u002F\u002Forcid.org\u002F0000-0002-1095-812X",{"name":237,"orcid":9},"Zhen Liu",{"name":239,"orcid":9},"Sitong Li",{"name":241,"orcid":9},"Jingru Peng",{"name":243,"orcid":9},"Sidi Chen",{"tldr":245,"method":246,"finding":247,"direction":54,"opportunity":248},"提出增强NightNet实现无人机夜间柑橘果实轻量检测。","基于NightNet改进的轻量网络，用于无人机夜间图像。","增强NightNet在夜间无人机图像中实现轻量且准确的柑橘检测。","夜间无人机果实检测的轻量化与低照度增强可进一步结合多光谱或时序信息。","2026-09-13T23:30:02.020275Z",{"id":251,"title":252,"url":253,"summary":254,"summary_zh":255,"content":9,"source_name":256,"source_url":253,"published_at":257,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":258,"score_detail":259,"sources":264,"tags":266,"search_phrases":268,"slug":270,"view_count":36,"doi":271,"paper":272,"created_at":289},1970,"Real-time hass avocado ripeness detection using close-range UAV imagery and YOLO-based deep learning","https:\u002F\u002Fdoi.org\u002F10.1016\u002Fj.atech.2026.102547","Rapid and accurate assessment of Hass avocado ripeness is essential for improving harvest planning, optimizing picking time, increasing productivity, and reducing postharvest losses. This study presents a real-time system for Hass avocado ripeness detection using close-range UAV imagery and YOLO-based deep learning models. A UAV equipped with a camera captured field images, which were transmitted in real time to a processing station through SCRCPY for fruit detection and ripeness classification. The proposed approach consists of two sequential stages. First, a YOLOv8 object detection model was trained using a dataset of 340 manually annotated images generated through real-world adaptation and labeled in Roboflow. The detector achieved a precision of 0.7951, a recall of 0.7849, an mAP@0.5 of 0.8388, and an mAP@0.5:0.95 of 0.6308, demonstrating reliable fruit localization under field conditions. Second, a YOLOv8 classification model was trained on a dataset of 20,231 cropped fruit images derived from an initial collection of 14,710 images categorized into five ripeness stages (unripe, midripe, ripe, fullripe, and overripe). The classification model achieved an accuracy of 0.79 , a macro-average F1-score of 0.77, a weighted-average F1-score of 0.78, a Top-1 accuracy of 0.78, and a Top-5 accuracy of 1.00, confirming its effectiveness for ripeness assessment. The complete system was evaluated under real agricultural conditions with variable illumination and partial fruit occlusion, demonstrating that the integration of close-range UAV imagery and deep learning provides an effective tool for automated avocado ripeness assessment in precision agriculture. The proposed approach was evaluated using close-range UAV imagery acquired in a commercial orchard; therefore, additional validation under different orchards, environmental conditions, and acquisition scenarios is recommended to further assess its generalization capability.","对哈斯鳄梨成熟度的快速准确评估对于改进采收规划、优化采摘时间、提高生产率以及减少采后损失至关重要。本研究提出了一种基于近距无人机影像和YOLO深度学习模型的哈斯鳄梨成熟度实时检测系统。搭载摄像头的无人机采集田间图像，并通过SCRCPY实时传输至处理站进行果实检测和成熟度分类。所提出的方法包括两个连续阶段。首先，利用340张通过实际场景适配生成并在Roboflow中标注的人工标注图像数据集训练YOLOv8目标检测模型。该检测器实现了0.7951的精确率、0.7849的召回率、0.8388的mAP@0.5和0.6308的mAP@0.5:0.95，表明其在田间条件下具有可靠的果实定位能力。其次，利用从初始14,710张图像中提取的20,231张裁剪果实图像数据集训练YOLOv8分类模型，这些图像被分为五个成熟度阶段（未熟、半熟、成熟、完熟和过熟）。分类模型实现了0.79的准确率、0.77的宏平均F1分数、0.78的加权平均F1分数、0.78的Top-1准确率和1.00的Top-5准确率，证实了其在成熟度评估中的有效性。整个系统在光照变化和果实部分遮挡的真实农业条件下进行了评估，结果表明近距无人机影像与深度学习的集成为精准农业中的鳄梨成熟度自动评估提供了有效工具。所提出的方法是在商业果园中获取的近距无人机影像上评估的，因此建议在不同果园、环境条件和采集场景下进行额外验证，以进一步评估其泛化能力。","Smart Agricultural Technology","2026-09-06T00:00:00Z",71,{"impact":260,"substance":115,"depth":72,"authority":261,"freshness":262,"relevant":22,"comment":263},15,12,4,"研究提出基于无人机和YOLO的鳄梨成熟度实时检测系统，方法新颖且数据详实，对精准农业有参考价值。",[265],{"name":256,"url":253},[77,27,28,29,267],"鳄梨",[269,225],"农业人工智能 智慧农业 目标检测 无人机","农业人工智能智慧农业目标检测无人机-1970","10.1016\u002Fj.atech.2026.102547",{"doi":271,"openalex_id":273,"authors":274,"venue":256,"cited_by_count":36,"oa_url":253,"card":283,"direction":287,"ingested_from":56},"W7210263003",[275,278,280],{"name":276,"orcid":277},"Verner Leonidas TUDELA TACO","https:\u002F\u002Forcid.org\u002F0009-0000-2214-3377",{"name":279,"orcid":9},"Wilson Emerson QUISPE ACCOSTUPA",{"name":281,"orcid":282},"Erasmo Sulla Espinoza","https:\u002F\u002Forcid.org\u002F0000-0002-1223-1223",{"tldr":284,"method":285,"finding":286,"direction":287,"opportunity":288},"用无人机近景图像和YOLO深度学习实时检测哈斯鳄梨成熟度，实现精准农业应用。","YOLOv8检测与分类模型，340张标注图像和20231张裁剪图像，通过SCRC","检测mAP@0.5达0.8388，分类准确率0.79，在田间条件下有效评估成熟度。","智慧农业 \u002F 农业物联网","可研究多果园、多环境下的泛化性，或结合多光谱图像提升成熟度分类精度。","2026-09-09T23:30:03.462565Z",{"id":291,"title":292,"url":293,"summary":294,"summary_zh":295,"content":9,"source_name":162,"source_url":293,"published_at":296,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":297,"score_detail":298,"sources":300,"tags":302,"search_phrases":305,"slug":308,"view_count":36,"doi":309,"paper":310,"created_at":337},1734,"Lightweight Near-Infrared Spectral Reconstruction from Red UAV Imagery Using Artificial Intelligence for Low-Cost Remote Sensing","https:\u002F\u002Fdoi.org\u002F10.3390\u002Frs18173015","Near-infrared imagery is essential for vegetation monitoring, precision agriculture, and environmental remote sensing, but multispectral UAV systems remain significantly more expensive and less accessible than conventional RGB imaging platforms. This study presents a lightweight artificial intelligence framework for reconstructing the NIR spectral band exclusively from the red spectral band acquired by a UAV. The proposed methodology formulates the reconstruction task as a pixel-wise nonlinear regression problem and employs a compact multilayer perceptron (MLP) containing only 609 trainable parameters, without exploiting spatial neighborhood information. The framework was developed and evaluated using 280 synchronized multispectral UAV image sets acquired with a DJI Phantom 4 Multispectral platform over a heterogeneous agricultural landscape in the Republic of Moldova. Of these, 252 image sets were used for model development, and 28 were reserved as a held-out within-mission test subset. Quantitative evaluation on a held-out test dataset from the same acquisition mission yielded a mean squared error of 0.010329, a root mean squared error of 0.101632, a mean absolute error of 0.079883, a coefficient of determination of 0.253383, and a Pearson correlation coefficient of 0.683637 between measured and reconstructed normalized NIR digital intensities. The results indicate that the model captures part of the red–NIR relationship under the evaluated acquisition conditions; however, the moderate coefficient of determination suggests that the reconstructed values are an approximation rather than a replacement for measured NIR observations. An illustrative NDVI-based assessment showed that broad spatial vegetation patterns remained identifiable. Rather than introducing a new neural network architecture, this work establishes a compact empirical baseline to investigate the practical performance and limitations of pixel-wise NIR reconstruction from a single red-band value with minimal model complexity.","近红外影像对于植被监测、精准农业和环境遥感至关重要，但多光谱无人机系统仍比传统RGB成像平台昂贵得多，且可及性较低。本研究提出了一种轻量级人工智能框架，仅从无人机获取的红色波段重建近红外光谱波段。所提方法将重建任务表述为逐像素非线性回归问题，并采用仅含609个可训练参数的紧凑型多层感知器（MLP），不利用空间邻域信息。该框架基于在摩尔多瓦共和国异质农业景观上使用DJI Phantom 4 Multispectral平台采集的280组同步多光谱无人机影像集进行开发和评估。其中，252组影像集用于模型开发，28组作为任务内保留测试子集。对来自同一采集任务的保留测试数据集进行的定量评估显示，实测与重建的归一化近红外数字强度之间的均方误差为0.010329，均方根误差为0.101632，平均绝对误差为0.079883，决定系数为0.253383，皮尔逊相关系数为0.683637。结果表明，在评估的采集条件下，模型捕捉了红–近红外关系的一部分；然而，中等的决定系数表明，重建值是对实测近红外观测的近似，而非替代。一项基于NDVI的示例评估显示，大尺度空间植被模式仍可辨识。本研究并非引入新的神经网络架构，而是建立了一个紧凑的经验基线，以在最小模型复杂度下探究基于单一红色波段值进行逐像素近红外重建的实际性能与局限性。","2026-09-04T00:00:00Z",58,{"impact":17,"substance":72,"depth":71,"authority":261,"freshness":262,"relevant":22,"comment":299},"研究用AI从红波段重建近红外，降低遥感成本，方法新颖但精度有限，对低成本农业监测有参考价值。",[301],{"name":162,"url":293},[27,28,30,303,304],"植被监测","近红外",[306,307],"农业人工智能 植被监测 无人机 近红外","农业人工智能 植被监测","农业人工智能植被监测无人机近红外-1734","10.3390\u002Frs18173015",{"doi":309,"openalex_id":311,"authors":312,"venue":162,"cited_by_count":36,"oa_url":293,"card":332,"direction":54,"ingested_from":56},"W7208716610",[313,316,318,321,323,326,329],{"name":314,"orcid":315},"Viorel Bostan","https:\u002F\u002Forcid.org\u002F0000-0002-2422-3538",{"name":317,"orcid":9},"Nicu Drumea",{"name":319,"orcid":320},"Viorel Cărbune","https:\u002F\u002Forcid.org\u002F0000-0002-1556-4453",{"name":322,"orcid":9},"Valeriu Seinic",{"name":324,"orcid":325},"Igor Calmîcov","https:\u002F\u002Forcid.org\u002F0000-0002-4017-3702",{"name":327,"orcid":328},"Adriana Ursu","https:\u002F\u002Forcid.org\u002F0009-0003-4263-2444",{"name":330,"orcid":331},"Maria Gutu","https:\u002F\u002Forcid.org\u002F0000-0002-2820-393X",{"tldr":333,"method":334,"finding":335,"direction":54,"opportunity":336},"用仅含609个参数的MLP从UAV红光波段重建近红外波段，低成本替代多光谱相机。","像素级非线性回归，紧凑MLP（609参数），280组多光谱UAV图像训练。","重建NIR与实测相关性中等（R²=0.253），可识别植被模式，但精度有限。","可探索引入空间信息或时序特征提升重建精度，或扩展至其他波段与作物类型。","2026-09-05T23:30:27.438287Z"]