[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"item-3017":3,"related-3017":57},{"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":23,"tags":25,"search_phrases":31,"slug":34,"view_count":35,"doi":36,"paper":37,"created_at":56},3017,"Automated seedling vigor estimation in cucumbers using digital image processing","https:\u002F\u002Fdoi.org\u002F10.65764\u002Ftjas.2026.267412","Background and Objective: Farmers often rely on experience rather than quantitative indicators to determine seedling quality, resulting in inefficiencies in resource allocation. This study aimed to (1) compare seedling vigor and growth characteristics between open-pollinated (OP) and F1 hybrid cucumber seed types, and (2) develop a non-destructive predictive model for seedling vigor estimation based on digital image analysis.Methodology: Seedling images were acquired under controlled LED lighting using a 5 MP digital camera positioned 30 cm above 14-day-old seedlings. A YOLOv8-based object detection model was applied to detect and isolate true leaf regions, from which pixel count and RGB color values were extracted as input features for a multiple linear regression model to predict the seedling vigor index (SVI).Main Results: The YOLOv8 object detection model achieved a mean average precision (mAP@0.5) of 96.00%, precision of 93.40%, and recall of 94.40% in detecting true leaves. Multiple linear regression analysis was conducted using the Scikit-learn library in Python. Scikit-learn provides regression-based machine learning algorithms, including multiple linear regression. The equation was then applied to the training and test sets, using the pixel values of true leaves as the independent variable and SVI as the dependent variable. Using multiple regression analysis, the trained model generated the equation SVI = -2.27 + (2.84 × 10-5 × pixel) + (3.48 × 10-5 × R) + (-7.49 × 10-2 × G) + (2.06 × 10-1 × B), R2 = 0.57 and RMSE = 1.10. The model achieved the test set, r = 0.79 and RMSE = 1.34. The model performance showed a correlation coefficient of 0.85 for OP data and 0.91 for F1 hybrid data.Conclusions: The results demonstrate that digital image processing combined with object detection provides a non-destructive and effective approach to estimate seedling vigor quality. The predictive models developed from OP and F1 hybrid datasets indicate potential application in precision agriculture for automated seedling quality assessment and transplanting decision support.","背景与目标：农民通常依赖经验而非定量指标来判断幼苗质量，导致资源配置效率低下。本研究旨在（1）比较开放授粉（OP）与F1杂交黄瓜种子类型之间的幼苗活力和生长特性，（2）基于数字图像分析开发一种用于幼苗活力评估的无损预测模型。方法：使用500万像素数码相机置于14日龄幼苗上方30 cm处，在受控LED光照下获取幼苗图像。应用基于YOLOv8的目标检测模型检测并分离真叶区域，从中提取像素计数和RGB颜色值作为多元线性回归模型的输入特征，以预测幼苗活力指数（SVI）。主要结果：YOLOv8目标检测模型在检测真叶时达到了96.00%的平均精度均值（mAP@0.5）、93.40%的精确率和94.40%的召回率。使用Python中的Scikit-learn库进行多元线性回归分析。Scikit-learn提供基于回归的机器学习算法，包括多元线性回归。随后将该方程应用于训练集和测试集，以真叶像素值作为自变量，SVI作为因变量。通过多元回归分析，训练模型生成的方程为SVI = -2.27 + (2.84 × 10-5 × 像素) + (3.48 × 10-5 × R) + (-7.49 × 10-2 × G) + (2.06 × 10-1 × B)，R2 = 0.57，RMSE = 1.10。该模型在测试集上达到r = 0.79，RMSE = 1.34。模型性能显示，OP数据的相关系数为0.85，F1杂交数据的相关系数为0.91。结论：结果表明，数字图像处理结合目标检测为评估幼苗活力质量提供了一种无损且有效的方法。基于OP和F1杂交数据集开发的预测模型表明，其在精准农业中具有用于自动化幼苗质量评估和移栽决策支持的潜在应用。",null,"Thai Journal of Agricultural Science","2026-09-19T00:00:00Z","论文",10,false,71,{"impact":17,"substance":18,"depth":19,"authority":17,"freshness":20,"relevant":21,"comment":22},12,21,17,9,1,"基于YOLOv8与多元回归的黄瓜幼苗活力无损估测，方法具体、指标完整，对智慧育苗有参考价值，但属细分作物研究，影响范围有限。",[24],{"name":10,"url":6},[26,27,28,29,30],"智慧农业","农业人工智能","无损检测","图像识别","黄瓜育苗",[32,33],"黄瓜 幼苗活力 图像处理","YOLOv8 幼苗 检测","黄瓜幼苗活力图像处理-3017",0,"10.65764\u002Ftjas.2026.267412",{"doi":36,"openalex_id":38,"authors":39,"venue":10,"cited_by_count":35,"oa_url":6,"card":48,"direction":54,"ingested_from":55},"W7213631983",[40,42,44,46],{"name":41,"orcid":9},"Thanabodee Withunchettanan",{"name":43,"orcid":9},"Raksak Sermsak",{"name":45,"orcid":9},"Pichittra Kaewsorn",{"name":47,"orcid":9},"Kriengkri Kaewtrakulpong",{"tldr":49,"method":50,"finding":51,"direction":52,"opportunity":53},"用YOLOv8检测黄瓜真叶并结合多元回归，实现幼苗活力指数无损预测。","LED下拍摄14天幼苗，YOLOv8分割真叶，提取像素与RGB做多元线性回归。","YOLOv8检测mAP@0.5达96%，模型测试r=0.79，F1杂交种相关性达0.91。","农业人工智能与决策模型","可扩展多品种、多环境数据，融合时序图像与深度学习提升活力预测泛化性。","数字乡村与农业信息化","openalex","2026-09-20T23:30:26.929607Z",{"total":58,"page":21,"page_size":58,"items":59},6,[60,92,138,179,221,262],{"id":61,"title":62,"url":63,"summary":64,"summary_zh":9,"content":9,"source_name":65,"source_url":9,"published_at":11,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":66,"score_detail":67,"sources":74,"tags":76,"search_phrases":79,"slug":82,"view_count":35,"doi":9,"paper":83,"created_at":91},3047,"基于CNN\u002FLR\u002FGA-BP的辣椒叶病识别稳定性与精度多维度视觉特征比较","https:\u002F\u002Fwww.mdpi.com\u002F2311-7524\u002F12\u002F9\u002F1176","塔里木大学Xueting Ma、Yifei Li等联合东北林业大学、南京农业大学，构建1260份辣椒叶数据集（健康、细菌性斑点病、黄化卷叶病），采用Lab b-channel、RGB super-green、Otsu-ACWE三种分割算法系统评估最优预处理方案，提取32个融合视觉特征并以随机森林消除7个低贡献冗余特征保留25个判别变量。20次独立重复试验表明：CNN测试平均精度97.67%、平均AUC 0.999，稳定性最佳；LR计算成本低适合资源受限场景；GA-BP非线性拟合能力弱、预测波动严重。该研究为辣椒叶病识别提供标准化实验框架。","MDPI Horticulturae 12(9):1176",78,{"impact":68,"substance":69,"depth":70,"authority":71,"freshness":72,"relevant":21,"comment":73},16,22,18,14,8,"基于1260份辣椒叶数据集系统比较CNN\u002FLR\u002FGA-BP三种模型，方法规范、结论可靠，对作物病害智能识别有参考价值。",[75],{"name":65,"url":63},[26,27,29,77,78],"辣椒叶病","病害诊断",[80,81],"塔里木大学 辣椒叶病 识别","CNN 辣椒叶病 视觉特征","塔里木大学辣椒叶病识别-3047",{"doi":9,"openalex_id":9,"authors":84,"venue":9,"cited_by_count":35,"oa_url":9,"card":85,"direction":52,"ingested_from":90},[],{"tldr":86,"method":87,"finding":88,"direction":52,"opportunity":89},"比较CNN、LR、GA-BP对辣椒叶病的识别精度与稳定性，并评估三种分割预处理方案。","1260份辣椒叶图像，Lab\u002FRGB\u002FOtsu-ACWE分割，32特征经随机森林","CNN平均精度97.67%、AUC 0.999且最稳定；LR成本低适合资源受限；GA-BP波动严重。","可探索轻量化CNN与LR融合的田间实时识别方案，并验证多作物、多病害下的泛化稳定性。","agent","2026-09-21T00:04:39.222838Z",{"id":93,"title":94,"url":95,"summary":96,"summary_zh":97,"content":9,"source_name":98,"source_url":95,"published_at":99,"category":12,"cover_url":9,"hotness":13,"is_selected":100,"score":101,"score_detail":102,"sources":107,"tags":109,"search_phrases":112,"slug":115,"view_count":21,"doi":116,"paper":117,"created_at":137},1755,"A non-destructive watermelon sweetness classification via vision transformer with cross-modal knowledge distillation","https:\u002F\u002Fdoi.org\u002F10.1016\u002Fj.compag.2026.112393","A non-destructive watermelon sweetness classification via vision transformer with cross-modal knowledge distillation。Computers and Electronics in Agriculture","基于视觉Transformer与跨模态知识蒸馏的非破坏性西瓜甜度分类方法。计算机与农业电子学","Computers and Electronics in Agriculture","2026-09-05T00:00:00Z",true,70,{"impact":103,"substance":104,"depth":70,"authority":103,"freshness":105,"relevant":21,"comment":106},15,20,2,"提出基于视觉Transformer与跨模态知识蒸馏的无损西瓜甜度分级方法，发表于权威期刊，方法新颖，对农产品品质检测有参考价值。",[108],{"name":98,"url":95},[26,27,28,110,111],"西瓜","模型蒸馏",[113,114],"农业人工智能 无损检测 智慧农业 模型蒸馏","农业人工智能 无损检测","农业人工智能无损检测智慧农业模型蒸馏-1755","10.1016\u002Fj.compag.2026.112393",{"doi":116,"openalex_id":118,"authors":119,"venue":98,"cited_by_count":35,"oa_url":9,"card":132,"direction":52,"ingested_from":55},"W7208829728",[120,123,126,129],{"name":121,"orcid":122},"Mustafa Kareem Hadi","https:\u002F\u002Forcid.org\u002F0000-0001-6469-3799",{"name":124,"orcid":125},"Siti Khairunniza Bejo","https:\u002F\u002Forcid.org\u002F0000-0002-4972-1701",{"name":127,"orcid":128},"Abdul Rashid Mohamed Shariff","https:\u002F\u002Forcid.org\u002F0000-0003-4626-4995",{"name":130,"orcid":131},"Nazmi Mat Nawi","https:\u002F\u002Forcid.org\u002F0000-0002-5916-5745",{"tldr":133,"method":134,"finding":135,"direction":52,"opportunity":136},"提出用视觉Transformer结合跨模态知识蒸馏，实现西瓜甜度的无损分类。","视觉Transformer与跨模态知识蒸馏，利用光谱数据辅助图像模型训练。","跨模态蒸馏可提升图像模型对西瓜甜度的分类精度，实现无损检测。","可探索将跨模态蒸馏用于其他水果内部品质（如糖度、酸度）的无损检测，或结合多传感器数据提升模型泛化性。","2026-09-06T23:30:01.479655Z",{"id":139,"title":140,"url":141,"summary":142,"summary_zh":143,"content":9,"source_name":144,"source_url":141,"published_at":145,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":146,"score_detail":147,"sources":150,"tags":152,"search_phrases":155,"slug":158,"view_count":35,"doi":159,"paper":160,"created_at":178},1730,"PAT: An Image Analysis Tool for Automated Scoring of Pollen in Alexander-Stained Anthers","https:\u002F\u002Fdoi.org\u002F10.1093\u002Fjxb\u002Ferag436","Abstract Quantitative pollen viability analysis is a critical but labor-intensive step in plant reproductive biology. Existing deep-learning Segment Anything Models (SAM) fail to reliably segment viable pollen in Alexander-stained anthers. To address this, we fine-tuned an existing Cellpose-SAM model for pollen segmentation. We integrated it into PAT (Pollen Analysis Tool), a cross-platform desktop application. PAT features instance segmentation with interactive quality control, an in-app model retraining module, and publication-ready statistical outputs. We deployed PAT in an EMS suppressor screen of semi-sterile Arabidopsis smg7-6 mutants, enabling efficient candidate prioritization for whole genome sequencing and mapping candidate mutation. This screen led to the identification of a point mutation in CAP-D2 ( capd2-2 ), a Condensin I subunit, that rescues the smg7-6 meiotic phenotype. Notably, mutation in a Condensin II subunits (CAP-D3 and CAP-H2) does not confer rescue. Further characterization suggests the capd2-2 allele is hypomorphic, showing no defects in vegetative growth, chromocenter compaction, or transposable element silencing. Collectively, we demonstrate that accessible AI tools have the potential to bridge gaps in plant phenotyping and accelerate the pace of biological discovery. Highlight We combined AI-powered image analysis with an easy-to-use desktop app to automate plant pollen counting, then used it to identify a new genetic suppressor of meiotic defects.","定量花粉活力分析是植物生殖生物学中一个关键但劳动密集的步骤。现有的深度学习分割一切模型（SAM）在亚历山大染色花药中无法可靠地分割有活力的花粉。为解决此问题，我们对现有的Cellpose-SAM模型进行了微调，用于花粉分割。我们将其整合到PAT（花粉分析工具）中，这是一个跨平台的桌面应用程序。PAT具备实例分割及交互式质量控制、应用内模型再训练模块以及可直接用于发表的统计输出功能。我们在半不育拟南芥smg7-6突变体的EMS抑制子筛选中部署了PAT，从而能够高效地对候选突变体进行优先级排序，以进行全基因组测序和候选突变定位。该筛选鉴定出CAP-D2（capd2-2）中的一个点突变，该基因编码凝缩蛋白I的一个亚基，该突变能够挽救smg7-6的减数分裂表型。值得注意的是，凝缩蛋白II亚基（CAP-D3和CAP-H2）的突变并未赋予挽救效果。进一步的特征分析表明，capd2-2等位基因属于亚效等位基因，在营养生长、染色质中心压缩或转座子沉默方面未显示任何缺陷。总体而言，我们证明了易于使用的人工智能工具具有弥合植物表型分析差距并加速生物学发现的潜力。亮点：我们将人工智能驱动的图像分析与易于使用的桌面应用相结合，实现了植物花粉计数的自动化，并利用该方法鉴定了一个新的减数分裂缺陷遗传抑制子。","Journal of Experimental Botany","2026-09-04T00:00:00Z",79,{"impact":70,"substance":69,"depth":70,"authority":71,"freshness":148,"relevant":21,"comment":149},7,"AI工具用于花粉活力自动分析，加速遗传筛选，对育种和表型分析有实用价值。",[151],{"name":144,"url":141},[26,27,153,154,29],"育种","表型分析",[156,157],"农业人工智能 图像识别 智慧农业 表型分析","农业人工智能 图像识别","农业人工智能图像识别智慧农业表型分析-1730","10.1093\u002Fjxb\u002Ferag436",{"doi":159,"openalex_id":161,"authors":162,"venue":144,"cited_by_count":35,"oa_url":141,"card":172,"direction":177,"ingested_from":55},"W7160664980",[163,166,169],{"name":164,"orcid":165},"Darya Volkava","https:\u002F\u002Forcid.org\u002F0009-0003-3578-6592",{"name":167,"orcid":168},"Karel Říha","https:\u002F\u002Forcid.org\u002F0000-0002-6124-0118",{"name":170,"orcid":171},"Vivek K. Raxwal","https:\u002F\u002Forcid.org\u002F0000-0002-5182-6377",{"tldr":173,"method":174,"finding":175,"direction":52,"opportunity":176},"开发了花粉分析工具PAT，自动化评估花粉活力，并用于发现新的减数分裂抑制因子。","微调Cellpose-SAM模型，集成到跨平台桌面应用PAT，支持交互式质量控制","PAT高效筛选拟南芥突变体，发现CAP-D2点突变可挽救smg7-6表型，而Condensin II","将AI图像分析工具扩展到其他作物表型，如种子活力、果实发育，结合交互式平台加速遗传筛选。","农业遥感与作物表型","2026-09-05T23:30:19.596351Z",{"id":180,"title":181,"url":182,"summary":183,"summary_zh":184,"content":9,"source_name":185,"source_url":182,"published_at":186,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":15,"score_detail":187,"sources":189,"tags":191,"search_phrases":193,"slug":195,"view_count":35,"doi":196,"paper":197,"created_at":220},1536,"A Non-Invasive Approach for Detecting Water Adulteration in Orange Juice Using Computer Vision and Deep Learning","https:\u002F\u002Fdoi.org\u002F10.1016\u002Fj.jfca.2026.109484","Food fraud due to water adulteration in orange juice is a significant concern for the beverage industry. This study aimed to develop a rapid and non-invasive computer-vision method based on deep learning to detect and classify water-adulteration levels ranging from 1% to 15% in three orange juice products representing freshly squeezed juice, juice from concentrate, and orange nectar. High-resolution images were acquired under two shutter-speed-based exposure conditions: the higher-exposure acquisition condition (1\u002F30 s) and the lower-exposure acquisition condition (1\u002F250 s), while maintaining constant scene illumination, and were analyzed using ResNet50 convolutional neural networks. In the image-level hold-out test, the model trained with images acquired at 1\u002F250 s achieved an accuracy of 88.3%, whereas the model trained at 1\u002F30 s reached 83.1%. To provide a more stringent assessment of sample-level transferability, the previously trained and fixed 1\u002F250 s model was subsequently evaluated using 240 independently prepared and blindly coded samples obtained from subsequent purchases of the same commercial products. This independent blind validation achieved a 24-class accuracy of 86.7%. Misclassifications occurred predominantly between adjacent or closely related water-adulteration levels within the same juice product. When the independent-validation predictions were collapsed into a binary pure-versus-adulterated screening task, the model achieved 100.0% sensitivity, 93.3% specificity, 99.2% accuracy, and 96.7% balanced accuracy. These findings support the feasibility of the method as a rapid pre-screening tool for detecting visible-image patterns associated with controlled water dilution and show that detecting the presence of adulteration is more reliable than assigning an exact value to closely spaced adulteration levels. The independent validation provides evidence of transferability to newly prepared samples from subsequent purchases of the same products. However, further validation across additional brands, production batches, orange origins, seasons, and acquisition environments is required before broader applicability can be established.","橙汁中因掺水导致的食品欺诈是饮料行业关注的重要问题。本研究旨在开发一种基于深度学习的快速、非侵入性计算机视觉方法，用于检测和分类三种橙汁产品（鲜榨汁、浓缩还原汁和橙汁饮料）中1%至15%的掺水水平。在保持场景照明恒定的条件下，分别以两种基于快门速度的曝光条件采集高分辨率图像：高曝光采集条件（1\u002F30秒）和低曝光采集条件（1\u002F250秒），并使用ResNet50卷积神经网络进行分析。在图像级留出测试中，以1\u002F250秒采集图像训练的模型准确率达到88.3%，而以1\u002F30秒训练的模型准确率为83.1%。为对样本级可迁移性进行更严格的评估，先前训练并固定的1\u002F250秒模型随后被用于评估240个独立制备、盲法编码的样本，这些样本来自后续购买的同款商业产品。该独立盲法验证实现了24类别86.7%的准确率。误分类主要发生在同一果汁产品内相邻或相近的掺水水平之间。当独立验证预测结果被合并为二元纯正与掺假筛查任务时，模型实现了100.0%的灵敏度、93.3%的特异度、99.2%的准确率和96.7%的平衡准确率。这些发现支持该方法作为快速预筛查工具的可行性，用于检测与受控水稀释相关的可见图像模式，并表明检测掺假的存在比精确赋值于间隔较小的掺水水平更为可靠。独立验证提供了该方法对后续购买同款产品新制备样本具有可迁移性的证据。然而，在确立更广泛的适用性之前，仍需在更多品牌、生产批次、橙子产地、季节和采集环境中进行进一步验证。","Journal of Food Composition and Analysis","2026-09-01T00:00:00Z",{"impact":17,"substance":104,"depth":70,"authority":71,"freshness":148,"relevant":21,"comment":188},"研究提出基于深度学习的橙汁掺水无损检测方法，独立验证准确率高，对食品安全监管有参考价值。",[190],{"name":185,"url":182},[26,27,28,192],"食品安全",[194,114],"农业人工智能 无损检测 智慧农业 食品安全","农业人工智能无损检测智慧农业食品安全-1536","10.1016\u002Fj.jfca.2026.109484",{"doi":196,"openalex_id":198,"authors":199,"venue":185,"cited_by_count":35,"oa_url":214,"card":215,"direction":52,"ingested_from":55},"W7204957864",[200,202,204,207,209,211],{"name":201,"orcid":9},"Ana M. Pérez-Calabuig",{"name":203,"orcid":9},"Sandra Pradana‐López",{"name":205,"orcid":206},"John C. Cancilla","https:\u002F\u002Forcid.org\u002F0000-0003-3645-7224",{"name":208,"orcid":9},"María Luz Mena",{"name":210,"orcid":9},"Carlos López-Pingarrón",{"name":212,"orcid":213},"José S. Torrecilla","https:\u002F\u002Forcid.org\u002F0000-0003-1209-203X","https:\u002F\u002Fwww.sciencedirect.com\u002Fscience\u002Farticle\u002Fpii\u002FS0889157526006277\u002Fpdf",{"tldr":216,"method":217,"finding":218,"direction":52,"opportunity":219},"用深度学习计算机视觉检测橙汁掺水，实现非侵入式快速筛查。","采集不同曝光图像，用ResNet50分类掺水等级，独立盲样验证。","掺水检测准确率高，区分精确等级较难，模型可迁移至新样本。","可扩展至其他食品掺假检测，需跨品牌、批次、季节等验证，或开发便携设备。","2026-09-03T23:30:50.596809Z",{"id":222,"title":223,"url":224,"summary":225,"summary_zh":226,"content":9,"source_name":227,"source_url":224,"published_at":186,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":15,"score_detail":228,"sources":230,"tags":232,"search_phrases":235,"slug":237,"view_count":35,"doi":238,"paper":239,"created_at":261},1381,"From Visual Words to Vision Transformers: Dual Approaches to Water Stress Classification in Maize","https:\u002F\u002Fdoi.org\u002F10.1016\u002Fj.atech.2026.102519","Early water stress in maize can quietly reduce growth and yield, making early detection from simple RGB images valuable for timely irrigation. This study compares two fundamentally different representation strategies for maize water stress classification: deep features learned from localized image regions using a custom Swin transformer and handcrafted visual features classified using a backpropagation neural network (BPNN). The Swin transformer learns stress-related representations from localized cropped image regions, with patch-level predictions aggregated into image-level classifications through majority voting, whereas the BPNN relies on handcrafted color- and texture-based descriptors extracted from segmented full-plant images. Both methods are evaluated on a maize water stress image dataset, using an independent test set that is not used during training or cross-validation for the final performance comparison. While the Swin transformer achieved 98% image-level accuracy after aggregation, the BPNN achieved 97% accuracy. Despite relying on fundamentally different representation strategies, both methods achieved high classification performance on the independently held-out test set. The results show that high-accuracy maize water stress classification can be achieved using low-cost RGB imagery and provide insight into the tradeoffs between learned and handcrafted representations in terms of classification performance and computational characteristics. These findings demonstrate the potential of computer-vision approaches for RGB-based agricultural water stress monitoring.","玉米早期水分胁迫会悄然降低生长和产量，因此利用简单的RGB图像进行早期检测对于及时灌溉具有重要意义。本研究比较了两种根本不同的玉米水分胁迫分类表示策略：一种是通过自定义Swin变换器从局部图像区域学习深度特征，另一种是利用反向传播神经网络（BPNN）对手工视觉特征进行分类。Swin变换器从局部裁剪图像区域学习与胁迫相关的表示，通过多数投票将补丁级预测聚合为图像级分类；而BPNN则依赖于从分割的全株图像中提取的基于颜色和纹理的手工描述符。两种方法均在玉米水分胁迫图像数据集上进行了评估，并使用独立测试集（训练或交叉验证中未使用）进行最终性能比较。尽管Swin变换器在聚合后实现了98%的图像级准确率，BPNN也达到了97%的准确率。尽管依赖根本不同的表示策略，两种方法在独立保留的测试集上均取得了较高的分类性能。结果表明，利用低成本RGB图像即可实现高精度的玉米水分胁迫分类，并揭示了学习表示与手工表示在分类性能和计算特性方面的权衡。这些发现展示了基于计算机视觉方法在RGB农业水分胁迫监测中的潜力。","Smart Agricultural Technology",{"impact":17,"substance":69,"depth":70,"authority":17,"freshness":148,"relevant":21,"comment":229},"研究对比Swin Transformer与手工特征在玉米水分胁迫分类中的表现，均达97%以上准确率，为低成本RGB图像监测提供新思路。",[231],{"name":227,"url":224},[26,27,233,234,29],"玉米","水分胁迫",[236,157],"农业人工智能 图像识别 智慧农业 水分胁迫","农业人工智能图像识别智慧农业水分胁迫-1381","10.1016\u002Fj.atech.2026.102519",{"doi":238,"openalex_id":240,"authors":241,"venue":227,"cited_by_count":35,"oa_url":255,"card":256,"direction":52,"ingested_from":55},"W7204901486",[242,244,246,249,252],{"name":243,"orcid":9},"Sumaira Ghazal",{"name":245,"orcid":9},"Sardar Ali Abbas",{"name":247,"orcid":248},"Arslan Munir","https:\u002F\u002Forcid.org\u002F0000-0002-3126-8945",{"name":250,"orcid":251},"Ignacio A. Ciampitti","https:\u002F\u002Forcid.org\u002F0000-0001-9619-5129",{"name":253,"orcid":254},"Waqar S. Qureshi","https:\u002F\u002Forcid.org\u002F0000-0003-0176-8145","https:\u002F\u002Fwww.sciencedirect.com\u002Fscience\u002Farticle\u002Fpii\u002FS2772375526007446\u002Fpdf",{"tldr":257,"method":258,"finding":259,"direction":52,"opportunity":260},"比较Swin Transformer与手工特征BPNN在玉米水分胁迫RGB图像分类中的表现，两者均达","使用自定义Swin Transformer学习局部特征，BPNN结合手工颜色纹理","Swin Transformer准确率98%，BPNN 97%，证明低成本RGB图像可实现高精度水分","可探索轻量化模型或融合两种表征策略，以平衡精度与计算效率，适应田间实时监测需求。","2026-09-02T23:30:04.449177Z",{"id":263,"title":264,"url":265,"summary":266,"summary_zh":9,"content":9,"source_name":267,"source_url":9,"published_at":268,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":146,"score_detail":269,"sources":272,"tags":274,"search_phrases":278,"slug":281,"view_count":35,"doi":9,"paper":282,"created_at":289},3048,"基于双路径注意力与多尺度融合的作物病虫害识别网络DPMFNet","https:\u002F\u002Fwww.mdpi.com\u002F1099-4300\u002F28\u002F9\u002F1032","盐城工学院Hong Zhang、Fagen Song等联合江苏开放大学提出DPMFNet轻量级双路径网络，集成空间-通道双注意力（SCDA）与多尺度深度可分离卷积（MDSC）模块，构建AttMDSCBlock残差结构。在PlantVillage与AI Challenger 2018数据集上DPMFNet仅14.24M参数和2.55G FLOPs，跨注意力机制融合局部细节与全局上下文，轻量化金字塔策略自适应整合多分辨率特征，在复杂农田场景下兼顾精度与可部署性，为嵌入式田间设备提供高性价比方案。","MDPI Entropy 28(9):1032","2026-09-20T00:00:00Z",{"impact":68,"substance":69,"depth":70,"authority":270,"freshness":13,"relevant":21,"comment":271},13,"轻量级双路径注意力网络在两大公开数据集上兼顾精度与可部署性，对嵌入式田间设备落地有实质参考价值。",[273],{"name":267,"url":265},[26,27,275,276,277],"病虫害识别","作物监测","轻量化模型",[279,280],"盐城工学院 DPMFNet 病虫害识别","PlantVillage AI Challenger 作物病害","盐城工学院DPMFNet病虫害识别-3048",{"doi":9,"openalex_id":9,"authors":283,"venue":9,"cited_by_count":35,"oa_url":9,"card":284,"direction":52,"ingested_from":90},[],{"tldr":285,"method":286,"finding":287,"direction":52,"opportunity":288},"提出轻量级双路径网络DPMFNet，实现复杂农田场景下的作物病虫害高精度识别。","空间-通道双注意力与多尺度深度可分离卷积，构建AttMDSCBlock残差结构。","仅14.24M参数、2.55G FLOPs，在PlantVillage与AI Challenger ","可探索真实田间多病虫害并发与边缘设备实时推理的轻量化自适应识别研究。","2026-09-21T00:04:39.305757Z"]