[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"taxonomy":3,"taxonomy-items:\u002Ftopic\u002Fhorticulture?page=4&technology=%E5%86%9C%E4%B8%9A%E4%BA%BA%E5%B7%A5%E6%99%BA%E8%83%BD":570,"taxonomy-facets:B:":1736},{"version":4,"topics":5},"2026-10-22.2",[6,71,127,199,270,331,387,433,479,533],{"code":7,"slug":8,"name":9,"path":10,"parent_code":11,"description":12,"count":13,"children":14},"I","digital-agriculture","数字农业与前沿技术","\u002Ftopic\u002Fdigital-agriculture",null,"浏览数字农业与前沿技术相关政策、产业动态与研究。",1341,[15,23,31,39,47,55,63],{"code":16,"slug":17,"name":18,"path":19,"parent_code":7,"description":20,"count":21,"children":22},"I1","smart-agriculture","智慧农业总论","\u002Ftopic\u002Fdigital-agriculture\u002Fsmart-agriculture","智慧农业总体路径、模式与数字化转型",112,[],{"code":24,"slug":25,"name":26,"path":27,"parent_code":7,"description":28,"count":29,"children":30},"I2","agricultural-ai","农业AI与大模型","\u002Ftopic\u002Fdigital-agriculture\u002Fagricultural-ai","人工智能算法、大模型与决策支持系统",634,[],{"code":32,"slug":33,"name":34,"path":35,"parent_code":7,"description":36,"count":37,"children":38},"I3","remote-sensing","遥感与农情监测","\u002Ftopic\u002Fdigital-agriculture\u002Fremote-sensing","卫星与无人机遥感、长势监测与产量估测",325,[],{"code":40,"slug":41,"name":42,"path":43,"parent_code":7,"description":44,"count":45,"children":46},"I4","gis","GIS与空间分析","\u002Ftopic\u002Fdigital-agriculture\u002Fgis","地理信息、空间统计与区域计量方法",69,[],{"code":48,"slug":49,"name":50,"path":51,"parent_code":7,"description":52,"count":53,"children":54},"I5","agricultural-iot","物联网传感与边缘计算","\u002Ftopic\u002Fdigital-agriculture\u002Fagricultural-iot","农业传感器、物联网组网与边缘计算",155,[],{"code":56,"slug":57,"name":58,"path":59,"parent_code":7,"description":60,"count":61,"children":62},"I6","digital-village","数字乡村与数据要素","\u002Ftopic\u002Fdigital-agriculture\u002Fdigital-village","数字乡村建设、数字鸿沟与数据治理",151,[],{"code":64,"slug":65,"name":66,"path":67,"parent_code":7,"description":68,"count":69,"children":70},"I7","research-methods","研究方法与评估","\u002Ftopic\u002Fdigital-agriculture\u002Fresearch-methods","综述、文献计量与因果识别方法",2,[],{"code":72,"slug":73,"name":74,"path":75,"parent_code":11,"description":76,"count":77,"children":78},"F","inputs-equipment","农业投入品与装备","\u002Ftopic\u002Finputs-equipment","浏览农业投入品与装备相关政策、产业动态与研究。",608,[79,87,95,103,111,119],{"code":80,"slug":81,"name":82,"path":83,"parent_code":72,"description":84,"count":85,"children":86},"F1","seed-industry","种业与品种管理","\u002Ftopic\u002Finputs-equipment\u002Fseed-industry","种质资源、品种审定与知识产权",162,[],{"code":88,"slug":89,"name":90,"path":91,"parent_code":72,"description":92,"count":93,"children":94},"F2","biotechnology","生物育种与基因技术","\u002Ftopic\u002Finputs-equipment\u002Fbiotechnology","分子设计育种、基因编辑与合成生物",63,[],{"code":96,"slug":97,"name":98,"path":99,"parent_code":72,"description":100,"count":101,"children":102},"F3","fertilizers","肥料与水肥一体化","\u002Ftopic\u002Finputs-equipment\u002Ffertilizers","施肥技术、化肥减量与新型肥料",88,[],{"code":104,"slug":105,"name":106,"path":107,"parent_code":72,"description":108,"count":109,"children":110},"F4","crop-protection","农药与绿色防控","\u002Ftopic\u002Finputs-equipment\u002Fcrop-protection","农药管理、绿色植保与生物防治",135,[],{"code":112,"slug":113,"name":114,"path":115,"parent_code":72,"description":116,"count":117,"children":118},"F5","machinery","农机装备与机械化","\u002Ftopic\u002Finputs-equipment\u002Fmachinery","农机装备、机械化作业与导航",121,[],{"code":120,"slug":121,"name":122,"path":123,"parent_code":72,"description":124,"count":125,"children":126},"F6","agricultural-robots","机器人与无人农场","\u002Ftopic\u002Finputs-equipment\u002Fagricultural-robots","农业机器人、无人农场与智能作业",36,[],{"code":128,"slug":129,"name":130,"path":131,"parent_code":11,"description":132,"count":133,"children":134},"H","environment","资源环境与气候变化","\u002Ftopic\u002Fenvironment","浏览资源环境与气候变化相关政策、产业动态与研究。",672,[135,143,151,159,167,175,183,191],{"code":136,"slug":137,"name":138,"path":139,"parent_code":128,"description":140,"count":141,"children":142},"H1","farmland","耕地保护与高标准农田","\u002Ftopic\u002Fenvironment\u002Ffarmland","耕地数量质量、高标准农田与农田建设",29,[],{"code":144,"slug":145,"name":146,"path":147,"parent_code":128,"description":148,"count":149,"children":150},"H2","soil","土壤健康与改良","\u002Ftopic\u002Fenvironment\u002Fsoil","土壤理化生物性状、盐碱地治理与固碳",207,[],{"code":152,"slug":153,"name":154,"path":155,"parent_code":128,"description":156,"count":157,"children":158},"H3","irrigation","水资源与灌溉","\u002Ftopic\u002Fenvironment\u002Firrigation","水资源配置、节水灌溉与旱作农业",150,[],{"code":160,"slug":161,"name":162,"path":163,"parent_code":128,"description":164,"count":165,"children":166},"H4","carbon","农业碳与绿色低碳","\u002Ftopic\u002Fenvironment\u002Fcarbon","碳排放核算、碳汇与绿色转型",44,[],{"code":168,"slug":169,"name":170,"path":171,"parent_code":128,"description":172,"count":173,"children":174},"H5","waste","面源污染与废弃物利用","\u002Ftopic\u002Fenvironment\u002Fwaste","面源污染治理与秸秆粪污等废弃物利用",8,[],{"code":176,"slug":177,"name":178,"path":179,"parent_code":128,"description":180,"count":181,"children":182},"H6","biodiversity","生态保护与生物多样性","\u002Ftopic\u002Fenvironment\u002Fbiodiversity","生态系统修复、生物多样性与土地利用变化",45,[],{"code":184,"slug":185,"name":186,"path":187,"parent_code":128,"description":188,"count":189,"children":190},"H7","climate","气候适应与防灾减灾","\u002Ftopic\u002Fenvironment\u002Fclimate","气候变化影响、气候韧性农业与灾害预警",197,[],{"code":192,"slug":193,"name":194,"path":195,"parent_code":128,"description":196,"count":197,"children":198},"H8","sustainable-agriculture","可持续与保护性农业","\u002Ftopic\u002Fenvironment\u002Fsustainable-agriculture","可持续生产模式与保护性耕作",40,[],{"code":200,"slug":201,"name":202,"path":203,"parent_code":11,"description":204,"count":205,"children":206},"A","field-crops","大田作物","\u002Ftopic\u002Ffield-crops","浏览大田作物相关政策、产业动态与研究。",1227,[207,215,223,231,239,247,255,262],{"code":208,"slug":209,"name":210,"path":211,"parent_code":200,"description":212,"count":213,"children":214},"A1","rice","水稻","\u002Ftopic\u002Ffield-crops\u002Frice","以水稻为对象的生产、育种、植保、机械化与产业内容",337,[],{"code":216,"slug":217,"name":218,"path":219,"parent_code":200,"description":220,"count":221,"children":222},"A2","wheat","小麦","\u002Ftopic\u002Ffield-crops\u002Fwheat","小麦及冬春麦区相关生产与科研",144,[],{"code":224,"slug":225,"name":226,"path":227,"parent_code":200,"description":228,"count":229,"children":230},"A3","maize","玉米","\u002Ftopic\u002Ffield-crops\u002Fmaize","玉米生产、育种、密植高产与复合种植",189,[],{"code":232,"slug":233,"name":234,"path":235,"parent_code":200,"description":236,"count":237,"children":238},"A4","soybean-oilseeds","大豆与油料","\u002Ftopic\u002Ffield-crops\u002Fsoybean-oilseeds","大豆、油菜、花生、向日葵、芝麻等油料与蛋白作物",166,[],{"code":240,"slug":241,"name":242,"path":243,"parent_code":200,"description":244,"count":245,"children":246},"A5","tubers-grains","薯类与杂粮","\u002Ftopic\u002Ffield-crops\u002Ftubers-grains","马铃薯、甘薯、高粱、谷子、荞麦等薯类杂粮",92,[],{"code":248,"slug":249,"name":250,"path":251,"parent_code":200,"description":252,"count":253,"children":254},"A6","cash-crops","棉麻糖与特色经作","\u002Ftopic\u002Ffield-crops\u002Fcash-crops","棉花、甘蔗、烟草、蚕桑等工业原料型作物",87,[],{"code":256,"slug":257,"name":258,"path":259,"parent_code":200,"description":260,"count":141,"children":261},"A7","forage-cropping","饲草与耕作制度","\u002Ftopic\u002Ffield-crops\u002Fforage-cropping","饲草产业与间套轮作、连作障碍等种植制度",[],{"code":263,"slug":264,"name":265,"path":266,"parent_code":200,"description":267,"count":268,"children":269},"A8","yield-food-security","单产提升与粮食安全","\u002Ftopic\u002Ffield-crops\u002Fyield-food-security","单产行动、产能建设与粮食安全、收购形势",143,[],{"code":271,"slug":272,"name":273,"path":274,"parent_code":11,"description":275,"count":276,"children":277},"B","horticulture","园艺作物","\u002Ftopic\u002Fhorticulture","浏览园艺作物相关政策、产业动态与研究。",531,[278,285,293,301,308,315,323],{"code":279,"slug":280,"name":281,"path":282,"parent_code":271,"description":283,"count":189,"children":284},"B1","vegetables","蔬菜","\u002Ftopic\u002Fhorticulture\u002Fvegetables","蔬菜育种、栽培、病虫害与产业",[],{"code":286,"slug":287,"name":288,"path":289,"parent_code":271,"description":290,"count":291,"children":292},"B2","fruit","果树","\u002Ftopic\u002Fhorticulture\u002Ffruit","苹果、柑橘、猕猴桃、葡萄、热带亚热带果树",147,[],{"code":294,"slug":295,"name":296,"path":297,"parent_code":271,"description":298,"count":299,"children":300},"B3","tea-coffee","茶叶与咖啡","\u002Ftopic\u002Fhorticulture\u002Ftea-coffee","茶叶、咖啡、可可等饮品原料作物",51,[],{"code":302,"slug":303,"name":304,"path":305,"parent_code":271,"description":306,"count":125,"children":307},"B4","flowers","花卉与观赏植物","\u002Ftopic\u002Fhorticulture\u002Fflowers","花卉产业、月季等观赏园艺",[],{"code":309,"slug":310,"name":311,"path":312,"parent_code":271,"description":313,"count":141,"children":314},"B5","mushrooms","食用菌","\u002Ftopic\u002Fhorticulture\u002Fmushrooms","食用菌菌种、栽培与产业",[],{"code":316,"slug":317,"name":318,"path":319,"parent_code":271,"description":320,"count":321,"children":322},"B6","medicinal-plants","中药材与药食同源","\u002Ftopic\u002Fhorticulture\u002Fmedicinal-plants","道地药材、药用植物与药食同源开发",22,[],{"code":324,"slug":325,"name":326,"path":327,"parent_code":271,"description":328,"count":329,"children":330},"B7","protected-horticulture","设施园艺与植物工厂","\u002Ftopic\u002Fhorticulture\u002Fprotected-horticulture","设施化保护地生产与植物工厂、无土栽培、都市农业",67,[],{"code":332,"slug":333,"name":334,"path":335,"parent_code":11,"description":336,"count":337,"children":338},"C","livestock","畜牧兽医","\u002Ftopic\u002Flivestock","浏览畜牧兽医相关政策、产业动态与研究。",198,[339,347,355,363,371,379],{"code":340,"slug":341,"name":342,"path":343,"parent_code":332,"description":344,"count":345,"children":346},"C1","pigs","生猪","\u002Ftopic\u002Flivestock\u002Fpigs","生猪养殖、产能与产业形势",14,[],{"code":348,"slug":349,"name":350,"path":351,"parent_code":332,"description":352,"count":353,"children":354},"C2","ruminants","草食牲畜","\u002Ftopic\u002Flivestock\u002Fruminants","奶牛、肉牛、羊养殖与奶业",47,[],{"code":356,"slug":357,"name":358,"path":359,"parent_code":332,"description":360,"count":361,"children":362},"C3","poultry","家禽","\u002Ftopic\u002Flivestock\u002Fpoultry","蛋鸡、肉鸡、水禽养殖",18,[],{"code":364,"slug":365,"name":366,"path":367,"parent_code":332,"description":368,"count":369,"children":370},"C4","feed","饲料与饲草","\u002Ftopic\u002Flivestock\u002Ffeed","饲料安全、蛋白饲料与饲草供给",35,[],{"code":372,"slug":373,"name":374,"path":375,"parent_code":332,"description":376,"count":377,"children":378},"C5","animal-health","疫病防控与兽药","\u002Ftopic\u002Flivestock\u002Fanimal-health","动物疫病、疫苗、兽药与人兽共患病",50,[],{"code":380,"slug":381,"name":382,"path":383,"parent_code":332,"description":384,"count":385,"children":386},"C6","smart-livestock","养殖环境与智慧畜牧","\u002Ftopic\u002Flivestock\u002Fsmart-livestock","粪污资源化、动物福利与智能化养殖",10,[],{"code":388,"slug":389,"name":390,"path":391,"parent_code":11,"description":392,"count":197,"children":393},"D","aquaculture","渔业水产","\u002Ftopic\u002Faquaculture","浏览渔业水产相关政策、产业动态与研究。",[394,402,410,418,426],{"code":395,"slug":396,"name":397,"path":398,"parent_code":388,"description":399,"count":400,"children":401},"D1","pond-factory-aquaculture","池塘与工厂化养殖","\u002Ftopic\u002Faquaculture\u002Fpond-factory-aquaculture","池塘养殖、循环水与设施渔业",4,[],{"code":403,"slug":404,"name":405,"path":406,"parent_code":388,"description":407,"count":408,"children":409},"D2","offshore-aquaculture","深远海养殖与海洋牧场","\u002Ftopic\u002Faquaculture\u002Foffshore-aquaculture","深远海养殖工船、海洋牧场与渔业资源",5,[],{"code":411,"slug":412,"name":413,"path":414,"parent_code":388,"description":415,"count":416,"children":417},"D3","integrated-aquaculture","稻渔种养与生态渔业","\u002Ftopic\u002Faquaculture\u002Fintegrated-aquaculture","稻渔综合种养、增殖渔业与生态养殖",12,[],{"code":419,"slug":420,"name":421,"path":422,"parent_code":388,"description":423,"count":424,"children":425},"D4","aquatic-health","水质与病害","\u002Ftopic\u002Faquaculture\u002Faquatic-health","养殖水质调控与水生动物病害防控",0,[],{"code":427,"slug":428,"name":429,"path":430,"parent_code":388,"description":431,"count":69,"children":432},"D5","smart-aquaculture","智慧渔业","\u002Ftopic\u002Faquaculture\u002Fsmart-aquaculture","渔业数字化、智能投喂与养殖装备",[],{"code":434,"slug":435,"name":436,"path":437,"parent_code":11,"description":438,"count":439,"children":440},"E","food-supply","农产品加工与流通","\u002Ftopic\u002Ffood-supply","浏览农产品加工与流通相关政策、产业动态与研究。",107,[441,449,457,464,471],{"code":442,"slug":443,"name":444,"path":445,"parent_code":434,"description":446,"count":447,"children":448},"E1","food-processing","加工与食品制造","\u002Ftopic\u002Ffood-supply\u002Ffood-processing","初加工、精深加工与食品制造工艺",25,[],{"code":450,"slug":451,"name":452,"path":453,"parent_code":434,"description":454,"count":455,"children":456},"E2","nutrition","营养与功能食品","\u002Ftopic\u002Ffood-supply\u002Fnutrition","膳食营养、功能成分与大食物观",1,[],{"code":458,"slug":459,"name":460,"path":461,"parent_code":434,"description":462,"count":416,"children":463},"E3","cold-chain","冷链仓储物流","\u002Ftopic\u002Ffood-supply\u002Fcold-chain","保鲜包装、冷链与粮食仓储物流",[],{"code":465,"slug":466,"name":467,"path":468,"parent_code":434,"description":469,"count":369,"children":470},"E4","food-safety","质量安全与溯源","\u002Ftopic\u002Ffood-supply\u002Ffood-safety","质量安全检测、农药残留与全链条追溯",[],{"code":472,"slug":473,"name":474,"path":475,"parent_code":434,"description":476,"count":477,"children":478},"E5","branding-commerce","品牌流通与电商","\u002Ftopic\u002Ffood-supply\u002Fbranding-commerce","品牌建设、产销对接与农村电商",33,[],{"code":480,"slug":481,"name":482,"path":483,"parent_code":11,"description":484,"count":485,"children":486},"G","rural-services","农业服务与乡村发展","\u002Ftopic\u002Frural-services","浏览农业服务与乡村发展相关政策、产业动态与研究。",344,[487,495,502,510,518,526],{"code":488,"slug":489,"name":490,"path":491,"parent_code":480,"description":492,"count":493,"children":494},"G1","extension-transfer","农技推广与成果转化","\u002Ftopic\u002Frural-services\u002Fextension-transfer","农技推广体系、科技特派员与科技小院",139,[],{"code":496,"slug":497,"name":498,"path":499,"parent_code":480,"description":500,"count":439,"children":501},"G2","farm-services","社会化服务与规模经营","\u002Ftopic\u002Frural-services\u002Ffarm-services","托管服务、土地流转与新型经营主体",[],{"code":503,"slug":504,"name":505,"path":506,"parent_code":480,"description":507,"count":508,"children":509},"G3","rural-finance","农村金融与保险","\u002Ftopic\u002Frural-services\u002Frural-finance","信贷担保、农业保险与财政支农",20,[],{"code":511,"slug":512,"name":513,"path":514,"parent_code":480,"description":515,"count":516,"children":517},"G4","rural-industry","乡村产业与县域经济","\u002Ftopic\u002Frural-services\u002Frural-industry","乡村产业、三产融合与县域经济",28,[],{"code":519,"slug":520,"name":521,"path":522,"parent_code":480,"description":523,"count":524,"children":525},"G5","rural-governance","乡村治理与人才","\u002Ftopic\u002Frural-services\u002Frural-governance","乡村治理、数字治理与人才培养",11,[],{"code":527,"slug":528,"name":529,"path":530,"parent_code":480,"description":531,"count":321,"children":532},"G6","rural-integration","帮扶与城乡融合","\u002Ftopic\u002Frural-services\u002Frural-integration","常态化帮扶、区域协作与农民增收",[],{"code":534,"slug":535,"name":536,"path":537,"parent_code":11,"description":538,"count":539,"children":540},"J","policy","政策与宏观","\u002Ftopic\u002Fpolicy","浏览政策与宏观相关政策、产业动态与研究。",132,[541,548,555,562],{"code":542,"slug":543,"name":544,"path":545,"parent_code":534,"description":546,"count":408,"children":547},"J1","policy-planning","政策制度与发展规划","\u002Ftopic\u002Fpolicy\u002Fpolicy-planning","部委政策、规划文件与解读",[],{"code":549,"slug":550,"name":551,"path":552,"parent_code":534,"description":553,"count":447,"children":554},"J2","modernization","农业强国与现代化","\u002Ftopic\u002Fpolicy\u002Fmodernization","农业强国、农业农村现代化与新质生产力",[],{"code":556,"slug":557,"name":558,"path":559,"parent_code":534,"description":560,"count":400,"children":561},"J3","global-agriculture","国际合作与全球农业","\u002Ftopic\u002Fpolicy\u002Fglobal-agriculture","国际农业合作与国别农业研究",[],{"code":563,"slug":564,"name":565,"path":566,"parent_code":534,"description":567,"count":568,"children":569},"J4","seasonal-agriculture","农时与节庆活动","\u002Ftopic\u002Fpolicy\u002Fseasonal-agriculture","农时生产调度、丰收节与消费季",98,[],{"total":571,"page":400,"page_size":508,"items":572},100,[573,657,713,776,864,926,975,1044,1109,1159,1198,1240,1311,1374,1428,1493,1545,1588,1637,1684],{"id":574,"title":575,"url":576,"summary":577,"summary_zh":578,"content":11,"source_name":579,"source_url":576,"published_at":580,"category":581,"cover_url":11,"hotness":385,"is_selected":582,"score":583,"score_detail":584,"sources":587,"tags":589,"primary_topic":595,"secondary_topics":596,"classification":597,"search_phrases":621,"slug":624,"view_count":69,"doi":625,"paper":626,"created_at":656},2741,"A tomato maturity detection method against occlusion and variable illumination","https:\u002F\u002Fdoi.org\u002F10.1016\u002Fj.compag.2026.112423","A tomato maturity detection method against occlusion and variable illumination。Computers and Electronics in Agriculture","一种抗遮挡和可变光照的番茄成熟度检测方法","Computers and Electronics in Agriculture","2026-09-16T00:00:00Z","论文",false,68,{"impact":416,"substance":361,"depth":585,"authority":345,"freshness":173,"relevant":455,"comment":586},16,"核心期刊论文，针对遮挡与光照变化下的番茄成熟度检测提出新方法，属农业人工智能细分领域实质进展，但应用范围有限，未达重大突破层级。",[588],{"name":579,"url":576},[590,591,592,593,594],"智慧农业","农业人工智能","目标检测","番茄","作物表型",{"code":279,"slug":280,"name":281,"path":282,"parent_code":271},[],{"status":598,"locked":582,"version":4,"method":599,"evidence":600},"classified","rules",{"candidates":601,"agricultural_evidence":604},[602,609,613,617],{"code":279,"score":603,"supported":604,"title":605,"abstract":607,"tags":608},15,true,[606],"tomato",[593,606],[593],{"code":24,"score":69,"supported":582,"title":610,"abstract":611,"tags":612},[],[],[591,592],{"code":16,"score":455,"supported":582,"title":614,"abstract":615,"tags":616},[],[],[590],{"code":32,"score":455,"supported":582,"title":618,"abstract":619,"tags":620},[],[],[594],[622,623],"农业人工智能 作物表型 智慧农业 目标检测","农业人工智能 作物表型","农业人工智能作物表型智慧农业目标检测-2741","10.1016\u002Fj.compag.2026.112423",{"doi":625,"openalex_id":627,"authors":628,"venue":579,"cited_by_count":424,"oa_url":11,"card":649,"direction":653,"ingested_from":655},"W7213429689",[629,632,635,638,641,644,646],{"name":630,"orcid":631},"Hao Meng","https:\u002F\u002Forcid.org\u002F0000-0001-7511-2910",{"name":633,"orcid":634},"Wenzhe Li","https:\u002F\u002Forcid.org\u002F0009-0008-1630-4697",{"name":636,"orcid":637},"Di Wang","https:\u002F\u002Forcid.org\u002F0000-0002-3911-8159",{"name":639,"orcid":640},"Hui Zhao","https:\u002F\u002Forcid.org\u002F0009-0005-4192-5776",{"name":642,"orcid":643},"Ximing Li","https:\u002F\u002Forcid.org\u002F0000-0003-4022-1273",{"name":645,"orcid":11},"Dongdong Cui",{"name":647,"orcid":648},"Fernando Auat Cheein","https:\u002F\u002Forcid.org\u002F0000-0002-6347-7696",{"tldr":650,"method":651,"finding":652,"direction":653,"opportunity":654},"提出一种抗遮挡和光照变化的番茄成熟度检测方法。","基于深度学习的图像检测，针对遮挡与光照变化优化。","该方法在遮挡和变光照下仍能准确检测番茄成熟度。","农业人工智能与决策模型","可探索多模态融合与轻量化部署，提升田间复杂场景实时检测鲁棒性。","openalex","2026-09-17T23:30:01.491379Z",{"id":658,"title":659,"url":660,"summary":661,"summary_zh":11,"content":11,"source_name":662,"source_url":11,"published_at":11,"category":581,"cover_url":11,"hotness":385,"is_selected":582,"score":583,"score_detail":663,"sources":667,"tags":669,"primary_topic":673,"secondary_topics":674,"classification":675,"search_phrases":699,"slug":702,"view_count":683,"doi":11,"paper":703,"created_at":712},2733,"Non-Destructive Sensing and Modeling for Biomass Estimation and Yield Prediction of Protected Vegetables——综述","https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F18\u002F1980","江苏大学Xiaodong Zhang等综述了设施蔬菜生物量与产量无损感知与建模研究进展,聚焦叶菜与果菜类设施栽培,总结了生物量与产量指标及地面真实测量方法,比较了RGB成像、三维视觉、光谱感知和环境数据的特点,综述了叶菜生物量估算、连续生长监测、收获预测以及果菜花果感知、果品计数、单果质量估算、分阶段收获产量预测等进展。","Agriculture (MDPI) 2026年第16卷18期",{"impact":585,"substance":508,"depth":664,"authority":665,"freshness":69,"relevant":455,"comment":666},17,13,"系统梳理设施蔬菜生物量与产量无损感知建模进展，方法覆盖RGB、三维视觉与光谱，专业参考价值较高，但属综述且时效性弱，适合主题聚合而非每日精选。",[668],{"name":662,"url":660},[590,591,670,671,672],"产量预测","设施蔬菜","无损感知",{"code":279,"slug":280,"name":281,"path":282,"parent_code":271},[],{"status":598,"locked":582,"version":4,"method":599,"evidence":676},{"candidates":677,"agricultural_evidence":604},[678,682,687,691,695],{"code":279,"score":400,"supported":604,"title":679,"abstract":680,"tags":681},[],[281],[671],{"code":324,"score":683,"supported":604,"title":684,"abstract":685,"tags":686},3,[],[671],[],{"code":16,"score":455,"supported":582,"title":688,"abstract":689,"tags":690},[],[],[590],{"code":24,"score":455,"supported":582,"title":692,"abstract":693,"tags":694},[],[],[591],{"code":32,"score":455,"supported":582,"title":696,"abstract":697,"tags":698},[],[],[670],[700,701],"农业人工智能 产量预测 无损感知 智慧农业","农业人工智能 产量预测","农业人工智能产量预测无损感知智慧农业-2733",{"doi":11,"openalex_id":11,"authors":704,"venue":11,"cited_by_count":424,"oa_url":11,"card":705,"direction":709,"ingested_from":711},[],{"tldr":706,"method":707,"finding":708,"direction":709,"opportunity":710},"综述设施蔬菜生物量与产量无损感知与建模研究进展。","综述RGB成像、三维视觉、光谱感知及环境数据等方法。","总结了叶菜与果菜生物量估算、生长监测和产量预测的技术进展。","智慧农业 \u002F 农业物联网","多模态感知融合与跨品种泛化模型是设施蔬菜无损估产的研究空白。","agent","2026-09-17T00:04:40.596719Z",{"id":714,"title":715,"url":716,"summary":717,"summary_zh":718,"content":11,"source_name":719,"source_url":716,"published_at":580,"category":581,"cover_url":11,"hotness":385,"is_selected":582,"score":720,"score_detail":721,"sources":723,"tags":725,"primary_topic":729,"secondary_topics":730,"classification":732,"search_phrases":760,"slug":763,"view_count":683,"doi":764,"paper":765,"created_at":775},2682,"Machine Learning-Based Decision Support System for Greenhouse Crop Management Under Mite Infestation Conditions","https:\u002F\u002Fdoi.org\u002F10.36099\u002Fjess.v1i3.001","This paper presents an integrated machine learning-based Decision Support System (DSS) for greenhouse crop management under mite infestation conditions, specifically designed for Sri Lankan agricultural contexts. The research addresses critical challenges faced by greenhouse farmers regarding pest management and crop productivity optimization through a comprehensive system combining NetLogo simulation for synthetic data generation, ensemble machine learning models, and a web-based interface. Field research with Sri Lankan greenhouse farmers revealed that mite infestations cause up to 40% crop losses, driving panic-induced pesticide overuse and knowledge gaps in pest management timing. The system integrates environmental monitoring, pest prediction, and crop yield forecasting to provide actionable recommendations for farmers. The mite infestation prediction model achieved 86% accuracy, while the system successfully addresses data scarcity challenges through agent-based modeling. The DSS demonstrates potential for transforming reactive farming practices into predictive, data-driven approaches while accommodating the technological constraints of developing agricultural contexts.","本文提出了一种基于机器学习的集成决策支持系统（DSS），用于螨虫侵染条件下的温室作物管理，专为斯里兰卡农业情境设计。该研究针对温室农户在害虫管理和作物生产力优化方面面临的关键挑战，通过一个综合系统加以解决，该系统结合了用于合成数据生成的NetLogo仿真、集成机器学习模型以及基于网络的界面。针对斯里兰卡温室农户的实地研究表明，螨虫侵染可导致高达40%的作物损失，进而引发恐慌性农药过度使用以及害虫管理时机方面的知识缺口。该系统整合了环境监测、害虫预测和作物产量预测，为农户提供可操作的推荐建议。螨虫侵染预测模型达到了86%的准确率，同时该系统通过基于智能体的建模成功应对了数据稀缺的挑战。该决策支持系统展现出将被动应对式耕作实践转变为预测性、数据驱动方法的潜力，同时兼顾了发展中农业情境的技术约束。","Journal of Environmental and Sustainability Science",74,{"impact":603,"substance":508,"depth":664,"authority":416,"freshness":385,"relevant":455,"comment":722},"将机器学习与智能体仿真结合用于温室螨害预测与决策支持，方法新颖、数据翔实，对设施农业植保信息化有参考价值。",[724],{"name":719,"url":716},[590,591,726,727,728],"设施农业","决策支持系统","病虫害预警",{"code":324,"slug":325,"name":326,"path":327,"parent_code":271},[731],{"code":24,"slug":25,"name":26,"path":27,"parent_code":7},{"status":598,"locked":582,"version":4,"method":599,"evidence":733},{"candidates":734,"agricultural_evidence":604},[735,743,749,756],{"code":24,"score":736,"supported":604,"title":737,"abstract":740,"tags":742},27,[738,739],"machine learning","decision support",[741,738,739],"机器学习",[591,727],{"code":324,"score":345,"supported":604,"title":744,"abstract":746,"tags":748},[745],"greenhouse",[747,745],"温室",[],{"code":104,"score":750,"supported":604,"title":751,"abstract":752,"tags":755},6,[],[753,754],"农药","pesticide",[],{"code":16,"score":455,"supported":582,"title":757,"abstract":758,"tags":759},[],[],[590],[761,762],"农业人工智能 决策支持系统 病虫害预警 智慧农业","农业人工智能 决策支持系统","农业人工智能决策支持系统病虫害预警智慧农业-2682","10.36099\u002Fjess.v1i3.001",{"doi":764,"openalex_id":766,"authors":767,"venue":719,"cited_by_count":424,"oa_url":716,"card":770,"direction":653,"ingested_from":655},"W7213235680",[768],{"name":769,"orcid":11},"S. Nasiketha",{"tldr":771,"method":772,"finding":773,"direction":653,"opportunity":774},"为斯里兰卡温室农户开发基于机器学习的决策支持系统，预测螨害并优化作物管理。","NetLogo仿真生成合成数据，集成机器学习模型与网页界面。","螨害预测准确率达86%，可缓解数据稀缺并减少农药滥用。","可探索小样本下合成数据与迁移学习结合，提升发展中国家温室病虫害预测泛化能力。","2026-09-16T23:30:51.573509Z",{"id":777,"title":778,"url":779,"summary":780,"summary_zh":781,"content":11,"source_name":782,"source_url":779,"published_at":783,"category":581,"cover_url":11,"hotness":385,"is_selected":582,"score":784,"score_detail":785,"sources":788,"tags":790,"primary_topic":794,"secondary_topics":795,"classification":797,"search_phrases":821,"slug":824,"view_count":455,"doi":825,"paper":826,"created_at":863},2656,"DeepPhenoTree-Apple Edition: a multi-site apple phenology RGB annotated dataset with deep learning baseline models","https:\u002F\u002Fdoi.org\u002F10.1186\u002Fs13007-026-01591-w","Abstract In machine learning-driven plant phenotyping, well-annotated image datasets are essential for developing robust models capable of capturing phenological variability across environments. Here, we introduce DeepPhenoTree-Apple Edition , a multi-site and multi-variety RGB image dataset dedicated to the detection of key phenological stages in apple trees. The dataset includes 48,320 time-stamped RGB images acquired across four European orchards of the Apple REFPOP consortium under biological, agronomic, and environmental variability, including differences in genotypes, orchard architectures, phenological development, temperature, and humidity conditions. From this large corpus, a carefully curated subset of 808 representative images was manually annotated. It includes 241,600 expert annotations covering developmental stages from dormant bud to fruit maturity. Images were acquired using a standardized tractor-mounted phenotyping platform equipped with active flash illumination. Active flash illumination was used to reduce illumination variability and homogenize exposure, shadows, and sunlight differences across sites. Phenological structures were annotated following the BBCH scale, with bounding boxes adapted to organ visibility and developmental stage. In addition to the dataset, we provide deep-learning baseline experiments to illustrate detection performance and detection performance across locations.","在机器学习驱动的植物表型分析中，标注良好的图像数据集对于开发能够捕捉不同环境下物候变异的稳健模型至关重要。在此，我们介绍DeepPhenoTree-Apple Edition，一个用于检测苹果树关键物候阶段的多地点、多品种RGB图像数据集。该数据集包含48,320张带有时间戳的RGB图像，采集自Apple REFPOP联盟的四个欧洲果园，涵盖了生物学、农艺学和环境变异，包括基因型、果园结构、物候发育、温度和湿度条件的差异。从这一大型语料库中，我们精心挑选了808张具有代表性的图像子集进行人工标注。该子集包含241,600条专家标注，覆盖从休眠芽到果实成熟的发育阶段。图像使用配备主动闪光照明（active flash illumination）的标准化拖拉机搭载表型分析平台采集。主动闪光照明用于减少光照变异性，并统一不同地点的曝光、阴影和阳光差异。物候结构按照BBCH scale（BBCH尺度）进行标注，边界框根据器官可见性和发育阶段进行调整。除数据集外，我们还提供了深度学习基线实验，以展示检测性能以及不同地点的检测性能。","Plant Methods","2026-09-15T00:00:00Z",81,{"impact":361,"substance":321,"depth":361,"authority":345,"freshness":786,"relevant":455,"comment":787},9,"多站点苹果物候RGB标注数据集，规模大、标注专业并附深度学习基线，对作物表型与智慧果园研究有实质参考价值。",[789],{"name":782,"url":779},[590,591,791,792,793],"苹果","表型组学","图像数据集",{"code":286,"slug":287,"name":288,"path":289,"parent_code":271},[796],{"code":24,"slug":25,"name":26,"path":27,"parent_code":7},{"status":598,"locked":582,"version":4,"method":599,"evidence":798},{"candidates":799,"agricultural_evidence":604},[800,806,812,817],{"code":286,"score":361,"supported":604,"title":801,"abstract":803,"tags":805},[802],"apple",[288,791,804,802],"orchard",[791],{"code":24,"score":361,"supported":604,"title":807,"abstract":809,"tags":811},[808],"deep learning",[810,741,738],"深度学习",[591],{"code":112,"score":683,"supported":604,"title":813,"abstract":814,"tags":816},[],[815],"tractor",[],{"code":16,"score":455,"supported":582,"title":818,"abstract":819,"tags":820},[],[],[590],[822,823],"农业人工智能 图像数据集 智慧农业 表型组学","农业人工智能 图像数据集","农业人工智能图像数据集智慧农业表型组学-2656","10.1186\u002Fs13007-026-01591-w",{"doi":825,"openalex_id":827,"authors":828,"venue":782,"cited_by_count":424,"oa_url":779,"card":857,"direction":861,"ingested_from":655},"W7213246841",[829,832,834,836,839,842,845,847,849,851,853,855],{"name":830,"orcid":831},"Herearii Metuarea","https:\u002F\u002Forcid.org\u002F0009-0008-2716-0617",{"name":833,"orcid":11},"Abdoul-Djalil Ousseini-Hamza",{"name":835,"orcid":11},"Walter Guerra",{"name":837,"orcid":838},"Francesca Zuffa","https:\u002F\u002Forcid.org\u002F0009-0009-8797-0205",{"name":840,"orcid":841},"Francesco Panzeri","https:\u002F\u002Forcid.org\u002F0009-0005-8774-5201",{"name":843,"orcid":844},"Andrea Patocchi","https:\u002F\u002Forcid.org\u002F0000-0002-0919-2702",{"name":846,"orcid":11},"Lidia Lozano",{"name":848,"orcid":11},"Shauny Van Hoye",{"name":850,"orcid":11},"François Laurens",{"name":852,"orcid":11},"Jeremy Labrosse",{"name":854,"orcid":11},"Pejman Rasti",{"name":856,"orcid":11},"David Rousseau",{"tldr":858,"method":859,"finding":860,"direction":861,"opportunity":862},"发布多站点多品种苹果物候RGB图像数据集，并给出深度学习基线检测模型。","采集4个欧洲果园48320张图像，精选808张按BBCH标框标注，用主动闪光和拖","数据集覆盖从休眠芽到果实成熟的241600个标注，基线模型可跨地点检测物候期。","农业遥感与作物表型","可基于该多站点数据研究跨环境域适应与物候期细粒度检测，提升模型泛化能力。","2026-09-16T23:30:24.293828Z",{"id":865,"title":866,"url":867,"summary":868,"summary_zh":869,"content":11,"source_name":870,"source_url":867,"published_at":580,"category":581,"cover_url":11,"hotness":385,"is_selected":582,"score":871,"score_detail":872,"sources":875,"tags":877,"primary_topic":879,"secondary_topics":880,"classification":881,"search_phrases":902,"slug":905,"view_count":683,"doi":906,"paper":907,"created_at":925},2638,"Full shuffle and p-rectified semi-inner powerful IoU-based chili pepper flower recognition with environment-aware YOLO11","https:\u002F\u002Fdoi.org\u002F10.3389\u002Ffpls.2026.1935123","As a common greenhouse-grown commercial crop, chili pepper yields have always been a focus of attention. Effective management is essential for both environmental sustainability and high productivity in greenhouses. Whether for growth monitoring, yield estimates, and automated production, precise identification of chili flowers is critical. To this goal, this paper works upon the YOLO11 object detection model, focusing on the precise detection of chili flowers. The diversity of samples across various scenarios is increased by utilizing a self-built dataset of chili flowers in greenhouses and integrating data augmentation. To address YOLO11’s limitations in this task, this paper proposes a unified framework that integrates three modules: (1) full shuffle to enhance information exchange between model channels and optimize weights; (2) Since YOLO11 cannot incorporate external environmental data during training, the environment-aware C3k2 is introduced; and (3) The proposed p-rectified Semi-inner Powerful IoU accelerates model convergence and flexibility. Experiments show that our method obtains 84.8% mAP50, 50.2% mAP50-95, 81.1% Precision, and 75.6% Recall, outperforming the baseline YOLO11’s 77.3% by 7.5 percentage points, as well as YOLOv8, YOLO12, RTDETR, and Faster R-CNN under the same experimental conditions. Therefore, in terms of performance, our method generally surpasses existing algorithms for chili flower detection. Research on detecting chili flowers in greenhouses remains limited. The study offers valuable insights for the advancement of smart agriculture. Future work will further explore the model’s generalization capabilities across multiple varieties and growth stages, as well as deploy it on embedded devices for practical application.","作为常见的温室商业化种植作物，辣椒的产量一直备受关注。有效的管理对于温室的環境可持续性和高生产力都至关重要。无论是生长监测、产量估算还是自动化生产，辣椒花朵的精准识别都至关重要。为实现这一目标，本文基于YOLO11目标检测模型，聚焦辣椒花朵的精准检测。通过利用自建的温室辣椒花朵数据集并结合数据增强，增加了不同场景下样本的多样性。针对YOLO11在此任务中的局限性，本文提出了一个集成三个模块的统一框架：（1）完全混洗（full shuffle），以增强模型通道间的信息交换并优化权重；（2）由于YOLO11在训练过程中无法纳入外部环境数据，引入了环境感知C3k2（environment-aware C3k2）；（3）提出的p-rectified Semi-inner Powerful IoU加速了模型收敛并提升了灵活性。实验表明，本方法取得了84.8%的mAP50、50.2%的mAP50-95、81.1%的精确率和75.6%的召回率，较基线YOLO11的77.3%提升了7.5个百分点，并在相同实验条件下优于YOLOv8、YOLO12、RTDETR和Faster R-CNN。因此，在性能方面，本方法总体上超越了现有的辣椒花朵检测算法。目前关于温室辣椒花朵检测的研究仍然有限。本研究为智慧农业的发展提供了有价值的见解。未来工作将进一步探索模型在多个品种和生长阶段上的泛化能力，并将其部署到嵌入式设备上以实现实际应用。","Frontiers in Plant Science",76,{"impact":603,"substance":873,"depth":664,"authority":665,"freshness":385,"relevant":455,"comment":874},21,"基于YOLO11的辣椒花检测新方法，mAP50提升7.5个百分点，对设施农业智能监测有参考价值，但属细分技术进展，影响范围有限。",[876],{"name":870,"url":867},[590,591,726,592,878],"辣椒",{"code":324,"slug":325,"name":326,"path":327,"parent_code":271},[],{"status":598,"locked":582,"version":4,"method":599,"evidence":882},{"candidates":883,"agricultural_evidence":604},[884,890,894,898],{"code":16,"score":885,"supported":604,"title":886,"abstract":887,"tags":889},7,[],[590,888],"smart agriculture",[590],{"code":324,"score":750,"supported":604,"title":891,"abstract":892,"tags":893},[],[747,745],[],{"code":24,"score":408,"supported":604,"title":895,"abstract":896,"tags":897},[],[592],[591,592],{"code":279,"score":400,"supported":604,"title":899,"abstract":900,"tags":901},[],[878],[878],[903,904],"农业人工智能 智慧农业 目标检测 设施农业","农业人工智能 智慧农业","农业人工智能智慧农业目标检测设施农业-2638","10.3389\u002Ffpls.2026.1935123",{"doi":906,"openalex_id":908,"authors":909,"venue":870,"cited_by_count":424,"oa_url":867,"card":920,"direction":709,"ingested_from":655},"W7213351973",[910,912,914,917],{"name":911,"orcid":11},"Cui-Ping Zhang",{"name":913,"orcid":11},"Zhi-Yong Wang",{"name":915,"orcid":916},"Xuewei Wang","https:\u002F\u002Forcid.org\u002F0000-0001-9604-3045",{"name":918,"orcid":919},"Zhi Li","https:\u002F\u002Forcid.org\u002F0000-0001-5571-0518",{"tldr":921,"method":922,"finding":923,"direction":653,"opportunity":924},"基于YOLO11改进，实现温室辣椒花精准检测。","自建数据集+数据增强，引入全混洗、环境感知C3k2和p校正IoU。","mAP50达84.8%，比基线YOLO11提升7.5个百分点，优于多个对比模型。","温室辣椒花检测研究少，可探索多品种、多生长期泛化及嵌入式部署。","2026-09-16T23:30:09.868848Z",{"id":927,"title":928,"url":929,"summary":930,"summary_zh":11,"content":931,"source_name":932,"source_url":11,"published_at":933,"category":934,"cover_url":11,"hotness":385,"is_selected":582,"score":871,"score_detail":935,"sources":937,"tags":939,"primary_topic":943,"secondary_topics":944,"classification":945,"search_phrases":970,"slug":973,"view_count":455,"doi":11,"paper":11,"created_at":974},2604,"崔运鹏研究员团队 农产品品质管控云边端一体化中间件研究进展","https:\u002F\u002Fwww.toutiao.com\u002Farticle\u002F7683885963974410792","发表于智慧农业(中英文)2026年第4期。研究阐述云边端协同技术在大田种植、设施园艺和畜牧养殖等典型农业场景中的应用模式与协同机制，综合分析边缘智能调度、语义互操作与可信协同等国内外研究进展。","**引用格式**：王婷, 王娜, 崔运鹏, 刘娟. 农产品品质管控云边端一体化中间件研究进展与展望[J]. 智慧农业(中英文), 2026, 8(4): 1-16.\n\nCitation: WANG Ting, WANG Na, CUI Yunpeng, LIU Juan. Research Progress and Prospects of Cloud-Edge-Device Integrated Middleware for Agricultural Product Quality Control[J]. Smart Agriculture, 2026, 8(4): 1-16.\n\nDOI： 10.12133\u002Fj.smartag.SA202510002\n\n**农产品品质管控云边端一体化中间件研究进展与展望**\n\n王婷1，2， 王娜3， 崔运鹏1，2*， 刘娟1，2\n\n（1.中国农业科学院农业信息研究所，北京 100081，中国； 2.农业农村部农业大数据重点实验室，北京 100081，中国； 3.96962部队，北京102206，中国）\n\n**摘要：［目的\u002F意义］**现代农业加速向精准化、智能化管控转型，对生产过程的实时响应能力和海量多源数据处理能力提出了更高要求，推动传统中心化云计算架构向云边端协同一体化演进。为支撑农产品品质的精准感知与智能调控，有必要系统梳理云边端协同在农业领域的关键技术与应用基础，为一体化中间件设计提供理论与技术支撑。**［进展］**阐述了云边端协同技术在大田种植、设施园艺和畜牧养殖等典型农业场景中的应用模式与协同机制，归纳了农业物联网在实际落地过程中普遍存在的异构设备协议不兼容、数据标准体系缺失、边缘资源调度失衡，以及隐私与安全风险等问题。在此基础上，综合分析了边缘智能调度、语义互操作与可信协同等国内外研究进展，重点梳理了面向农产品品质管控的云边端一体化适配中间件的框架结构与关键功能模块，包括品质数据标准化映射、异构协议转换引擎、基于农产品特性的智能计算调度，以及模型热更新机制等，并进一步讨论了“设备–作物–模型”三位一体协同机制在屏蔽底层硬件差异、提升系统弹性适配能力方面的作用。**［结论\u002F展望］**云边端一体化协同及其自适应中间件是支撑农产品品质智能管控的关键基础设施，可为农产品品质精准感知、自动控制与智能决策系统的开发与工程化实施提供可复用的架构思路与技术参考。未来需在跨场景数据与协议标准体系构建、边缘智能与大模型融合应用、隐私保护与可信计算机制完善，以及大规模工程化示范等方面开展深入研究，进一步提升农产品品质管控系统的智能化、可靠性与可推广性。\n\n**关键词：**物联网；云边端一体化；中间件；农产品品质；智能管控\n\n**0 引 言**\n\n在消费升级与健康观念普及的双重驱动下，市场对农产品的关注点已从基础的营养安全，延伸至新鲜度、口感风味及品牌信誉等更丰富的维度。与此同时，国家层面正通过一系列标准化规划严控源头质量，推动农产品品质管控体系加速向精细化和智能化转型。2024年，农业农村部印发《关于大力发展智慧农业的指导意见》，明确指出要“集成应用‘四情’监测、精准水肥药施用、智能农机装备、无人驾驶航空器和智能决策系统等技术，构建主要作物大面积单产提升的数字化种植技术体系”。同期发布的《全国智慧农业行动计划（2024—2028年）》提出实施“智慧农业重点领域应用拓展行动”，目标到2026年底“探索一批主要作物大面积单产提升智能化解决方案和智慧农（牧、渔）场技术模式，农业生产信息化率达到30%以上”。此外，《中共中央 国务院关于进一步深化农村改革 扎实推进乡村全面振兴的意见》（2025年中央一号文件）进一步强调“支持发展智慧农业，拓展人工智能、数据、低空等技术应用场景”，为农业生产向精准化、智能化转型指明了清晰的发展方向与行动路径。\n\n物联网技术的广泛应用推动了农产品品质管控过程中各类传感器和智能设备的接入，这些设备能够实时采集土壤湿度、温度、光照、病虫害等关键数据，用于监测作物生长环境和品质指标，通过数据的远程传输和智能分析，实现作物的生长周期管理、环境和投入品监测、智能灌溉、病虫害防治等。虽然云计算的出现解决了传统物联网设备计算能力有限的问题，但随着设备种类和布设数量的增加产生越来越多的数据，给网络带来非常大的负担，并且由于云平台通常远离设备端，单纯依靠云计算已难以满足低延时、大数据量、本地处理等实际需求。因此，云边端一体化架构应运而生，即传感器等终端设备进行数据采集和初步处理，在边缘设备进行数据汇聚、智能分析和临时存储，最终将关键数据和决策传送到云平台，实现智能化管控和大数据驱动的决策支持。这种模式能够提升农产品品质管控的实时性、可靠性和灵活性，在保障农产品安全、降低管理成本等方面发挥着重要的作用。\n\n本文从农业领域云边端一体化应用现状出发，分析其在农产品智能管控方面的不同应用场景，深入探讨其存在的问题，包括协议异构适配难、数据互操作性差、算力资源分配失衡、云边协同调度低效等多维度技术难题，并基于人工智能、联邦计算、区块链等技术从资源配置、边缘计算、数据安全等多角度提出云边端技术优化策略。同时，本文面向农产品品质智能管控，设计了集成农产品品质数据标准化映射、异构设备协议转换、智能计算任务调度与负载均衡、多层次数据缓存、模型热更新和联邦部署等核心功能模块的云边端一体化技术框架。最后，结合新兴信息技术发展趋势，前瞻性分析了人工智能前沿技术在农产品品质管控方向的发展前景。本文旨在为农产品智能化品质管控体系建设提供可借鉴的理论指导和技术实施方案。\n\n**1 云边端技术在农产品品质管控领域的应用**\n\n**1.1 云边端常用架构分析**\n\n根据任务卸载机制的不同，云边端主要可以划分为云-边-端、云-（边）-端、云-边-端-端架构，为多样化的农产品品质管控场景与业务需求的适配提供了不同的架构参考，具体案例如表1所示。\n\n表1 云边端不同架构具体案例\n\nTable 1 Specific cases of different cloud-edge-device architectures\n\n![Image 1](https:\u002F\u002Fp3-sign.toutiaoimg.com\u002Ftos-cn-i-axegupay5k\u002Fd17831d41c38416180fcb2883acb61cb~tplv-tt-origin-web:gif.jpeg?_iz=58558&from=article.pc_detail&lk3s=953192f4&x-expires=1790121856&x-signature=gf8Rtfn3PrSNi0ORE8cjRLKhf1o%3D)\n（1）云-边-端架构。云-边-端架构是一种三层服务体系，其结构如图1所示。在该模式下，终端设备产生的计算任务可通过任务调度机制，将部分任务转交给边缘服务器以利用其计算资源。当自身资源无法满足终端设备的计算需求时，边缘服务器可进一步向云平台发起资源请求，并将部分任务转发至云端进行处理。此类设计将云平台对终端用户进行了隐藏，即终端用户不直接与云端进行交互，而是以边缘服务器为唯一上层交互对象。\n\n![Image 2](https:\u002F\u002Fp3-sign.toutiaoimg.com\u002Ftos-cn-i-tjoges91tu\u002F41d2aab83e55f1dac8f18dd665f9ebae~tplv-tt-origin-web:gif.jpeg?_iz=58558&from=article.pc_detail&lk3s=953192f4&x-expires=1790121856&x-signature=sjAAhIh9eN03g3WfSEAbvAFoaYA%3D)\n图1 云-边-端架构\n\nFig. 1 Cloud-Edge-Device architecture\n\n2）云-（边）-端架构。云-（边）-端架构是一种两层服务体系结构，如图2所示。在该模式下，终端设备能够直接访问云平台资源。当终端设备生成需要处理的任务时，可以自主选择将任务分别卸载至云端或边缘服务器，无需经过边缘服务器作为中介进行中转。此时，云平台对终端设备是可见的，终端既可以将任务分配给边缘服务器，也能够直接调用云端的计算资源。\n\n![Image 3](https:\u002F\u002Fp3-sign.toutiaoimg.com\u002Ftos-cn-i-tjoges91tu\u002Fa634e26f4528d9045d564c681fab8b47~tplv-tt-origin-web:gif.jpeg?_iz=58558&from=article.pc_detail&lk3s=953192f4&x-expires=1790121856&x-signature=Ir2bXp6H1Tu%2BDtXRETE%2B6epksVA%3D)\n图2 云-（边）-端架构\n\nFig. 2 Cloud-（Edge）-Device architecture\n\n（3）云-边-端-端架构。随着人工智能技术的演化升级，智能驾驶、无人机、机器人等对数据延迟、系统响应时间、可靠性要求较高的智能终端设备得到广泛应用。这些设备通常允许本地处理和存储，却由于资源和能力的限制，可能出现丢失、错误或损坏的问题，降低了数据的完整性和可靠性。为了解决这个问题，已有研究在云-边-端架构的基础上增加了端-端协同机制，通过分布式存储和任务副本，在多个终端之间实现数据采集与处理的冗余与互备。如图3所示，云平台位于中心负责全局模型训练与策略下发，边缘节点承担区域内的数据聚合与实时推理，各类终端设备既与边缘节点交互，又通过端-端协同实现本地感知数据与控制指令的共享，从而提升系统在复杂农业场景下的鲁棒性与可扩展性。\n\n![Image 4](https:\u002F\u002Fp3-sign.toutiaoimg.com\u002Ftos-cn-i-tjoges91tu\u002Ff9c75f6b03a2cdcb2d721ddd8782b968~tplv-tt-origin-web:gif.jpeg?_iz=58558&from=article.pc_detail&lk3s=953192f4&x-expires=1790121856&x-signature=WKW8A5Fy3NcA0oBoP29zWTaepFs%3D)\n图3 云-边-端-端架构\n\nFig. 3 Cloud-edge-device-device architecture\n\n**1.2 云边端技术在农产品品质管控领域的****应用场景分析**\n\n农业生产具有显著的地域差异，同一种农产品品质管控需求在不同地形地貌下，会呈现出不同的数据形态、通信条件与算力供给。因此，本文从地理地形角度分析不同农业场景下云-边-端一体化技术的具体应用。\n\n**1.2.1 基于不同地理地形的农业应用场景**\n\n（1）面向平原大田的规模化作业场景。平原大田场景以连片地块与高强度机械化作业为特征，品质管控的核心在于“精准投入”与“产出预测”。面对气象、墒情等高频环境数据，以及时相遥感、无人机巡检等大体量影像数据，系统的挑战在于如何实现广域覆盖下的统一调度与关键农时（如灌溉、病害爆发期）的快速响应。在云-边-端架构设计上，侧重于“广域汇聚与分层处理”：端设备（传感器、农机终端）专注于数据的连续采集与预处理；边缘端扮演“过滤器”角色，负责清洗高频数据、提取关键地块影像并触发即时告警；云端则发挥算力优势，针对跨地块、跨季节的长周期数据进行分析，通过大样本训练优化产量预测与投入品评估模型。\n\n（2）面向丘陵山地的果茶园作业场景。丘陵山地地形破碎、遮挡严重，网络信号盲区多且供电难，是典型的“弱网弱电”环境。此场景下，病虫害监测往往依赖图像与视频传输，但是却严重受限于本地网络带宽。因此，必须构建“端边协同”的韧性架构：端设备需满足低功耗免维护的能力；边缘端则需具备更强的自治能力，支持断网缓存与本地推理。针对高价值的病害报警，采取“可解释性优先”策略——边缘端先上传告警摘要与最小证据集（如缩略图），待网络恢复后再异步补传完整数据，确保在恶劣环境下关键信息依然可达。\n\n（3）面向设施园艺的实时闭环场景。日光温室与植物工厂属于环境高度可控的封闭系统，作物对环境波动极度敏感，很容易导致高温胁迫等突发事件，这对控制系统的实时性提出了毫秒级要求。此外，果径、色泽等作物表型分析产生了大量实时视频流。在此场景下，云边分工需满足“边缘保生存、云端促生长”：各类执行器与传感器在端侧实现协议互通；边缘端作为核心控制大脑，全权负责系统闭环控制、本地视频流分析及异常熔断保护，确保即使外部断网，温室仍能独立维持自治运行；云端则从繁琐的实时控制中解脱出来，专注于跨温室的生长模型迭代与能效策略下发。\n\n（4）面向集约化畜牧的隐私计算场景。畜牧养殖的对象是具有复杂行为的活体群体，防疫要求比较严苛。系统不仅要处理步态、体温、进食量等海量非接触式监测数据，还面临高并发视频分析带来的算力瓶颈，以及养殖场对核心生产数据（如存栏、配方）的隐私保护需求。因此，该场景宜采用“边云混合”的隐私计算架构：穿戴设备与巡检机器人在端侧全天候感知；边缘端部署高性能计算单元，就地“消化”海量视频数据，实时识别发情、跛行等异常行为并推送告警，仅上传结构化日志；云端则利用联邦学习技术，在不触碰各场原始数据的前提下，汇聚多方参数优化全局疫病预警与遗传评估模型，实现数据安全与算法进化的平衡。\n\n**1.2.2 从具体应用场景到云边端分工的映射**\n\n地理地形与业务特征的差异导致单一的云边端架构无法覆盖所有农业场景。本文从网络条件、数据形态和控制时效三个维度，总结出以下分工准则。\n\n首先是网络条件，以网络稳定性决定控制权归属。在丘陵山地或深远海等网络不稳的“弱连接”地带，必须保证边缘自治优先。边缘节点不能只是传声筒，而应具备完整的本地数据库与断点续传能力。面对断网风险，要确保“高价值数据先走”——比如优先上传一张病害报警的缩略图，而不是全量视频流。此时，云端负责归档与模型分发中心，无需介入实时控制回路，以防因网络波动导致业务瘫痪。\n\n其次是数据形态，以数据与算力的匹配度决定“计算重心”落点。对于设施园艺或集约化养殖这类产生海量视频流的场景，带宽是最大瓶颈。因此，应采取“算力下沉”策略，把视频结构化、异常行为检测等消耗大量带宽的任务留在边缘端消化，只往云端传回处理后的日志或异常片段。反之，对于大田种植这种以温湿度等低频小数据为主的场景，边缘端只需做简单的聚合转发，把复杂的分析算力保留在云端，从而大幅降低边缘硬件的铺设成本。\n\n最后是控制时效，以时效性刚需划定“控制红线”。在温室环境调控或农机自动驾驶等场景中，毫秒级的响应延迟直接关乎作物存活或设备安全。因此，需要满足“控制不出园”：凡是涉及通风降温、紧急避障等生命攸关的闭环控制，必须在端侧或边缘端独立完成，绝不能依赖云端指令。云端则专注于慢变量的管理，如基于长周期历史数据优化灌溉模型或能效策略，从而实现快慢控制流的物理分离。\n\n综上所述，农业云边端一体化不应是一个僵化的标准模板，而是一套能随环境动态伸缩的弹性体系。这也正是本文设计自适应中间件的初衷——通过软件层面的协议自适应、资源可编排和模型可热插拔，在底层硬件异构与上层多变的农艺需求之间，搭建一座可配置、可落地的桥梁。\n\n**2 农业云边端一体化的关键问题与挑战**\n\n在农业领域实际应用中，云边端一体化在资源协同调度机制、数据标准化、协议兼容、普及应用及安全等多个方面存在问题，需要进一步优化。\n\n**2.1 资源调度缺陷与边缘AI适配瓶颈**\n\n云边端一体化技术在农业场景的纵深发展，当前主要受制于资源协同调度机制的非自适应性与边缘端人工智能（Artificial Intelligence, AI）模型部署的效能瓶颈。\n\n首先，现有的云边协同调度策略普遍存在感知粒度粗糙的问题。由于缺乏对野外节点能耗状态、网络波动及任务优先级的实时感知，系统难以在动态变化的农业环境中实现精准的负载均衡。在大规模并发场景（如大面积果园巡检）中，这种静态调度模式极易引发资源竞争失序，导致高优先级控制指令因队列阻塞而延迟，甚至造成数据流中断与“长尾”掉队现象。这不仅削弱了系统在大规模部署下的鲁棒性，也严重阻碍了云边计算效能的线性扩展。\n\n其次，面向农业的AI技术在边缘端落地时，面临着严峻的“算力-算法-环境”适配矛盾：一是农业边缘网关受限的计算资源无法匹配主流视觉模型庞大的参数量与计算开销，导致推理延迟激增，难以满足农机避障、精准采摘等场景对毫秒级实时响应的刚性需求。二是农业场景具有高度的非结构化特征，比如复杂光照、遮挡、形态变异等，使得在理想数据集上训练的模型在迁移至田间实地时，常因特征分布偏移而出现明显的性能退化。三是当前架构大多使用“云端离线训练-边缘静态部署”的方法，缺乏边缘端增量学习与持续优化的能力。当新品种引入或病虫害发生变异时，边缘设备无法利用本地样本进行实时微调，这种依赖云端全参重新训练与下发的长周期迭代模式，既增加了网络带宽开销，也限制了AI系统对农业生产动态变化的适应生命力。\n\n**2.2 标准化缺失**\n\n终端数据采集设备和边缘网关设备品牌众多，采用的通信协议标准不统一，在数据传输格式和消息封装机制方面存在明显的差异化，造成云边端协议栈适配复杂度很高。同时，这些设备的数据接口规范不同，导致不同厂商设备之间的数据兼容性差，存在严重的互操作性障碍。\n\n由于农业领域缺乏行业统一的物联网数据标准、接口规范和元数据描述，来自气象站、土壤传感器、高清航拍影像等传感器、终端设备与数据平台在数据采集与处理环节采用了各自独立的规范和格式，因此数据格式、采样频率、语义定义等各方面差异巨大，阻碍了云边端各层之间的数据交换和融合分析。以温湿度和光照强度为例，不同厂家生产的传感器在测量精度、量程范围、校准基准及报告频率上各不相同：部分设备将湿度表示为相对湿度百分比，而另一些设备则以绝对湿度（g\u002Fm3）输出，光照值亦存在勒克斯（lux）、瓦特每平方米（W\u002Fm2）等多种计量单位。终端设备、边缘设备和云平台未能兼顾多源数据的自动识别与统一转换，导致数据在端-边-云流转过程中需要频繁进行手动映射和二次开发，从而加剧系统的集成难度与运维成本，使得跨设备、跨平台的数据融合与联合分析变得繁复且易出错，严重影响了农产品品质管控算法的精准性和云边端协同效率。\n\n**2.3 异构设备协议不兼容**\n\n首先，物联网设备与云平台之间缺乏统一的接口协议，多数传感器和网关采用各自厂商定义的私有协议或不完全遵循消息队列遥测传输协议（Message Queuing Telemetry Transport, MQTT）、受限应用协议（Constrained Application Protocol, CoAP）等主流标准，导致数据推送或订阅时协议解析失败。其次，终端设备种类繁多且更新换代周期长，不同厂商在传输层安全、消息格式、数据编码和握手机制等方面标准不一，使得同一应用场景下的多设备融合集成需要进行复杂的二次开发或协议网关适配。再次，由于各类协议转换层往往依赖于专有固件或边缘中间件，任何一端的固件升级或平台迭代都可能引发接口断裂，进而造成数据采集中断与互联互通困难。这些异构协议的不兼容性不仅增加了系统部署和维护成本，也制约了云边端协同、智能管控及规模化推广的效率。\n\n**2.4 成本与普及障碍**\n\n由于部署成本问题，云边端一体化技术在农业领域的推广和普遍应用面临很大困难。首先，中小农场的部署成本高，高性能传感器、边缘计算节点和专用网关设备价格不菲，再加上专业安装、网络接入及后续运维费用，使得初始投入对小规模经营者而言难以承担。其次，系统集成与定制化开发需求增加了项目复杂度和服务成本，尤其是在网络覆盖不佳的农村地区，需要额外的基础设施改造和带宽保障，进一步抬升了总成本。更重要的是，云边端一体化解决方案通常需要较长的投资回报周期——从设备安装到数据优化、模型训练再到生产效益提升，往往需要数个生长季才能显现经济效益，这使得资金回收和收益预期不确定性加剧，从而削弱了中小农场主的投资意愿，成为制约云边端技术大规模推广的重要难题。\n\n**2.5 开放架构下的安全与隐私多维挑战**\n\n云-边-端架构的引入打破了传统农业控制系统的封闭性，其分布式的拓扑结构与异构的通信环境，使农业数据安全面临着从物理感知层到业务应用层的全链路挑战。\n\n首先，与部署在物理隔离机房的传统服务器不同，水肥一体机、野外监控终端等农业终端设备长期置于无人值守的开放环境中，天然缺乏物理防御屏障。这种物理暴露性使得设备极易遭受暴力破解或接口劫持。攻击者一旦攻陷边缘节点，便可实施虚假数据注入攻击，例如伪造极寒低温警报，诱导上层控制系统做出错误的决策，从而造成不可逆的作物损毁。\n\n其次，在云边端的数据流转路径中，由于底层短距协议与上层传输控制协议\u002F网际协议（Transmission Control Protocol\u002FInternet Protocol, TCP\u002FIP）协议栈存在标准差异，难以构建端到端的全链路加密隧道。边缘网关作为协议转换的中枢，往往需要在内存中对数据进行“解密-重封装”操作。这一明文窗口期使得网关成为整个信任链条中的薄弱环节，极易遭受中间人攻击，导致关键控制指令被窃听甚至篡改，破坏了系统的完整性保护。\n\n再次，农业领域的云边端系统汇聚了高精度的环境参数、独家水肥配方及作物生长模型，这些数据本质上构成了农业企业的核心生产工艺与知识产权。区别于一般用户隐私，此类生产数据的泄露将导致商业机密穿透：竞争对手可通过分析底层数据反向推导企业的经营成本结构与技术壁垒。若缺乏完善的隐私计算或联邦学习机制，单纯的数据上云将使企业面临巨大的市场博弈风险。\n\n同时，农业物联网设备品牌繁杂、固件版本碎片化严重，导致统一的安全补丁分发与漏洞管理极其困难。单一终端（如弱口令摄像头）的安全短板极易被利用成为僵尸网络的入侵跳板，进而通过横向移动瘫痪整个农场的自动化控制网络，引发级联失效。\n\n**3 云边端一体化技术优化策略**\n\n第2节系统梳理了农业云边端应用中面临的技术瓶颈、数据标准缺失、异构协议壁垒、覆盖成本高及数据安全隐患等挑战。围绕这些问题，本节基于对前沿学术成果的深度调研，从边缘智能协同调度、异构数据语义适配、存储优化及跨主体可信安全协同四个维度出发，深入讨论人工智能、元数据映射、编码技术及联邦学习在农产品品质管控场景中的适用性与优化思路，并为第4节中间件架构中的对应功能模块设计提供理论支撑。\n\n**3.1 边缘智能协同调度策略**\n\n针对农业边缘节点算力分布不均、品质管控实时性要求较高、单一云端调度面临高延迟的问题，现有研究普遍主张采用多接入边缘计算（Multi-access Edge Computing, MEC）架构下的分层协同调度机制。\n\n面对动态变化的农业任务负载，尤其是突发情况，比如病虫害暴发，静态调度算法难以适应。KHANI等深入探讨了深度强化学习算法（Deep Reinforcement Learning, DRL）在MEC资源分配中的应用，通过构建基于DRL的双阶段优化框架，克服了传统静态策略缺乏适应性的缺陷，在动态变化的网络负载下实现了计算资源与通信带宽的实时最优匹配，显著降低了任务处理延迟。进一步地，针对云边协同场景中复杂的任务依赖关系，WU等提出了多维度协同调度模型。该研究不仅仅局限于单一的计算卸载，而是联合优化了服务器放置、服务部署及任务路由，利用分支定界与启发式算法相结合的策略，解决了农业虚拟现实（Virtual Reality, VR）\u002F增强现实（Augmented Reality, AR）巡检等高计算密集型任务在受限边缘节点上的负载均衡问题，有效防止了资源过载或闲置。此外，在处理具有强依赖关系的复杂农业任务（如先采集图像再进行病害识别）时，最新的DRL方法也展现出优越性。WANG等提出了一种基于马尔可夫决策过程的任务卸载方案，通过智能体自主学习任务间的拓扑结构与能耗约束，实现了比传统 Deep Q-Network方法更低的时延与能耗开销，验证了在复杂物联网工作流中引入智能卸载机制的必要性。\n\n为解决边缘节点间的资源竞争问题，LUO等 综述了非合作博弈模型在多用户边缘计算中的应用，证明了纳什均衡策略在分布式调度中的有效性。ABDI等提出了一种基于Q-learning的任务卸载算法，在考虑任务排队时间的同时最大化任务完成率，特别适用于高并发的农情监测场景。\n\n**3.2 语义标准化与协议适配策略**\n\n针对多源异构数据难以深度融合的问题，KALYANI等在云-雾-边协同架构中引入了基于本体的数字孪生技术。该研究构建了一套标准化的农业数字模型，能够自动将来自不同厂商传感器的原始数据流映射为统一的语义实体，有效屏蔽了底层数据格式的差异，验证了其在跨平台数据集成中的互操作性优势。LI等的综述进一步强调，缺乏统一的农业领域本体是限制AI模型泛化能力的核心瓶颈，并建议通过建立开放的农业知识图谱标准，实现跨区域、跨设备的协同感知与模型迁移。\n\n在解决异构设备接入难题方面，MAHBUB和SHUBAIR建议采用软件定义网络思想重构边缘网关，通过动态加载协议插件实现对MQTT、CoAP及私有Modbus协议的兼容。YU等指出，传统的硬编码网关已无法适应农业设备快速迭代的需求。未来的主流趋势是基于容器技术的软件定义网关。通过将不同协议栈封装为独立的微服务插件，边缘网关可实现协议解析模块的动态加载与热更新，真正实现农业终端设备的“即插即用”与系统的弹性扩展。\n\n**3.3 边缘存储与传输优化策略**\n\n针对丘陵山地弱网环境下的数据传输瓶颈及传统多副本策略带来的高昂存储成本，现有研究提出利用编码计算与边缘智能压缩技术来构建低冗余、高可靠的边缘存储传输体系。\n\nKREKOVIC等设计了一种基于边缘预测滤波器的双模型数据缩减机制，在边缘设备上运行轻量级长短期记忆网络模型实时预测传感器读数，仅当实际采集值与预测值的偏差超过预设阈值时，才触发数据上传。仿真实验表明，在保证数据重建精度的前提下，该方法可减少高达90%的冗余传输，显著延长了电池寿命，特别适用于带宽受限的偏远农场环境。\n\n此外，面对日益增长的图像与视频分析需求，JIANG等提出了“Farm-LightSeek”边缘中心化分析框架。该框架将轻量级多模态大模型部署于边缘节点，能够在本地完成跨模态数据（如无人机影像与气象数据）的融合推理与病害检测，仅向云端回传结构化的决策建议而非原始高带宽数据，从而有效规避了弱网环境下的传输拥塞。\n\n针对复杂地形下的网络覆盖难题，YU等在综述中指出，采用“设备-边缘中继-云”的链式协同架构是解决山区弱网传输的关键。通过利用中继边缘节点进行数据的多级聚合与压缩转发，该架构在保证长距离通信稳定性的同时，大幅降低了回传链路的带宽压力。\n\n**3.4 可信协同策略**\n\n针对农业跨主体协作中面临的“数据孤岛”效应及传统加密机制在确权激励上的局限性，学术界普遍主张融合联邦学习与区块链技术，构建“数据不出域、价值可流转”的可信协同计算范式。\n\n现有研究表明，针对农业数据非独立同分布及物联网终端算力参差不齐的特征，自适应异步联邦学习架构已成为解决模型收敛迟滞的主流方案。MUGHAL等提出的“贡献度加权聚合”机制，通过动态评估边缘节点的本地数据质量与计算能力来分配聚合权重，验证了其在提升模型泛化能力方面的有效性；同时，异步梯度更新策略被广泛证实能有效规避低配节点对整体训练进度的拖累，即聚合服务器一旦接收到设定比例的局部梯度便更新全局模型，从而在保障精度的前提下显著提升跨农场病害识别模型的迭代效率。GARRO等通过系统文献综述，详细论证了联邦学习在跨农场病害识别中的应用潜力。研究表明，通过在各农场边缘节点本地训练病害检测模型，仅向云端上传模型梯度参数而非原始图像数据，能够有效实现“数据不出域”的隐私保护目标。\n\n进一步地，针对联邦学习中的激励缺失问题，REN等提出了FLCoin架构，将区块链共识机制与联邦学习过程深度耦合，实现了边缘节点间的去中心化模型聚合与自动化激励分配。实验表明，相比传统的实用拜占庭容错协议（Practical Byzantine Fault Tolerance, PBFT）共识，FLCoin架构在通信开销上降低了90%，训练时间减少了35%，且共识延迟时间始终保持在3 s以下，完全满足农业实时调度的时效性需求。\n\n在精准农业的安全防护方面，SHARMA等［31］构建了基于以太坊区块链与星际文件系统（InterPlanetary File System, IPFS）分布式存储的联邦学习框架，专门用于农业物联网（Internet of Things, IoT）设备的网络攻击检测。该框架利用智能合约自动化管理模型更新的上链与验证流程，确保了模型训练全流程的可审计性与防篡改性，为农业数据的安全共享提供了可信技术保障。\n\n**4 农产品品质管控云边端一体化适配中间件框架**\n\n面向农产品品质管控中多源异构设备接入难、资源调度不灵活及数据安全隐患等问题，本文基于现有文献与工程实践的架构性设计与功能抽象，提出了适用于大田种植、设施园艺及畜牧养殖等多场景的云边端一体化适配中间件架构的设计思路。该中间件作为连接底层物理设备与上层管控应用的枢纽，集成了农产品品质数据标准化映射、异构设备协议转换引擎、基于农产品特性的智能计算调度、多层次缓存策略，以及模型热更新等功能模块，通过软件定义的方式屏蔽了底层硬件的复杂性与异构性，为实现农产品全生命周期的精准感知、自动控制与智能决策提供了标准化的技术底座与工程化实现路径。\n\n**4.1 农产品品质管控系统**\n\n基于云边端一体化架构的农产品品质管控系统总体框架包括4个层次：应用场景层、终端设备层、云边端一体化中间件层，以及数据云平台和边缘服务器层，其对应关系如图4所示。首先，在应用场景层，系统面向设施园艺、大田种植和畜禽养殖三大典型场景，分别以番茄、油菜、绿茶、柑橘等作物，以及鸡、羊等养殖动物为代表，用于反映不同空间尺度、网络条件和品质管控实时性要求下的多样化农业应用需求。其次，在终端设备层，系统部署环境传感器、微控制器和智能设备等数据采集终端，用于实时感知温度、湿度、光照、气体浓度等环境因子，以及水、肥、农药等投入品信息，并驱动风机、灌溉、补光和投喂等执行过程。再次，在云边端一体化中间件层，中间件按照“数据适配接入-数据标准化映射-设备自动发现与注册-智能模型分发-多层次数据缓存”的处理流程组织各功能模块。数据适配接入模块负责接入不同协议和数据格式的终端数据；数据标准化映射模块基于农产品品质数据字典，将多源数据统一转换为标准化质量特征；设备自动发现与注册模块通过异构协议解析与转换，将新接入设备映射到统一的设备信息模型；智能模型分发模块将模型知识库中的品质管控模型按场景和算力条件分配到边缘节点或终端设备；多层次数据缓存模块在端、边、云之间分级存储原始数据、特征数据和预测结果，以支撑在线推理和历史追溯。综合考虑农产品生长环境与投入品管控的时延要求、品质管控模型的算力需求，以及云边端之间传输参数的规模，系统采用“云-（边）-端”架构模式，可根据具体任务选择在边缘节点或云端完成存储与计算，以实现资源与任务的弹性匹配。最后，在数据云平台和边缘服务器层，边缘服务器依托就近部署和较低时延，对来自终端的局部原始数据进行预处理、特征提取与实时推理，仅将关键特征或复杂任务上传至云端；云端数据平台则承担大规模数据管理、农产品品质大数据建模与智能管控模型的训练、评估和版本迭代下发，为边缘和终端提供统一的服务管理与决策支持能力。\n\n![Image 5](https:\u002F\u002Fp3-sign.toutiaoimg.com\u002Ftos-cn-i-tjoges91tu\u002F3a924bf70bb12f5127f40e07ab916bc5~tplv-tt-origin-web:gif.jpeg?_iz=58558&from=article.pc_detail&lk3s=953192f4&x-expires=1790121856&x-signature=L%2FzqnjWcuyEVgnwLbilx%2FypXggw%3D)\n图4 农产品品质管控系统总体框架\n\nFig. 4 Overall framework of agricultural product quality control system\n\n中间件进一步针对3类典型场景支持差异化的边缘计算范式与交互逻辑：在设施园艺场景中，采用“边缘自治与实时闭环范式”，由边缘网关作为本地控制中枢，内置轻量级推理引擎，根据温湿度、光照等数据直接驱动风机、湿帘和补光灯，实现毫秒级闭环控制，仅在网络空闲时向云端异步同步环境摘要数据，以保证外网中断时温室仍能维持稳定运行；在大田种植场景中，采用“云-边协同与稀疏传输范式”，边缘端负责对无人机巡检影像进行清洗和初筛，只上传识别出倒伏、干旱或病虫害的关键帧，由云端利用强算力完成长周期产量预测模型训练，并将更新后的推理模型定期下发至边缘端，实现“边端过滤、云端精算”的协同；在畜禽养殖场景中，采用“混合代理与隐私隔离范式”，边缘端一方面作为高性能数据通道实时转发体温、运动量等高频异常告警信息，另一方面作为隐私防火墙，对涉及养殖规模、饲料配方等敏感经营数据进行本地脱敏或联邦加密后再上传，确保通过公网传输数据时企业核心商业机密得到有效保护。\n\n**4.2 云边端一体化适配中间件**\n\n针对农业物联网环境网络覆盖不稳定、终端设备异构性强，以及农产品品质指标随生长周期动态变化等特征，本文所述的中间件采用“设备-作物-模型”三位一体的协同管理方法，具备数据标准化映射、异构协议转换引擎、智能计算任务调度与负载均衡、模型热更新与联邦部署，以及多层次缓存功能。\n\n**4.2.1 农产品品质数据标准化映射**\n\n多源异构农业大数据治理架构中，通常通过元数据管理、统一数据字典和语义模型来支撑跨平台数据融合与治理。在许政等提出的“基于云-边-端的多源异构大数据治理架构”思路基础上，本文进一步将上述方法细化至农产品品质数据字典以及设备\u002F模型参数字典层面，实现对品质指标与关联设备、模型参数的统一描述与标准化映射。\n\n（1）数据字典。针对大田种植、设施种植、畜牧养殖等不同生长场景下终端设备采集的农产品品质相关数据，分别构建可动态更新的农产品生产数据字典、品质指标字典、管控模型字典和设备信息字典，并对数据进行定义和格式规范，如图5所示。\n\n![Image 6](https:\u002F\u002Fp3-sign.toutiaoimg.com\u002Ftos-cn-i-tjoges91tu\u002F0f6c1b9a3fc502d1eb5796c0942b3e16~tplv-tt-origin-web:gif.jpeg?_iz=58558&from=article.pc_detail&lk3s=953192f4&x-expires=1790121856&x-signature=onMP3g%2FaSLR1vTlz%2BzKOeMKTbBM%3D)\n图5 农产品品质管控数据字典结构\n\nFig. 5 Agricultural product quality control data dictionary structure\n\n农产品生产数据字典包括环境、生物\u002F畜禽健康、投入品、农事操作等不同主题数据。其中，环境数据包括温度、湿度、光照、土壤成分等；生物健康数据包括植物生长状况、病虫害监测数据等；畜禽健康数据包括体温、体重等；投入品数据包括化肥、饲料等；农事操作数据包括种植时间、种植密度等。品质指标字典包括农产品的营养品质、外观品质、风味品质等指标，例如番茄的番茄红素、油菜的油酸、青稞的直链淀粉含量等。管控模型字典包括模型基本信息、超参数、管控参数等。设备信息字典包括设备厂家、设备型号、协议类型等。具体示例如表2和表3所示。\n\n表2 农产品品质管控模型基本信息字典示例\n\nTable 2 Example of basic information dictionary for agricultural product quality control models\n\n![Image 7](https:\u002F\u002Fp3-sign.toutiaoimg.com\u002Ftos-cn-i-tjoges91tu\u002Fcadb0e4323da8075f13d8f4e1c06d9a0~tplv-tt-origin-web:gif.jpeg?_iz=58558&from=article.pc_detail&lk3s=953192f4&x-expires=1790121856&x-signature=KdImL2Qa1HZ559Rh0dUAWtdxyA4%3D)\n表3 农产品品质管控设备参数字典示例\n\nTable 3 Example of device parameter dictionary for agricultural product quality control\n\n![Image 8](https:\u002F\u002Fp3-sign.toutiaoimg.com\u002Ftos-cn-i-tjoges91tu\u002F38a674be5aec78f32db2f1bb6c2ac685~tplv-tt-origin-web:gif.jpeg?_iz=58558&from=article.pc_detail&lk3s=953192f4&x-expires=1790121856&x-signature=wbB885Jh471tVC9H59naEoCyzGw%3D)\n（2）数据适配接口。根据数据结构不同，农产品品质管控数据可以分为结构化类、半结构化类和非结构化类，具体如图6所示。\n\n![Image 9](https:\u002F\u002Fp3-sign.toutiaoimg.com\u002Ftos-cn-i-tjoges91tu\u002Fd72f9b0d4adb4f0d29bf092784b36813~tplv-tt-origin-web:gif.jpeg?_iz=58558&from=article.pc_detail&lk3s=953192f4&x-expires=1790121856&x-signature=cLkWeAVxeoBJbMljb0uKPOpxspQ%3D)\n图6 农产品品质数据接入方式\n\nFig. 6 Agricultural product quality data access methods\n\n结构化数据通常通过关系型数据库，常用的包括MySQL、PostgreSQL、SQL Server等；非关系型数据库，常用的包括MongoDB、InfluxDB等，文件CSV、EXCEL等，非结构化数据一般包括图像、视频等；半结构化数据通常通过JSON、XML 等格式的数据文件传输。针对数据不同接入方式，设置不同数据接口，并根据数据字典进行数据语义统一。\n\n（3）数据标准化映射。针对不同设备数据格式不一致的问题，根据数据字典，对温度、湿度、光照等多类数据进行自动识别和格式转换，实现来自不同设备的品质数据的语义统一，提高对不同厂商的同类传感数据的标准化准确率，有效解决数据异构问题。\n\n（4）数据清洗与补全。针对农业传感器易受泥土覆盖、水汽腐蚀导致数据漂移的问题，利用邻近节点的历史数据，对异常数据进行实时清洗和缺失值填补。\n\n**4.2.2 面向复杂农业场景的多维资源协同管理**\n\n针对农业品质管控中“设备多源异构、作物轮作频繁、算法模型迭代快”的应用现状，中间件可以采用“农产品品质管控设备-农产品-农产品品质管控模型”三位一体的协同管理方法。通过协议虚拟化、配置模板化和部署容器化技术，实现对底层硬件资源与上层农艺逻辑的灵活编排。\n\n（1）基于虚拟化映射的异构屏蔽实现多设备管理。面对大华、海康、精讯畅通等不同厂商设备共存的现状，在协议适配层之上引入设备虚拟化层。首先将所有物理设备的能力抽象为标准化的功能接口。例如，将不同品牌的云台摄影机统一映射为视觉传感器虚拟对象，具备图像采集、云台旋转等方法；将各类土壤传感器映射为土壤探针对象。然后，当上层应用下发采集全景信息的指令时，中间件根据设备字典中的驱动映射表，自动将其翻译为特定设备的私有二进制指令，从而实现“一次开发，多端兼容”。\n\n（2）基于作物配置文件的动态场景切换实现多作物管理。针对设施农业中常见的轮作场景，利用作物配置文件构建不同场景的定义载体，采用JSON格式，封装农产品在全生命周期不同生长阶段的环境阈值、关键农事指标、对应的传感器采样策略，以及模型依赖。当完成换茬轮作时，管理员仅需在管控设备加载新的作物配置文件，中间件即自动重置报警阈值、更新采样频率，并触发关联模型的自动拉取，无需修改任何底层代码即可实现“一键换茬”式的场景动态重构，能够极大降低系统对多作物场景的适配成本。\n\n（3）基于容器化的模型仓库与动态部署实现多模型管理。为解决病虫害识别等AI模型在不同作物间难以复用的问题，构建云边协同的模型仓库与动态容器化部署机制。云端维护一个分层级的算法库，包含通用基础模型（如目标识别模型）和专用微调模型（如番茄叶霉病识别模型），且每个模型镜像均包含完整的依赖环境。基于动态容器调度方法，当系统加载“番茄配置文件”时，编排引擎自动从仓库中拉取对应的番茄病虫害检测容器并启动；当切换为绿茶时，自动销毁旧容器并启动绿茶生长监测容器。同时，利用Docker容器的隔离特性，支持多模型在同一边缘网关上并行运行且互不干扰。\n\n**4.2.3 支持弱网的异构设备协议转换引擎**\n\n农业传感器常用的通信协议根据传输方式和应用场景的不同，可以分为有线、无线、应用层协议三大类，具体如表4所示。通过构建异构设备协议转换引擎自动识别并适配多种农业传感设备通信协议，实现农产品品质生产数据采集的“即插即用”。\n\n表4 农业领域常用协议类型\n\nTable 4 Common protocol types in the agricultural domain\n\n![Image 10](https:\u002F\u002Fp3-sign.toutiaoimg.com\u002Ftos-cn-i-tjoges91tu\u002Fef9579de81cd719a13ded89b1d77a5ca~tplv-tt-origin-web:gif.jpeg?_iz=58558&from=article.pc_detail&lk3s=953192f4&x-expires=1790121856&x-signature=n38G%2BeZTbHFrCIOFRtB6Tyo3yBY%3D)\n（1）设备自动发现与注册。针对农业物联网设备种类繁多、标准不一的特点，构建农产品智能管控设备能力自描述模型，使设备能够在接入系统时自动上报其数据格式、通信协议和功能特性等信息。中间件基于设备字典，通过识别这些信息，自动完成设备注册和配置，实现设备接入的“零配置”。基于设备协议与参数信息，更新设备字典，并采用轻量级的设备描述语言。模型支持包括土壤墒情传感器、气象监测设备、图像采集装置等在内的多种农业物联网设备，从而减少设备接入配置时间，大幅降低系统部署难度。\n\n（2）协议解析与转换。设计可扩展的协议转换引擎，支持农业物联网常用的多种通信协议，基于插件架构，可根据实际需求动态加载协议解析模块。首先内置基本协议解析库，集成主流工业协议的驱动和解析模块，支持设备即插即用。其次，针对老旧设备或私有协议，支持自定义脚本或配置工具进行协议适配，降低升级改造成本。同时，在协议转换过程中引入数据缓冲机制，确保在网络不稳定的农业环境中数据传输的可靠性。\n\n（3）面向农村弱网环境的“存储-转发”机制。针对农村4G\u002F5G信号覆盖盲区多的问题，在边缘侧设置断网容灾缓冲区。首先，当检测到网络上行链路中断时，自动将采集的数据写入本地数据库，待网络恢复后，根据数据优先级进行分级回传，避免瞬间突发的流量导致网络拥塞。同时，采用差分压缩算法处理温度这种连续平稳数据，减少无效数据传输，适应低带宽通信网络。\n\n**4.2.4 基于农产品特性的智能计算任务调度与负载均衡**\n\n由于农产品品质管控模型的计算任务具有明显的季节性波动和实时性，传统静态负载分配策略缺乏对时空异质性的适应能力，难以同时满足多场景的需求。本文综合考虑任务特性、资源状态和网络条件，设计面向农产品品质管控的动态负载均衡策略，以实现云边端三层架构的自适应任务调度。\n\n（1）农业任务分类与优先级定义。基于农产品管控需求，将模型计算任务划分为四个优先级类别。\n\n1）第一优先级别：实时响应任务，包括环境异常告警、设备故障检测等，要求响应时间\u003C100 ms，优先在端设备或边缘节点处理。\n\n2）第二优先级别：准实时任务，包括病虫害识别、生长状态监测等，要求响应时间\u003C5 s，主要在边缘层处理。\n\n3）第三优先级别：批处理任务，包括产量预测、品质评估等，要求响应时间\u003C30 min，可在边缘或云端处理。\n\n4）第四优先级别：离线分析任务，包括历史数据分析、模型训练等，没有严格的时间限制，主要在云端处理。\n\n（2）基于农业物候期的预测性资源调度。为了解决通用物联网中间件在农业全周期监测中面临的算力冗余与关键期数据缺失的问题，根据农产品的生长周期和管理需求，针对每一阶段的农艺特征，差异化配置边缘节点的采样频率、传输优先级与AI模型的加载策略，预测性地调整资源分配策略。比如，在播种期重点保障土壤监测和播种设备控制任务的计算资源；在开花坐果期或病虫害高发期，自动提高图像采集频率和病虫害识别模型的优先级，调用更多边缘算力进行病虫害检测；在收获期集中资源处理产量评估和品质检测任务；在作物苗期或休眠期，自动降低采样频率，仅保留基础环境监控进程，释放边缘计算资源。 表5以番茄为例给出了不同物候期下系统资源调度配置的具体实例，从生长周期、关注重点、采样频率到边缘模型和算力分配等维度，直观展示了上述预测性资源调度策略在实际生产中的配置方式。\n\n表5 不同物候期下的系统资源调度配置表（以番茄为例）\n\nTable 5 Resource scheduling configuration across phenological stages (a case of tomato)\n\n![Image 11](https:\u002F\u002Fp3-sign.toutiaoimg.com\u002Ftos-cn-i-tjoges91tu\u002F91b6d98854a682367fd15d5196edaeb9~tplv-tt-origin-web:gif.jpeg?_iz=58558&from=article.pc_detail&lk3s=953192f4&x-expires=1790121856&x-signature=KmADZ1FdJPSeyazx4BpMzwowDHE%3D)\n（3）能量感知的任务调整。针对大量依赖太阳能供电的野外边缘端节点，实时检测电池剩余电量和当前光照强度，当光照充足且电量高时，尽可能在本地边缘端执行复杂的视频分析任务，减少向云端传输大文件带来的通信能耗。当连续阴雨或夜间电量低时，自动切换至轻量级模型，仅上传提取后的关键特征数据，比如病虫害的文本告警替代原图，并拒绝非紧急的远程更新请求，防止节点因耗尽电量而离线。\n\n（4）边缘节点资源不足时的动态迁移。当边缘节点出现资源不足的问题时，使用三级迁移方法：第一级为边缘内迁移。实时监测边缘集群内各节点的中央处理器（Central Processing Unit, CPU）使用率、内存占用率和网络带宽，当某节点资源使用率超过80%阈值时，将低优先级任务迁移至负载较轻的邻近边缘节点。第二级为边缘到云迁移。保留关键的实时任务在边缘处理，确保农业生产的连续性；当边缘集群整体负载超过90%时，将优先级为3和4的任务上移至云端处理。第三级为紧急降级策略。在极端情况下，暂停非关键的数据分析任务，将所有计算资源优先分配给安全相关的环境监控任务。启动备用的简化算法，降低计算复杂度但保证基本功能。\n\n**4.2.5 资源受限下的模型热更新与联邦部署**\n\n为解决农业边缘网关资源有限、网络带宽不稳，以及数据隐私敏感等问题，本文所述的中间件使用模型热更新机制与“数据不出域、模型多地”的联邦部署架构，通过","智慧农业资讯 2026-09-10","2026-09-10T12:42:00Z","报道",{"impact":361,"substance":508,"depth":361,"authority":345,"freshness":750,"relevant":455,"comment":936},"国家级科研机构在核心期刊发表的云边端一体化中间件综述，系统梳理架构、关键模块与未来方向，专业深度与信息增量突出，对智慧农业工程化落地有参考价值。",[938],{"name":932,"url":929},[590,591,940,941,942],"农业物联网","农产品品质","云边端协同",{"code":324,"slug":325,"name":326,"path":327,"parent_code":271},[],{"status":598,"locked":582,"version":4,"method":599,"evidence":946},{"candidates":947,"agricultural_evidence":604},[948,952,957,962,966],{"code":16,"score":400,"supported":604,"title":949,"abstract":950,"tags":951},[],[590],[590],{"code":324,"score":683,"supported":604,"title":953,"abstract":954,"tags":956},[],[955],"设施园艺",[],{"code":332,"score":683,"supported":604,"title":958,"abstract":959,"tags":961},[],[960],"畜牧",[],{"code":24,"score":455,"supported":582,"title":963,"abstract":964,"tags":965},[],[],[591],{"code":48,"score":455,"supported":582,"title":967,"abstract":968,"tags":969},[],[],[940],[971,972],"农业人工智能 云边端协同 农业物联网 农产品品质","农业人工智能 云边端协同","农业人工智能云边端协同农业物联网农产品品质-2604","2026-09-16T00:03:51.069819Z",{"id":976,"title":977,"url":978,"summary":979,"summary_zh":980,"content":11,"source_name":579,"source_url":978,"published_at":981,"category":581,"cover_url":11,"hotness":385,"is_selected":582,"score":982,"score_detail":983,"sources":987,"tags":989,"primary_topic":993,"secondary_topics":994,"classification":995,"search_phrases":1014,"slug":1017,"view_count":400,"doi":1018,"paper":1019,"created_at":1043},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。","2026-09-14T00:00:00Z",83,{"impact":361,"substance":984,"depth":985,"authority":345,"freshness":786,"relevant":455,"comment":986},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。",[988],{"name":579,"url":978},[590,591,990,991,992],"边缘计算","库尔勒香梨","水果分级",{"code":286,"slug":287,"name":288,"path":289,"parent_code":271},[],{"status":598,"locked":582,"version":4,"method":599,"evidence":996},{"candidates":997,"agricultural_evidence":604},[998,1002,1006,1010],{"code":286,"score":683,"supported":604,"title":999,"abstract":1000,"tags":1001},[],[804],[],{"code":16,"score":455,"supported":582,"title":1003,"abstract":1004,"tags":1005},[],[],[590],{"code":24,"score":455,"supported":582,"title":1007,"abstract":1008,"tags":1009},[],[],[591],{"code":48,"score":455,"supported":582,"title":1011,"abstract":1012,"tags":1013},[],[],[990],[1015,1016],"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":1018,"openalex_id":1020,"authors":1021,"venue":579,"cited_by_count":424,"oa_url":978,"card":1038,"direction":709,"ingested_from":655},"W7212908812",[1022,1024,1027,1029,1031,1033,1035],{"name":1023,"orcid":11},"Bingyu Cao",{"name":1025,"orcid":1026},"Peng Zhou","https:\u002F\u002Forcid.org\u002F0000-0002-6345-5307",{"name":1028,"orcid":11},"Zhikai Yang",{"name":1030,"orcid":11},"Wei Chen",{"name":1032,"orcid":11},"Yingchao Wang",{"name":1034,"orcid":11},"Mingqi Kan",{"name":1036,"orcid":1037},"Haiyong Chen","https:\u002F\u002Forcid.org\u002F0000-0002-5262-4208",{"tldr":1039,"method":1040,"finding":1041,"direction":709,"opportunity":1042},"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":1045,"title":1046,"url":1047,"summary":1048,"summary_zh":11,"content":11,"source_name":1049,"source_url":11,"published_at":1050,"category":581,"cover_url":11,"hotness":385,"is_selected":582,"score":1051,"score_detail":1052,"sources":1054,"tags":1056,"primary_topic":1061,"secondary_topics":1062,"classification":1065,"search_phrases":1097,"slug":1100,"view_count":69,"doi":11,"paper":1101,"created_at":1108},2494,"露地白萝卜定植收获农机农艺垂直大模型调优方法研究(NRA-LoRA, Qwen3-4B LLM-Metric 84.32)","http:\u002F\u002Fwww.qikanvip.com\u002Fqkml\u002F166270.html","山东农业大学机械与电子工程学院李扬、褚化星、李天华、赵维松、潘浩晨、王善平,山东省设施园艺智慧生产技术装备重点实验室,农业农村部南京农业机械化研究所,山东省果树研究所联合发表。针对通用大语言模型在露地白萝卜定植与收获场景中专业性不足、农机农艺协同能力弱等问题,提出融合农机农艺知识垂直大模型调优方法。构建包含农机、农艺、农机农艺关联和定植收获关联 4 类知识专用数据集,共 14524 条数据;提出非均匀 Rank 分配的 NRA-LoRA 微调方法。Qwen3-4B(NRA-LoRA)表现最优,LLM-Metric 达 84.32,BLEU-4\u002FROUGE-1\u002F2\u002FL 分别 30.95\u002F52.68\u002F23.84\u002F41.96。","农机化研究 2026(11) 2026-09-10","2026-09-09T16:00:00Z",75,{"impact":603,"substance":321,"depth":361,"authority":345,"freshness":750,"relevant":455,"comment":1053},"面向露地白萝卜定植收获的农机农艺垂直大模型微调研究，方法新颖、数据规模可观，属细分领域实质性技术进展，但应用面较窄，适合作为主题聚合素材而非头条精选。",[1055],{"name":1049,"url":1047},[590,591,1057,1058,1059,1060],"智能农机","大语言模型","农机农艺融合","白萝卜",{"code":279,"slug":280,"name":281,"path":282,"parent_code":271},[1063,1064],{"code":112,"slug":113,"name":114,"path":115,"parent_code":72},{"code":24,"slug":25,"name":26,"path":27,"parent_code":7},{"status":598,"locked":582,"version":4,"method":599,"evidence":1066},{"candidates":1067,"agricultural_evidence":604},[1068,1075,1080,1085,1089,1093],{"code":112,"score":985,"supported":604,"title":1069,"abstract":1071,"tags":1074},[1070],"农机",[1070,1072,1073],"农业机械","机械化",[1059,1057],{"code":24,"score":665,"supported":604,"title":1076,"abstract":1078,"tags":1079},[1077],"大模型",[1077],[591,1058],{"code":279,"score":524,"supported":604,"title":1081,"abstract":1083,"tags":1084},[1082],"萝卜",[1082],[],{"code":286,"score":683,"supported":604,"title":1086,"abstract":1087,"tags":1088},[],[288],[],{"code":324,"score":683,"supported":604,"title":1090,"abstract":1091,"tags":1092},[],[955],[],{"code":16,"score":455,"supported":582,"title":1094,"abstract":1095,"tags":1096},[],[],[590],[1098,1099],"农业人工智能 农机农艺融合 大语言模型 智慧农业","农业人工智能 农机农艺融合","农业人工智能农机农艺融合大语言模型智慧农业-2494",{"doi":11,"openalex_id":11,"authors":1102,"venue":11,"cited_by_count":424,"oa_url":11,"card":1103,"direction":653,"ingested_from":711},[],{"tldr":1104,"method":1105,"finding":1106,"direction":653,"opportunity":1107},"提出NRA-LoRA微调方法，构建农机农艺知识数据集，提升白萝卜定植收获垂直大模型性能。","构建14524条农机农艺知识数据集，采用非均匀Rank分配的NRA-LoRA微调","Qwen3-4B(NRA-LoRA)最优，LLM-Metric达84.32，BLEU-4\u002FROUGE","可探索非均匀Rank分配在更多作物农机农艺场景的泛化，及知识图谱增强的垂直大模型。","2026-09-15T00:04:27.326828Z",{"id":1110,"title":1111,"url":1112,"summary":1113,"summary_zh":11,"content":11,"source_name":1114,"source_url":11,"published_at":1050,"category":581,"cover_url":11,"hotness":385,"is_selected":582,"score":1115,"score_detail":1116,"sources":1118,"tags":1120,"primary_topic":1122,"secondary_topics":1123,"classification":1124,"search_phrases":1148,"slug":1150,"view_count":683,"doi":11,"paper":1151,"created_at":1158},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",80,{"impact":361,"substance":984,"depth":361,"authority":665,"freshness":173,"relevant":455,"comment":1117},"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。",[1119],{"name":1114,"url":1112},[590,591,990,592,1121],"蓝莓",{"code":286,"slug":287,"name":288,"path":289,"parent_code":271},[],{"status":598,"locked":582,"version":4,"method":599,"evidence":1125},{"candidates":1126,"agricultural_evidence":604},[1127,1131,1136,1140,1144],{"code":286,"score":786,"supported":604,"title":1128,"abstract":1129,"tags":1130},[804],[],[1121],{"code":56,"score":683,"supported":604,"title":1132,"abstract":1133,"tags":1135},[],[1134],"农业大数据",[],{"code":24,"score":69,"supported":582,"title":1137,"abstract":1138,"tags":1139},[],[],[591,592],{"code":16,"score":455,"supported":582,"title":1141,"abstract":1142,"tags":1143},[],[],[590],{"code":48,"score":455,"supported":582,"title":1145,"abstract":1146,"tags":1147},[],[],[990],[1149,904],"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目标检测边缘计算-2490",{"doi":11,"openalex_id":11,"authors":1152,"venue":11,"cited_by_count":424,"oa_url":11,"card":1153,"direction":653,"ingested_from":711},[],{"tldr":1154,"method":1155,"finding":1156,"direction":653,"opportunity":1157},"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",{"id":1160,"title":1161,"url":1162,"summary":1163,"summary_zh":11,"content":1164,"source_name":1165,"source_url":11,"published_at":1166,"category":934,"cover_url":11,"hotness":385,"is_selected":582,"score":1051,"score_detail":1167,"sources":1169,"tags":1171,"primary_topic":1175,"secondary_topics":1176,"classification":1177,"search_phrases":1193,"slug":1196,"view_count":683,"doi":11,"paper":11,"created_at":1197},2454,"中国农科院资划所构建可解释 AI 模型实现猕猴桃产地精准溯源(AI in Agriculture)","https:\u002F\u002Fnews.sciencenet.cn\u002Fhtmlnews\u002F2026\u002F9\u002F571168.shtm","中国农科院农业资源与农业区划研究所智慧农业创新团队利用光谱成像技术与多模型组合策略,以中国 6 个地区猕猴桃为对象构建产地溯源模型,并运用可解释性分析方法对最优组合模型进行分层和整体解读。在提升溯源准确率的同时,揭示了光谱数据背后判断产地的内在逻辑。该研究为破解光谱分析中数据分析模型难以解释的难题提供了系统解决思路,在关键波长选择、模型稳定性提升及预测结果可信度增强等方面具有参考价值。成果发表在《Artificial Intelligence in Agriculture》。","[![Image 1: 科学网新闻频道](https:\u002F\u002Fnews.sciencenet.cn\u002Fimages\u002Fnews.jpg)](https:\u002F\u002Fnews.sciencenet.cn\u002F)\n\n[生命科学](https:\u002F\u002Fwww.sciencenet.cn\u002Flife\u002F) | [医学科学](https:\u002F\u002Fwww.sciencenet.cn\u002Fmedicine\u002F) | [化学科学](https:\u002F\u002Fwww.sciencenet.cn\u002Fchemistry\u002F) | [工程材料](https:\u002F\u002Fwww.sciencenet.cn\u002Fmaterial\u002F) | [信息科学](https:\u002F\u002Fwww.sciencenet.cn\u002Finformation\u002F) | [地球科学](https:\u002F\u002Fwww.sciencenet.cn\u002Fearth\u002F) | [数理科学](https:\u002F\u002Fwww.sciencenet.cn\u002Fmathematics\u002F) | [管理综合](https:\u002F\u002Fwww.sciencenet.cn\u002Fpolicy\u002F)[站内规定](https:\u002F\u002Fblog.sciencenet.cn\u002Fblog-45-1064777.html) | [手机版](https:\u002F\u002Fwap.sciencenet.cn\u002F)\n\n[首页](https:\u002F\u002Fwww.sciencenet.cn\u002F) | [新闻](https:\u002F\u002Fnews.sciencenet.cn\u002F) | [博客](https:\u002F\u002Fblog.sciencenet.cn\u002Fblog.php) | [院士](https:\u002F\u002Fnews.sciencenet.cn\u002Fys\u002F) | [人才](https:\u002F\u002Ftalent.sciencenet.cn\u002F) | [会议](https:\u002F\u002Fmeeting.sciencenet.cn\u002F) | [基金·项目](https:\u002F\u002Ffund.sciencenet.cn\u002F) | [论文](https:\u002F\u002Fpaper.sciencenet.cn\u002F) | [绘图](https:\u002F\u002Fvisual-web.sciencenet.cn\u002F) | [视频·直播](https:\u002F\u002Fblog.sciencenet.cn\u002Fblog.php?mod=video) | [小柯机器人](https:\u002F\u002Fpaper.sciencenet.cn\u002FAInews) | [医学科普](https:\u002F\u002Fyxkp.sciencenet.cn\u002F)\n\n作者：李晨,姬悦 来源：中国科学报 发布时间：2026\u002F9\u002F9 21:39:55 选择字号：小 中 大 新模型实现猕猴桃产地精准溯源 近日，中国农业科学院农业资源与农业区划研究所智慧农业创新团队根据可解释人工智能构建农产品产地溯源模型，在提升溯源准确率的同时揭示了光谱数据判断产地的内在逻辑，相关研究成果发表在《农业人工智能》（_**Artificial Intelligence in Agriculture**_）上。 利用光谱数据判断农产品产地时，现有方法多依赖单一算法，往往只追求预测准确度，却忽略了模型决策逻辑的解释，限制了可靠溯源模型的开发和对判断过程的理解。 科研团队以中国六个地区的猕猴桃为对象，融合光谱成像技术与多模型组合策略构建产地溯源模型，并运用可解释性分析方法对最优组合模型进行分层和整体解读，在提升溯源准确率的同时，揭示了光谱数据背后判断产地的内在逻辑。 该研究为破解光谱分析中数据分析模型难以解释的难题提供了系统解决思路，在关键波长选择、模型稳定性提升及预测结果可信度增强等方面具有参考价值，有助于推动光谱技术在农业生产与品质监管中的实际应用。研究得到国家重点研发计划、中国农业科学院科技创新工程等项目支持。 相关论文信息：https:\u002F\u002Fdoi.org\u002F10.1016\u002Fj.aiia.2026.05.010 版权声明：凡本网注明“来源：中国科学报、科学网、科学新闻杂志”的所有作品，网站转载，请在正文上方注明来源和作者，且不得对内容作实质性改动；微信公众号、头条号等新媒体平台，转载请联系授权。邮箱：shouquan@stimes.cn。\n\n打印 发E-mail给：\n\n以下评论只代表网友个人观点，不代表科学网观点。\n\n[![Image 2](https:\u002F\u002Fnews.sciencenet.cn\u002Fimages\u002Fnewcomm.gif)](https:\u002F\u002Fnews.sciencenet.cn\u002Fhtml\u002Fcomment.aspx?id=571168)\n\n相关新闻 相关论文\n*   1\n*   [基金委发布2个2026年度项目指南](https:\u002F\u002Fnews.sciencenet.cn\u002Fhtmlnews\u002F2026\u002F9\u002F571167.shtm)\n\n*   2\n*   [2026年度 “澳门青年学者计划”获选结果公布](https:\u002F\u002Fnews.sciencenet.cn\u002Fhtmlnews\u002F2026\u002F9\u002F571166.shtm)\n\n*   3\n*   [不收论文，一份陶哲轩参与的新期刊诞生了](https:\u002F\u002Fnews.sciencenet.cn\u002Fhtmlnews\u002F2026\u002F9\u002F571162.shtm)\n\n*   4\n*   [浙江多所高校人事调整](https:\u002F\u002Fnews.sciencenet.cn\u002Fhtmlnews\u002F2026\u002F9\u002F571160.shtm)\n\n*   5\n*   [我国天文学界倡议建设“新型智能社区”](https:\u002F\u002Fnews.sciencenet.cn\u002Fhtmlnews\u002F2026\u002F9\u002F571156.shtm)\n\n*   6\n*   [OpenAI称破解数学“千禧年难题”](https:\u002F\u002Fnews.sciencenet.cn\u002Fhtmlnews\u002F2026\u002F9\u002F571153.shtm)\n\n*   7\n*   [基金委发布3个国家自然科学基金专项项目申请指南](https:\u002F\u002Fnews.sciencenet.cn\u002Fhtmlnews\u002F2026\u002F9\u002F571152.shtm)\n\n*   8\n*   [我国建成全球首套年产500吨碳酸铷铯示范线](https:\u002F\u002Fnews.sciencenet.cn\u002Fhtmlnews\u002F2026\u002F9\u002F571151.shtm)\n\n*   1\n*   [Energies 期刊青年编委招募——邀您共同促进学术期刊发展！](https:\u002F\u002Fnews.sciencenet.cn\u002Fhtmlpaper\u002F2026\u002F9\u002F202699142522865155325.shtm)\n\n*   2\n*   [文献清单：肝脏病学2024–2025热门高访问论文| MDPI Livers](https:\u002F\u002Fnews.sciencenet.cn\u002Fhtmlpaper\u002F2026\u002F9\u002F20269915317833155324.shtm)\n\n*   3\n*   [从土地覆被变化看彩虹鱼保护：西新几内亚鸟头半岛的遥感证据 | MDPI Land](https:\u002F\u002Fnews.sciencenet.cn\u002Fhtmlpaper\u002F2026\u002F9\u002F20269915133880155323.shtm)\n\n*   4\n*   [JLPEA期刊客座编辑招募 | 邀您共同促进学术期刊发展](https:\u002F\u002Fnews.sciencenet.cn\u002Fhtmlpaper\u002F2026\u002F9\u002F202699145758302155322.shtm)\n\n*   5\n*   [攻克国家二级保护动物繁育瓶颈：斑鳠性别分化机制研究取得重要进展 | MDPI Fishes](https:\u002F\u002Fnews.sciencenet.cn\u002Fhtmlpaper\u002F2026\u002F9\u002F2026991345299155321.shtm)\n\n*   6\n*   [从防栓到精准抗凝：抗凝治疗的临床平衡与未来趋势](https:\u002F\u002Fnews.sciencenet.cn\u002Fhtmlpaper\u002F2026\u002F9\u002F202699145724255155320.shtm)\n\n*   7\n*   [文献清单：“提升老年人护理质量”方向 | MDPI Healthcare](https:\u002F\u002Fnews.sciencenet.cn\u002Fhtmlpaper\u002F2026\u002F9\u002F20269914565411155319.shtm)\n\n*   8\n*   [Foods“食品包装与保鲜”栏目——金征宇院士领衔，聚焦食品保藏与包装前沿研究](https:\u002F\u002Fnews.sciencenet.cn\u002Fhtmlpaper\u002F2026\u002F9\u002F202699145459758155318.shtm)\n\n[图片新闻](https:\u002F\u002Fnews.sciencenet.cn\u002Fimgnews.aspx)\n\n[![Image 3: 改造人体细胞能否重建骨骼](https:\u002F\u002Fnews.sciencenet.cn\u002F\u002Fupload\u002Fnews\u002Fimages\u002F2026\u002F9\u002F2026913172457210.png)](https:\u002F\u002Fnews.sciencenet.cn\u002F\u002Fhtmlnews\u002F2026\u002F9\u002F571360.shtm)[![Image 4: 我国科学家实现量子安全定位](https:\u002F\u002Fnews.sciencenet.cn\u002F\u002Fupload\u002Fnews\u002Fimages\u002F2026\u002F9\u002F20269112252315680.png)](https:\u002F\u002Fnews.sciencenet.cn\u002F\u002Fhtmlnews\u002F2026\u002F9\u002F571313.shtm)\n[![Image 5: 科学网2026年8月十佳博文榜单公布](https:\u002F\u002Fnews.sciencenet.cn\u002F\u002Fupload\u002Fnews\u002Fimages\u002F2026\u002F9\u002F20269101430125990.png)](https:\u002F\u002Fnews.sciencenet.cn\u002Fhttps:\u002F\u002Fblog.sciencenet.cn\u002Fblog-45-1551789.html)[![Image 6: 新技术可为千年木构古建“无痛”加固](https:\u002F\u002Fnews.sciencenet.cn\u002F\u002Fupload\u002Fnews\u002Fimages\u002F2026\u002F9\u002F2026952235515210.jpg)](https:\u002F\u002Fnews.sciencenet.cn\u002F\u002Fhtmlnews\u002F2026\u002F9\u002F570943.shtm)\n[>>更多](https:\u002F\u002Fnews.sciencenet.cn\u002Fimgnews.aspx)\n\n[一周新闻排行](https:\u002F\u002Fnews.sciencenet.cn\u002Fzphnews.aspx)\n*   1\n*   [到2050年，全球癌症病例可能激增67%](https:\u002F\u002Fnews.sciencenet.cn\u002F\u002Fhtmlnews\u002F2026\u002F9\u002F571066.shtm)\n\n*   2\n*   [2025年度世界一流科技期刊目录发布](https:\u002F\u002Fnews.sciencenet.cn\u002F\u002Fhtmlnews\u002F2026\u002F9\u002F571326.shtm)\n\n*   3\n*   [直播回放|清华团队“种”出光通信新材料](https:\u002F\u002Fnews.sciencenet.cn\u002Fhttps:\u002F\u002Fweibo.com\u002Fl\u002Fwblive\u002Fp\u002Fshow\u002F1022:2321325338760983478439)\n\n*   4\n*   [科学家首次捕捉到中子星从星团走向银河的踪迹](https:\u002F\u002Fnews.sciencenet.cn\u002F\u002Fhtmlnews\u002F2026\u002F9\u002F571018.shtm)\n\n*   5\n*   [OpenAI称破解数学“千禧年难题”](https:\u002F\u002Fnews.sciencenet.cn\u002F\u002Fhtmlnews\u002F2026\u002F9\u002F571153.shtm)\n\n*   6\n*   [AGI时代到来与否仍待定 教育变革已应开始](https:\u002F\u002Fnews.sciencenet.cn\u002F\u002Fhtmlnews\u002F2026\u002F9\u002F571038.shtm)\n\n*   7\n*   [广东省二医成立顽固性痛经多学科诊疗中心](https:\u002F\u002Fnews.sciencenet.cn\u002F\u002Fhtmlnews\u002F2026\u002F9\u002F571287.shtm)\n\n*   8\n*   [日本诺奖得主遭质疑多篇论文存在图片造假](https:\u002F\u002Fnews.sciencenet.cn\u002F\u002Fhtmlnews\u002F2026\u002F9\u002F571033.shtm)\n\n*   9\n*   [校企联合完成三维粘性挂谷猜想形式化验证](https:\u002F\u002Fnews.sciencenet.cn\u002F\u002Fhtmlnews\u002F2026\u002F9\u002F571198.shtm)\n\n*   10\n*   [“诺奖风向标”：2026年拉斯克奖揭晓](https:\u002F\u002Fnews.sciencenet.cn\u002F\u002Fhtmlnews\u002F2026\u002F9\u002F571247.shtm)\n\n[编辑部推荐博文](http:\u002F\u002Fblog.sciencenet.cn\u002Fblog.php?mod=recommend)\n*   ![Image 7](https:\u002F\u002Fnews.sciencenet.cn\u002Fimages\u002Ft11.gif)\n*   [科学网2026年8月十佳博文榜单公布！](https:\u002F\u002Fblog.sciencenet.cn\u002Fblog-45-1551789.html)\n\n*   ![Image 8](https:\u002F\u002Fnews.sciencenet.cn\u002Fimages\u002Ft11.gif)\n*   [封面文章 | 《盐湖研究》：锂镁分离性能研究](https:\u002F\u002Fblog.sciencenet.cn\u002Fblog-45-1551710.html)\n\n*   ![Image 9](https:\u002F\u002Fnews.sciencenet.cn\u002Fimages\u002Ft11.gif)\n*   [博士生论坛开放申请！“光化学研究：面向多学科交叉的前沿探索” | MDP ...](https:\u002F\u002Fblog.sciencenet.cn\u002Fblog-3516770-1552450.html)\n\n*   ![Image 10](https:\u002F\u002Fnews.sciencenet.cn\u002Fimages\u002Ft11.gif)\n*   [永久收藏！IEEE Transactions 全刊盘点，选刊精准不盲投！工科投稿再也 ...](https:\u002F\u002Fblog.sciencenet.cn\u002Fblog-3462448-1552447.html)\n\n*   ![Image 11](https:\u002F\u002Fnews.sciencenet.cn\u002Fimages\u002Ft11.gif)\n*   [为什么年轻女性多数都有纤细的腰身？](https:\u002F\u002Fblog.sciencenet.cn\u002Fblog-374215-1552436.html)\n\n*   ![Image 12](https:\u002F\u002Fnews.sciencenet.cn\u002Fimages\u002Ft11.gif)\n*   [科学家正尝试把药厂搬到太空](https:\u002F\u002Fblog.sciencenet.cn\u002Fblog-41174-1552162.html)\n\n*   [更多>>](https:\u002F\u002Fblog.sciencenet.cn\u002Fblog.php?mod=recommend)","中国科学报 2026-09-09","2026-09-08T16:00:00Z",{"impact":361,"substance":508,"depth":664,"authority":345,"freshness":750,"relevant":455,"comment":1168},"国家级科研团队用可解释AI提升猕猴桃产地溯源准确率并揭示光谱判别逻辑，方法新颖、结论可靠，属智慧农业细分领域实质进展，值得进入每日精选。",[1170],{"name":1165,"url":1162},[590,591,1172,1173,1174],"农产品溯源","猕猴桃","光谱成像",{"code":286,"slug":287,"name":288,"path":289,"parent_code":271},[],{"status":598,"locked":582,"version":4,"method":599,"evidence":1178},{"candidates":1179,"agricultural_evidence":604},[1180,1184,1188],{"code":286,"score":416,"supported":604,"title":1181,"abstract":1182,"tags":1183},[1173],[1173],[1173],{"code":16,"score":400,"supported":604,"title":1185,"abstract":1186,"tags":1187},[],[590],[590],{"code":24,"score":400,"supported":604,"title":1189,"abstract":1190,"tags":1192},[],[1191],"artificial intelligence",[591],[1194,1195],"农业人工智能 农产品溯源 光谱成像 智慧农业","农业人工智能 农产品溯源","农业人工智能农产品溯源光谱成像智慧农业-2454","2026-09-15T00:04:22.091088Z",{"id":1199,"title":1200,"url":1201,"summary":1202,"summary_zh":11,"content":11,"source_name":1203,"source_url":11,"published_at":1204,"category":581,"cover_url":11,"hotness":385,"is_selected":582,"score":1205,"score_detail":1206,"sources":1208,"tags":1210,"primary_topic":1215,"secondary_topics":1216,"classification":1217,"search_phrases":1228,"slug":1231,"view_count":683,"doi":11,"paper":1232,"created_at":1239},2396,"[预印本]Multi-Task Learning for Sparsely-Labeled Time Series: A Case Study on Cold-Hardiness Modeling","https:\u002F\u002Farxiv.org\u002Fabs\u002F2609.09062","arXiv 2609.09062（2026-09-10）。研究针对葡萄抗寒性的时间序列过程建模问题（数据有限且标签时序稀疏），提出基于RNN的多任务学习方法。实验证明某些MTL架构优于单任务学习和现有科学模型，并在相关萌芽期预测任务上展现良好潜力。","arXiv | 2026-09-10","2026-09-10T00:00:00Z",66,{"impact":416,"substance":361,"depth":585,"authority":416,"freshness":173,"relevant":455,"comment":1207},"预印本提出RNN多任务学习用于葡萄抗寒性稀疏标签时序建模，方法新颖但尚属早期研究，产业影响有限，可作为农业AI前沿动态收录。",[1209],{"name":1203,"url":1201},[591,1211,1212,1213,1214],"预印本","葡萄","抗寒育种","时间序列",{"code":286,"slug":287,"name":288,"path":289,"parent_code":271},[],{"status":598,"locked":582,"version":4,"method":599,"evidence":1218},{"candidates":1219,"agricultural_evidence":604},[1220,1224],{"code":286,"score":400,"supported":604,"title":1221,"abstract":1222,"tags":1223},[],[1212],[1212],{"code":24,"score":455,"supported":582,"title":1225,"abstract":1226,"tags":1227},[],[],[591],[1229,1230],"农业人工智能 抗寒育种 时间序列 预印本","农业人工智能 抗寒育种","农业人工智能抗寒育种时间序列预印本-2396",{"doi":11,"openalex_id":11,"authors":1233,"venue":11,"cited_by_count":424,"oa_url":11,"card":1234,"direction":653,"ingested_from":711},[],{"tldr":1235,"method":1236,"finding":1237,"direction":653,"opportunity":1238},"提出RNN多任务学习方法，用于标签稀疏的葡萄抗寒时间序列建模。","基于RNN的多任务学习，处理有限且时序稀疏的标签数据。","部分MTL架构优于单任务学习和现有科学模型，并利于萌芽期预测。","可探索多任务学习在更多稀疏标签农业时序场景（如病虫害、物候）中的迁移与泛化。","2026-09-14T00:06:33.014967Z",{"id":1241,"title":1242,"url":1243,"summary":1244,"summary_zh":1245,"content":11,"source_name":870,"source_url":1243,"published_at":1246,"category":581,"cover_url":11,"hotness":385,"is_selected":582,"score":1115,"score_detail":1247,"sources":1249,"tags":1251,"primary_topic":1254,"secondary_topics":1255,"classification":1256,"search_phrases":1281,"slug":1283,"view_count":455,"doi":1284,"paper":1285,"created_at":1310},2301,"A crop cultivation monitoring platform for evaluating the early growth of cucumbers","https:\u002F\u002Fdoi.org\u002F10.3389\u002Ffpls.2026.1869455","The precise temporal characterization of early growth dynamics in crops under abiotic stress is critical for stress resistance management in smart agriculture. However, traditional manual measurement methods struggle to achieve high-throughput and accurate quantification of multiple phenotypic parameters over continuous time series. To address this, we developed a full-time-series crop growth monitoring system that continuously collects fine-grained image data of cucumber seed germination and seedling growth under soil culture conditions, constructing an annotated dataset for identifying germination status and quantifying cotyledon area. The YOLOv8n-F_SGX and YOLOv8-seg models were developed and deployed, achieving detection and segmentation accuracies of 0.978 and 0.987, respectively, which enabled automatic monitoring of germination rate and precise extraction of cotyledon area. To further investigate the mitigating effects of nanomaterial priming on salt stress, we conducted germination and seedling growth experiments on cucumber seeds using a series of ZnO-NPs (particle size 30 nm) suspension concentrations at different salt stress levels. The results demonstrated that under salt stress conditions of 0–150 mmol·L −1 , priming with 100 mg·L −1 ZnO-NPs resulted in optimal germination dynamics (highest germination rate and fastest germination speed), along with the largest cotyledon area and highest growth rate in seedlings. In contrast, when the ZnO-NPs concentration increased to 400 mg·L −1 or higher, it significantly inhibited seed germination and seedling growth.","在智慧农业的抗逆管理中，精确表征作物在非生物胁迫下早期生长动态的时间特征至关重要。然而，传统的人工测量方法难以实现对连续时间序列上多个表型参数的高通量、准确量化。为解决这一问题，我们开发了一套全时间序列作物生长监测系统，持续采集土培条件下黄瓜种子萌发和幼苗生长的细粒度图像数据，构建了用于识别萌发状态和量化子叶面积的标注数据集。我们开发并部署了YOLOv8n-F_SGX和YOLOv8-seg模型，分别达到0.978和0.987的检测与分割精度，从而实现了萌发率的自动监测和子叶面积的精确提取。为进一步探究纳米材料引发对盐胁迫的缓解效应，我们在不同盐胁迫水平下，使用一系列浓度ZnO-NPs（粒径30 nm）悬浮液对黄瓜种子进行了萌发和幼苗生长实验。结果表明，在0–150 mmol·L⁻¹盐胁迫条件下，100 mg·L⁻¹ ZnO-NPs引发处理可获得最佳的萌发动态（最高萌发率和最快萌发速度），同时幼苗子叶面积最大、生长速率最高。相反，当ZnO-NPs浓度增至400 mg·L⁻¹或更高时，则显著抑制了种子萌发和幼苗生长。","2026-09-11T00:00:00Z",{"impact":361,"substance":321,"depth":361,"authority":345,"freshness":173,"relevant":455,"comment":1248},"构建黄瓜早期生长全时序监测系统，YOLOv8检测分割精度达0.978\u002F0.987，并给出ZnO-NPs缓解盐胁迫的最佳浓度，方法新颖、数据扎实，对智慧育种与抗逆栽培有参考价值。",[1250],{"name":870,"url":1243},[590,591,1252,594,1253],"盐胁迫","黄瓜育种",{"code":279,"slug":280,"name":281,"path":282,"parent_code":271},[],{"status":598,"locked":582,"version":4,"method":599,"evidence":1257},{"candidates":1258,"agricultural_evidence":604},[1259,1265,1269,1273,1277],{"code":279,"score":885,"supported":604,"title":1260,"abstract":1261,"tags":1264},[],[1262,1263],"黄瓜","cucumber",[1253],{"code":16,"score":885,"supported":604,"title":1266,"abstract":1267,"tags":1268},[],[590,888],[590],{"code":144,"score":683,"supported":604,"title":1270,"abstract":1271,"tags":1272},[],[145],[],{"code":24,"score":455,"supported":582,"title":1274,"abstract":1275,"tags":1276},[],[],[591],{"code":32,"score":455,"supported":582,"title":1278,"abstract":1279,"tags":1280},[],[],[594],[1282,623],"农业人工智能 作物表型 智慧农业 黄瓜育种","农业人工智能作物表型智慧农业黄瓜育种-2301","10.3389\u002Ffpls.2026.1869455",{"doi":1284,"openalex_id":1286,"authors":1287,"venue":870,"cited_by_count":424,"oa_url":1243,"card":1305,"direction":709,"ingested_from":655},"W7212301419",[1288,1290,1292,1294,1296,1298,1301,1303],{"name":1289,"orcid":11},"Deyi Lei",{"name":1291,"orcid":11},"Zaibiao Zhu",{"name":1293,"orcid":11},"Shen Penghong",{"name":1295,"orcid":11},"Zhibo Zhong",{"name":1297,"orcid":11},"Mohamed Ahmed Moustafa",{"name":1299,"orcid":1300},"Jieyu Xian","https:\u002F\u002Forcid.org\u002F0000-0002-2526-6644",{"name":1302,"orcid":11},"Hongbin Wu",{"name":1304,"orcid":11},"Xiuqing Fu",{"tldr":1306,"method":1307,"finding":1308,"direction":709,"opportunity":1309},"构建黄瓜早期生长监测平台，用YOLO模型自动量化发芽与子叶面积，并评估ZnO-NPs缓解盐胁迫效果。","连续时序图像采集与标注，YOLOv8n-F_SGX检测和YOLOv8-seg分割","100 mg·L−1 ZnO-NPs在0–150 mmol·L−1盐胁迫下促进发芽和幼苗生长，400","可扩展至多作物多胁迫场景，融合时序表型与深度学习实现早期胁迫预警和纳米材料精准调控。","2026-09-13T23:30:09.699346Z",{"id":1312,"title":1313,"url":1314,"summary":1315,"summary_zh":1316,"content":11,"source_name":579,"source_url":1314,"published_at":1246,"category":581,"cover_url":11,"hotness":385,"is_selected":582,"score":1317,"score_detail":1318,"sources":1320,"tags":1322,"primary_topic":1326,"secondary_topics":1327,"classification":1328,"search_phrases":1348,"slug":1350,"view_count":69,"doi":1351,"paper":1352,"created_at":1373},2284,"Lightweight nighttime citrus fruit detection in UAV imagery using an enhanced NightNet","https:\u002F\u002Fdoi.org\u002F10.1016\u002Fj.compag.2026.112378","Lightweight nighttime citrus fruit detection in UAV imagery using an enhanced NightNet。Computers and Electronics in Agriculture","基于增强型NightNet的无人机图像轻量化夜间柑橘果实检测。《计算机与电子农业》",77,{"impact":585,"substance":873,"depth":664,"authority":345,"freshness":786,"relevant":455,"comment":1319},"核心期刊论文，提出轻量化夜间柑橘检测方法，方法新颖且面向无人机夜间作业场景，具备细分领域参考价值。",[1321],{"name":579,"url":1314},[590,1323,591,1324,1325],"无人机","遥感","柑橘",{"code":286,"slug":287,"name":288,"path":289,"parent_code":271},[],{"status":598,"locked":582,"version":4,"method":599,"evidence":1329},{"candidates":1330,"agricultural_evidence":604},[1331,1336,1340,1344],{"code":286,"score":603,"supported":604,"title":1332,"abstract":1334,"tags":1335},[1333],"citrus",[1325,1333],[1325],{"code":16,"score":455,"supported":582,"title":1337,"abstract":1338,"tags":1339},[],[],[590],{"code":24,"score":455,"supported":582,"title":1341,"abstract":1342,"tags":1343},[],[],[591],{"code":32,"score":455,"supported":582,"title":1345,"abstract":1346,"tags":1347},[],[],[1324],[1349,904],"农业人工智能 智慧农业 无人机 柑橘","农业人工智能智慧农业无人机柑橘-2284","10.1016\u002Fj.compag.2026.112378",{"doi":1351,"openalex_id":1353,"authors":1354,"venue":579,"cited_by_count":424,"oa_url":11,"card":1368,"direction":861,"ingested_from":655},"W7212284586",[1355,1357,1360,1362,1364,1366],{"name":1356,"orcid":11},"Zhixian Wu",{"name":1358,"orcid":1359},"Juntao Xiong","https:\u002F\u002Forcid.org\u002F0000-0002-1095-812X",{"name":1361,"orcid":11},"Zhen Liu",{"name":1363,"orcid":11},"Sitong Li",{"name":1365,"orcid":11},"Jingru Peng",{"name":1367,"orcid":11},"Sidi Chen",{"tldr":1369,"method":1370,"finding":1371,"direction":861,"opportunity":1372},"提出增强NightNet实现无人机夜间柑橘果实轻量检测。","基于NightNet改进的轻量网络，用于无人机夜间图像。","增强NightNet在夜间无人机图像中实现轻量且准确的柑橘检测。","夜间无人机果实检测的轻量化与低照度增强可进一步结合多光谱或时序信息。","2026-09-13T23:30:02.020275Z",{"id":1375,"title":1376,"url":1377,"summary":1378,"summary_zh":1379,"content":11,"source_name":579,"source_url":1377,"published_at":1246,"category":581,"cover_url":11,"hotness":385,"is_selected":582,"score":1051,"score_detail":1380,"sources":1382,"tags":1384,"primary_topic":1386,"secondary_topics":1387,"classification":1388,"search_phrases":1405,"slug":1407,"view_count":683,"doi":1408,"paper":1409,"created_at":1427},2282,"Mushroom cap phenotype extraction based on YOLOv11-SMD","https:\u002F\u002Fdoi.org\u002F10.1016\u002Fj.compag.2026.112402","Mushroom cap phenotype extraction based on YOLOv11-SMD。Computers and Electronics in Agriculture","基于YOLOv11-SMD的蘑菇菌盖表型提取。《农业计算机与电子》",{"impact":603,"substance":508,"depth":361,"authority":345,"freshness":173,"relevant":455,"comment":1381},"核心期刊论文，提出YOLOv11-SMD用于蘑菇菌盖表型提取，方法有创新且面向食用菌智能化生产，但属细分技术进展，影响力有限。",[1383],{"name":579,"url":1377},[590,591,592,311,1385],"表型分析",{"code":309,"slug":310,"name":311,"path":312,"parent_code":271},[],{"status":598,"locked":582,"version":4,"method":599,"evidence":1389},{"candidates":1390,"agricultural_evidence":604},[1391,1397,1401],{"code":309,"score":603,"supported":604,"title":1392,"abstract":1394,"tags":1396},[1393],"mushroom",[1395,1393],"蘑菇",[311],{"code":24,"score":69,"supported":582,"title":1398,"abstract":1399,"tags":1400},[],[],[591,592],{"code":16,"score":455,"supported":582,"title":1402,"abstract":1403,"tags":1404},[],[],[590],[1406,904],"农业人工智能 智慧农业 目标检测 表型分析","农业人工智能智慧农业目标检测表型分析-2282","10.1016\u002Fj.compag.2026.112402",{"doi":1408,"openalex_id":1410,"authors":1411,"venue":579,"cited_by_count":424,"oa_url":11,"card":1422,"direction":861,"ingested_from":655},"W7212266027",[1412,1415,1417,1419],{"name":1413,"orcid":1414},"Yang Zhou","https:\u002F\u002Forcid.org\u002F0000-0002-0100-2652",{"name":1416,"orcid":11},"Quanlong Miao",{"name":1418,"orcid":11},"Yuefeng Chen",{"name":1420,"orcid":1421},"Jiajia Yu","https:\u002F\u002Forcid.org\u002F0000-0002-8764-8429",{"tldr":1423,"method":1424,"finding":1425,"direction":861,"opportunity":1426},"提出YOLOv11-SMD模型，用于蘑菇菌盖表型自动提取。","基于YOLOv11改进的SMD模型，结合图像数据提取菌盖表型。","YOLOv11-SMD能有效提取蘑菇菌盖表型，提升检测精度。","可探索多品种、多生长阶段菌盖表型动态监测与三维重建。","2026-09-13T23:30:01.859046Z",{"id":1429,"title":1430,"url":1431,"summary":1432,"summary_zh":1433,"content":11,"source_name":579,"source_url":1431,"published_at":1434,"category":581,"cover_url":11,"hotness":385,"is_selected":582,"score":784,"score_detail":1435,"sources":1437,"tags":1439,"primary_topic":1442,"secondary_topics":1443,"classification":1445,"search_phrases":1465,"slug":1468,"view_count":400,"doi":1469,"paper":1470,"created_at":1492},2276,"PT-GNN: A physiological topology-aware graph neural network for multimodal early detection of cucumber downy mildew using hyperspectral and chlorophyll fluorescence sensing","https:\u002F\u002Fdoi.org\u002F10.1016\u002Fj.compag.2026.112433","PT-GNN: A physiological topology-aware graph neural network for multimodal early detection of cucumber downy mildew using hyperspectral and chlorophyll fluorescence sensing。Computers and Electronics in Agriculture","PT-GNN：一种生理拓扑感知图神经网络，利用高光谱和叶绿素荧光传感进行黄瓜霜霉病的多模态早期检测。","2026-09-12T00:00:00Z",{"impact":361,"substance":321,"depth":361,"authority":345,"freshness":786,"relevant":455,"comment":1436},"提出生理拓扑感知图神经网络融合高光谱与叶绿素荧光实现黄瓜霜霉病早期检测，方法新颖且发表于农业信息领域核心期刊，具备较强专业参考价值。",[1438],{"name":579,"url":1431},[590,591,1262,1440,1441],"病害检测","高光谱遥感",{"code":279,"slug":280,"name":281,"path":282,"parent_code":271},[1444],{"code":32,"slug":33,"name":34,"path":35,"parent_code":7},{"status":598,"locked":582,"version":4,"method":599,"evidence":1446},{"candidates":1447,"agricultural_evidence":604},[1448,1452,1457,1461],{"code":279,"score":603,"supported":604,"title":1449,"abstract":1450,"tags":1451},[1263],[1262,1263],[1262],{"code":32,"score":416,"supported":604,"title":1453,"abstract":1455,"tags":1456},[1454],"hyperspectral",[1454],[1441],{"code":16,"score":455,"supported":582,"title":1458,"abstract":1459,"tags":1460},[],[],[590],{"code":24,"score":455,"supported":582,"title":1462,"abstract":1463,"tags":1464},[],[],[591],[1466,1467],"农业人工智能 高光谱遥感 智慧农业 病害检测","农业人工智能 高光谱遥感","农业人工智能高光谱遥感智慧农业病害检测-2276","10.1016\u002Fj.compag.2026.112433",{"doi":1469,"openalex_id":1471,"authors":1472,"venue":579,"cited_by_count":424,"oa_url":11,"card":1487,"direction":861,"ingested_from":655},"W7212310174",[1473,1476,1478,1480,1482,1484],{"name":1474,"orcid":1475},"Yibin Li","https:\u002F\u002Forcid.org\u002F0000-0002-5906-5074",{"name":1477,"orcid":11},"Zonghuan Han",{"name":1479,"orcid":11},"Yong Wang",{"name":1481,"orcid":11},"Wei Gao",{"name":1483,"orcid":11},"Yiding Zhang",{"name":1485,"orcid":1486},"Lingxian Zhang","https:\u002F\u002Forcid.org\u002F0000-0002-8665-7075",{"tldr":1488,"method":1489,"finding":1490,"direction":861,"opportunity":1491},"提出生理拓扑感知图神经网络PT-GNN，融合高光谱与叶绿素荧光实现黄瓜霜霉病早期检测。","构建生理拓扑图神经网络，融合高光谱与叶绿素荧光多模态传感数据。","PT-GNN能有效利用生理拓扑关系，提升黄瓜霜霉病早期检测精度。","可探索生理拓扑图构建的通用性，迁移至其他作物病害及多模态传感器融合场景。","2026-09-13T23:30:01.409569Z",{"id":1494,"title":1495,"url":1496,"summary":1497,"summary_zh":11,"content":11,"source_name":1498,"source_url":11,"published_at":1166,"category":581,"cover_url":11,"hotness":385,"is_selected":582,"score":1317,"score_detail":1499,"sources":1501,"tags":1503,"primary_topic":1506,"secondary_topics":1507,"classification":1509,"search_phrases":1533,"slug":1536,"view_count":69,"doi":11,"paper":1537,"created_at":1544},2274,"ArXiv 2609.10469 AgroVisNet:轻量化卷积网络及萝卜-马铃薯-葫芦病害诊断基准","https:\u002F\u002Farxiv.org\u002Fabs\u002F2609.10469","提出AgroVisNet紧凑型卷积网络及专家验证基准BD-PlantDX,包含孟加拉国Bogura和Nilphamari地区12类萝卜、土豆、葫芦共12432张田间图像。模型参数仅29万,测试精度99.52%,加权F1 99.52%,部署后量化0.46MB,CPU推理8.40ms\u002F张。","arXiv 2609.10469",{"impact":361,"substance":321,"depth":361,"authority":665,"freshness":750,"relevant":455,"comment":1500},"提出29万参数轻量卷积网络与万余张田间病害基准数据集，量化后仅0.46MB、CPU单张8.4ms，对低成本边缘部署的作物病害诊断有实用参考价值。",[1502],{"name":1498,"url":1496},[590,591,990,1504,1505],"植物病害识别","轻量化模型",{"code":279,"slug":280,"name":281,"path":282,"parent_code":271},[1508],{"code":240,"slug":241,"name":242,"path":243,"parent_code":200},{"status":598,"locked":582,"version":4,"method":599,"evidence":1510},{"candidates":1511,"agricultural_evidence":604},[1512,1516,1521,1525,1529],{"code":279,"score":524,"supported":604,"title":1513,"abstract":1514,"tags":1515},[1082],[1082],[],{"code":240,"score":173,"supported":604,"title":1517,"abstract":1519,"tags":1520},[1518],"马铃薯",[],[],{"code":24,"score":683,"supported":582,"title":1522,"abstract":1523,"tags":1524},[],[],[591,1504,1505],{"code":16,"score":455,"supported":582,"title":1526,"abstract":1527,"tags":1528},[],[],[590],{"code":48,"score":455,"supported":582,"title":1530,"abstract":1531,"tags":1532},[],[],[990],[1534,1535],"农业人工智能 植物病害识别 轻量化模型 智慧农业","农业人工智能 植物病害识别","农业人工智能植物病害识别轻量化模型智慧农业-2274",{"doi":11,"openalex_id":11,"authors":1538,"venue":11,"cited_by_count":424,"oa_url":11,"card":1539,"direction":653,"ingested_from":711},[],{"tldr":1540,"method":1541,"finding":1542,"direction":653,"opportunity":1543},"提出轻量卷积网络AgroVisNet及萝卜、马铃薯、葫芦病害诊断基准BD-PlantDX。","构建12432张田间图像基准，设计29万参数紧凑CNN并量化部署。","测试精度99.52%，量化后仅0.46MB，CPU推理8.40ms\u002F张。","可探索跨地区跨作物泛化、田间复杂光照下的轻量模型鲁棒性与边缘部署。","2026-09-13T00:04:05.295687Z",{"id":1546,"title":1547,"url":1548,"summary":1549,"summary_zh":11,"content":11,"source_name":1550,"source_url":11,"published_at":1551,"category":581,"cover_url":11,"hotness":385,"is_selected":582,"score":1552,"score_detail":1553,"sources":1555,"tags":1557,"primary_topic":1559,"secondary_topics":1560,"classification":1561,"search_phrases":1576,"slug":1579,"view_count":400,"doi":11,"paper":1580,"created_at":1587},2270,"华中农大实现柑橘黄龙病早期无损诊断 提LLM辅助拉曼光谱相似度融合框架","https:\u002F\u002Fwww.toutiao.com\u002Farticle\u002F7682626369448165922","华中农业大学工学院联合园艺林学学院在Agriculture Communications发表研究,提出LLM辅助的多通道光谱相似度融合框架(LAMSSF),依托少量经qPCR验证的参考样本,实现柑橘HLB早期无损诊断,降低模型对大规模标注数据的依赖。","华中农业大学 Agriculture Communications","2026-09-06T16:00:00Z",82,{"impact":321,"substance":321,"depth":361,"authority":345,"freshness":750,"relevant":455,"comment":1554},"华中农大提出LLM辅助拉曼光谱融合框架，实现柑橘黄龙病早期无损诊断，方法新颖且降低标注依赖，具备行业推广价值。",[1556],{"name":1550,"url":1548},[590,591,1440,1325,1558],"光谱遥感",{"code":286,"slug":287,"name":288,"path":289,"parent_code":271},[],{"status":598,"locked":582,"version":4,"method":599,"evidence":1562},{"candidates":1563,"agricultural_evidence":604},[1564,1568,1572],{"code":286,"score":416,"supported":604,"title":1565,"abstract":1566,"tags":1567},[1325],[1325],[1325],{"code":16,"score":455,"supported":582,"title":1569,"abstract":1570,"tags":1571},[],[],[590],{"code":24,"score":455,"supported":582,"title":1573,"abstract":1574,"tags":1575},[],[],[591],[1577,1578],"农业人工智能 光谱遥感 智慧农业 病害检测","农业人工智能 光谱遥感","农业人工智能光谱遥感智慧农业病害检测-2270",{"doi":11,"openalex_id":11,"authors":1581,"venue":11,"cited_by_count":424,"oa_url":11,"card":1582,"direction":653,"ingested_from":711},[],{"tldr":1583,"method":1584,"finding":1585,"direction":653,"opportunity":1586},"提出LLM辅助多通道光谱相似度融合框架，实现柑橘黄龙病早期无损诊断。","拉曼光谱多通道相似度融合，LLM辅助，少量qPCR验证参考样本。","少量标注样本即可实现HLB早期无损诊断，降低对大规模标注数据依赖。","可探索LLM辅助光谱融合在更多作物病害早期诊断中的迁移与少样本泛化能力。","2026-09-13T00:04:04.913308Z",{"id":1589,"title":1590,"url":1591,"summary":1592,"summary_zh":11,"content":11,"source_name":1593,"source_url":11,"published_at":1166,"category":581,"cover_url":11,"hotness":385,"is_selected":582,"score":1115,"score_detail":1594,"sources":1596,"tags":1598,"primary_topic":1600,"secondary_topics":1601,"classification":1602,"search_phrases":1625,"slug":1628,"view_count":424,"doi":11,"paper":1629,"created_at":1636},2264,"南京农业大学等单位3MSP2多簇番茄序贯采摘规划在Computers and Electronics in Agriculture发表","https:\u002F\u002Fpubscholar.cn\u002Fheadlines\u002F1fe11dc4f9cf2a7db5fd265751febd525e0c4ea77d8edb8ce4d6232ff27da05120c65eab10d9c0dbda65b8d76671d51f","山东农业大学学者在Computers and Electronics in Agriculture发表3MSP2方法,结合\"眼在臂上\"和\"眼在手上\"双RGB-D相机视角,从簇级排序到单果姿态估计的完整决策链,实现温室大果型番茄\"低损伤高效率\"连续采收。","PubScholar 智慧设施种养研究专题",{"impact":361,"substance":321,"depth":361,"authority":345,"freshness":173,"relevant":455,"comment":1595},"双RGB-D视角簇级序贯采摘规划方法发表于农业工程核心期刊，方法新颖、决策链完整，对温室番茄自动化采收有实质参考价值。",[1597],{"name":1593,"url":1591},[590,591,1599,593,955],"采摘机器人",{"code":279,"slug":280,"name":281,"path":282,"parent_code":271},[],{"status":598,"locked":582,"version":4,"method":599,"evidence":1603},{"candidates":1604,"agricultural_evidence":604},[1605,1609,1613,1617,1621],{"code":279,"score":416,"supported":604,"title":1606,"abstract":1607,"tags":1608},[593],[593],[593],{"code":324,"score":400,"supported":604,"title":1610,"abstract":1611,"tags":1612},[],[747],[955],{"code":120,"score":455,"supported":582,"title":1614,"abstract":1615,"tags":1616},[],[],[1599],{"code":16,"score":455,"supported":582,"title":1618,"abstract":1619,"tags":1620},[],[],[590],{"code":24,"score":455,"supported":582,"title":1622,"abstract":1623,"tags":1624},[],[],[591],[1626,1627],"农业人工智能 采摘机器人 智慧农业 设施园艺","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设施园艺-2264",{"doi":11,"openalex_id":11,"authors":1630,"venue":11,"cited_by_count":424,"oa_url":11,"card":1631,"direction":653,"ingested_from":711},[],{"tldr":1632,"method":1633,"finding":1634,"direction":653,"opportunity":1635},"提出3MSP2方法，实现温室大果番茄从簇级排序到单果姿态估计的序贯采摘规划。","双RGB-D相机（眼在臂上+眼在手上）与簇级排序、单果姿态估计决策链。","该方法可实现温室大果型番茄低损伤、高效率的连续采收。","可探索多臂协同采摘与动态障碍物下的实时重规划，提升复杂温室环境适应性。","2026-09-13T00:04:03.523312Z",{"id":1638,"title":1639,"url":1640,"summary":1641,"summary_zh":11,"content":11,"source_name":1642,"source_url":11,"published_at":1643,"category":581,"cover_url":11,"hotness":385,"is_selected":582,"score":1644,"score_detail":1645,"sources":1647,"tags":1649,"primary_topic":1651,"secondary_topics":1652,"classification":1654,"search_phrases":1673,"slug":1675,"view_count":455,"doi":11,"paper":1676,"created_at":1683},2215,"中国农业大学马韫韬教授团队:攻克温室黄瓜冠层点云分割难题 助力设施数字化精准管理","https:\u002F\u002Fh.mffb.com.cn\u002Fa\u002F569056.html","马韫韬教授团队构建Cuc-HiSeg分割框架,通过尺度变换、角度旋转、边缘扰动等策略扩充数据集,显著缓解三维标注样本稀缺难题,使深度学习模型分割F1值提升5.6%-6.4%;研发结合形态学优化与K-Means++二次分割的改进区域生长算法,有效克服叶片重叠与点云缺失干扰,叶片实例分割F1值最高达95.0%,在结果期复杂场景下仍保持93.9%的高精度。该方法在苗期、开花初期、结果期均表现出优异的稳定性与适应性,可高效完成日光温室黄瓜冠层点云的高精度解析。","农业信息网 2026年9月8日","2026-09-08T00:00:00Z",78,{"impact":361,"substance":321,"depth":361,"authority":345,"freshness":750,"relevant":455,"comment":1646},"中国农大团队提出黄瓜冠层点云分割新框架，F1值提升显著且多生育期稳定，属设施农业数字化领域有实质增量的技术突破。",[1648],{"name":1642,"url":1640},[590,591,726,1262,1650],"点云分割",{"code":279,"slug":280,"name":281,"path":282,"parent_code":271},[1653],{"code":324,"slug":325,"name":326,"path":327,"parent_code":271},{"status":598,"locked":582,"version":4,"method":599,"evidence":1655},{"candidates":1656,"agricultural_evidence":604},[1657,1661,1665,1669],{"code":279,"score":416,"supported":604,"title":1658,"abstract":1659,"tags":1660},[1262],[1262],[1262],{"code":324,"score":524,"supported":604,"title":1662,"abstract":1663,"tags":1664},[747],[747],[],{"code":24,"score":400,"supported":604,"title":1666,"abstract":1667,"tags":1668},[],[810],[591],{"code":16,"score":455,"supported":582,"title":1670,"abstract":1671,"tags":1672},[],[],[590],[1674,904],"农业人工智能 智慧农业 点云分割 设施农业","农业人工智能智慧农业点云分割设施农业-2215",{"doi":11,"openalex_id":11,"authors":1677,"venue":11,"cited_by_count":424,"oa_url":11,"card":1678,"direction":861,"ingested_from":711},[],{"tldr":1679,"method":1680,"finding":1681,"direction":861,"opportunity":1682},"构建Cuc-HiSeg框架，实现温室黄瓜冠层点云高精度叶片实例分割。","数据增强扩充点云样本，结合改进区域生长与K-Means++二次分割。","叶片分割F1最高95.0%，结果期复杂场景仍达93.9%，各生育期稳定。","可迁移至其他设施作物冠层解析，并探索点云分割与生长量、产量性状的自动关联。","2026-09-12T00:06:40.512918Z",{"id":1685,"title":1686,"url":1687,"summary":1688,"summary_zh":1689,"content":11,"source_name":870,"source_url":1687,"published_at":1204,"category":581,"cover_url":11,"hotness":385,"is_selected":582,"score":784,"score_detail":1690,"sources":1692,"tags":1694,"primary_topic":1697,"secondary_topics":1698,"classification":1700,"search_phrases":1718,"slug":1721,"view_count":683,"doi":1722,"paper":1723,"created_at":1735},2158,"EfficientNet-CBAM-prototype: an attention-guided EfficientNet framework with dynamic prototype representation for tomato disease classification","https:\u002F\u002Fdoi.org\u002F10.3389\u002Ffpls.2026.1918175","Introduction In the field of precision agriculture, one of the major hurdles is the early and accurate identification of plant diseases. Farmers may face serious irreversible loss in yield if there is a delay in diagnosis by even a few days. The CNN model has helped in improving the classification of plant diseases but faces difficulty in distinguishing fine-grained symptoms, has limited generalizability, and is not easily interpretable. Though the deep CNN model performs well in classifying plant diseases, the current methods have several flaws. Most importantly, existing algorithms misclassify visually complex samples due to their limited spatial differentiation of disease-specific morphological features, including lesion borders and necrotic regions. Classification reliability is further compromised by low inter-class embedding separability for visually comparable illness phenotypes. Apart from these representational problems, training instability is still a major problem for attention-based models. Combining randomly initialized attention modules with pretrained backbone networks causes this instability, which still limits practical application. Methods We suggest EfficientNet-CBAM-Prototype (ECP-Net), a unique end-to-end deep learning architecture, to overcome these constraints. Three complementary techniques are combined into a single framework by ECP-Net. First, parameter-efficient multi-scale feature extraction is done using an EfficientNetB0 backbone. Second, joint channel-wise and spatial feature recalibration is performed using a stabilized convolutional block attention module (CBAM). Third, a dynamic prototype memory layer uses cosine similarity-based categorization and exponential moving average (EMA) updates to maintain class-representative embedding vectors. To overcome the instability caused by randomly initialized attention weights, we further propose a two-phase training strategy wherein CBAM is frozen during phase 1 to allow prototype stabilization and then jointly fine-tuned with the learning rate in phase 2. Results Evaluated on the PlantVillage tomato subset comprising 10 disease classes across a class-balanced split of 10,000 training, 500 validation, and 500 test samples, ECP-Net achieves 98.6% test accuracy, 98.59% F1-score, and 98.65% precision with only 4.80M parameters and 85.04 ms average inference time. These results outperformed baselines including VGG16 (97.00%), ResNet50 (81.20%), MobileNetV2 (81.20%), and CNN (70.00%). Discussion Generalization is further validated on 35 real-field tomato leaf images captured under natural, uncontrolled conditions, confirming practical deployment potential.","引言 在精准农业领域，植物病害的早期准确识别是主要难题之一。若诊断延迟哪怕几天，农民可能面临严重的不可逆产量损失。CNN模型有助于改进植物病害分类，但在区分细粒度症状方面存在困难，泛化能力有限，且不易解释。尽管深度CNN模型在植物病害分类中表现良好，当前方法仍存在若干缺陷。最重要的是，现有算法由于对病害特异性形态特征（包括病斑边缘和坏死区域）的空间区分能力有限，会对视觉上复杂的样本产生误分类。对于视觉上相似的病害表型，类间嵌入可分性低，进一步损害了分类可靠性。除这些表征问题外，训练不稳定性仍是基于注意力模型的主要问题。随机初始化的注意力模块与预训练骨干网络相结合会导致这种不稳定性，这仍然限制了实际应用。方法 我们提出EfficientNet-CBAM-Prototype（ECP-Net），一种新颖的端到端深度学习架构，以克服这些限制。ECP-Net将三种互补技术结合到一个统一框架中。首先，使用EfficientNetB0骨干网络进行参数高效的多尺度特征提取。其次，使用稳定化的卷积块注意力模块（CBAM）进行通道与空间的联合特征重校准。第三，动态原型记忆层使用基于余弦相似度的分类和指数移动平均（EMA）更新来维护类代表性嵌入向量。为克服随机初始化注意力权重引起的不稳定性，我们进一步提出两阶段训练策略，其中CBAM在第一阶段冻结以使原型稳定，然后在第二阶段以学习率联合微调。结果 在PlantVillage番茄子集上评估，该子集包含10个病害类别，按类别均衡划分的10,000个训练样本、500个验证样本和500个测试样本，ECP-Net达到98.6%的测试准确率、98.59%的F1分数和98.65%的精确率，仅需4.80M参数和85.04 ms平均推理时间。这些结果优于基线，包括VGG16（97.00%）、ResNet50（81.20%）、MobileNetV2（81.20%）和CNN（70.00%）。",{"impact":361,"substance":321,"depth":361,"authority":345,"freshness":786,"relevant":455,"comment":1691},"提出注意力引导与动态原型结合的番茄病害识别框架，在公开数据集上取得98.6%准确率并完成真实田间图像验证，方法新颖、数据扎实，对作物病害智能诊断有实用参考价值。",[1693],{"name":870,"url":1687},[590,591,810,1695,1696],"病虫害识别","番茄种植",{"code":279,"slug":280,"name":281,"path":282,"parent_code":271},[1699],{"code":24,"slug":25,"name":26,"path":27,"parent_code":7},{"status":598,"locked":582,"version":4,"method":599,"evidence":1701},{"candidates":1702,"agricultural_evidence":604},[1703,1708,1712],{"code":279,"score":361,"supported":604,"title":1704,"abstract":1705,"tags":1707},[606],[593,1706,606],"茄子",[1696],{"code":24,"score":786,"supported":604,"title":1709,"abstract":1710,"tags":1711},[],[810,808],[591,810,1695],{"code":16,"score":885,"supported":604,"title":1713,"abstract":1714,"tags":1717},[],[1715,1716],"精准农业","precision agriculture",[590],[1719,1720],"农业人工智能 病虫害识别 智慧农业 深度学习","农业人工智能 病虫害识别","农业人工智能病虫害识别智慧农业深度学习-2158","10.3389\u002Ffpls.2026.1918175",{"doi":1722,"openalex_id":1724,"authors":1725,"venue":870,"cited_by_count":424,"oa_url":1687,"card":1730,"direction":861,"ingested_from":655},"W7212120374",[1726,1728],{"name":1727,"orcid":11},"E Jansi",{"name":1729,"orcid":11},"Kavitha BR",{"tldr":1731,"method":1732,"finding":1733,"direction":653,"opportunity":1734},"提出ECP-Net，用注意力与动态原型提升番茄病害分类精度与可解释性。","EfficientNetB0+稳定CBAM+动态原型记忆层，两阶段训练策略。","在PlantVillage番茄10类上达98.6%准确率，仅4.80M参数，优于多个基线。","可探索原型可解释性在田间复杂背景与跨域病害识别中的泛化能力。","2026-09-11T23:30:25.043704Z",{"technologies":1737,"keywords":1778},[1738,1739,1740,1742,1743,1744,1745,1746,1748,1749,1751,1753,1755,1757,1758,1760,1762,1764,1765,1766,1768,1770,1771,1772,1774,1776],{"name":591,"count":571},{"name":1715,"count":984},{"name":1741,"count":585},"农业机器人",{"name":1324,"count":603},{"name":592,"count":385},{"name":990,"count":786},{"name":810,"count":173},{"name":1747,"count":173},"遥感监测",{"name":940,"count":750},{"name":1750,"count":408},"分子育种",{"name":1752,"count":408},"数字孪生",{"name":1754,"count":408},"生物育种",{"name":1756,"count":400},"农业遥感",{"name":1172,"count":400},{"name":1759,"count":400},"图像识别",{"name":1761,"count":400},"物联网",{"name":1763,"count":400},"计算机视觉",{"name":594,"count":683},{"name":741,"count":683},{"name":1767,"count":69},"图像分割",{"name":1769,"count":69},"基因编辑",{"name":1058,"count":69},{"name":1441,"count":69},{"name":1773,"count":455},"AI",{"name":1775,"count":455},"强化学习",{"name":1777,"count":455},"高光谱",[1779,1781,1784,1785,1787,1789,1791,1793,1795,1796,1798,1800],{"name":590,"count":1780},174,{"name":1782,"count":1783},"种业振兴",73,{"name":311,"count":447},{"name":1786,"count":984},"成果转化",{"name":1788,"count":361},"乡村振兴",{"name":1790,"count":361},"种质资源",{"name":1792,"count":664},"院地合作",{"name":1794,"count":603},"数字农业",{"name":593,"count":665},{"name":1797,"count":665},"科技特派员",{"name":1799,"count":524},"咖啡产业",{"name":1801,"count":524},"花卉产业"]