[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"item-1647":3,"related-1647":81},{"id":4,"title":5,"url":6,"summary":7,"summary_zh":8,"source_name":9,"source_url":6,"published_at":10,"category":11,"cover_url":12,"hotness":13,"is_selected":14,"embedding_ready":15,"score":16,"sources":17,"tags":19,"primary_topic":12,"secondary_topics":21,"search_phrases":22,"slug":23,"view_count":24,"doi":25,"paper":26,"created_at":53,"content":12,"score_detail":54,"classification":56},1647,"Underwater Small Object Detection: A Survey","https:\u002F\u002Fdoi.org\u002F10.26599\u002Fair.2026.9150010","Abstract Underwater small object detection is one of the emerging core technologies at the intersection of computer vision and underwater imaging detection, which aims to achieve efficient and accurate detection and identification of faint, small targets in underwater environments, and has been widely applied in fields of aquatic organism detection, mineral resource exploration and other fields. Among them, underwater object detection methods based on optical images, with their high resolution and good flexibility, have demonstrated significant advantages in underwater biological detection, recognition and other scenarios. However, there is currently a relative lack of comprehensive reviews on existing research regarding underwater small object detection. To address this gap, this paper presents a systematic review of small object detection in underwater optical images. Firstly, at the level of problems and data, the specific problems faced by underwater small detection are comprehensively sorted out and classified into two categories: the inherent characteristics of underwater small objects, and the degradation characteristics of underwater images. Then 16 existing typical datasets are sorted out and analyzed in detail. Subsequently, at the methodological level, existing underwater small object detection algorithms are systematically summarized from the perspectives of traditional machine learning and deep learning, and an in-depth analysis of improved deep learning-based strategies is conducted by covering five aspects: data augmentation and pre-processing, multi-scale feature fusion, attention mechanism introduction, loss function optimization, and lightweight design. Finally, the challenges faced by underwater small object detection at present are pointed out, and the potential directions for future development are explored, aiming to provide valuable references for subsequent related research.","水下小目标检测是计算机视觉与水下成像探测交叉领域的新兴核心技术之一，旨在实现对水下环境中微弱、小型目标的高效精准检测与识别，已广泛应用于水生生物检测、矿产资源勘探等领域。其中，基于光学图像的水下目标检测方法凭借其高分辨率和良好的灵活性，在水下生物检测、识别等场景中展现出显著优势。然而，目前关于水下小目标检测的现有研究尚缺乏较为全面的综述。为弥补这一不足，本文对水下光学图像中的小目标检测进行了系统性综述。首先，在问题与数据层面，全面梳理了水下小目标检测面临的具体问题，并将其划分为两类：水下小目标自身固有特性以及水下图像退化特性。随后，对现有16个典型数据集进行了详细整理与分析。其次，在方法层面，从传统机器学习和深度学习两个视角系统总结了现有水下小目标检测算法，并从数据增强与预处理、多尺度特征融合、注意力机制引入、损失函数优化以及轻量化设计五个方面对基于深度学习的改进策略进行了深入剖析。最后，指出了当前水下小目标检测所面临的挑战，并探讨了未来潜在的发展方向，旨在为后续相关研究提供有价值的参考。","CAAI Artificial Intelligence Research","2026-09-02T00:00:00Z","论文",null,10,false,true,0,[18],{"name":9,"url":6},[20],"智慧农业",[],[],"UnderwaterSmallObjectDetection:ASurvey-1647",1,"10.26599\u002Fair.2026.9150010",{"doi":25,"openalex_id":27,"authors":28,"venue":9,"cited_by_count":16,"oa_url":44,"card":45,"direction":51,"ingested_from":52},"W7204992344",[29,31,33,35,37,38,39,41],{"name":30,"orcid":12},"Peixian Zhuang",{"name":32,"orcid":12},"Yuchun Wang",{"name":34,"orcid":12},"Yihang Wang",{"name":36,"orcid":12},"Fei Liu",{"name":34,"orcid":12},{"name":36,"orcid":12},{"name":40,"orcid":12},"Yanchen Guo",{"name":42,"orcid":43},"Zhenqi Fu","https:\u002F\u002Forcid.org\u002F0000-0003-2950-7190","https:\u002F\u002Fwww.sciopen.com\u002Flocal\u002Farticle_pdf\u002F10.26599\u002FAIR.2026.9150010.pdf",{"tldr":46,"method":47,"finding":48,"direction":49,"opportunity":50},"综述水下光学图像中的小目标检测，梳理问题、数据集、算法，并展望未来方向。","系统综述传统机器学习与深度学习算法，分析数据增强、多尺度融合、注意力机制等改进策","水下小目标检测面临目标特征弱与图像退化双重挑战，深度学习改进策略是主流。","农业人工智能与决策模型","可借鉴水下小目标检测方法，应用于水产养殖中的鱼苗或病害监测，探索轻量化模型与数据增强技术。","农业遥感与作物表型","openalex","2026-09-04T23:30:33.706301Z",{"impact":16,"substance":16,"depth":16,"authority":16,"freshness":16,"relevant":16,"comment":55},"该论文主题为水下小目标检测，与三农、农业信息化等领域无直接关联，不进入每日精选。",{"status":57,"locked":14,"version":58,"method":59,"evidence":60},"needs_review","2026-10-22.2","rules",{"candidates":61,"agricultural_evidence":14,"reason":80},[62,75],{"code":63,"score":64,"supported":15,"title":65,"abstract":66,"tags":74},"I2",9,[],[67,68,69,70,71,72,73],"目标检测","深度学习","机器学习","计算机视觉","machine learning","deep learning","computer vision",[],{"code":76,"score":24,"supported":14,"title":77,"abstract":78,"tags":79},"I1",[],[],[20],"农业相关性需要复核",{"total":82,"page":24,"page_size":82,"items":83},6,[84,113,141,176,204,236],{"id":85,"title":86,"url":87,"summary":88,"summary_zh":12,"source_name":89,"source_url":12,"published_at":90,"category":11,"cover_url":12,"hotness":13,"is_selected":14,"embedding_ready":15,"score":91,"sources":92,"tags":94,"primary_topic":12,"secondary_topics":99,"search_phrases":100,"slug":103,"view_count":16,"doi":12,"paper":104,"created_at":112},5658,"基于混合卷积Transformer与知识自蒸馏的香蕉品种跨等级分类泛化增强","https:\u002F\u002Farxiv.org\u002Fabs\u002F2610.29008","预印本，未同行评审。针对香蕉品种分类中训练等级与实际品质等级的域偏移，用31678张图像、三个品种三个等级数据并采用留一等级交叉验证，比较卷积网络、视觉Transformer与双流架构；提出的混合模型加自蒸馏平均准确率达81.25%，优于两类基线。","arXiv 2026-10-21","2026-10-20T16:00:00Z",56,[93],{"name":89,"url":87},[20,95,96,97,98],"农业人工智能","作物分类","香蕉","领域泛化",[],[101,102],"香蕉品种 跨等级分类","混合卷积Transformer 知识自蒸馏","香蕉品种跨等级分类-5658",{"doi":12,"openalex_id":12,"authors":105,"venue":12,"cited_by_count":16,"oa_url":12,"card":106,"direction":49,"ingested_from":111},[],{"tldr":107,"method":108,"finding":109,"direction":49,"opportunity":110},"提出混合卷积Transformer加知识自蒸馏，提升香蕉品种跨等级分类泛化能力。","31678张图像、三品种三等级，留一等级交叉验证，对比CNN、ViT与双流架构。","混合模型加自蒸馏平均准确率81.25%，优于卷积网络和视觉Transformer基线。","可探索知识自蒸馏在更多农产品跨域分级中的应用，并研究轻量化部署与真实分拣场景验证。","agent","2026-10-23T00:32:42.976104Z",{"id":114,"title":115,"url":116,"summary":117,"summary_zh":12,"source_name":118,"source_url":12,"published_at":119,"category":11,"cover_url":12,"hotness":13,"is_selected":14,"embedding_ready":15,"score":120,"sources":121,"tags":123,"primary_topic":12,"secondary_topics":128,"search_phrases":129,"slug":132,"view_count":16,"doi":12,"paper":133,"created_at":140},5657,"ADA 2025：甜菜水分亏缺胁迫检测多模态数据集","https:\u002F\u002Farxiv.org\u002Fabs\u002F2610.23609","预印本，未同行评审。发布超过500GB的甜菜缺水胁迫多模态公开数据集，融合热红外、高光谱、彩色影像与深度影像，附环境元数据及相对含水量、叶厚两类真值，填补作物水分胁迫领域公开数据稀缺的空白；仅展示数据有效性，未系统评测模型性能。","arXiv 2026-10-16","2026-10-15T16:00:00Z",75,[122],{"name":118,"url":116},[20,124,125,126,127],"甜菜","遥感监测","水分胁迫","多模态数据",[],[130,131],"甜菜 水分亏缺 多模态数据集","甜菜 高光谱 热红外 胁迫检测","甜菜水分亏缺多模态数据集-5657",{"doi":12,"openalex_id":12,"authors":134,"venue":12,"cited_by_count":16,"oa_url":12,"card":135,"direction":51,"ingested_from":111},[],{"tldr":136,"method":137,"finding":138,"direction":51,"opportunity":139},"发布超500GB甜菜水分亏缺多模态公开数据集，含热红外、高光谱、彩色与深度影像及两类真值。","融合热红外、高光谱、彩色、深度影像与环境元数据，附相对含水量和叶厚真值。","数据集有效可用，填补作物水分胁迫公开数据稀缺空白，但未系统评测模型性能。","可基于该多模态数据集系统评测水分胁迫检测模型，探索多源影像融合与跨模态特征学习。","2026-10-23T00:32:42.559583Z",{"id":142,"title":143,"url":144,"summary":145,"summary_zh":12,"source_name":146,"source_url":12,"published_at":147,"category":11,"cover_url":12,"hotness":13,"is_selected":14,"embedding_ready":15,"score":148,"sources":149,"tags":151,"primary_topic":156,"secondary_topics":162,"search_phrases":163,"slug":166,"view_count":16,"doi":12,"paper":167,"created_at":175},5655,"面向节能温室运行的LED照明控制策略比较研究","https:\u002F\u002Farxiv.org\u002Fabs\u002F2610.28570","预印本，未同行评审。在同一温室并行对比常开、启发式与基于马尔可夫链太阳预报的预测补光控制，2026年2—4月番茄两重复试验显示预测策略达到目标日光积分且成本较常开降低65%，单位电费干物质18.976克、约为常开两倍多；仅单温室验证。","arXiv 2026-10-20","2026-10-19T16:00:00Z",66,[150],{"name":146,"url":144},[20,152,153,154,155],"LED补光","温室环控","节能控制","番茄栽培",{"code":157,"slug":158,"name":159,"path":160,"parent_code":161},"B7","protected-horticulture","设施园艺与植物工厂","\u002Ftopic\u002Fhorticulture\u002Fprotected-horticulture","B",[],[164,165],"温室 LED补光 节能控制","马尔可夫链 太阳预报 补光","温室LED补光节能控制-5655",{"doi":12,"openalex_id":12,"authors":168,"venue":12,"cited_by_count":16,"oa_url":12,"card":169,"direction":173,"ingested_from":111},[],{"tldr":170,"method":171,"finding":172,"direction":173,"opportunity":174},"在同一温室对比三种LED补光控制策略，预测策略节能且增产。","并行对比常开、启发式与马尔可夫链太阳预报预测控制，番茄两重复试验。","预测策略达标日光积分且成本降65%，单位电费干物质约为常开两倍多。","智慧农业 \u002F 农业物联网","仅单温室验证，可扩展多温室多作物验证，并融合实时传感与电价优化控制。","2026-10-23T00:32:42.146061Z",{"id":177,"title":178,"url":179,"summary":180,"summary_zh":12,"source_name":181,"source_url":12,"published_at":182,"category":11,"cover_url":12,"hotness":13,"is_selected":14,"embedding_ready":15,"score":183,"sources":184,"tags":186,"primary_topic":12,"secondary_topics":191,"search_phrases":192,"slug":195,"view_count":16,"doi":12,"paper":196,"created_at":203},5654,"MultiAltitudeSR：基于物理地面采样距离条件化的多航高农业无人机图像超分辨率","https:\u002F\u002Farxiv.org\u002Fabs\u002F2610.24992","预印本，未同行评审。针对农业无人机航高与空间分辨率的矛盾，提出以源与目标地面采样距离为条件的超分框架，用10—70米真实航拍图联合训练，21种航高组合免微调即可支持1.18至6.03倍放大；下游仅做零样本分割初步验证。","arXiv 2026-10-18","2026-10-17T16:00:00Z",70,[185],{"name":181,"url":179},[20,187,188,189,190],"无人机","农业遥感","作物监测","图像超分辨率",[],[193,194],"农业无人机 超分辨率","地面采样距离 无人机遥感","农业无人机超分辨率-5654",{"doi":12,"openalex_id":12,"authors":197,"venue":12,"cited_by_count":16,"oa_url":12,"card":198,"direction":51,"ingested_from":111},[],{"tldr":199,"method":200,"finding":201,"direction":51,"opportunity":202},"提出按源与目标地面采样距离条件化的超分框架，实现多航高农业无人机图像免微调放大。","以源\u002F目标地面采样距离为条件，用10—70米真实航拍图联合训练超分模型。","21种航高组合免微调支持1.18至6.03倍放大，下游仅零样本分割初步验证。","下游任务仅零样本分割初验，可系统评估超分对作物表型与分割精度的影响及跨航高泛化。","2026-10-23T00:32:42.046879Z",{"id":205,"title":206,"url":207,"summary":208,"summary_zh":12,"source_name":209,"source_url":12,"published_at":210,"category":11,"cover_url":12,"hotness":13,"is_selected":14,"embedding_ready":15,"score":211,"sources":212,"tags":214,"primary_topic":218,"secondary_topics":223,"search_phrases":224,"slug":227,"view_count":16,"doi":12,"paper":228,"created_at":235},5653,"蔬菜田半自动驾驶精准喷药系统多速度田间性能评价","https:\u002F\u002Farxiv.org\u002Fabs\u002F2610.24893","预印本，未同行评审。用五个生长季10599张图、84311个标注实例训练目标检测模型，结合多目标跟踪与12喷头装置在萝卜田开展三档速度田间试验。最低速平均精度达94.6%，提速后杂草命中率由78.3%降至35.0%，未达预设商业门槛。","arXiv 2026-10-17","2026-10-16T16:00:00Z",64,[213],{"name":209,"url":207},[20,95,215,216,217],"杂草识别","田间试验","精准喷药",{"code":219,"slug":220,"name":221,"path":222,"parent_code":161},"B1","vegetables","蔬菜","\u002Ftopic\u002Fhorticulture\u002Fvegetables",[],[225,226],"萝卜田 精准喷药 田间试验","半自动驾驶 喷药 杂草检测","萝卜田精准喷药田间试验-5653",{"doi":12,"openalex_id":12,"authors":229,"venue":12,"cited_by_count":16,"oa_url":12,"card":230,"direction":173,"ingested_from":111},[],{"tldr":231,"method":232,"finding":233,"direction":173,"opportunity":234},"在萝卜田测试半自动精准喷药系统，评估三档速度下的检测与喷施性能。","五个生长季10599张图训练目标检测，结合多目标跟踪与12喷头装置开展田间试验。","最低速精度94.6%，提速后杂草命中率由78.3%降至35.0%，未达商业门槛。","高速下跟踪与喷施协同失效是瓶颈，可研究速度自适应跟踪与喷头时序补偿算法。","2026-10-23T00:32:41.788532Z",{"id":237,"title":238,"url":239,"summary":240,"summary_zh":12,"source_name":241,"source_url":12,"published_at":242,"category":11,"cover_url":12,"hotness":13,"is_selected":14,"embedding_ready":15,"score":243,"sources":244,"tags":246,"primary_topic":250,"secondary_topics":251,"search_phrases":252,"slug":255,"view_count":16,"doi":12,"paper":256,"created_at":263},5652,"Bombadil：面向植物电生理信号的基础模型","https:\u002F\u002Farxiv.org\u002Fabs\u002F2610.26597","预印本，未同行评审。以100个植物物种、十类环境（实验室、温室、商业果园、天然林）超400万小时植物电信号无标签预训练基础模型，可从同一信号追踪七项胁迫生理指标变化，并覆盖训练中无标签的物种与胁迫类型；跨场景田间验证仍有限。","arXiv 2026-10-19","2026-10-18T16:00:00Z",74,[245],{"name":241,"url":239},[20,95,247,248,249],"基础模型","植物电生理","胁迫监测",{"code":157,"slug":158,"name":159,"path":160,"parent_code":161},[],[253,254],"Bombadil 植物电信号 基础模型","植物电生理 胁迫监测","Bombadil植物电信号基础模型-5652",{"doi":12,"openalex_id":12,"authors":257,"venue":12,"cited_by_count":16,"oa_url":12,"card":258,"direction":49,"ingested_from":111},[],{"tldr":259,"method":260,"finding":261,"direction":49,"opportunity":262},"构建植物电生理信号基础模型，用超400万小时无标签数据预训练。","100个物种、十类环境超400万小时植物电信号无标签预训练。","同一信号可追踪七项胁迫生理指标，并泛化到未见物种与胁迫类型。","跨场景田间验证不足，可研究多环境迁移与胁迫早期预警的落地评估。","2026-10-23T00:32:41.455307Z"]