[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"item-3254":3,"related-3254":68},{"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":22,"tags":24,"search_phrases":30,"slug":33,"view_count":34,"doi":35,"paper":36,"created_at":67},3254,"End-edge-cloud collaborative group phenotype recognition and diagnosis for rice seedling in vertical rice seedling cultivation","https:\u002F\u002Fdoi.org\u002F10.1016\u002Fj.compag.2026.112412","End-edge-cloud collaborative group phenotype recognition and diagnosis for rice seedling in vertical rice seedling cultivation。Computers and Electronics in Agriculture","端边云协同的水稻秧苗群体表型识别与诊断及其在立体水稻育秧中的应用。",null,"Computers and Electronics in Agriculture","2026-09-23T00:00:00Z","论文",10,false,80,{"impact":17,"substance":18,"depth":17,"authority":19,"freshness":13,"relevant":20,"comment":21},18,20,14,1,"核心期刊论文，提出端边云协同的水稻育苗群体表型识别诊断方法，方法新颖、时效性强，对智慧育苗有参考价值。",[23],{"name":10,"url":6},[25,26,27,28,29],"智慧农业","农业人工智能","边缘计算","水稻育苗","表型识别",[31,32],"水稻育苗 群体表型 识别","垂直水稻育苗 端边云协同","水稻育苗群体表型识别-3254",0,"10.1016\u002Fj.compag.2026.112412",{"doi":35,"openalex_id":37,"authors":38,"venue":10,"cited_by_count":34,"oa_url":9,"card":60,"direction":64,"ingested_from":66},"W7214024203",[39,41,44,46,48,50,52,54,57],{"name":40,"orcid":9},"Zheng Ma",{"name":42,"orcid":43},"Ruohan Wang","https:\u002F\u002Forcid.org\u002F0009-0009-7656-5132",{"name":45,"orcid":9},"Yuxuan Cai",{"name":47,"orcid":9},"Donghong Wen",{"name":49,"orcid":9},"Long Zhou",{"name":51,"orcid":9},"Yunzhe Huang",{"name":53,"orcid":9},"Yihua Liu",{"name":55,"orcid":56},"Zhongbin Su","https:\u002F\u002Forcid.org\u002F0000-0002-8966-8933",{"name":58,"orcid":59},"Hongbo Li","https:\u002F\u002Forcid.org\u002F0000-0001-7711-5744",{"tldr":61,"method":62,"finding":63,"direction":64,"opportunity":65},"提出端边云协同的水稻秧苗群体表型识别与诊断方法，用于垂直育秧场景。","端边云协同架构结合深度学习，对垂直育秧水稻秧苗群体图像进行表型识别与诊断。","端边云协同可实现水稻秧苗群体表型高效识别与诊断，支撑垂直育秧精准管理。","农业遥感与作物表型","可探索端边云协同下多尺度群体表型实时诊断与育秧环境调控闭环，弥补边缘算力与模型轻量化研究空白。","openalex","2026-09-23T23:30:01.424596Z",{"total":69,"page":20,"page_size":69,"items":70},6,[71,103,148,190,236,264],{"id":72,"title":73,"url":74,"summary":75,"summary_zh":9,"content":9,"source_name":76,"source_url":9,"published_at":77,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":78,"score_detail":79,"sources":83,"tags":85,"search_phrases":88,"slug":91,"view_count":34,"doi":92,"paper":93,"created_at":102},2852,"面向边缘部署的温室串番茄采摘机器人:YOLOv8n-BiFPN-WIoU+ROS分布式控制——Frontiers in Plant Science","https:\u002F\u002Fwww.frontiersin.org\u002Fjournals\u002Fplant-science\u002Farticles\u002F10.3389\u002Ffpls.2026.1926482\u002Ffull","论文提出了一种可在边缘部署的温室串番茄采摘机器人,集成YOLOv8n-BiFPN-WIoU视觉检测模型与基于ROS的分布式控制架构。BiFPN加权特征融合结构增强了跨尺度特征表示,WIoU改进了部分遮挡和重叠目标的定位稳健性。在独立测试集上,该模型Precision达88.534%、Recall 89.377%、F1 88.954%、mAP@0.5为92.338%、mAP@0.5:0.95为71.368%,相比基线YOLOv8n分别提升2.143、2.779、2.460、4.587和5.675个百分点。视觉检测器与ROS机器人系统集成,协调目标感知、轨道站点间运动和机械臂操作。","Frontiers in Plant Science","2026-09-14T00:00:00Z",81,{"impact":17,"substance":80,"depth":17,"authority":19,"freshness":81,"relevant":20,"comment":82},22,9,"方法改进与实测指标扎实、面向边缘部署的温室串番茄采摘机器人研究，对智慧农业具参考价值，但属细分技术进展，未达产业级突破。",[84],{"name":76,"url":74},[25,26,27,86,87],"串番茄","采摘机器人",[89,90],"农业人工智能 采摘机器人 智慧农业 边缘计算","农业人工智能 采摘机器人","农业人工智能采摘机器人智慧农业边缘计算-2852","10.3389\u002Ffpls.2026.1926482\u002Ffull",{"doi":92,"openalex_id":9,"authors":94,"venue":9,"cited_by_count":34,"oa_url":9,"card":95,"direction":99,"ingested_from":101},[],{"tldr":96,"method":97,"finding":98,"direction":99,"opportunity":100},"提出边缘部署的温室串番茄采摘机器人，融合改进YOLOv8n检测与ROS分布式控制。","YOLOv8n结合BiFPN加权特征融合与WIoU损失，集成ROS分布式控制架构","模型mAP@0.5达92.338%，较基线YOLOv8n提升4.587个百分点，可完成采摘。","智慧农业 \u002F 农业物联网","可探索轻量化模型在更多边缘设备上的泛化部署，及多机器人协同采摘调度优化。","agent","2026-09-18T00:03:30.570025Z",{"id":104,"title":105,"url":106,"summary":107,"summary_zh":108,"content":9,"source_name":76,"source_url":106,"published_at":109,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":110,"score_detail":111,"sources":116,"tags":118,"search_phrases":121,"slug":124,"view_count":34,"doi":125,"paper":126,"created_at":147},2774,"SpatioFormer: spatial perception enhancement for lightweight agricultural pest and disease detection","https:\u002F\u002Fdoi.org\u002F10.3389\u002Ffpls.2026.1925867","Introduction In precision agriculture, accurate and efficient detection of crop pests and diseases is crucial. However, existing models in complex environments are prone to insufficient spatial perception and attenuation of disease texture features, making it difficult to balance recognition accuracy and lightweighting. Methods To address this, this study proposes a lightweight spatial perception enhancement hybrid architecture, SpatioFormer. First, a Pixel-level Detail Retrieval (PDR) mechanism is designed. This mechanism leverages cross-layer dynamic routing to facilitate the fusion of deep semantic features with shallow texture features, significantly enhancing the capability to capture disease features. Second, we design a Spatially Adaptive Modulation Attention (SA-SHMA) mechanism, which utilizes large-kernel depthwise convolution to capture contextual information and combines dynamic modulation maps for fine-grained focusing, efficiently recovering spatial details, and suppressing background noise. Furthermore, this paper introduces a Context-Guided Asymmetric Gated Linear Unit (CGA-GLU), which utilizes an asymmetric design focusing on the gating branch and incorporates contextual information for guidance, enhancing the inter-channel representation capability with minimal computational overhead. Results Finally, extensive experiments on the PDDD and Tomato-Village datasets validated the effectiveness of the proposed model. The proposed model achieves a Top-1 accuracy of 81.05% on the PDDD dataset and an AP 50 of 61.53% on the Tomato-Village dataset, with testing latency on edge devices being highly competitive among existing models. Discussion Compared to existing lightweight hybrid models, SpatioFormer effectively recovers shallow spatial details and precisely suppresses complex background noise under an extremely low parameter budget. Consequently, it achieves a superior balance between practical disease localization capability and inference latency on resource-constrained edge devices.","引言 在精准农业中，准确高效地检测作物病虫害至关重要。然而，复杂环境下的现有模型容易出现空间感知不足和病害纹理特征衰减的问题，难以兼顾识别精度与轻量化。方法 为解决这一问题，本研究提出了一种轻量级空间感知增强混合架构——SpatioFormer。首先，设计了像素级细节检索（Pixel-level Detail Retrieval，PDR）机制。该机制利用跨层动态路由，促进深层语义特征与浅层纹理特征的融合，显著增强了对病害特征的捕捉能力。其次，设计了空间自适应调制注意力（Spatially Adaptive Modulation Attention，SA-SHMA）机制，该机制利用大核深度卷积捕获上下文信息，并结合动态调制图进行细粒度聚焦，高效恢复空间细节并抑制背景噪声。此外，本文引入了上下文引导非对称门控线性单元（Context-Guided Asymmetric Gated Linear Unit，CGA-GLU），其采用聚焦门控分支的非对称设计，并融入上下文信息进行引导，以极小的计算开销增强了通道间表征能力。结果 最后，在PDDD和Tomato-Village数据集上的大量实验验证了所提模型的有效性。所提模型在PDDD数据集上取得了81.05%的Top-1准确率，在Tomato-Village数据集上取得了61.53%的AP 50，其在边缘设备上的测试延迟在现有模型中极具竞争力。讨论 与现有轻量级混合模型相比，SpatioFormer在极低的参数预算下有效恢复了浅层空间细节，并精确抑制了复杂背景噪声。因此，它在实际病害定位能力与资源受限边缘设备上的推理延迟之间实现了更优的平衡。","2026-09-16T00:00:00Z",77,{"impact":112,"substance":113,"depth":17,"authority":114,"freshness":81,"relevant":20,"comment":115},16,21,13,"提出轻量化空间感知增强架构，在边缘设备上兼顾检测精度与推理延迟，方法新颖、实验扎实，对农业病虫害智能识别有参考价值。",[117],{"name":76,"url":106},[25,26,27,119,120],"番茄","病虫害检测",[122,123],"农业人工智能 病虫害检测 智慧农业 边缘计算","农业人工智能 病虫害检测","农业人工智能病虫害检测智慧农业边缘计算-2774","10.3389\u002Ffpls.2026.1925867",{"doi":125,"openalex_id":127,"authors":128,"venue":76,"cited_by_count":34,"oa_url":140,"card":141,"direction":99,"ingested_from":66},"W7213437661",[129,131,134,136,138],{"name":130,"orcid":9},"Wenbo Ma",{"name":132,"orcid":133},"Hao Sun","https:\u002F\u002Forcid.org\u002F0000-0002-6983-8149",{"name":135,"orcid":9},"Kun Zhou",{"name":137,"orcid":9},"Meichun Wang",{"name":139,"orcid":9},"Rui Fu","https:\u002F\u002Fwww.frontiersin.org\u002Fjournals\u002Fplant-science\u002Farticles\u002F10.3389\u002Ffpls.2026.1925867\u002Fpdf",{"tldr":142,"method":143,"finding":144,"direction":145,"opportunity":146},"提出轻量混合架构SpatioFormer，提升复杂环境下农作物病虫害检测的空间感知能力。","设计PDR跨层动态路由、SA-SHMA大核注意力与CGA-GLU门控，在PDDD","在极低参数量下恢复浅层空间细节并抑制背景噪声，边缘设备延迟具竞争力。","农业人工智能与决策模型","可探索将空间感知增强机制迁移至多作物多病害场景，并研究边缘端实时部署的能效优化。","2026-09-17T23:30:14.148727Z",{"id":149,"title":150,"url":151,"summary":152,"summary_zh":153,"content":9,"source_name":154,"source_url":151,"published_at":109,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":155,"score_detail":156,"sources":160,"tags":162,"search_phrases":165,"slug":168,"view_count":34,"doi":169,"paper":170,"created_at":189},2771,"Edge-AI Based Smart Pet Monitoring Framework: A Rabbit Case Study","https:\u002F\u002Fdoi.org\u002F10.64643\u002Fijirt.208534-459","Continuous Monitoring of animals is hard when the people who take care of them are not around.Changes in how they are eaten, how much they drink, the way they stand and what they do during the day can show if their health is changing or if their normal routine is different.This paper talks about an Edge AI-based Smart Pet Monitoring Framework.The goal is watching the pet's behaviour and the environment around them in time.The system keeps privacy in mind.It uses computer vision, IoT, sensors and computing that happens close to the data not in the cloud This reduces the need for internet access.A Raspberry Pi serves as the edge device running a Python and OpenCV image processing chain that uses a YOLOv9 detector to spot feeding, drinking, posture and activity states.Temperature, humidity, ultrasonic distance and water-level sensors give environment data while an Arduino control layer handles alerts using an LCD, buzzer, LEDs and a servo.Local inference gives real-time processing and improves privacy, data ownership and response time.We tested than 200 images in a controlled setting and found it works but the test is still early.This framework offers a flexible base for smart pet monitoring, which could help at home in shelters and, for vets once more tests are done.","当照顾动物的人不在身边时，对动物进行持续监测是很困难的。它们进食方式、饮水量、站立姿态以及白天活动的变化，可以反映其健康状况是否发生变化或日常规律是否出现异常。本文讨论了一种基于边缘人工智能的智能宠物监测框架。其目标是及时观察宠物的行为及其周围环境。该系统注重隐私保护。它使用计算机视觉、物联网、传感器以及靠近数据端而非云端进行的计算，这减少了对互联网接入的需求。树莓派作为边缘设备，运行Python和OpenCV图像处理流程，并使用YOLOv9检测器识别进食、饮水、姿态和活动状态。温度、湿度、超声波距离和水位传感器提供环境数据，而Arduino控制层通过LCD、蜂鸣器、LED和舵机处理警报。本地推理实现了实时处理，并改善了隐私、数据所有权和响应时间。我们在受控环境中测试了200多张图像，发现系统可以工作，但测试仍处于早期阶段。该框架为智能宠物监测提供了一个灵活的基础，在完成更多测试后，可能有助于家庭、收容所以及兽医使用。","International Journal of Innovative Research in Technology",49,{"impact":157,"substance":19,"depth":114,"authority":158,"freshness":81,"relevant":20,"comment":159},8,5,"边缘AI宠物监测框架，与农业信息化关联偏弱且测试样本仅200张，属早期探索性论文，不宜进入每日精选。",[161],{"name":154,"url":151},[25,26,27,163,164],"物联网","动物监测",[166,167],"农业人工智能 动物监测 智慧农业 边缘计算","农业人工智能 动物监测","农业人工智能动物监测智慧农业边缘计算-2771","10.64643\u002Fijirt.208534-459",{"doi":169,"openalex_id":171,"authors":172,"venue":154,"cited_by_count":34,"oa_url":183,"card":184,"direction":99,"ingested_from":66},"W7213273522",[173,175,177,179,181],{"name":174,"orcid":9},"Vinit Masale",{"name":176,"orcid":9},"Suyog Mamankar",{"name":178,"orcid":9},"Prathamesh Sawant",{"name":180,"orcid":9},"Mhaboob Ali",{"name":182,"orcid":9},"Prof. Jyoti Shrote","https:\u002F\u002Fijirt.org\u002Fpublishedpaper\u002FIJIRT208534_PAPER.pdf",{"tldr":185,"method":186,"finding":187,"direction":99,"opportunity":188},"提出基于边缘AI的宠物监测框架，用树莓派和YOLOv9识别兔子行为与环境。","树莓派边缘计算、YOLOv9、OpenCV、Arduino及温湿度超声波传感器。","200张图像测试验证了框架可行性，但需更多测试才能实际应用。","可扩展至畜禽行为健康监测，解决边缘设备算力与多目标识别精度问题。","2026-09-17T23:30:10.728310Z",{"id":191,"title":192,"url":193,"summary":194,"summary_zh":195,"content":9,"source_name":10,"source_url":193,"published_at":77,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":196,"score_detail":197,"sources":201,"tags":203,"search_phrases":206,"slug":209,"view_count":20,"doi":210,"paper":211,"created_at":235},2499,"Real-time on-tree Korla pear grading with a shared ordinal–anomaly descriptor field","https:\u002F\u002Fdoi.org\u002F10.1016\u002Fj.compag.2026.112371","On-tree grading of Korla fragrant pears couples two decisions of different nature: a gradual Grade-A\u002FGrade-B appearance transition and a Grade-C surface-defect judgement, both made under illumination change, occlusion and scale variation. Detectors that treat the fruit as a single class, and graders that place the three grades on one flat classification axis, optimise a small A\u002FB deviation and an A\u002FC defect error through the same geometry and leave no explicit account of what ordinal and defect information each spatial location should preserve. We address this by adding an explicit ordinal–anomaly descriptor field to the shared feature hierarchy. A shallow module generates three channels – ordinal tendency, defect tendency and foreground confidence – that condition four downstream detector stages, and a hierarchical head converts the two kinds of evidence into normalised grade probabilities. On the orchard-disjoint KPR-3 dataset (3826 images, 12,458 instances, three commercial orchards), the model reaches 94.0% mAP@50 and 92.6% matched-instance grade accuracy, and lowers the unsafe-pick rate from 4.8% to 2.1%. On a Jetson Orin Nano at 15 W, the INT8 engine runs at 93.1% mAP@50 and sustains 42 FPS of end-to-end pipelined throughput. Monocular sensing remains a key limitation under severe front–back fruit overlap.","库尔勒香梨的树上分级耦合了两个性质不同的决策：渐变的A级\u002FB级外观过渡，以及C级表面缺陷判定，二者均在光照变化、遮挡和尺度变化下做出。将果实视为单一类别的检测器，以及将三个等级置于同一扁平分类轴上的分级器，通过相同的几何结构优化较小的A\u002FB偏差和A\u002FC缺陷误差，却未明确说明每个空间位置应保留何种序数信息和缺陷信息。我们通过在共享特征层级中增加一个显式的序数–异常描述符场来解决这一问题。一个浅层模块生成三个通道——序数倾向、缺陷倾向和前景置信度——用以调节四个下游检测器阶段，一个层级式头部将两类证据转换为归一化等级概率。在果园不相交的KPR-3数据集（3826幅图像、12458个实例、三个商业果园）上，模型达到94.0% mAP@50和92.6%的匹配实例等级准确率，并将不安全采摘率从4.8%降至2.1%。在15 W的Jetson Orin Nano上，INT8引擎以93.1% mAP@50运行，并维持42 FPS的端到端流水线吞吐量。在严重前后果实重叠情况下，单目感知仍是主要局限。",83,{"impact":17,"substance":198,"depth":199,"authority":19,"freshness":81,"relevant":20,"comment":200},23,19,"提出序数-异常描述子场的树上香梨实时分级方法，在自建KPR-3数据集与Jetson边缘端验证，方法新颖、数据扎实，对果园采摘机器人有直接参考价值。",[202],{"name":10,"url":193},[25,26,27,204,205],"库尔勒香梨","水果分级",[207,208],"农业人工智能 库尔勒香梨 智慧农业 水果分级","农业人工智能 库尔勒香梨","农业人工智能库尔勒香梨智慧农业水果分级-2499","10.1016\u002Fj.compag.2026.112371",{"doi":210,"openalex_id":212,"authors":213,"venue":10,"cited_by_count":34,"oa_url":193,"card":230,"direction":99,"ingested_from":66},"W7212908812",[214,216,219,221,223,225,227],{"name":215,"orcid":9},"Bingyu Cao",{"name":217,"orcid":218},"Peng Zhou","https:\u002F\u002Forcid.org\u002F0000-0002-6345-5307",{"name":220,"orcid":9},"Zhikai Yang",{"name":222,"orcid":9},"Wei Chen",{"name":224,"orcid":9},"Yingchao Wang",{"name":226,"orcid":9},"Mingqi Kan",{"name":228,"orcid":229},"Haiyong Chen","https:\u002F\u002Forcid.org\u002F0000-0002-5262-4208",{"tldr":231,"method":232,"finding":233,"direction":99,"opportunity":234},"提出共享序数-异常描述子场，实现树上库尔勒香梨实时分级。","在共享特征层级加入序数、缺陷、前景三通道描述子，用KPR-3数据集训练。","mAP@50达94.0%，分级准确率92.6%，不安全采摘率从4.8%降至2.1%。","可探索多模态或深度传感融合，解决严重前后遮挡下的单目感知局限。","2026-09-15T23:30:01.440708Z",{"id":237,"title":238,"url":239,"summary":240,"summary_zh":9,"content":9,"source_name":241,"source_url":9,"published_at":242,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":243,"score_detail":244,"sources":247,"tags":249,"search_phrases":252,"slug":255,"view_count":34,"doi":9,"paper":256,"created_at":263},2498,"Bangladesh AI Model AgroVisNet Spots Crop Blight on Phones(轻量级 99.52% 准确率 0.46 MB 量化部署)","https:\u002F\u002Fagritechinsights.com\u002Findex.php\u002F2026\u002F09\u002F10\u002Fbangladesh-ai-model-spots-crop-blight-on-phones","孟加拉国研究团队发表《AgroVisNet: A lightweight Convolutional Network and the BD-PlantDX Expert-Validated Benchmark for Radish, Potato and Pointed Gourd Disease Classification》。BD-PlantDX 基准数据集包含 12432 张高分辨率田间图像,涵盖 12 类(萝卜、马铃薯、尖瓜的健康与患病状态)。AgroVisNet 采用分组瓶颈残差块配以序列通道和空间注意力机制,在 BD-PlantDX 上达到 99.52% 测试准确率和相同加权 F1 分数。仅 290572 可训练参数,比所评估的 ImageNet 预训练轻量级骨干少 8.7-16.8 倍参数量。量化部署仅 0.46 MB,单 CPU 8.40 毫秒即可分类一张图像。","Agritech Insights 2026-09-10","2026-09-09T16:00:00Z",78,{"impact":17,"substance":80,"depth":17,"authority":245,"freshness":157,"relevant":20,"comment":246},12,"孟加拉国团队提出轻量级CNN模型AgroVisNet并发布万张级田间病害基准数据集，99.52%准确率、0.46MB量化模型可在手机端8.4毫秒推理，对发展中国家小农户病害识别具有实质参考价值。",[248],{"name":241,"url":239},[25,26,27,250,251],"小农农业","作物病害识别",[253,254],"作物病害识别 农业人工智能 小农农业 智慧农业","作物病害识别 农业人工智能","作物病害识别农业人工智能小农农业智慧农业-2498",{"doi":9,"openalex_id":9,"authors":257,"venue":9,"cited_by_count":34,"oa_url":9,"card":258,"direction":145,"ingested_from":101},[],{"tldr":259,"method":260,"finding":261,"direction":145,"opportunity":262},"提出轻量网络AgroVisNet与BD-PlantDX数据集，实现手机端作物病害高精度识别。","构建12432张12类田间图像基准，用分组瓶颈残差块与注意力机制训练。","测试准确率99.52%，仅29万参数，量化后0.46MB，CPU单张8.4毫秒。","可探索跨作物跨区域泛化、田间复杂光照鲁棒性及边缘设备实时多病害检测。","2026-09-15T00:04:27.679059Z",{"id":265,"title":266,"url":267,"summary":268,"summary_zh":9,"content":9,"source_name":269,"source_url":9,"published_at":242,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":15,"score_detail":270,"sources":272,"tags":274,"search_phrases":277,"slug":280,"view_count":34,"doi":9,"paper":281,"created_at":288},2490,"Lightweight Detection of Blueberries at Different Maturity Stages in Complex Orchard Environments(YOLOv12n+SHSA2C2f+知识蒸馏)","https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F18\u002F1949","云南农业大学大数据学院联合云南省农业大数据工程技术研究中心 2026 年 9 月 10 日在《Agriculture》发表。针对小果、密集、遮挡、色彩过渡细微等复杂田间条件下的蓝莓成熟期检测难题,构建 4680 图像 60573 标注框田间数据集,重新设计 YOLOv12n 的 P3\u002FP4\u002FP5 检测结构为 P2\u002FP3\u002FP4 架构,引入轻量化 SHSA2C2f 单头自注意力特征增强模块,DINOv3 (ViT-S\u002F16) 引导的非对称知识蒸馏。精度导向 M4 模型 mAP@0.5 达 92.68%±0.04%、mAP@0.5:0.95 86.38%±0.10%,参数量仅 0.79 M;经 TensorRT FP16 转换后保留 92.75% mAP@0.5,可在 Jetson Orin Nano 上以 15.9 FPS 运行。","MDPI Agriculture 2026-09-10",{"impact":17,"substance":198,"depth":17,"authority":114,"freshness":157,"relevant":20,"comment":271},"面向复杂果园的蓝莓成熟期轻量化检测研究，方法新颖、数据规模扎实且给出边缘部署实测性能，具备较高专业参考价值。",[273],{"name":269,"url":267},[25,26,27,275,276],"目标检测","蓝莓",[278,279],"农业人工智能 智慧农业 目标检测 边缘计算","农业人工智能 智慧农业","农业人工智能智慧农业目标检测边缘计算-2490",{"doi":9,"openalex_id":9,"authors":282,"venue":9,"cited_by_count":34,"oa_url":9,"card":283,"direction":145,"ingested_from":101},[],{"tldr":284,"method":285,"finding":286,"direction":145,"opportunity":287},"提出轻量YOLOv12n改进模型，实现复杂果园中多成熟期蓝莓的实时检测。","4680张田间图像数据集，P2\u002FP3\u002FP4结构、SHSA2C2f模块与DINOv","M4模型mAP@0.5达92.68%，仅0.79M参数，Jetson Orin Nano上15.9 ","可探索多作物通用轻量检测框架，或结合成熟度计数用于产量预测与采摘决策。","2026-09-15T00:04:27.061523Z"]