[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"item-2782":3,"related-2782":101},{"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,"primary_topic":30,"secondary_topics":35,"classification":36,"search_phrases":71,"slug":74,"view_count":75,"doi":76,"paper":77,"created_at":100},2782,"Development of a Wearable Sensor-PAM for Continuous Monitoring of Photosynthetic Dynamics","https:\u002F\u002Fdoi.org\u002F10.64898\u002F2026.09.12.751108","Continuous monitoring of photosynthetic performance is essential for understanding plant responses to fluctuating environments and has important applications in plant physiology, field phenotyping, and digital agriculture. However, conventional pulse-amplitude modulation (PAM) fluorometers are primarily designed for point measurements and are not suitable for long-term monitoring while attached to intact leaves. Here, we developed Sensor-PAM, a wearable chlorophyll fluorescence measurement system capable of continuously monitoring photosynthetic dynamics from the abaxial side of a leaf. The system combines a commercially available color sensor with blue LEDs in a compact, low-cost optical design to perform PAM measurements. Chlorophyll fluorescence measured from the abaxial leaf surface showed a strong correlation with conventional adaxial measurements and accurately reflected changes in photosynthetic performance induced by chilling and high-light stress. Measurements obtained using Sensor-PAM also showed good agreement with those from a commercial PAM fluorometer across diverse plant species. Furthermore, the wearable system enabled continuous monitoring of the effective quantum yield of photosystem II [Y(II)] from the same position on the same strawberry leaf for three days under both greenhouse and outdoor conditions, successfully capturing photosynthetic responses to changing irradiance and temperature in real time. These findings establish Sensor-PAM as a wearable platform for continuous chlorophyll fluorescence monitoring, extending conventional PAM fluorometry from point-based measurements to long-term monitoring of photosynthetic dynamics under natural environmental conditions.","连续监测光合性能对于理解植物对波动环境的响应至关重要，在植物生理学、田间表型分析和数字农业中具有重要应用。然而，传统的脉冲振幅调制（PAM）荧光仪主要设计用于单点测量，不适合在附着于完整叶片的情况下进行长期监测。在此，我们开发了Sensor-PAM，一种可穿戴叶绿素荧光测量系统，能够从叶片远轴面连续监测光合动态。该系统将市售颜色传感器与蓝色LED相结合，采用紧凑、低成本的光学设计来执行PAM测量。从叶片远轴面测得的叶绿素荧光与传统近轴面测量结果高度相关，并准确反映了低温胁迫和高光胁迫引起的光合性能变化。使用Sensor-PAM获得的测量结果在不同植物物种中也与商用PAM荧光仪的结果吻合良好。此外，该可穿戴系统能够在温室和户外条件下，对同一片草莓叶片同一位置连续三天监测光系统II有效量子产率[Y(II)]，成功实时捕捉了光合作用对光照和温度变化的响应。这些发现确立了Sensor-PAM作为一种可穿戴叶绿素荧光监测平台，将传统PAM荧光测定从单点测量扩展到自然环境下光合动态的长期监测。",null,"bioRxiv (Cold Spring Harbor Laboratory)","2026-09-17T00:00:00Z","论文",10,false,81,{"impact":17,"substance":18,"depth":17,"authority":19,"freshness":13,"relevant":20,"comment":21},18,22,13,1,"可穿戴叶绿素荧光监测系统实现叶片光合动态连续监测，方法新颖、数据扎实，对作物表型与数字农业有实用价值，值得进入每日精选。",[23],{"name":10,"url":6},[25,26,27,28,29],"数字农业","智慧农业","叶绿素荧光","可穿戴传感器","光合表型",{"code":31,"slug":32,"name":33,"path":34,"parent_code":9},"B","horticulture","园艺作物","\u002Ftopic\u002Fhorticulture",[],{"status":37,"locked":14,"version":38,"method":39,"evidence":40},"partial","2026-10-22.2","rules",{"candidates":41,"agricultural_evidence":45,"reason":70},[42,50,58,65],{"code":43,"score":44,"supported":45,"title":46,"abstract":47,"tags":49},"I1",8,true,[],[25,48],"digital agriculture",[25,26],{"code":51,"score":52,"supported":45,"title":53,"abstract":54,"tags":57},"B1",6,[],[55,56],"草莓","strawberry",[],{"code":59,"score":52,"supported":45,"title":60,"abstract":61,"tags":64},"B7",[],[62,63],"温室","greenhouse",[],{"code":66,"score":20,"supported":14,"title":67,"abstract":68,"tags":69},"I5",[],[],[28],"一级领域明确，二级方向需要细分",[72,73],"可穿戴传感器 叶绿素荧光 光合表型 数字农业","可穿戴传感器 叶绿素荧光","可穿戴传感器叶绿素荧光光合表型数字农业-2782",2,"10.64898\u002F2026.09.12.751108",{"doi":76,"openalex_id":78,"authors":79,"venue":10,"cited_by_count":92,"oa_url":6,"card":93,"direction":97,"ingested_from":99},"W7213448370",[80,83,86,89],{"name":81,"orcid":82},"Ko‐ichiro Miyamoto","https:\u002F\u002Forcid.org\u002F0000-0001-7889-6714",{"name":84,"orcid":85},"Kentaro Ifuku","https:\u002F\u002Forcid.org\u002F0000-0003-0241-8008",{"name":87,"orcid":88},"Tatsuo Yoshinobu","https:\u002F\u002Forcid.org\u002F0000-0001-7993-716X",{"name":90,"orcid":91},"Kaori Kohzuma","https:\u002F\u002Forcid.org\u002F0000-0003-3899-543X",0,{"tldr":94,"method":95,"finding":96,"direction":97,"opportunity":98},"开发了可穿戴叶绿素荧光系统Sensor-PAM，实现叶片背面光合动态连续监测。","低成本颜色传感器与蓝光LED集成，进行PAM测量并验证。","背面测量与常规正面测量强相关，可连续三天监测草莓叶片Y(II)动态。","农业遥感与作物表型","可穿戴荧光传感器为田间高通量表型提供新手段，可探索多物种长期监测与数据融合。","openalex","2026-09-17T23:30:29.467723Z",{"total":52,"page":20,"page_size":52,"items":102},[103,182,276,327,386,451],{"id":104,"title":105,"url":106,"summary":107,"summary_zh":9,"content":9,"source_name":108,"source_url":9,"published_at":109,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":110,"score_detail":111,"sources":116,"tags":118,"primary_topic":123,"secondary_topics":124,"classification":125,"search_phrases":167,"slug":170,"view_count":171,"doi":9,"paper":172,"created_at":181},4199,"A digital framework for smart greenhouse tomato cultivation（智能温室番茄种植的数字化框架）","https:\u002F\u002Fwh.nstl.gov.cn\u002Fpaper_detail.html?id=c531f437ff2bcc18494127a2b1b452ec","Guo Jianjun、Zhang Beibei、Zhou Piao、Xu Yudian、Luo Shangwen、Ji Jingyu、Zhuo Yuhan、Xu Xin、Lin Linjun（仲恺农业工程学院）。针对传统温室番茄栽培中管理滞后、信息化程度低、劳动力依赖度高的问题，构建了一套融合物联网（IoT）、大数据分析、人工智能（AI）和数字孪生技术的综合数字系统。该系统能实时采集并处理温室环境数据、作物生长数据，集成专家经验知识库，通过模型驱动决策优化灌溉、施肥、病虫害防控等关键生产环节，实现温室番茄种植的精细化管理和提质增效。","Smart Agricultural Technology 2026","2026-10-01T00:00:00Z",77,{"impact":17,"substance":112,"depth":113,"authority":19,"freshness":114,"relevant":20,"comment":115},20,17,9,"核心期刊论文，融合IoT、AI与数字孪生的温室番茄数字化框架，方法新颖且具产业应用价值，值得入选每日精选。",[117],{"name":108,"url":106},[26,119,120,121,122],"农业人工智能","数字孪生","智能温室","番茄种植",{"code":31,"slug":32,"name":33,"path":34,"parent_code":9},[],{"status":37,"locked":14,"version":38,"method":39,"evidence":126},{"candidates":127,"agricultural_evidence":45,"reason":70},[128,134,138,144,149,156,162],{"code":51,"score":112,"supported":45,"title":129,"abstract":132,"tags":133},[130,131],"番茄","tomato",[130],[122],{"code":59,"score":112,"supported":45,"title":135,"abstract":136,"tags":137},[62,63],[62],[121],{"code":66,"score":52,"supported":45,"title":139,"abstract":140,"tags":143},[],[141,142],"物联网","IoT",[],{"code":43,"score":145,"supported":45,"title":146,"abstract":147,"tags":148},5,[],[120],[120,26],{"code":150,"score":151,"supported":45,"title":152,"abstract":153,"tags":155},"F3",3,[],[154],"施肥",[],{"code":157,"score":151,"supported":45,"title":158,"abstract":159,"tags":161},"H3",[],[160],"灌溉",[],{"code":163,"score":20,"supported":14,"title":164,"abstract":165,"tags":166},"I2",[],[],[119],[168,169],"仲恺农业工程学院 智能温室 番茄","数字孪生 温室 番茄种植","仲恺农业工程学院智能温室番茄-4199",4,{"doi":9,"openalex_id":9,"authors":173,"venue":9,"cited_by_count":92,"oa_url":9,"card":174,"direction":178,"ingested_from":180},[],{"tldr":175,"method":176,"finding":177,"direction":178,"opportunity":179},"构建融合IoT、大数据、AI与数字孪生的智能温室番茄种植数字化框架。","集成物联网实时采集环境与作物数据，结合专家知识库与AI模型驱动决策。","实现灌溉、施肥、病虫害防控的精细化管理，降低劳动力依赖并提质增效。","智慧农业 \u002F 农业物联网","可探索数字孪生模型与作物生长机理耦合的实时闭环调控及跨温室迁移验证。","agent","2026-10-05T00:04:56.224544Z",{"id":183,"title":184,"url":185,"summary":186,"summary_zh":187,"content":9,"source_name":188,"source_url":185,"published_at":189,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":15,"score_detail":190,"sources":193,"tags":195,"primary_topic":200,"secondary_topics":201,"classification":202,"search_phrases":243,"slug":246,"view_count":151,"doi":247,"paper":248,"created_at":275},3350,"Precision agriculture for water saving: The case of processing tomato and table grape","https:\u002F\u002Fdoi.org\u002F10.1002\u002Fppp3.70253","Societal Impact Statement Agriculture faces increasing pressure to produce high‐quality food while reducing water use under intensifying climate change and water scarcity. This study compared sensor‐based precision irrigation, drone monitoring, and near‐infrared spectroscopy with conventional farmer management in table grape and processing tomato production systems. Precision agriculture improved crop water status, photosynthetic performance, yield, and marketable quality, while reducing irrigation by 8%–15% in table grape and approximately 15% in processing tomato. These findings support wider adoption of integrated digital tools through targeted incentives and farmer training, offering a scalable strategy to strengthen long‐term water security, farm resilience, and sustainable food production globally. Summary The study aimed to evaluate the effectiveness of precision agriculture (PA) technologies, specifically sensor‐based irrigation and drones, on table grape (cv. Allison ) and processing tomato (cv. Taylor ) production compared to traditional farming methods (control). The research involved field experiments in two locations in the Puglia region, southeastern Italy, in 2023 and 2024. For table grape and processing tomato, two different irrigation managements (PA vs. Control\u002FFarmer) were compared, monitoring physiological, morphological, yield, and quality parameters. For processing tomato, drone imagery, and ground measurements were also conducted. Predictive models for fruit ripeness and quality traits of both species were also developed using near‐infrared (NIR) spectroscopy data, preprocessing techniques, and PLS regression. For table grape, the PA vines showed greater water potential stability, more uniform stomatal conductance, and higher chlorophyll content, resulting in higher and more consistent production with 8%–15% water savings. For processing tomato, PA management improved plant vegetative indicators, total and marketable yields, and reduced water consumption by approximately 15%. Three out of four calibrated models using NIR showed predictive performance suitable for future practical applications. The findings highlight the potential of PA to improve water resource utilization, crop development, yield, and fruit quality, contributing to more sustainable agricultural systems in regions facing water scarcity and climate change. Moreover, this work demonstrates how sensor‐driven irrigation of the two crops (with more equilibrated plants) directly influenced high‐accuracy predictive modelling. The integration of environmental sensors (for irrigation) and optical sensors (for quality) will represent the core of modern smart farming.","社会影响声明 在气候变化加剧和水资源短缺的背景下，农业面临着在生产高质量食品的同时减少用水量的日益增大的压力。本研究在鲜食葡萄和加工番茄生产系统中，将基于传感器的精准灌溉、无人机监测和近红外光谱技术与传统农户管理进行了比较。精准农业改善了作物水分状况、光合性能、产量和商品品质，同时在鲜食葡萄中减少了8%–15%的灌溉用水，在加工番茄中减少了约15%。这些发现支持通过有针对性的激励措施和农户培训来更广泛地采用集成数字工具，为增强全球长期水资源安全、农场韧性和可持续食品生产提供了一种可推广的策略。摘要 本研究旨在评估精准农业（PA）技术，特别是基于传感器的灌溉和无人机，在鲜食葡萄（品种Allison）和加工番茄（品种Taylor）生产中相较于传统耕作方法（对照）的有效性。研究于2023年和2024年在意大利东南部普利亚地区的两个地点进行了田间试验。对于鲜食葡萄和加工番茄，比较了两种不同的灌溉管理方式（PA vs. 对照\u002F农户），监测了生理、形态、产量和品质参数。对于加工番茄，还进行了无人机影像采集和地面测量。研究还利用近红外（NIR）光谱数据、预处理技术和PLS回归，开发了两种作物果实成熟度和品质性状的预测模型。对于鲜食葡萄，PA处理的葡萄藤表现出更强的水势稳定性、更均匀的气孔导度和更高的叶绿素含量，从而实现了更高且更稳定的产量，并节约了8%–15%的用水。对于加工番茄，PA管理改善了植株营养指标、总产量和商品产量，并减少了约15%的耗水量。使用NIR校准的四个模型中有三个显示出适合未来实际应用的预测性能。研究结果凸显了PA在改善水资源利用、作物发育、产量和果实品质方面的潜力，有助于在水资源短缺和气候变化地区构建更可持续的农业系统。此外，本研究还表明，两种作物的传感器驱动灌溉（使植株更加均衡）……","Plants People Planet","2026-09-23T00:00:00Z",{"impact":17,"substance":18,"depth":17,"authority":191,"freshness":114,"relevant":20,"comment":192},14,"意大利田间试验证实传感器与无人机精准灌溉可节水8%–15%并提升产量品质，对缺水地区智慧农业推广有实证参考价值。",[194],{"name":188,"url":185},[26,196,197,198,199],"农业遥感","精准灌溉","节水农业","近红外光谱",{"code":31,"slug":32,"name":33,"path":34,"parent_code":9},[],{"status":37,"locked":14,"version":38,"method":39,"evidence":203},{"candidates":204,"agricultural_evidence":45,"reason":70},[205,212,216,223,231,238],{"code":43,"score":206,"supported":45,"title":207,"abstract":209,"tags":211},15,[208],"precision agriculture",[210,208],"精准农业",[26],{"code":51,"score":191,"supported":45,"title":213,"abstract":214,"tags":215},[131],[130,131],[],{"code":217,"score":191,"supported":45,"title":218,"abstract":220,"tags":222},"B2",[219],"grape",[221,219],"葡萄",[],{"code":157,"score":224,"supported":45,"title":225,"abstract":226,"tags":230},11,[],[160,227,228,229],"水资源","irrigation","water resource",[197,198],{"code":232,"score":52,"supported":45,"title":233,"abstract":234,"tags":237},"H7",[],[235,236],"气候变化","climate change",[],{"code":239,"score":20,"supported":14,"title":240,"abstract":241,"tags":242},"I3",[],[],[196],[244,245],"加工番茄 精准灌溉 节水","鲜食葡萄 无人机 灌溉","加工番茄精准灌溉节水-3350","10.1002\u002Fppp3.70253",{"doi":247,"openalex_id":249,"authors":250,"venue":188,"cited_by_count":92,"oa_url":185,"card":270,"direction":178,"ingested_from":99},"W7214079746",[251,254,257,259,261,264,267],{"name":252,"orcid":253},"Giuseppe Ferrara","https:\u002F\u002Forcid.org\u002F0000-0002-2129-6723",{"name":255,"orcid":256},"Alessandro Pesole","https:\u002F\u002Forcid.org\u002F0009-0007-1292-4517",{"name":258,"orcid":9},"Rita De Marco",{"name":260,"orcid":9},"Sara Bisceglie",{"name":262,"orcid":263},"Giovanni Popeo","https:\u002F\u002Forcid.org\u002F0009-0003-6796-3880",{"name":265,"orcid":266},"Simone Pascuzzi","https:\u002F\u002Forcid.org\u002F0000-0002-6699-3485",{"name":268,"orcid":269},"Luigi Tedone","https:\u002F\u002Forcid.org\u002F0000-0003-4398-3820",{"tldr":271,"method":272,"finding":273,"direction":178,"opportunity":274},"对比传感器精准灌溉、无人机与近红外光谱和传统管理在葡萄与番茄上的节水增产效果。","意大利普利亚两年田间试验，传感器灌溉、无人机监测与NIR光谱PLS建模。","精准农业改善水分与光合状态，葡萄节水8%–15%、番茄约15%，并提高产量与品质。","可探索多源传感与NIR模型跨品种跨区域迁移，并量化农户采纳激励与培训的长期节水效益。","2026-09-24T23:30:10.070697Z",{"id":277,"title":278,"url":279,"summary":280,"summary_zh":9,"content":9,"source_name":281,"source_url":9,"published_at":282,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":110,"score_detail":283,"sources":285,"tags":287,"primary_topic":291,"secondary_topics":292,"classification":293,"search_phrases":315,"slug":318,"view_count":151,"doi":9,"paper":319,"created_at":326},3247,"From Phenotyping to Supervised Agentic Decision Support: A Review of Sensing and Artificial Intelligence for Greenhouse Strawberry Cultivation（从表型到监督智能体决策支持：温室草莓栽培的传感与人工智能综述）","https:\u002F\u002Fm2.mtmt.hu\u002Fapi\u002Fpublication\u002F37426142","MDPI Horticulturae 12(7) 765 发表综述：温室草莓栽培越来越依赖传感技术、人工智能和决策支持基础设施，但其在园艺上的价值取决于异构测量能否转化为生物学上有意义的作物状态和实际管理决策。本综述综合草莓表型、多模态传感、AI作物状态解释和监督智能体协调，提出表型到行动的温室草莓栽培框架。综述研究表明，通过成像、光谱、环境、根区和建模方法在测量和解释营养、生殖、果实品质、压力和环境作物状态方面取得实质性进展。然而，大部分文献仍强调测量精度、模型性能或基础设施能力，而较少研究验证AI输出是否改善作物响应、管理决策、工作流、资源使用或生产结果。综述区分数据采集传感技术与AI解释预测方法，并研究作物状态信息如何与实际温室决策相连接。","MDPI Horticulturae","2026-09-21T00:00:00Z",{"impact":17,"substance":112,"depth":113,"authority":19,"freshness":114,"relevant":20,"comment":284},"系统梳理温室草莓从表型感知到智能体决策支持的研究进展，指出AI输出缺乏生产结果验证这一关键缺口，对智慧园艺研究有参考价值。",[286],{"name":281,"url":279},[26,119,288,289,290],"决策支持","温室草莓","多模态传感",{"code":31,"slug":32,"name":33,"path":34,"parent_code":9},[],{"status":37,"locked":14,"version":38,"method":39,"evidence":294},{"candidates":295,"agricultural_evidence":45,"reason":70},[296,301,305,311],{"code":51,"score":297,"supported":45,"title":298,"abstract":299,"tags":300},19,[55,56],[55],[],{"code":59,"score":297,"supported":45,"title":302,"abstract":303,"tags":304},[62,63],[62],[],{"code":163,"score":17,"supported":45,"title":306,"abstract":309,"tags":310},[307,308],"artificial intelligence","decision support",[],[119,288],{"code":43,"score":20,"supported":14,"title":312,"abstract":313,"tags":314},[],[],[26],[316,317],"温室草莓 人工智能 决策支持","草莓表型 多模态传感","温室草莓人工智能决策支持-3247",{"doi":9,"openalex_id":9,"authors":320,"venue":9,"cited_by_count":92,"oa_url":9,"card":321,"direction":178,"ingested_from":180},[],{"tldr":322,"method":323,"finding":324,"direction":178,"opportunity":325},"综述温室草莓从表型感知到监督智能体决策支持的传感与AI技术，提出表型到行动框架。","综合草莓表型、多模态传感、AI状态解释与监督智能体协调的文献综述。","感知与AI解释进展显著，但少有研究验证AI输出是否真正改善作物响应与管理决策。","可研究AI决策输出对温室草莓实际管理、资源利用与产量结果的闭环验证与效果评估。","2026-09-23T00:04:33.259764Z",{"id":328,"title":329,"url":330,"summary":331,"summary_zh":9,"content":9,"source_name":332,"source_url":9,"published_at":333,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":15,"score_detail":334,"sources":336,"tags":338,"primary_topic":342,"secondary_topics":343,"classification":344,"search_phrases":373,"slug":376,"view_count":52,"doi":377,"paper":378,"created_at":385},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",{"impact":17,"substance":18,"depth":17,"authority":191,"freshness":114,"relevant":20,"comment":335},"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。",[337],{"name":332,"url":330},[26,119,339,340,341],"边缘计算","串番茄","采摘机器人",{"code":31,"slug":32,"name":33,"path":34,"parent_code":9},[],{"status":37,"locked":14,"version":38,"method":39,"evidence":345},{"candidates":346,"agricultural_evidence":45,"reason":70},[347,353,357,361,365,369],{"code":348,"score":349,"supported":45,"title":350,"abstract":351,"tags":352},"F6",12,[341],[341],[341],{"code":51,"score":224,"supported":45,"title":354,"abstract":355,"tags":356},[130],[130],[],{"code":59,"score":224,"supported":45,"title":358,"abstract":359,"tags":360},[62],[62],[],{"code":43,"score":20,"supported":14,"title":362,"abstract":363,"tags":364},[],[],[26],{"code":163,"score":20,"supported":14,"title":366,"abstract":367,"tags":368},[],[],[119],{"code":66,"score":20,"supported":14,"title":370,"abstract":371,"tags":372},[],[],[339],[374,375],"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":377,"openalex_id":9,"authors":379,"venue":9,"cited_by_count":92,"oa_url":9,"card":380,"direction":178,"ingested_from":180},[],{"tldr":381,"method":382,"finding":383,"direction":178,"opportunity":384},"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。","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。","2026-09-18T00:03:30.570025Z",{"id":387,"title":388,"url":389,"summary":390,"summary_zh":9,"content":391,"source_name":392,"source_url":9,"published_at":393,"category":394,"cover_url":9,"hotness":13,"is_selected":14,"score":395,"score_detail":396,"sources":398,"tags":400,"primary_topic":404,"secondary_topics":405,"classification":406,"search_phrases":446,"slug":449,"view_count":75,"doi":9,"paper":9,"created_at":450},1338,"池州市印发《人工智能产业发展行动方案(2026-2028)》：聚焦九华黄精、竹业、池州鳜鱼等\"人工智能+\"现代农业","https:\u002F\u002Fwww.chizhou.gov.cn\u002FOpennessContent\u002Fshow\u002F1752209.html","方案围绕九华黄精、竹业、池州鳜鱼、富硒、食用菌、茶叶等特色主导产业，以 AI 技术赋能农业全产业链，推动\"数字皖农\"从示范试点走向全域覆盖；以 AI 和数字技术赋能乡村治理和便民服务，缩小城乡\"数字鸿沟\"。","池政办〔2026〕14 号\n\n各 县、区人民政府，江南新兴产业集中区、九华山风景区、开发区、平天湖风景区管委会，市政府各部门、各直属机构：\n\n《池州市人工智能产业发展行动方案（2026-2028 年）》业经市政府 第 61 次 常务会议审议通过，现印发给你们，请认真贯彻落实。\n\n池州市人民政府办公室\n\n2026 年 8 月 20 日\n\n（此件公开发布）\n\n池州市人工智能产业发展行动方案（2026-2028 年）\n\n为贯彻落实 安徽省加快发展人工智能产业部署，推进全市人工智能 产业 高质量发展，力争 到 2028 年底，全市人工智能产业规模突破 135 亿元，拥有规模以上工业企业 80 家，培育产值 1 亿元以上企业 30 家，人工智能创新应用场景数量达到 180 个以上，特 制定本方案。\n\n一、实施底层能力强基行动\n\n**（一）****加强新型基础设施建设。**鼓励多路线、多模式建设一批城市算力基础设施，创新智能算力基础设施运营模式，鼓励发展标准化、可扩展的算力云服务。优化升级网络基础设施，合理规划千兆、万兆光网，有序推进 5G 网络向 5G-A 升级演进，前瞻布局 6G 网络。推动公共基础设施物联网感知建设，夯实 数智化发展底座。（责任单位：市数据资源局、市工业和信息化局、电信池州分公司、移动池州分公司、联通池州分公司）\n\n**（二）筑牢半导体产业发展根基。**锚定大尺寸晶圆制造、第三代半导体、先进封测、部件材料配套等细分领域，推进差异化发展，力争在人工智能芯片领域开辟特色发展新赛道。（责任单位：市发展改革委）\n\n**（三）加快行业垂类模型引育。**利用现有通用基础大模型，打造具有池州辨识度、行业竞争力的行业垂类模型。围绕先进制造、现代服务业、现代农业、政务服务等领域，依托重点企业应用场景，做精做强一批具有行业特色的专用模型。（责任单位：市科技局、市工业和信息化局、市发展改革委、市农业农村局、市数据资源局，有关市直部门）\n\n**（四）深化数据资源开发利用。**深化“数据要素×”行动，推进数据产权登 记。加快数据资产化进程，推动合规入账入表与市场化流通，拓宽数据产权登记凭证质押融资等应用场景，释放数据要素价值。培育数据标注产业，建设公共领域、重点行业高质量训练数据集。鼓励企业通过可信数据空间共享和使用产业链上下游企业数据，支持在商贸物流、医疗健康、现代农业等领域建设企业、行业可信数据空间。培育数据企业，提升数据产业整体竞争力。（责任单位：市数据资源局、市财政局、市发展改革委、市工业和信息化局、市市场监管局、市投控集团）\n\n二、实施应用赋能拓展行动\n\n**（五）强化人工智能场景创新应用。**重点在制造、商贸、物流、文化、交通、医疗、教育、农业、养老等领域谋划建设重大应用场景，鼓励创建省级人工智能应用典型案例、创新项目、优秀应用，形成一批标志性创新应用场景。（责任单位：市科技局、市发展改革委、市工业和信息化局、市数据资源局、市商务局，有关市直部门）\n\n**（六）“人工智能+”先进制造。**聚焦我市重点链主产业和未来产业，制定 AI 赋能“一业一方案”；聚焦高端化、智能化、绿色化三大方向，推动多模态大模型深度融入产业核心环节，促进 制造业转型升级；开展“AI+机器人”行动，推广智能质检、预测性维护、智能供应链、快速研发设计等技术；谋划布局具身智能、低空经济等产业，发挥我市轻量化材料、半导体、高端装备制造等多元场景应用优势。（责任单位：市发展改革委、市工业和信息化局）\n\n**（七）“人工智能+”现代农业。**围绕九华黄精、竹业、池州鳜鱼、富硒、食用菌、茶叶等特色主导产业，以 AI 技术赋能农业全产业链，推动“数字皖农”建设从示范试点走向全域覆盖，助力池州农业从“传统种养”向“智慧农业”转型。以 AI 和数字技术赋能乡村治理和便民服务，缩小城乡“数字鸿沟”，助力和美乡村建设。（责任单位：市农业农村局）\n\n**（八）“人工智能+”文旅康养。**统筹推进环九华山、主城区、沿秋浦河、环升金湖四大休闲度假康养片区 AI 赋能协同发展。将人工智能技术融入文旅康养产业，创设“AI+文旅”应用场景，创新培育数字沉浸体验、数智康养产品开发、低空经济融合新业态。加强人工智能在文旅管理中的应用，赋能 A 级旅游景区等场所的安全生产。（责任单位：市文化和旅游局）\n\n**（九）“人工智能+”教育教学。**探索人工智能与教育领域的融合创新，重点推进智能备课系统、智能教学平台、数字教师、AI 助教、自主学习系统建设。深入推进人工智能在教、学、管、研、评等方面的应用。实施教师人工智能素养提升专项计划，构建“线上与线下相融合”的培养体系。（责任单位：市教育和体育局）\n\n**（十）“人工智能+”医疗卫生。**深化“智联网医院”应用，依托医疗自然语言理解、医疗诊断治疗方案推荐等技术，提高诊疗效率、精度和服务能力。推进全市二级及以上公立医院电子病历和智慧医院建设，加强医院智慧服务与管理水平。推进 AI 医学影像辅助诊断在全市医疗机构的应用，提高诊断准确性和时效性。拓展智慧医疗应用场景，打造“智医助理”池州模式。（责任单位：市卫生健康委）\n\n**（十一）“人工智能+”城市治理。**迭代升级城市智能预警中枢，推动智能政务、应急处置、公共安全治理等方面的创新应用，打造新型智慧城市。建立跨部门协同的智能管理平台，聚焦城市级交通调度、线网优化等，打造一批典型应用场景；打造 12345 热线数据智能应用，探索构建“池州政务”智能体。推进人工智能在城市管理、应急处置等方面的应用，以 AI 技术推动“好房子、好小区、好社区、好城区”建设，全面提升城市功能品质活力和安全韧性水平。（责任单位：市数据资源局、市政府办公室、市公安局、市住房城乡建设局、市城市管理局、市应急管理局）\n\n**（十二）“人工智能+”其他。**加快 推进人工智能在 养老、司法、自然资源和规划、生态环境、林业、水利、人力资源和社会保障、市场监管、金融等领域应用。（责任单位：市民政局、市司法局、市自然资源和规划局、市生态环境局、市林业局、市水利局、市人力资源社会保障局、市市场监管局、人行池州市分行，有关市直部门）\n\n三、实施产业基础夯实行动\n\n**（十三）优化人工智能产业布局。**建立差异化特色化发展的人工智能创新格局。贵池区重点发展 AI 赋能新材料、智能制造、智慧农业、低空经济产业；东至县重点推进传统产业“智改数转网联绿色”智能化改造和产业链优化升级；石台县重点发展“AI+文旅”“AI+康养”等产业；青阳县重点发展轻量化材料、新能源电池在 AI 产品中的应用；江南新兴产业集中区重点发展 AI 在商业航天、具身智能中的应用；九华山风景区 围绕 打造世界级旅游景区和世界级文化胜地，重点发展 AI 文旅 新业态及示范场景打造；市经开区 布局 AI 芯片 产业；平天湖风景区重点发展数字文创等赛道。（责任单位：各县区人民政府、管委会）\n\n**（十四）育强人工智能企业。**建立人工智能企业培育机制，按“初创科技型企业-科技型中小企业-高新技术企业-专精特新企业-专精特新‘小巨人’企业-科技领军企业（上市培育企业）”梯次培育，壮大 人工智能 产业 创新主体。支持半导体、机器人、智能物流、智能政务等领域的人工智能企业，培育赋能行业解决方案供应商、数据服务商、人工智能服务商以及数据要素型企业。（责任单位：市科技局、市工业和信息化局）\n\n**（十五）大力招引人工智能企业。**完善人工智能产业链招商图谱 和人工智能场景清单，建立动态更新的强链、补链、延链企业库。聚焦细分赛道，采用以算引商、以数招商、定向招商、场景招商等方式，招引一批人工智能产业链关键环节、拥有自主核心技术的企业。（责任单位：市新质生产力促进服务中心）\n\n**（十六）完善创新平台体系。**坚持以平天湖科创湾为引领，打造 AI“研发+孵化+中试+产业化”全链条创新载体。围绕具身智能、半导体等前沿领域，完善省重点实验室、产业技术工程化中心、企业研发（技术）中心、科技成果转化概念验证中心、科技成果转化中试基地、制造业中试平台、科技型企业孵化器等全链条创新平台体系，加快科技成果从实验室走向产业化应用。推进与香港科技大学、上海交通大学、安徽工程大学 等高校院所的 产学研合作，围绕具身智能、先进材料、半导体领域共建产业创新平台。抢抓上海（长三角）国际科创中心扩围机遇，发挥池州（上海）科创中心平台作用，提升质效，招引人工智能领域项目和人才（团队）。（责任单位：市科技局、市新质生产力促进服务中心、市发展改革委、市工业和信息化局）\n\n四、实施发展生态构建行动\n\n**（十七）加大金融政策支持。**积极向上争取国家、省级专项资金支持，用好省人工智能池州子基金、天使基金，引导多元社会资本支持人工智能产业发展。统筹池州市新兴产业引导基金、本地科创基金等各类 基 金，加强对人工智能关键产业链项目的资金支持，精准服务企业发展需求。（责任单位：市科技局、市新质生产力促进服务中心、市投控集团）\n\n**（十八）强化高端人才供给。**通过实施重大科技项目、建设高能级科创平台等方式，吸引和汇聚一批国内外高水平创新团队。依托“万马奔池”计划，精准引进高层次人才团队。以人工智能技术应用为契机，创造更多 AI 训练师、AI 系统测试工程师、数据标注师等新职业岗位。鼓励企业与在池高校联合建设产业学院、研发中心、实训基地等，深度开展企业和高校人才交流互派。（责任单位：市委人才工作局、市人力资源社会保障局、市教育和体育局、市科技局）\n\n**（十九）提升全社会人工智能素养。**开展覆盖全学段、面向全社会的通识教育，承办省级以上人工智能展会、论坛等 各类活动，提升全民人工智能素养。强化人工智能教育培训，在党校主体班次开设人工智能课程，提升干部数字化素养与智能应用能力。开展数字化管理、智能化升级等新技能培训，培养一批人工智能专业技术人才。（责任单位：市科协、市教育和体育局、市科技局、市人力资源社会保障局、市委组织部、市委党校）\n\n**（二十）探索建立新型生产关系。**深化数据要素市场化配置改革，落实 完善公共数据授权运营和价格形成机制。探索城市“智能空间”复合开发模式，适应人工智能轻型生产、轻型研发等特征，探索建设智能产业综合体。围绕网络安全、数据安全、科技伦理等领域建立风险防控应对机制。加强人工智能产业市场监管和知识产权保护，营造公平、健康、有序的人工智能产业市场环境。（责任单位：市数据资源局、市发展改革委、市投控集团、市委网信办、市公安局、市科技局、市市场监管局）\n\n加强全市人工智能高质量发展工作统筹，由市科技局协调推进人工智能产业各项工作。各县区 政府、管委会 和 市 有关 部门要 立足行业和区域特点，强化政策支持和协同联动，推动各项任务落地落实。强化典型示范引领，注重总结工作中的好经验好做法，充分利用各类媒体加强宣传推广，营造加快“人工智能+”创新应用的浓厚氛围。\n\n如果内容不能正常显示：请安装pdf软件， 或下载本PDF文档 [[点击下载](https:\u002F\u002Fwww.chizhou.gov.cn\u002Fdownload?siteId=1&id=707664)]。","池州市人民政府 · 2026-09-01","2026-09-01T00:00:00Z","政策",70,{"impact":206,"substance":112,"depth":206,"authority":349,"freshness":44,"relevant":20,"comment":397},"池州市发布人工智能产业行动方案，明确以AI赋能九华黄精、竹业、鳜鱼等特色农业，推动智慧农业全域覆盖，政策具体、目标清晰，对地方农业数字化转型有指导意义。",[399],{"name":392,"url":389},[26,119,401,402,403],"特色产业","人工智能+","数字皖农",{"code":31,"slug":32,"name":33,"path":34,"parent_code":9},[],{"status":37,"locked":14,"version":38,"method":39,"evidence":407},{"candidates":408,"agricultural_evidence":45,"reason":70},[409,415,421,427,433,437,442],{"code":410,"score":151,"supported":45,"title":411,"abstract":412,"tags":414},"B3",[],[413],"茶叶",[],{"code":416,"score":151,"supported":45,"title":417,"abstract":418,"tags":420},"B5",[],[419],"食用菌",[],{"code":422,"score":151,"supported":45,"title":423,"abstract":424,"tags":426},"G5",[],[425],"乡村治理",[],{"code":428,"score":151,"supported":45,"title":429,"abstract":430,"tags":432},"I6",[],[431],"数字鸿沟",[],{"code":163,"score":75,"supported":14,"title":434,"abstract":435,"tags":436},[],[],[402,119],{"code":438,"score":20,"supported":14,"title":439,"abstract":440,"tags":441},"G4",[],[],[401],{"code":43,"score":20,"supported":14,"title":443,"abstract":444,"tags":445},[],[],[26],[447,448],"农业人工智能 人工智能+ 数字皖农 智慧农业","农业人工智能 人工智能+","农业人工智能人工智能+数字皖农智慧农业-1338","2026-09-02T00:05:03.287161Z",{"id":452,"title":453,"url":454,"summary":455,"summary_zh":9,"content":456,"source_name":457,"source_url":9,"published_at":458,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":459,"score_detail":460,"sources":463,"tags":465,"primary_topic":467,"secondary_topics":468,"classification":469,"search_phrases":489,"slug":492,"view_count":151,"doi":9,"paper":9,"created_at":493},706,"FSA-GhostNet:面向部署导向型作物生长监测的频域-空间自适应轻量化网络","https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1743","针对温室图像光照变化与遮挡难题,内蒙古农业大学团队构建黄瓜生长数据集 CGD 并提出轻量化骨干 FSA-GhostNet(参数2.07M,RTX 3060 上 13.96 ms\u002F张);在 CGD 上生长阶段分类准确率 98.22%、播后天数回归 MAE 0.93 天、R² 0.9962,ImageNet-100 上准确率 78.98%,为温室智慧农业边缘部署提供算力-精度均衡方案。","## 1. Introduction\n\nSmart agriculture increasingly relies on computer vision and Agricultural Internet of Things (IoT) systems to convert continuous field or greenhouse observations into actionable management information [[1](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1743#B1-agriculture-16-01743),[2](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1743#B2-agriculture-16-01743),[3](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1743#B3-agriculture-16-01743),[4](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1743#B4-agriculture-16-01743)]. Recent studies show that visual sensing platforms, including fixed cameras and unmanned aerial systems, can support crop growth monitoring, phenotyping, and production management in agricultural environments [[5](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1743#B5-agriculture-16-01743),[6](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1743#B6-agriculture-16-01743)]. In such settings, recognition models must remain computationally efficient while delivering reliable performance, because agricultural monitoring systems often operate under tight constraints on computing, memory, and energy [[7](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1743#B7-agriculture-16-01743),[8](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1743#B8-agriculture-16-01743),[9](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1743#B9-agriculture-16-01743)]. Edge-side analysis may also offer practical advantages over cloud-only processing in greenhouse monitoring [[10](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1743#B10-agriculture-16-01743),[11](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1743#B11-agriculture-16-01743)], including lower bandwidth demand for image transmission, shorter end-to-end latency [[12](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1743#B12-agriculture-16-01743)], and greater continuity when network connectivity is unstable. These conditions motivate the development of compact visual backbones, even though full hardware-level deployment validation remains beyond the scope of the present study.\n\nWithin this broader setting, greenhouse crop growth monitoring remains challenging [[13](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1743#B13-agriculture-16-01743),[14](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1743#B14-agriculture-16-01743)]. Among protected-cultivation vegetables, cucumber is a particularly relevant target for visual monitoring [[15](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1743#B15-agriculture-16-01743)] because commercial production depends on repeated assessment of plant condition, including leaf development, flowering, and fruit set, all of which inform management decisions. In the greenhouse scenario considered here, cucumber images exhibit illumination variation, partial occlusion, and cluttered backgrounds composed of soil, irrigation facilities, and reflective plastic surfaces [[16](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1743#B16-agriculture-16-01743)]. Adjacent growth states may also differ only through subtle morphological and textural cues [[17](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1743#B17-agriculture-16-01743)]. Growth monitoring is therefore sensitive to both background interference and fine-grained temporal variation. The practical question addressed in this study is how to preserve reliable growth recognition and temporal estimation within a compact visual model.\n\nExisting lightweight convolutional neural networks (CNNs), including ShuffleNet, MobileNetV3, and GhostNet-based variants, have shown favorable accuracy–efficiency trade-offs on generic benchmarks [[7](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1743#B7-agriculture-16-01743),[8](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1743#B8-agriculture-16-01743),[18](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1743#B18-agriculture-16-01743)]. More recent lightweight architectures have incorporated transformer-based or frequency-related mechanisms to strengthen contextual modeling [[19](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1743#B19-agriculture-16-01743),[20](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1743#B20-agriculture-16-01743),[21](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1743#B21-agriculture-16-01743)]. At the same time, deep learning has been widely applied in smart agriculture for disease recognition, target detection, yield estimation, and crop monitoring [[5](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1743#B5-agriculture-16-01743),[6](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1743#B6-agriculture-16-01743),[22](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1743#B22-agriculture-16-01743),[23](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1743#B23-agriculture-16-01743),[24](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1743#B24-agriculture-16-01743),[25](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1743#B25-agriculture-16-01743),[26](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1743#B26-agriculture-16-01743)]. Even so, comparatively few studies have considered lightweight backbone design, dense greenhouse crop growth monitoring, and the joint prediction of discrete growth categories and continuous developmental progress within a single framework. This gap matters in practical greenhouse monitoring, where compactness and sensitivity to temporal change are both required.\n\nAgainst this background, this study aimed to develop a lightweight visual backbone for greenhouse crop growth monitoring that improves feature representation under complex backgrounds while preserving computational compactness. Greenhouse cucumber was selected as the application scenario, and the proposed model was evaluated on both ImageNet-100 and a self-constructed Cucumber Growth Dataset (CGD). A joint learning setting was further examined for simultaneous growth-stage classification and continuous Days After Planting (DAP) regression. The main contributions of this study are as follows: (1) a Frequency–Spatial Adaptive (FSA) attention mechanism was introduced and incorporated into a lightweight backbone to support frequency–spatial feature modulation; (2) an FSA-GhostNet lightweight network was constructed by integrating FSA with enhanced lightweight modules and evaluated in terms of the accuracy–efficiency relationship; (3) a Cucumber Growth Dataset (CGD) was established for dense greenhouse cucumber growth monitoring and used for domain-specific assessment; and (4) a joint prediction framework was developed and validated for simultaneous classification and regression to support finer-grained crop growth analysis.\n\n## 2. Materials and Methods\n\nThis section describes the research design, data sources and sampling strategy, variables and instruments, model components, analytical procedure, validation strategy, and evaluation criteria used to assess FSA-GhostNet in both general-domain and agricultural crop-monitoring scenarios.\n\n### 2.1. Study Design\n\nA comparative design with two linked evaluation stages was adopted. In the first stage, ImageNet-100 served as a general-domain benchmark for assessing the classification capacity and generalization behavior of lightweight backbones under a unified protocol. In the second stage, the self-constructed Cucumber Growth Dataset (CGD) was used to evaluate performance in greenhouse crop monitoring and to extend the analysis from single-task classification to joint classification-regression prediction.\n\nThe target operating condition was periodic image-based greenhouse monitoring under resource-constrained agricultural IoT conditions. More specifically, the intended scenario involves repeated RGB image acquisition in a greenhouse environment, limited local computing and memory resources, and the need for timely model response without full dependence on continuous cloud communication. The overall workflow comprised CGD acquisition and labeling, model construction, general-domain benchmarking, greenhouse-domain validation, and joint task analysis. In the agricultural task, the analytical unit was a single image, and each image was associated with either a discrete timestamp category, a continuous Days After Planting (DAP) value, or both, depending on the experiment. All benchmark and domain-validation experiments were conducted under the same data partitioning scheme, augmentation strategy, training schedule, and computing environment.\n\n### 2.2. Dataset and Sampling\n\nTwo data sources were used: the self-constructed Cucumber Growth Dataset (CGD) for the agricultural task and ImageNet-100 for general-domain benchmark comparison.\n\n#### 2.2.1. Cucumber Growth Dataset (CGD)\n\nCGD was established to address the shortage of dense time-series image data for crop growth monitoring under realistic greenhouse conditions. Cucumber was selected as the target crop because it is an important protected-cultivation crop in the study region, exhibits clear phenotypic change across the production cycle, and requires timely management during cultivation. The dataset was therefore designed to support dense temporal observation of crop development and downstream prediction of both growth-stage category and continuous DAP. Data were collected in Shandong Province, China, a major vegetable production region with substantial demand for smart agricultural technologies. The acquisition site was a standardized temperature-controlled greenhouse of Shandong Ruyuan Chunqiu Gengdu Culture Co., Ltd., located in Xizhou Town, Qufu City, Shandong Province, China (117.056° E, 35.537° N).\n\nImages were acquired from December 2025 to February 2026 using a REDMI Turbo 4 smartphone, manufactured by Xiaomi Communications Co., Ltd., Beijing, China. Here, the smartphone served only as an image collection device under routine greenhouse conditions rather than as the target deployment platform for inference. A total of 16,840 images were collected across 22 independent acquisition sessions spanning from the first day after planting (DAP1) to the 71st day after planting (DAP71), yielding a dense temporal dataset with 22 timestamped categories. To improve comparability across time points, images were captured from fixed observation positions, primarily at eye level and from top-down views, and each image was aligned with the actual chronological growth age of the cucumber plants.\n\nFor downstream analysis, CGD provides both discrete and continuous labels. The timestamped categories were used as classification labels for growth-stage-related recognition, whereas the corresponding DAP values were retained as continuous targets for regression in the joint prediction setting. In the experimental implementation, the regression target was parsed directly from the numeric DAP value encoded in each timestamp folder name, so the target represented actual chronological age rather than an ordinal stage index. Because neighboring acquisition sessions in CGD were typically separated by short but irregular temporal intervals, most commonly about 2–4 days, the regression evaluation also considered whether a prediction fell within an acceptable 3-day tolerance window. In practical terms, the dataset was intended to support visual estimation of greenhouse crop developmental status and thereby provide a basis for time-sensitive cultivation decisions such as growth assessment, management scheduling, and harvesting planning.\n\n#### 2.2.2. ImageNet-100\n\nImageNet-100 is a randomly selected subset of ImageNet-1K (ILSVRC 2012) [[27](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1743#B27-agriculture-16-01743)] containing 100 object categories. In this study, it served as a standardized reference benchmark for head-to-head architectural comparison under a unified computational budget. The use of this dataset enabled the proposed backbone improvements to be assessed in a general visual recognition setting before domain-specific validation on greenhouse crop imagery.\n\nFor both datasets, samples were partitioned into training, validation, and testing subsets at a ratio of 6:2:2. A fixed global random seed (42) was used during data partitioning and data loading so that all comparison models were evaluated under the same underlying data distribution. For the joint prediction experiment, the same CGD split was retained and paired with the corresponding continuous DAP labels.\n\n### 2.3. Variables, Instruments, and Model Components\n\nThe independent variable was model configuration, including backbone choice in the benchmark comparisons and module composition in the ablation analysis. The dependent variables were classification accuracy, trainable parameter count, mean absolute error (MAE), root mean square error (RMSE), coefficient of determination ($R^{2}$), and the 3-Day Tolerance Rate (3DTR), depending on the experimental setting.\n\nTwo categories of instruments were involved. The first was the image acquisition device used to construct CGD, namely the REDMI Turbo 4 smartphone described above. The second was the training and evaluation platform used for model development and testing. [Table 1](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1743#table_body_display_agriculture-16-01743-t001) summarizes the hardware configuration and runtime environment used in this study. This platform corresponds to the model development and evaluation environment rather than to a physical agricultural edge device.\n\n**Table 1.** Training Hardware and Runtime Environment.\n\nThe model examined in this study was FSA-GhostNet, a lightweight visual backbone for agricultural temporal monitoring. Its main components are a hierarchical GhostNet-style backbone, the Frequency–Spatial Adaptive (FSA) attention module, an enhanced Ghost block, two context-aware activation functions, and a joint task prediction head.\n\n#### 2.3.1. FSA-GhostNet Architecture\n\nFSA-GhostNet follows a hierarchical feature extraction structure. The input image tensor is defined as\n\n$$\nX \\in \\mathbb{R}^{B \\times 3 \\times H \\times W} ,\n$$\n\n(1)\n\nwhere B denotes batch size and H and W denote the spatial height and width of the input image. The network first performs shallow feature mapping through a standard $3 \\times 3$ convolutional stem. A $3 \\times 3$ kernel was used at this stage to capture local spatial structure in the raw image while preserving a sufficiently broad receptive field for early visual encoding. The backbone then comprises four stacked FSA-GhostNetBlock stages, which progressively perform spatial downsampling and channel expansion to 32, 64, 128, and 256 channels, respectively, thereby producing multi-scale feature representations.\n\nAfter feature extraction, a global adaptive average pooling layer compresses the two-dimensional feature map into a one-dimensional representation, which is passed to a shared feature layer. The network then branches into two task-specific heads: a classification head for discrete Days After Planting (DAP) category prediction and a regression head for continuous DAP estimation. A parallel two-head design was adopted rather than a serial design so that the regression branch could learn continuous developmental progress directly from the shared representation rather than from discretized class predictions. This design reduces the dependence of regression on accumulated classification errors while preserving joint optimization of the two related tasks. The overall model structure is shown in [Figure 1](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1743#fig_body_display_agriculture-16-01743-f001).\n\n**Figure 1.** Overall workflow and model structure of FSA-GhostNet. The figure shows the global processing pipeline, while detailed module operations are described mathematically in [Section 2.3.2](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1743#sec2dot3dot2-agriculture-16-01743), [Section 2.3.3](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1743#sec2dot3dot3-agriculture-16-01743), [Section 2.3.4](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1743#sec2dot3dot4-agriculture-16-01743) and [Section 2.3.5](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1743#sec2dot3dot5-agriculture-16-01743).\n\n#### 2.3.2. Frequency–Spatial Adaptive Attention\n\nThe Frequency–Spatial Adaptive Attention (FSA) module was used to incorporate frequency-domain information into local convolutional processing. Given the input feature tensor defined above, with C denoting the number of channels, the module first applies channel compression using a bias-free $1 \\times 1$ convolution to obtain the compressed feature $X_{c}$. A $1 \\times 1$ kernel was selected here because it performs lightweight channel mixing and dimensionality reduction without introducing additional spatial aggregation, thereby reducing the computational burden of the subsequent frequency-domain transform. To improve numerical stability during mixed-precision (FP16) training, $X_{c}$ is converted to single precision (FP32) before the two-dimensional Fast Fourier Transform (2D-FFT) is applied. The mapping to the complex frequency domain is written as:\n\n$$\n\\mathcal{F} \\left(\\right. u , v \\left.\\right) = \\sum_{h = 0}^{H - 1} \\sum_{w = 0}^{W - 1} X_{c} \\left(\\right. h , w \\left.\\right) e^{- j 2 \\pi \\left(\\frac{u h}{H}+\\frac{v w}{W}\\right)} .\n$$\n\n(2)\n\nHere, $\\mathcal{F} \\left(\\right. u , v \\left.\\right)$ denotes the complex-valued frequency response at frequency coordinate $\\left(\\right. u , v \\left.\\right)$, h and w are spatial indices, and j is the imaginary unit. In physical terms, this transformation separates broad structural information and slowly varying illumination patterns from sharper local transitions such as leaf edges, vein textures, and background boundaries. In the frequency domain, the features are decomposed into a magnitude spectrum $\\mathcal{M} = \\left|\\right. \\mathcal{F} \\left|\\right.$ and a phase spectrum $\\mathcal{P} = \\angle \\mathcal{F}$. The FSA mechanism then generates adaptive low-frequency and high-frequency weights, denoted as $w_{low}$ and $w_{high}$, via Global Average Pooling (GAP) and a lightweight Multi-Layer Perceptron (MLP):\n\n$$\n\\left[w_{low},w_{high}\\right] = Softmax \\left(Conv_{1 \\times 1}\\left(ReLU\\left(Conv_{1 \\times 1}\\left(GAP\\left(\\right.X_{c}\\left.\\right)\\right)\\right)\\right)\\right) .\n$$\n\n(3)\n\nIn this expression, GAP denotes global average pooling and $w_{low}$ and $w_{high}$ denote adaptive weights for the low-frequency and high-frequency components, respectively. Based on a predefined frequency retention ratio $\\rho$ (set to 0.25 in this study), a low-frequency mask matrix $M_{low}$ and its complementary high-frequency mask matrix $M_{high} = 1 - M_{low}$ are constructed. The adaptive modulation process of the magnitude spectrum is expressed as:\n\n$$\n\\mathcal{M}^{'} = \\left(\\right. \\mathcal{M} \\bigodot M_{low} \\left.\\right) \\cdot w_{low} + \\left(\\right. \\mathcal{M} \\bigodot M_{high} \\left.\\right) \\cdot w_{high} .\n$$\n\n(4)\n\nHere, $\\mathcal{M}^{'}$ is the modulated magnitude spectrum and ⊙ denotes the Hadamard product. The modulated magnitude spectrum is then combined with the original phase spectrum $\\mathcal{P}$, and $X_{spatial}$ is reconstructed in the spatial domain via the inverse Fast Fourier Transform (iFFT). The output feature $Y_{FSA}$ is then obtained through spatial attention generation and depth-wise convolutional gating:\n\n$$\nY_{FSA} = \\left(X\\bigodot\\sigma\\left(Conv_{1 \\times 1}\\left(\\right. X_{spatial} \\left.\\right)\\right)\\right) \\bigodot \\sigma \\left(DWConv\\left(\\right.X\\left.\\right)\\right) .\n$$\n\n(5)\n\nIn this equation, $\\sigma \\left(\\right. \\cdot \\left.\\right)$ denotes the Sigmoid activation function and $DWConv \\left(\\right. \\cdot \\left.\\right)$ denotes depth-wise convolution.\n\n#### 2.3.3. Enhanced Ghost Module\n\nAn enhanced Ghost module was used to strengthen interaction between the primary and cheap branches. Given the input features of an intermediate layer, the primary branch generates intrinsic features $F_{pri}$ through standard convolution, whereas the cheap branch generates redundant features $F_{cheap}$ through depth-wise separable convolution. Rather than concatenating the two branches directly, $F_{cheap}$ is passed through a lightweight Feed-Forward Network (FFN) to generate an enhancement weight matrix $W_{enhance}$ with the same dimensions as $F_{pri}$. The refinement of the primary features is expressed as:\n\n$$\nF_{enhanced} = F_{pri} \\bigodot \\left(\\right. 1 + W_{enhan","Agriculture 2026, 16(16), 1743 (MDPI) 2026-08-14","2026-08-14T00:00:00Z",62,{"impact":349,"substance":17,"depth":461,"authority":349,"freshness":171,"relevant":20,"comment":462},16,"面向作物监测的轻量化网络研究，方法新颖但应用范围有限，时效性一般。",[464],{"name":457,"url":454},[26,119,210,62,466],"轻量化网络",{"code":31,"slug":32,"name":33,"path":34,"parent_code":9},[],{"status":37,"locked":14,"version":38,"method":39,"evidence":470},{"candidates":471,"agricultural_evidence":45,"reason":70},[472,476,481,485],{"code":43,"score":145,"supported":45,"title":473,"abstract":474,"tags":475},[],[26],[26,210],{"code":51,"score":151,"supported":45,"title":477,"abstract":478,"tags":480},[],[479],"黄瓜",[],{"code":59,"score":151,"supported":45,"title":482,"abstract":483,"tags":484},[],[62],[],{"code":163,"score":20,"supported":14,"title":486,"abstract":487,"tags":488},[],[],[119],[490,491],"农业人工智能 轻量化网络 智慧农业 精准农业","农业人工智能 轻量化网络","农业人工智能轻量化网络智慧农业精准农业-706","2026-08-18T23:59:19.085089Z"]