[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"item-2141":3,"related-2141":71},{"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":9,"secondary_topics":26,"classification":27,"search_phrases":45,"slug":48,"view_count":49,"doi":50,"paper":51,"created_at":70},2141,"Next-Generation Deep Learning: A Comprehensive Survey on Explainable, Efficient, Privacy-Preserving and Multimodal Artificial Intelligence","https:\u002F\u002Fdoi.org\u002F10.17148\u002Fijarcce.2026.15912","The continuous evolution of deep learning has significantly expanded the capabilities of artificial intelligence, enabling intelligent systems to solve increasingly complex problems across healthcare, computer vision, natural language processing, cybersecurity, finance, autonomous systems, and smart environments.Beginning with artificial neural networks and progressing through convolutional and recurrent networks to Transformers, Graph Neural Networks, Vision Transformers, and Large Language Models, deep learning has achieved remarkable improvements in feature representation, prediction accuracy, and knowledge transfer.Nevertheless, the growing complexity of these models introduces several challenges, including limited model transparency, high computational and energy requirements, data privacy risks, and the effective utilization of heterogeneous multimodal information.Unlike conventional surveys that primarily classify deep learning according to network architectures or application domains, this work presents a capability-oriented taxonomy that organizes recent developments into five major research directions: Explainable Deep Learning, Efficient Deep Learning, Privacy-Preserving Deep Learning, Green Deep Learning and Multimodal Deep Learning.Based on this framework, representative architectures are critically examined with respect to their operating principles, strengths, limitations, and suitability for diverse real-world applications.The survey also analyses emerging application areas, identifies unresolved challenges related to fairness, robustness, scalability, and trustworthy artificial intelligence, and discusses promising research opportunities involving Federated Large Language Models, Edge AI, Green AI, Self-supervised Learning, and Multimodal Foundation Models.The proposed framework provides a structured understanding of current advances while offering practical insights for the design of transparent, efficient, secure, and sustainable next-generation deep learning systems.","深度学习的持续演进显著拓展了人工智能的能力，使智能系统能够在医疗健康、计算机视觉、自然语言处理、网络安全、金融、自主系统和智能环境等领域解决日益复杂的问题。从人工神经网络起步，历经卷积网络和循环网络，发展到Transformer、图神经网络、视觉Transformer和大语言模型，深度学习在特征表示、预测精度和知识迁移方面取得了显著提升。然而，这些模型日益增长的复杂性也带来了若干挑战，包括模型透明度有限、计算和能耗需求高、数据隐私风险以及异构多模态信息的有效利用等问题。与主要按网络架构或应用领域对深度学习进行分类的传统综述不同，本文提出了一种以能力为导向的分类体系，将近期发展归纳为五大研究方向：可解释深度学习、高效深度学习、隐私保护深度学习、绿色深度学习和多模态深度学习。基于该框架，本文对代表性架构进行了批判性审视，分析其运行原理、优势、局限性以及对多样化实际应用的适用性。本综述还分析了新兴应用领域，指出了在公平性、鲁棒性、可扩展性和可信人工智能方面尚未解决的挑战，并讨论了涉及联邦大语言模型、边缘人工智能、绿色人工智能、自监督学习和多模态基础模型等有前景的研究机遇。所提出的框架为理解当前进展提供了结构化视角，同时为设计透明、高效、安全和可持续的下一代深度学习系统提供了实践指导。",null,"IJARCCE","2026-09-10T00:00:00Z","论文",10,false,0,{"impact":15,"substance":15,"depth":15,"authority":15,"freshness":15,"relevant":15,"comment":17},"通用深度学习综述，未涉及农业场景，与三农及农业信息化无直接关联，不建议进入每日精选。",[19],{"name":10,"url":6},[21,22,23,24,25],"可解释AI","深度学习","多模态","绿色AI","隐私计算",[],{"status":28,"locked":14,"version":29,"method":30,"evidence":31},"needs_review","2026-10-22.2","rules",{"candidates":32,"agricultural_evidence":14,"reason":44},[33],{"code":34,"score":35,"supported":36,"title":37,"abstract":40,"tags":43},"I2",28,true,[38,39],"deep learning","artificial intelligence",[22,41,38,39,42],"计算机视觉","computer vision",[21,23,22],"农业相关性需要复核",[46,47],"深度学习 隐私计算 可解释AI 多模态","深度学习 隐私计算","深度学习隐私计算可解释AI多模态-2141",1,"10.17148\u002Fijarcce.2026.15912",{"doi":50,"openalex_id":52,"authors":53,"venue":10,"cited_by_count":15,"oa_url":6,"card":62,"direction":68,"ingested_from":69},"W7212125020",[54,56,58,60],{"name":55,"orcid":9},"K. Ramalakshmi",{"name":57,"orcid":9},"S. Prema",{"name":59,"orcid":9},"M. Ajithaveni",{"name":61,"orcid":9},"M Ramya",{"tldr":63,"method":64,"finding":65,"direction":66,"opportunity":67},"综述可解释、高效、隐私保护、绿色与多模态五大深度学习方向，提出能力导向分类框架。","文献综述，按能力分类代表性架构，分析原理、优缺点与应用。","五大方向各有挑战，未来机会在联邦大模型、边缘AI、绿色AI、自监督与多模态基础模型。","农业人工智能与决策模型","可将可解释、高效、隐私保护与多模态深度学习迁移到农业场景，构建可信农业AI系统。","智慧农业 \u002F 农业物联网","openalex","2026-09-11T23:30:10.233616Z",{"total":72,"page":49,"page_size":72,"items":73},6,[74,176,256,310,372,424],{"id":75,"title":76,"url":77,"summary":78,"summary_zh":79,"content":9,"source_name":80,"source_url":77,"published_at":81,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":82,"score_detail":83,"sources":90,"tags":92,"primary_topic":96,"secondary_topics":101,"classification":113,"search_phrases":146,"slug":149,"view_count":15,"doi":150,"paper":151,"created_at":175},4898,"An Interpretable Deep Learning Framework for Spatial Prediction of Soil Salinity Using Multi-Source Remote Sensing and Environmental Data: A Case Study of Cangzhou, Hebei Province, China","https:\u002F\u002Fdoi.org\u002F10.3390\u002Fs26206465","Soil salinity poses a significant challenge to agricultural productivity and sustainable land management in coastal and semi-arid regions. Soil salinity in coastal agricultural regions is controlled by complex interactions among surface conditions, groundwater processes, soil properties, and terrain characteristics. Accurate spatial prediction requires integrating remote sensing observations with environmental information that represents these processes. This study presents a multi-source deep learning framework that integrates Sentinel-2-derived spectral indices with environmental covariates for regional soil salinity spatial prediction under the environmental conditions of the 2024 sampling period. SHapley Additive exPlanations (SHAP) is employed to quantify the contribution of environmental variables within the prediction framework and improve model interpretability. A convolutional neural network (CNN) incorporating a squeeze-and-excitation (SE) attention mechanism is developed as a nonlinear feature representation framework for integrating heterogeneous predictors, while the Crested Porcupine Optimizer (CPO) is used to optimize model hyperparameters. The proposed framework achieved the highest predictive performance among the benchmark models, with an R2 of 0.785, an RMSE of 0.928, and an MAE of 0.720. Integrating environmental covariates with spectral features further improves the prediction performance, increasing the R2 from 0.701 to 0.785. The results indicate the potential of integrating multi-source environmental information with interpretable deep learning for spatial prediction of near-surface soil salinity under similar coastal agricultural conditions.","土壤盐渍化对沿海和半干旱地区的农业生产力和可持续土地管理构成了重大挑战。沿海农业区域的土壤盐渍化受地表条件、地下水过程、土壤属性和地形特征之间复杂相互作用的控制。准确的空间预测需要将遥感观测与表征这些过程的环境信息相结合。本研究提出了一种多源深度学习框架，该框架将Sentinel-2衍生的光谱指数与环境协变量相结合，用于2024年采样期环境条件下的区域土壤盐渍化空间预测。采用SHapley加法解释（SHAP）方法量化预测框架中环境变量的贡献并提高模型可解释性。开发了一种融合压缩-激励（SE）注意力机制的卷积神经网络（CNN）作为非线性特征表示框架，用于整合异质性预测因子，同时使用冠豪猪优化器（CPO）优化模型超参数。所提出的框架在基准模型中取得了最高的预测性能，R²为0.785，RMSE为0.928，MAE为0.720。将环境协变量与光谱特征相结合进一步提高了预测性能，R²从0.701提升至0.785。结果表明，在类似的沿海农业条件下，将多源环境信息与可解释深度学习相结合用于近地表土壤盐渍化空间预测具有潜力。","Sensors","2026-10-13T00:00:00Z",75,{"impact":84,"substance":85,"depth":86,"authority":87,"freshness":88,"relevant":49,"comment":89},16,21,17,13,8,"方法新颖、指标可靠，将多源遥感与可解释深度学习结合用于盐渍化空间预测，对沿海农业区土壤治理有参考价值。",[91],{"name":80,"url":77},[93,94,21,22,95],"智慧农业","农业遥感","土壤盐渍化",{"code":34,"slug":97,"name":98,"path":99,"parent_code":100},"agricultural-ai","农业AI与大模型","\u002Ftopic\u002Fdigital-agriculture\u002Fagricultural-ai","I",[102,107],{"code":103,"slug":104,"name":105,"path":106,"parent_code":100},"I3","remote-sensing","遥感与农情监测","\u002Ftopic\u002Fdigital-agriculture\u002Fremote-sensing",{"code":108,"slug":109,"name":110,"path":111,"parent_code":112},"H2","soil","土壤健康与改良","\u002Ftopic\u002Fenvironment\u002Fsoil","H",{"status":114,"locked":14,"version":29,"method":30,"evidence":115},"classified",{"candidates":116,"agricultural_evidence":36},[117,121,128,134,141],{"code":34,"score":84,"supported":36,"title":118,"abstract":119,"tags":120},[38],[22,38],[21,22],{"code":103,"score":122,"supported":36,"title":123,"abstract":125,"tags":127},15,[124],"remote sensing",[126,124],"遥感",[94],{"code":108,"score":129,"supported":36,"title":130,"abstract":131,"tags":133},14,[109],[132,109],"土壤",[],{"code":135,"score":136,"supported":36,"title":137,"abstract":138,"tags":140},"H7",3,[],[139],"干旱",[],{"code":142,"score":49,"supported":14,"title":143,"abstract":144,"tags":145},"I1",[],[],[93],[147,148],"沧州 土壤盐渍化 遥感","Sentinel-2 土壤盐分 预测","沧州土壤盐渍化遥感-4898","10.3390\u002Fs26206465",{"doi":150,"openalex_id":152,"authors":153,"venue":80,"cited_by_count":15,"oa_url":77,"card":169,"direction":173,"ingested_from":69},"W7221715654",[154,157,160,162,165,167],{"name":155,"orcid":156},"Mingyuan Ye","https:\u002F\u002Forcid.org\u002F0009-0001-6434-5505",{"name":158,"orcid":159},"Peng Chen","https:\u002F\u002Forcid.org\u002F0000-0002-2549-6765",{"name":161,"orcid":9},"Junchi Ma",{"name":163,"orcid":164},"Bingxia Liu","https:\u002F\u002Forcid.org\u002F0000-0001-7958-7112",{"name":166,"orcid":9},"Jiahao Qi",{"name":168,"orcid":9},"Yuan Jing",{"tldr":170,"method":171,"finding":172,"direction":173,"opportunity":174},"构建融合多源遥感与环境变量的可解释深度学习框架，预测沧州土壤盐分空间分布。","Sentinel-2光谱指数+环境协变量，CNN-SE注意力，CPO优化超参数，","环境协变量与光谱特征结合将R²从0.701提升至0.785，模型性能最优。","农业遥感与作物表型","可迁移至其他滨海盐渍区，并探索时序动态预测与多深度土壤盐分协同反演。","2026-10-14T23:30:27.313838Z",{"id":177,"title":178,"url":179,"summary":180,"summary_zh":181,"content":9,"source_name":182,"source_url":179,"published_at":183,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":184,"score_detail":185,"sources":190,"tags":192,"primary_topic":195,"secondary_topics":196,"classification":197,"search_phrases":226,"slug":229,"view_count":15,"doi":230,"paper":231,"created_at":255},4811,"Explainable deep learning framework for real-time insect pest detection in smart agriculture","https:\u002F\u002Fdoi.org\u002F10.1007\u002Fs44163-026-02109-y","Abstract Global losses of crop production due to insect pest infestations are estimated to be around 20–40% annually, which represents one of the highest threats to agricultural productivity, farmer livelihoods and food security. Current manual scouting is labor intensive, subjective and non-scalable, and many automated scouting methods suffer when they come to deal with different field conditions and offer little insight into decision making. To tackle these challenges, this study proposes Crop Protection in Smart Agriculture—Real-time Insect Pest Detection using Continual Spatio-Temporal Graph Convolutional Network (CPSA-RIPD-CSTGCN), a unified and interpretable deep learning framework that combines the four synergistic components: (i) the Ensemble Transform Schmidt–Kalman Filter (ETSKF) for image preprocessing, including resizing, normalization, and noise suppression; (ii) data augmentation using rotation, flipping, scaling, shearing, and contrast adjustment, which enhances generalization and reduces overfitting; (iii) the Continual Spatio-Temporal Graph Convolutional Network (CSTGCN) that jointly models the spatial dependency across crop regions and temporal evolution of pest dynamics to achieve robust feature extraction and classification; and (iv) the Revolution Optimization Algorithm (ROA) for hyperparameter optimization of CSTGCN weights to adaptively adjust the network. Further, Gradient-weighted Class Activation Mapping (Grad-CAM) visualization is introduced to give post-hoc explainability as it shows regions that are relevant to the biological process and areas that are affected by pests. This framework was tested on IP102 benchmark dataset with 60%-40%-split train-test with 10 economically important classes of pests with 97% accuracy, 98% precision, 98% recall and 98% F1 score, mean ± 0.32% accuracy, across 5-fold cross validation. Among other baseline models, these results consistently outperform VGG16 (82%), InceptionV3 (85%), MobileNetV2 (87%), DenseNet121 (88%), EfficientNetB0 (89%) and task-specific models like EDCP-DIA-RNN(Early Disease and Crop Pest—Diagnosis\u002FDetection using Image Analysis RNN) (94%), EPCY-TPA-ANN (Efficient Prediction of Crop Yield—Through Parameter Analysis ANN) (91%) and ICPM-ESA-CNN (Interactive Crop and Pest Management Enhanced Soil Analysis CNN) (90%). The results of ablation studies showed the solitary effects of each module of the Ensemble Transform Schmidt–Kalman Filter pre-processing and Revolution Optimization Algorithm optimization, and the confusion matrix analysis demonstrated high per-class reliability and only a small amount of misclassification, implying that the proposed method could be applied to real-time, large-scale smart agriculture deployments.","摘要 由于害虫侵扰导致的全球作物产量损失估计每年约为20–40%，这是对农业生产力、农民生计和粮食安全的最大威胁之一。当前的人工田间巡查劳动强度大、主观性强且难以规模化，而许多自动化巡查方法在面对不同田间条件时表现不佳，且对决策制定提供的洞见有限。为应对这些挑战，本研究提出了智慧农业中的作物保护——基于持续时空图卷积网络的实时害虫检测（CPSA-RIPD-CSTGCN），这是一个统一且可解释的深度学习框架，融合了四个协同组件：（i）集成变换施密特-卡尔曼滤波器（ETSKF），用于图像预处理，包括调整尺寸、归一化和噪声抑制；（ii）使用旋转、翻转、缩放、剪切和对比度调整进行数据增强，以增强泛化能力并减少过拟合；（iii）持续时空图卷积网络（CSTGCN），联合建模作物区域间的空间依赖性和害虫动态的时间演化，以实现鲁棒的特征提取和分类；（iv）革命优化算法（ROA），用于CSTGCN权重的超参数优化，以自适应调整网络。此外，引入梯度加权类激活映射（Grad-CAM）可视化以提供事后可解释性，因为它显示了与生物过程相关的区域以及受害虫影响的区域。该框架在IP102基准数据集上进行了测试，采用60%-40%的训练-测试划分，涵盖10个具有经济重要性的害虫类别，在5折交叉验证中达到了97%的准确率、98%的精确率、98%的召回率和98%的F1分数，准确率均值±0.32%。在其他基线模型中，这些结果持续优于VGG16（82%）、InceptionV3（85%）、MobileNetV2（87%）、DenseNet121（88%）、EfficientNetB0（89%）以及特定任务模型，如EDCP-DIA-RNN（基于图像分析的早期病害和作物害虫诊断\u002F检测RNN）（94%）、EPCY-TPA-ANN（通过参数分析人工神经网络高效预测作物产量）（91%）和ICPM-ESA-CNN（交互式作物和害虫管理增强土壤分析CNN）（90%）。消融研究的结果显示了集成变换施密特-卡尔曼滤波器预处理和革命优化算法各模块的单独效果。","Discover Artificial Intelligence","2026-10-12T00:00:00Z",80,{"impact":186,"substance":187,"depth":186,"authority":87,"freshness":188,"relevant":49,"comment":189},18,22,9,"提出可解释时空图卷积虫害实时检测框架，在IP102基准上达97%准确率并优于多种主流模型，方法新颖、数据扎实，对智慧农业植保具有参考价值。",[191],{"name":182,"url":179},[93,193,21,22,194],"农业人工智能","病虫害检测",{"code":34,"slug":97,"name":98,"path":99,"parent_code":100},[],{"status":114,"locked":14,"version":29,"method":30,"evidence":198},{"candidates":199,"agricultural_evidence":36},[200,204,209,216,220],{"code":34,"score":86,"supported":36,"title":201,"abstract":202,"tags":203},[38],[22,38],[193,21,22,194],{"code":142,"score":122,"supported":36,"title":205,"abstract":207,"tags":208},[206],"smart agriculture",[93,206],[93],{"code":210,"score":72,"supported":36,"title":211,"abstract":212,"tags":215},"A8",[],[213,214],"粮食安全","food security",[],{"code":108,"score":72,"supported":36,"title":217,"abstract":218,"tags":219},[],[132,109],[],{"code":221,"score":136,"supported":36,"title":222,"abstract":223,"tags":225},"F4",[],[224],"crop protection",[],[227,228],"IP102 害虫检测","CSTGCN 虫害识别","IP102害虫检测-4811","10.1007\u002Fs44163-026-02109-y",{"doi":230,"openalex_id":232,"authors":233,"venue":182,"cited_by_count":15,"oa_url":179,"card":250,"direction":68,"ingested_from":69},"W7221565168",[234,237,240,243,246,248],{"name":235,"orcid":236},"J. Briskilal","https:\u002F\u002Forcid.org\u002F0000-0003-4849-3249",{"name":238,"orcid":239},"M. K. Vidhyalakshmi","https:\u002F\u002Forcid.org\u002F0000-0002-6879-9897",{"name":241,"orcid":242},"Aswathy K Cherian","https:\u002F\u002Forcid.org\u002F0009-0002-3679-2583",{"name":244,"orcid":245},"S. Neduncheliyan","https:\u002F\u002Forcid.org\u002F0009-0002-0904-8695",{"name":247,"orcid":9},"Ajay Kumar",{"name":249,"orcid":9},"Rakesh Kumar",{"tldr":251,"method":252,"finding":253,"direction":66,"opportunity":254},"提出可解释深度学习框架CPSA-RIPD-CSTGCN，实现智慧农业中害虫实时检测。","用CSTGCN时空图卷积、ROA超参优化、ETSKF预处理与Grad-CAM，在","10类害虫检测准确率97%，优于VGG16、EfficientNetB0及多种任务专用模型。","可探索轻量化模型在边缘设备部署，并融合多模态田间数据提升跨域泛化能力。","2026-10-13T23:30:10.529316Z",{"id":257,"title":258,"url":259,"summary":260,"summary_zh":9,"content":9,"source_name":261,"source_url":9,"published_at":262,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":82,"score_detail":263,"sources":265,"tags":267,"primary_topic":269,"secondary_topics":275,"classification":277,"search_phrases":297,"slug":300,"view_count":49,"doi":9,"paper":301,"created_at":309},2491,"DASO-RiceNet: 细粒度水稻病害与损伤分类的双注意力序列优化网络(准确率 96.5%)","https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F18\u002F1945","台湾朝阳科技大学 Yung-Fa Huang 等 2026 年 9 月 9 日在《Agriculture》发表(专刊'Applying Artificial Intelligence to Sustainable Crop Protection: Managing Pests and Diseases')。提出 DASO-RiceNet(Dual-Attention Semantic Optimization Network)深度学习框架,采用多阶段残差骨干进行特征提取,序列双注意力模块融合通道与空间注意力以强调诊断相关特征并抑制背景信息。在 10 类水稻病害与损伤数据集上,DASO-RiceNet 准确率 0.965、宏精确率 0.963、宏召回率 0.966、宏 F1 0.964,优于所对比的 CNN 与 Transformer 基线模型。Grad-CAM 和 LIME 提供了模型预测的定性洞察。","MDPI Agriculture 2026-09-09","2026-09-08T16:00:00Z",{"impact":84,"substance":85,"depth":86,"authority":87,"freshness":88,"relevant":49,"comment":264},"细粒度水稻病害识别新框架，指标扎实且具可解释性，对作物病虫害智能诊断有参考价值。",[266],{"name":261,"url":259},[93,193,21,22,268],"水稻病害",{"code":270,"slug":271,"name":272,"path":273,"parent_code":274},"A1","rice","水稻","\u002Ftopic\u002Ffield-crops\u002Frice","A",[276],{"code":34,"slug":97,"name":98,"path":99,"parent_code":100},{"status":114,"locked":14,"version":29,"method":30,"evidence":278},{"candidates":279,"agricultural_evidence":36},[280,285,289,293],{"code":270,"score":281,"supported":36,"title":282,"abstract":283,"tags":284},12,[272],[272],[268],{"code":34,"score":188,"supported":36,"title":286,"abstract":287,"tags":288},[],[22,39],[193,21,22],{"code":221,"score":136,"supported":36,"title":290,"abstract":291,"tags":292},[],[224],[],{"code":142,"score":49,"supported":14,"title":294,"abstract":295,"tags":296},[],[],[93],[298,299],"农业人工智能 智慧农业 水稻病害 深度学习","农业人工智能 智慧农业","农业人工智能智慧农业水稻病害深度学习-2491",{"doi":9,"openalex_id":9,"authors":302,"venue":9,"cited_by_count":15,"oa_url":9,"card":303,"direction":66,"ingested_from":308},[],{"tldr":304,"method":305,"finding":306,"direction":66,"opportunity":307},"提出双注意力序列优化网络DASO-RiceNet，实现10类水稻病害与损伤的高精度分类。","多阶段残差骨干+序列双注意力（通道与空间），在10类水稻病害数据集上训练评估。","准确率96.5%、宏F1 0.964，优于CNN与Transformer基线，Grad-CAM\u002FLI","可探索轻量化部署与田间复杂背景下的跨域泛化，并结合多模态数据提升早期病害预警能力。","agent","2026-09-15T00:04:27.118669Z",{"id":311,"title":312,"url":313,"summary":314,"summary_zh":315,"content":9,"source_name":316,"source_url":313,"published_at":317,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":318,"score_detail":319,"sources":321,"tags":323,"primary_topic":326,"secondary_topics":332,"classification":334,"search_phrases":352,"slug":355,"view_count":15,"doi":356,"paper":357,"created_at":371},5387,"LIGHTWEIGHT FEATURE FUSED ATTENTION DRIVEN MULTI LEVEL HYBRID MOBILENET V2 FRAMEWORK FOR EARLY TOMATO LEAF DISEASE DETECTION","https:\u002F\u002Fdoi.org\u002F10.62441\u002Factainnovations.vi.1918","It is crucial to find and treat tomato crop diseases early and accurately to improve crop efficiency and make sure there is no scarcity of crop. However, computerized analysis is provocative due to lighting differences, compound backgrounds, and graphic likenesses among infection regions. Research work proposed a lightweight Attention Driven Multi Level Hybrid MobileNet-V2 algorithm that fuses the handcrafted texture structures with Deep Learning (DL) features, then image preprocessing is incorporated using Gaussian, Bilateral, and Guided filters to minimize noise by preserving the disease lesions. LAB color model along with Contrast Limited Adaptive Histogram Equalization (CLAHE) are used to improve the contrast. Otsu thresholding is involved to effectively separate the leaf segment from diseased parts. Local Binary Pattern (LBP) and Gray Level Co-occurrence Matrix (GLCM) descriptors are used to obtain fine-grained features, while high level features are gained by pretrained MobileNet-V2 model. Features fusion is fed to the fully connected layers of deep learning model for training and validation purpose. Results have revealed that the suggested model attains better classification performance as compared to existing models, making it an effective explanation for real time tomato leaf disease detection for smart agriculture applications.","早期准确发现并治疗番茄作物病害，对于提高作物效率、确保作物不短缺至关重要。然而，由于光照差异、复杂背景以及感染区域之间的图形相似性，计算机化分析具有挑战性。研究提出了一种轻量级注意力驱动多层混合MobileNet-V2算法，该算法将手工纹理结构与深度学习（DL）特征相融合，然后结合高斯、双边和引导滤波器进行图像预处理，以在保留病害病变的同时最小化噪声。采用LAB颜色模型和对比度受限自适应直方图均衡化（CLAHE）来提高对比度。引入Otsu阈值分割以有效分离叶片片段与病害部分。使用局部二值模式（LBP）和灰度共生矩阵（GLCM）描述符获取细粒度特征，同时通过预训练的MobileNet-V2模型获取高层特征。特征融合后输入深度学习模型的全连接层进行训练和验证。结果表明，与现有模型相比，所提出的模型取得了更好的分类性能，使其成为智慧农业应用中实时番茄叶片病害检测的有效解决方案。","Acta Innovations","2026-10-18T00:00:00Z",64,{"impact":281,"substance":186,"depth":84,"authority":13,"freshness":88,"relevant":49,"comment":320},"提出融合手工纹理特征与MobileNet-V2的轻量番茄叶病检测方法，有一定方法新颖性，但属单篇论文、应用范围有限，未达每日精选门槛。",[322],{"name":316,"url":313},[93,193,22,324,325],"番茄病害","图像识别",{"code":327,"slug":328,"name":329,"path":330,"parent_code":331},"B1","vegetables","蔬菜","\u002Ftopic\u002Fhorticulture\u002Fvegetables","B",[333],{"code":34,"slug":97,"name":98,"path":99,"parent_code":100},{"status":114,"locked":14,"version":29,"method":30,"evidence":335},{"candidates":336,"agricultural_evidence":36},[337,343,347],{"code":327,"score":122,"supported":36,"title":338,"abstract":340,"tags":342},[339],"tomato",[341,339],"番茄",[324],{"code":34,"score":188,"supported":36,"title":344,"abstract":345,"tags":346},[],[22,38],[193,325,22],{"code":142,"score":348,"supported":36,"title":349,"abstract":350,"tags":351},7,[],[93,206],[93],[353,354],"MobileNet-V2 番茄叶片病害","番茄叶病 深度学习 检测","MobileNet-V2番茄叶片病害-5387","10.62441\u002Factainnovations.vi.1918",{"doi":356,"openalex_id":358,"authors":359,"venue":316,"cited_by_count":15,"oa_url":365,"card":366,"direction":68,"ingested_from":69},"W7222977034",[360,362],{"name":361,"orcid":9},"Mrs. Vijetalaxmi A Hegde",{"name":363,"orcid":364},"Abhay Ramakant Kasetwar","https:\u002F\u002Forcid.org\u002F0000-0002-5800-8764","https:\u002F\u002Factainnovations.com\u002Findex.php\u002Fpub\u002Farticle\u002Fdownload\u002F1918\u002F981",{"tldr":367,"method":368,"finding":369,"direction":66,"opportunity":370},"提出轻量注意力多级混合MobileNet-V2框架，融合手工纹理与深度学习特征实现番茄叶片病害早期检","高斯\u002F双边\u002F引导滤波去噪，LAB+CLAHE增强，Otsu分割，LBP与GLCM","所提模型分类性能优于现有模型，可有效用于智慧农业实时番茄叶片病害检测。","可探索多作物跨域泛化、田间复杂光照下轻量模型部署与可解释性，及病害早期分级检测。","2026-10-20T23:30:10.424511Z",{"id":373,"title":374,"url":375,"summary":376,"summary_zh":377,"content":9,"source_name":378,"source_url":375,"published_at":317,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":379,"score_detail":380,"sources":382,"tags":384,"primary_topic":386,"secondary_topics":387,"classification":388,"search_phrases":410,"slug":413,"view_count":15,"doi":414,"paper":415,"created_at":423},5385,"A Comprehensive Review of Deep Learning and Computer Vision Techniques for Fruit Quality Assessment and Classification","https:\u002F\u002Fdoi.org\u002F10.25258\u002Fijddt.16.75s.151","Fruit image analysis has become an essential research area in precision agriculture, enabling automated fruit classification, quality grading, maturity estimation, disease diagnosis, yield prediction, and intelligent harvesting through advances in computer vision and artificial intelligence (AI).Recent progress in imaging technologies, large-scale benchmark datasets, and deep learning has substantially improved the accuracy, robustness, and scalability of automated fruit analysis systems.This survey presents a comprehensive and systematic review of image processing and AI techniques for fruit image analysis.It covers the complete image analysis pipeline, including image acquisition, preprocessing, enhancement, color space transformation, segmentation, feature extraction, and handcrafted texture descriptors.Furthermore, conventional machine learning algorithms are reviewed alongside state-of-the-art deep learning architectures, including Convolutional Neural Networks (CNNs), Vision Transformers (ViTs), and Vision Mamba models, for applications such as fruit recognition, quality assessment, disease detection, ripeness estimation, and object detection.Publicly available benchmark datasets, evaluation metrics, and performance comparison strategies are also summarized to facilitate objective assessment of existing approaches.In addition, the survey discusses key research challenges, including illumination variations, occlusion, complex orchard environments, limited annotated datasets, computational complexity, and domain adaptation, while highlighting emerging research trends such as self-supervised learning, multimodal learning, explainable AI, lightweight deep networks, foundation models, and edge intelligence for real-time deployment.By providing a unified review of methodologies, datasets, evaluation protocols, challenges, and future research directions, this survey offers a comprehensive reference for researchers and practitioners developing intelligent and scalable fruit image analysis systems for precision agriculture.","水果图像分析已成为精准农业中的一个重要研究领域，借助计算机视觉和人工智能（AI）的进步，实现了果实自动分类、品质分级、成熟度估计、病害诊断、产量预测和智能采收。成像技术、大规模基准数据集和深度学习的最新进展显著提升了自动化水果分析系统的准确性、鲁棒性和可扩展性。本综述对用于水果图像分析的图像处理和AI技术进行了全面而系统的回顾。内容涵盖完整的图像分析流程，包括图像采集、预处理、增强、颜色空间变换、分割、特征提取和手工纹理描述符。此外，综述了传统机器学习算法以及最先进的深度学习架构，包括卷积神经网络（CNN）、视觉Transformer（ViT）和视觉Mamba模型，用于果实识别、品质评估、病害检测、成熟度估计和目标检测等应用。同时总结了公开可用的基准数据集、评价指标和性能比较策略，以便对现有方法进行客观评估。此外，本综述讨论了关键研究挑战，包括光照变化、遮挡、复杂果园环境、标注数据集有限、计算复杂度和域适应，并重点介绍了新兴研究趋势，如自监督学习、多模态学习、可解释AI、轻量级深度网络、基础模型和面向实时部署的边缘智能。通过提供方法、数据集、评价协议、挑战和未来研究方向的统一综述，本综述为开发面向精准农业的智能且可扩展的水果图像分析系统的研究人员和实践者提供了全面的参考。","International Journal of Drug Delivery Technology",56,{"impact":88,"substance":186,"depth":84,"authority":72,"freshness":88,"relevant":49,"comment":381},"系统性综述覆盖CNN、ViT、Vision Mamba等前沿方法，信息量足，但期刊权威性一般且偏学术综述，公共价值有限，适合主题聚合而非每日精选头条。",[383],{"name":378,"url":375},[93,193,41,22,385],"水果品质检测",{"code":34,"slug":97,"name":98,"path":99,"parent_code":100},[],{"status":114,"locked":14,"version":29,"method":30,"evidence":389},{"candidates":390,"agricultural_evidence":36},[391,398,404],{"code":34,"score":35,"supported":36,"title":392,"abstract":393,"tags":397},[38,42],[394,22,395,41,396,38,39,42],"目标检测","机器学习","machine learning",[193,22,41],{"code":142,"score":348,"supported":36,"title":399,"abstract":400,"tags":403},[],[401,402],"精准农业","precision agriculture",[93],{"code":405,"score":136,"supported":36,"title":406,"abstract":407,"tags":409},"B2",[],[408],"orchard",[],[411,412],"水果品质 计算机视觉 深度学习","农业人工智能 水果品质检测 计算机视觉 智慧农业","水果品质计算机视觉深度学习-5385","10.25258\u002Fijddt.16.75s.151",{"doi":414,"openalex_id":416,"authors":417,"venue":378,"cited_by_count":15,"oa_url":375,"card":418,"direction":68,"ingested_from":69},"W7222892831",[],{"tldr":419,"method":420,"finding":421,"direction":173,"opportunity":422},"系统综述深度学习与计算机视觉在水果品质评估与分级中的全流程方法。","综述图像采集到分割、特征提取及CNN、ViT、Vision Mamba等模型与公","深度学习显著提升水果分析精度与鲁棒性，但光照、遮挡和小样本仍是主要瓶颈。","可探索自监督与轻量化模型在复杂果园遮挡、小样本场景下的边缘实时部署。","2026-10-20T23:30:10.066042Z",{"id":425,"title":426,"url":427,"summary":428,"summary_zh":429,"content":9,"source_name":430,"source_url":427,"published_at":317,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":431,"score_detail":432,"sources":434,"tags":436,"primary_topic":439,"secondary_topics":440,"classification":445,"search_phrases":461,"slug":464,"view_count":15,"doi":465,"paper":466,"created_at":483},5320,"Machine Learning for Drought Analysis and Forecasting: A Critical Review of Data Fusion, Extreme-Event Sensitivity, Explainability and Generalization","https:\u002F\u002Fdoi.org\u002F10.17485\u002Fijst\u002Fv19i36.1263","Background: Machine learning is increasingly applied to drought monitoring and forecasting, but studies differ widely in how drought is defined, what data are used, and how models are validated, which makes reported performance difficult to compare directly. Objectives: To explain why reported machine-learning drought-model performance varies across definitions, data sources and validation schemes, and to assess whether recent studies adequately address rare severe events and cross-domain transferability. Method: Journal articles published January 2018 to August 2026 were reviewed, starting from a 70-record high-relevance subset of a 359-record discovery export and extended through citation chaining. Publisher and DOI verification produced a final evidence base of 55 articles: 48 primary studies and 7 reviews. Target definition, input data, model family, validation design, metrics, class-imbalance treatment, explainability and external testing were compared across studies. Findings: Reported performance varied with the modelling task rather than with any single algorithm: Random Forest reached 97.7% AUC-ROC for agricultural-drought mapping in Queensland, SVM reached average R²=0.90 for hydrological drought in Algeria, and a CatBoost stacking model reached R²=0.9065 for remote-sensing-based drought monitoring. These scores are not directly comparable because targets, data and validation protocols differ. Recent studies increasingly test cross-product and cross-region transfer, use SHAP\u002FLIME explainability, and adopt extreme-aware losses, yet class-wise reporting of severe and extreme drought, ordinal-aware evaluation, and leakage-resistant temporal validation remain inconsistently applied across the 48 primary studies. Novelty: The review jointly evaluates predictive accuracy alongside rare-event sensitivity, ordinal drought severity, leakage-resistant validation, explainability, uncertainty and cross-domain generalization, criteria largely absent from earlier machine-learning drought reviews. Keywords: Drought Forecasting, Machine Learning, Remote Sensing, Explainable AI, Extreme Drought","背景：机器学习正日益应用于干旱监测与预报，但各研究在干旱定义、所用数据和模型验证方式上差异很大，导致所报告的模型性能难以直接比较。目标：解释为何已报告的机器学习干旱模型性能会因定义、数据来源和验证方案的不同而变化，并评估近期研究是否充分处理了罕见的严重事件和跨域可迁移性问题。方法：综述了2018年1月至2026年8月发表的期刊论文，起始于359条发现导出记录中70条高相关性记录的子集，并通过引文追溯加以扩展。经出版商和DOI核验，最终证据基础为55篇文章：48篇原始研究和7篇综述。比较了各研究的目标定义、输入数据、模型族、验证设计、指标、类别不平衡处理、可解释性和外部测试。发现：所报告的性能随建模任务而非任何单一算法而变化：随机森林在昆士兰农业干旱制图中达到97.7%的AUC-ROC，SVM在阿尔及利亚水文干旱中达到平均R²=0.90，CatBoost堆叠模型在基于遥感的干旱监测中达到R²=0.9065。这些分数不能直接比较，因为目标、数据和验证方案不同。近期研究越来越多地测试跨产品和跨区域迁移，使用SHAP\u002FLIME可解释性，并采用极端事件感知损失函数，然而在48篇原始研究中，严重和极端干旱的分类别报告、序数感知评估以及防泄漏的时间验证仍应用不一致。新颖性：本综述将预测准确性与罕见事件敏感性、序数干旱严重程度、防泄漏验证、可解释性、不确定性和跨域泛化等标准共同评估，这些标准在以往的机器学习干旱综述中大多缺失。关键词：干旱预报，机器学习，遥感，可解释人工智能，极端干旱","Indian Journal of Science and Technology",71,{"impact":281,"substance":85,"depth":86,"authority":281,"freshness":188,"relevant":49,"comment":433},"系统综述机器学习干旱分析与预测，指出评估口径不一、极端事件敏感性与可解释性不足，对农业旱情监测信息化有参考价值，但属学术综述、影响面偏细分领域。",[435],{"name":430,"url":427},[193,21,437,126,438],"极端天气","干旱监测",{"code":34,"slug":97,"name":98,"path":99,"parent_code":100},[441],{"code":135,"slug":442,"name":443,"path":444,"parent_code":112},"climate","气候适应与防灾减灾","\u002Ftopic\u002Fenvironment\u002Fclimate",{"status":114,"locked":14,"version":29,"method":30,"evidence":446},{"candidates":447,"agricultural_evidence":36},[448,452,457],{"code":34,"score":84,"supported":36,"title":449,"abstract":450,"tags":451},[396],[395,396],[193,21],{"code":135,"score":122,"supported":36,"title":453,"abstract":455,"tags":456},[454],"drought",[139,454],[438],{"code":103,"score":348,"supported":36,"title":458,"abstract":459,"tags":460},[],[126,124],[126],[462,463],"机器学习 干旱预测","遥感 干旱监测","机器学习干旱预测-5320","10.17485\u002Fijst\u002Fv19i36.1263",{"doi":465,"openalex_id":467,"authors":468,"venue":430,"cited_by_count":15,"oa_url":427,"card":478,"direction":173,"ingested_from":69},"W7222882764",[469,472,475],{"name":470,"orcid":471},"Ritik Mallik","https:\u002F\u002Forcid.org\u002F0009-0000-2785-0265",{"name":473,"orcid":474},"Dayal Kumar Behera","https:\u002F\u002Forcid.org\u002F0000-0002-9740-8066",{"name":476,"orcid":477},"Satarupa Mohanty","https:\u002F\u002Forcid.org\u002F0000-0002-1434-7158",{"tldr":479,"method":480,"finding":481,"direction":66,"opportunity":482},"综述55篇文献，比较机器学习干旱监测预报在定义、数据与验证上的差异。","系统综述2018-2026年55篇文献，比较目标定义、数据、模型、验证与可解释性","模型性能差异主要源于任务与验证设计而非算法，极端事件与泄漏防护评估仍不足。","可构建面向极端干旱的序数感知评估与抗泄漏时序验证框架，提升跨区域泛化。","2026-10-19T23:30:22.840943Z"]