[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"item-143":3,"related-143":75},{"id":4,"title":5,"url":6,"summary":7,"summary_zh":8,"content":9,"source_name":10,"source_url":8,"published_at":11,"category":12,"cover_url":8,"hotness":13,"is_selected":14,"score":15,"score_detail":16,"sources":24,"tags":26,"primary_topic":31,"secondary_topics":37,"classification":44,"search_phrases":68,"slug":71,"view_count":72,"doi":73,"paper":8,"created_at":74},143,"CD-YOLO: 基于改进YOLOv11s的田间胡萝卜苗检测方法","https:\u002F\u002Fwww.smartag.net.cn\u002FCN\u002F10.12133\u002Fj.smartag.SA202511008","刘浩然、王宇、赵学观、吴华瑞、付豪、庞树杰、翟长远(北京市农林科学院智能装备技术研究中心\u002F国家农业智能装备工程技术研究中心)针对枝叶遮挡下田间胡萝卜苗精准识别难题,提出基于改进YOLOv11s的轻量化检测方法CD-YOLO:将主干网络部分标准卷积替换为深度可分离卷积降低模型复杂度,在C3k2模块引入高效多尺度注意力机制EMA构建C3k2_EMA模块增强关键特征动态感知,采用动态检测头强化多尺度特征融合与感知,显著提升检测鲁棒性。",null,"欢迎您访问《智慧农业（中英文）》官方网站！[English](https:\u002F\u002Fwww.smartag.net.cn\u002FEN\u002F2096-8094\u002Fhome.shtml)\n\n[![Image 1](https:\u002F\u002Fwww.smartag.net.cn\u002Fimages\u002F2096-8094\u002Fimages\u002Fbanner.png)](https:\u002F\u002Fwww.smartag.net.cn\u002FCN\u002F2096-8094\u002Fhome.shtml)\n\n[![Image 2](https:\u002F\u002Fwww.smartag.net.cn\u002Fimages\u002F2096-8094\u002Fimages\u002Fbanner_mobile.png)](https:\u002F\u002Fwww.smartag.net.cn\u002FCN\u002F2096-8094\u002Fhome.shtml)\n\n[![Image 3](https:\u002F\u002Fwww.smartag.net.cn\u002Fimages\u002F2096-8094\u002Fimages\u002Fyyzkj.png)](https:\u002F\u002Fpano.bookan.com.cn\u002F?id=60704&cid=352)\n\nToggle navigation[智慧农业](https:\u002F\u002Fwww.smartag.net.cn\u002FCN\u002F10.12133\u002Fj.smartag.SA202511008)\n\n*   [网站首页](https:\u002F\u002Fwww.smartag.net.cn\u002FCN\u002F2096-8094\u002Fhome.shtml)\n*   [期刊介绍](https:\u002F\u002Fwww.smartag.net.cn\u002FCN\u002Fcolumn\u002Fcolumn133.shtml)\n*   [编委会](https:\u002F\u002Fwww.smartag.net.cn\u002FCN\u002Fcolumn\u002Fcolumn153.shtml)\n*   [投稿指南](https:\u002F\u002Fwww.smartag.net.cn\u002FCN\u002Fcolumn\u002Fcolumn135.shtml)\n*   [期刊服务](https:\u002F\u002Fwww.smartag.net.cn\u002FCN\u002Fcolumn\u002Fcolumn124.shtml)\n*   [下载中心](https:\u002F\u002Fwww.smartag.net.cn\u002FCN\u002Fcolumn\u002Fcolumn126.shtml)\n*   [联系我们](https:\u002F\u002Fwww.smartag.net.cn\u002FCN\u002Fcolumn\u002Fcolumn127.shtml)\n\n[Smart Agriculture](https:\u002F\u002Fwww.smartag.net.cn\u002F) ›› [2026](https:\u002F\u002Fwww.smartag.net.cn\u002FCN\u002Farticle\u002FshowTenYearVolumnDetail.do?nian=2026), [Vol. 8](https:\u002F\u002Fwww.smartag.net.cn\u002FCN\u002Farticle\u002FshowTenYearVolumnDetail.do?nian=2026) ›› [Issue (2)](https:\u002F\u002Fwww.smartag.net.cn\u002FCN\u002Fvolumn\u002Fvolumn_838.shtml): 158-174.doi: [10.12133\u002Fj.smartag.SA202511008](https:\u002F\u002Fdoi.org\u002F10.12133\u002Fj.smartag.SA202511008)\n\n• 信息处理与决策 •[上一篇](https:\u002F\u002Fwww.smartag.net.cn\u002FCN\u002Fabstract\u002Fabstract22648.shtml)\n\n### CD-YOLO：基于改进YOLOv11s的田间胡萝卜苗检测方法\n\n刘浩然 1,2, 王宇 1, 赵学观 2,4, 吴华瑞 3, 付豪 2,4, 庞树杰 5, 翟长远 2,4([![Image 4](https:\u002F\u002Fwww.smartag.net.cn\u002Fimages\u002Femail.png)](mailto:zhaicy@nercita.org.cn))\n\n1.   1. 辽宁科技大学机械工程与自动化学院，辽宁 鞍山 114051，中国\n\n2. 北京市农林科学院智能装备技术研究中心，北京 100097，中国\n\n3. 北京市农林科学院信息技术研究中心，北京 100097，中国\n\n4. 国家农业智能装备工程技术研究中心，北京 100097，中国\n\n5. 中国农业机械化科学研究院集团有限公司，北京 100083，中国\n\n*   `收稿日期:`2025-11-09 `出版日期:`2026-03-30 \n*   `基金项目:`北京市农林科学院院创新能力建设专项(KJCX20230409); 国家自然科学基金(32201647); 改革与发展项目(GGFZ20250205) \n*   `作者简介:`刘浩然，硕士研究生，研究方向为智能化施药装备研发。E-mail：[lhr13646345271@outlook.com](mailto:lhr13646345271@outlook.com) \n*   `通信作者:`翟长远，博士，研究员，研究方向为智能化施药装备研发。E-mail：[zhaicy@nercita.org.cn](mailto:zhaicy@nercita.org.cn) \n\n### CD-YOLO: A Method for Detecting Carrot Seedlings in Field Based on Improved YOLOv11s\n\nLIU Haoran 1,2, WANG Yu 1, ZHAO Xueguan 2,4, WU Huarui 3, FU Hao 2,4, PANG Shujie 5, ZHAI Changyuan 2,4([![Image 5](https:\u002F\u002Fwww.smartag.net.cn\u002Fimages\u002Femail.png)](mailto:zhaicy@nercita.org.cn))\n\n1.   1. School of Mechanical Engineering and Automation, University of Science and Technology Liaoning, Anshan 114051, China\n\n2. Intelligent Equipment Research Center, Beijing Academy of Agriculture and Forestry Sciences, Beijing 100097, China\n\n3. Information Technology Research Center, Beijing Academy of Agriculture and Forestry Sciences, Beijing 100097, China\n\n4. National Engineering Research Center of Intelligent Equipment for Agriculture (NERCIEA), Beijing 100097, China\n\n5. Chinese Academy of Agricultural Mechanization Sciences Group Co. , Ltd. , Beijing 100083, China\n\n*   `Received:`2025-11-09 `Online:`2026-03-30 \n*   `Foundation items:`Beijing Academy of Agriculture and Forestry Sciences Innovation Capacity Building Project(KJCX20230409); National Natural Science Foundation of China(32201647); Reform and Development Project(GGFZ20250205) \n*   `About author:`LIU Haoran, E-mail: [lhr13646345271@outlook.com](mailto:lhr13646345271@outlook.com) \n*   `Corresponding author:`ZHAI Changyuan, E-mail: [zhaicy@nercita.org.cn](mailto:zhaicy@nercita.org.cn) \n\n## [在线阅读 38](https:\u002F\u002Fwww.smartag.net.cn\u002FCN\u002F10.12133\u002Fj.smartag.SA202511008#)\n\n## [知网下载](https:\u002F\u002Fwww.smartag.net.cn\u002FCN\u002F10.12133\u002Fj.smartag.SA202511008)\n\n## [本地下载 42](https:\u002F\u002Fwww.smartag.net.cn\u002FCN\u002F10.12133\u002Fj.smartag.SA202511008#1)\n\n[_![Image 6](https:\u002F\u002Fwww.smartag.net.cn\u002Fimages\u002Fknowledge\\_map1.png)_ ## 可视化 0](https:\u002F\u002Fwww.smartag.net.cn\u002FCN\u002Flexeme\u002FshowArticleByLexeme.do?articleID=22679)\n\n#### [摘要\u002FAbstract](https:\u002F\u002Fwww.smartag.net.cn\u002FCN\u002F10.12133\u002Fj.smartag.SA202511008)\n\n**摘要：**\n\n**【目的\u002F意义】** 枝叶遮挡是田间胡萝卜苗精准识别的主要挑战，对此开发一种轻量、高效的检测模型，以实现田间遮挡条件下胡萝卜苗的精准鲁棒识别。 **【方法】** 提出了一种基于改进YOLOv11s（You Only Look Once）的轻量化检测方法CD-YOLO（Carrot Detection-YOLO）。首先，为降低模型复杂度，将主干网络中的部分标准卷积替换为深度可分离卷积，以减少计算量与参数量；其次，在C3k2模块中引入高效多尺度注意力机制（Efficient Multi-Scale Attention, EMA），构建C3k2_EMA模块，增强模型对关键特征的动态感知能力，从而抑制背景与遮挡噪声；最后，采用动态检测头，强化模型对不同尺度特征的融合与感知，进一步提升检测鲁棒性。 **【结果和讨论】** 结果表明：CD-YOLO模型的计算量和模型大小相较于YOLOv11s分别降低了6.2 GFLOPs和4.8 MB，单张图像推理速度提升了4.7 ms，准确率、召回率与平均精度分别提高了3.0、1.5和2.4个百分点；在应对枝叶遮挡方面，CD-YOLO的漏识率为13.4%，较YOLOv11s下降了5.7个百分点；将模型部署于边缘算力设备，在随机选取的胡萝卜苗图像上进行测试，漏识率和误识率分别为5.1%和2.7%。 **【结论】** CD-YOLO模型在保持轻量化的同时，有效提升了田间胡萝卜苗的检测精度与速度，特别是在应对枝叶遮挡方面表现优异。该研究可为胡萝卜自动化育苗过程中的出苗率统计与精准管理提供可靠的技术支持。\n\n**关键词:**[胡萝卜苗,](https:\u002F\u002Fwww.smartag.net.cn\u002FCN\u002F10.12133\u002Fj.smartag.SA202511008#)[遮挡,](https:\u002F\u002Fwww.smartag.net.cn\u002FCN\u002F10.12133\u002Fj.smartag.SA202511008#)[目标检测,](https:\u002F\u002Fwww.smartag.net.cn\u002FCN\u002F10.12133\u002Fj.smartag.SA202511008#)[YOLOv11s,](https:\u002F\u002Fwww.smartag.net.cn\u002FCN\u002F10.12133\u002Fj.smartag.SA202511008#)[轻量化](https:\u002F\u002Fwww.smartag.net.cn\u002FCN\u002F10.12133\u002Fj.smartag.SA202511008#)\n\n**Abstract:**\n\n**[Objective]** In field environments under natural conditions, leaf occlusion and mutual plant shading pose significant challenges to the accurate identification of carrot seedlings. Furthermore, practical agricultural applications often rely on edge devices with limited computational power, necessitating a detection model that combines lightweight design, high accuracy, and robust anti-occlusion capability. The purpose of this research is to develop a robust recognition method for carrot seedlings suitable for complex field conditions, thereby enhancing the accuracy and efficiency of seedling emergence statistics in automated seedling raising processes and providing reliable technical support for precise farm management. **[Methods]** The CD-YOLO (Carrot Detection-You Only Look Once), a lightweight detection model was proposed based on an improved YOLOv11s. First, to reduce model complexity, several standard convolutions in the backbone network were replaced with depthwise separable convolutions (DWConv), thereby decreasing floating-point operations (FLOPs) and the number of parameters, establishing a lightweight foundation for edge deployment. Secondly, the efficient multi scale attention (EMA) mechanism was embedded into the critical feature extraction module C3k2, constructing a C3k2_EMA module. This module enhanced dynamic perception of local key features and reconstructed cross-scale contextual dependencies broken by occlusion through its parallel multi-branch structure, effectively suppressing background and occlusion noise. Finally, the DynamicHead detection head was introduced. Leveraging its scale-aware and spatial-aware mechanisms, it achieved a dynamic fusion of multi-level features and adaptive weight adjustment, further improving the model's decision-making robustness in complex scenes. To comprehensively evaluate model performance, a carrot seedling dataset covering various field scenarios was independently constructed. Through offline data augmentation, the original 1 274 images were expanded to 4 796, which were then split into training, validation, and test sets in an 8:1:1 ratio. Meanwhile, to systematically quantify the model's anti-occlusion performance, an occlusion severity assessment criterion based on the overlapping area of bounding boxes was proposed. Targets were categorized into three occlusion levels: mild, moderate, and severe. Based on this, a dedicated \"Occlusion Test Subset\" was separated from the main test set, providing an objective and reproducible benchmark for evaluating the model's anti-occlusion capability. **[Results and Discussions]** Experimental results on the custom dataset demonstrated that CD-YOLO comprehensively improved detection performance while maintaining its lightweight characteristics. Compared to the baseline model YOLOv11s, CD-YOLO reduced computational load by 6.2 GFLOPs (a 28.8% decrease), decreased model size by 4.8 MB (a 25.0% reduction), improved single-image inference speed by 4.7 ms, reaching 9.6 ms. Concurrently, precision, recall, and mean average precision (mAP 0.5) increased by 3.0, 1.5, and 2.4 percentage points, respectively, ultimately reaching 81.2%, 76.4%, and 84.0%. In comparisons with other lightweight backbone networks like MobileNetv3 and ShuffleNetv2, CD-YOLO consistently outperformed them on the accuracy-speed comprehensive metric, validating the effectiveness of its improvement strategies. In occlusion performance tests, the missed detection rate of CD-YOLO on the occlusion test subset was 13.4%, a 5.7 percentage points decrease compared to YOLOv11s. Its mAP 0.5 on the occlusion subset reached 80.6%, a 5.1 percentage points improvement over the baseline, whereas the improvement on the regular subset was 1.8 percentage points, proving the model's enhanced efficacy in occlusion scenarios. After deploying the model on an NVIDIA Jetson Orin NX edge device and accelerating it with TensorRT, the inference frame rate increased to 32.5 f\u002Fs. On random test images, CD-YOLO achieved missed detection and false detection rates of 5.1% and 2.7%, respectively, representing decreases of 7.7% and 2.6% compared to YOLOv11s, demonstrating promising practical application potential. Ablation studies and feature map visualizations further indicated that DWConv, C3k2_EMA, and DynamicHead formed a synergistic optimization loop: DWConv achieved computational compression, freeing up computational budget for subsequent modules; C3k2_EMA enhanced local perception and contextual reconstruction of occluded targets during the feature extraction stage; and DynamicHead performed dynamic fusion of multi-scale features at the decision-making end. Together, they ensured high-precision detection of incomplete targets under limited computational resources. **[Conclusions]** Through the synergistic design of \"lightweighting, feature enhancement, and dynamic fusion\", the CD-YOLO model achieved an excellent balance between computational efficiency, detection accuracy, and anti-occlusion capability. The model not only significantly reduced reliance on the computational power of edge devices but also effectively improved robustness and adaptability in complex field environments through structured attention and dynamic fusion mechanisms.\n\n**Key words:**[carrot seedlings,](https:\u002F\u002Fwww.smartag.net.cn\u002FCN\u002F10.12133\u002Fj.smartag.SA202511008#)[occlusion,](https:\u002F\u002Fwww.smartag.net.cn\u002FCN\u002F10.12133\u002Fj.smartag.SA202511008#)[object detection,](https:\u002F\u002Fwww.smartag.net.cn\u002FCN\u002F10.12133\u002Fj.smartag.SA202511008#)[YOLOv11s,](https:\u002F\u002Fwww.smartag.net.cn\u002FCN\u002F10.12133\u002Fj.smartag.SA202511008#)[lightweight](https:\u002F\u002Fwww.smartag.net.cn\u002FCN\u002F10.12133\u002Fj.smartag.SA202511008#)\n\n**中图分类号:**\n\n*   [TP391.4](https:\u002F\u002Fwww.smartag.net.cn\u002FCN\u002F10.12133\u002Fj.smartag.SA202511008#)\n\n*   [S24](https:\u002F\u002Fwww.smartag.net.cn\u002FCN\u002F10.12133\u002Fj.smartag.SA202511008#)\n\n#### 引用本文\n\n刘浩然, 王宇, 赵学观, 吴华瑞, 付豪, 庞树杰, 翟长远. CD-YOLO：基于改进YOLOv11s的田间胡萝卜苗检测方法[J]. 智慧农业(中英文), 2026, 8(2): 158-174.\n\nLIU Haoran, WANG Yu, ZHAO Xueguan, WU Huarui, FU Hao, PANG Shujie, ZHAI Changyuan. CD-YOLO: A Method for Detecting Carrot Seedlings in Field Based on Improved YOLOv11s[J]. Smart Agriculture, 2026, 8(2): 158-174.\n\n#### 使用本文\n\n**[](https:\u002F\u002Fwww.smartag.net.cn\u002FCN\u002F10.12133\u002Fj.smartag.SA202511008#) \u002F[推荐](javascript:void(null))**\n**导出引用管理器**[EndNote](https:\u002F\u002Fwww.smartag.net.cn\u002FCN\u002Farticle\u002FgetTxtFile.do?fileType=EndNote&id=22679)|[Ris](https:\u002F\u002Fwww.smartag.net.cn\u002FCN\u002Farticle\u002FgetTxtFile.do?fileType=Ris&id=22679)|[BibTeX](https:\u002F\u002Fwww.smartag.net.cn\u002FCN\u002Farticle\u002FgetTxtFile.do?fileType=BibTeX&id=22679)\n\n**链接本文:**[https:\u002F\u002Fwww.smartag.net.cn\u002FCN\u002F10.12133\u002Fj.smartag.SA202511008](https:\u002F\u002Fwww.smartag.net.cn\u002FCN\u002F10.12133\u002Fj.smartag.SA202511008)\n\n[https:\u002F\u002Fwww.smartag.net.cn\u002FCN\u002FY2026\u002FV8\u002FI2\u002F158](https:\u002F\u002Fwww.smartag.net.cn\u002FCN\u002FY2026\u002FV8\u002FI2\u002F158)\n\n#### [图\u002F表 20](javascript:;)\n\n[![Image 7](https:\u002F\u002Fwww.smartag.net.cn\u002Ffileup\u002F2096-8094\u002FFIGURE\u002F2026-8-2\u002FImages\u002F2096-8094-2026-8-2-158\u002Fthumbnail\u002F78D45F7B-E79C-4c5c-80B6-F28383C3B503-F001.jpg)](https:\u002F\u002Fwww.smartag.net.cn\u002FCN\u002F10.12133\u002Fj.smartag.SA202511008#figureClass13750)\n\n**图1**\n\n胡萝卜苗采集图像实例\n\n![Image 8](https:\u002F\u002Fwww.smartag.net.cn\u002Ffileup\u002F2096-8094\u002FFIGURE\u002F2026-8-2\u002FImages\u002F2096-8094-2026-8-2-158\u002F78D45F7B-E79C-4c5c-80B6-F28383C3B503-F001.jpg)\n\n 图1 \n\n[![Image 9](https:\u002F\u002Fwww.smartag.net.cn\u002Ffileup\u002F2096-8094\u002FFIGURE\u002F2026-8-2\u002FImages\u002F2096-8094-2026-8-2-158\u002Fthumbnail\u002F78D45F7B-E79C-4c5c-80B6-F28383C3B503-F002.jpg)](https:\u002F\u002Fwww.smartag.net.cn\u002FCN\u002F10.12133\u002Fj.smartag.SA202511008#figureClass13751)\n\n**图2**\n\n胡萝卜苗遮挡程度量化图\n\n![Image 10](https:\u002F\u002Fwww.smartag.net.cn\u002Ffileup\u002F2096-8094\u002FFIGURE\u002F2026-8-2\u002FImages\u002F2096-8094-2026-8-2-158\u002F78D45F7B-E79C-4c5c-80B6-F28383C3B503-F002.jpg)\n\n 图2 \n\n[![Image 11](https:\u002F\u002Fwww.smartag.net.cn\u002Fimages\u002Ftable-icon.gif)](https:\u002F\u002Fwww.smartag.net.cn\u002FCN\u002F10.12133\u002Fj.smartag.SA202511008#inline_content13752)\n\n**表1**\n\n胡萝卜苗数据集分布表\n\n| 类别 | 图像数量\u002F张 | 标签数量\u002F张 |\n| :---: | :---: | :---: |\n| 训练集 | 3 837 | 43 880 |\n| 验证集 | 480 | 6 156 |\n| 测试集 | 479 | 6 867 |\n\n 表1 \n\n[![Image 12](https:\u002F\u002Fwww.smartag.net.cn\u002Ffileup\u002F2096-8094\u002FFIGURE\u002F2026-8-2\u002FImages\u002F2096-8094-2026-8-2-158\u002Fthumbnail\u002F78D45F7B-E79C-4c5c-80B6-F28383C3B503-F003.jpg)](https:\u002F\u002Fwww.smartag.net.cn\u002FCN\u002F10.12133\u002Fj.smartag.SA202511008#figureClass13753)\n\n**图3**\n\nCD-YOLO结构图\n\n![Image 13](https:\u002F\u002Fwww.smartag.net.cn\u002Ffileup\u002F2096-8094\u002FFIGURE\u002F2026-8-2\u002FImages\u002F2096-8094-2026-8-2-158\u002F78D45F7B-E79C-4c5c-80B6-F28383C3B503-F003.jpg)\n\n 图3 \n\n[![Image 14](https:\u002F\u002Fwww.smartag.net.cn\u002Ffileup\u002F2096-8094\u002FFIGURE\u002F2026-8-2\u002FImages\u002F2096-8094-2026-8-2-158\u002Fthumbnail\u002F78D45F7B-E79C-4c5c-80B6-F28383C3B503-F004.jpg)](https:\u002F\u002Fwww.smartag.net.cn\u002FCN\u002F10.12133\u002Fj.smartag.SA202511008#figureClass13754)\n\n**图4**\n\nEMA模块\n\n![Image 15](https:\u002F\u002Fwww.smartag.net.cn\u002Ffileup\u002F2096-8094\u002FFIGURE\u002F2026-8-2\u002FImages\u002F2096-8094-2026-8-2-158\u002F78D45F7B-E79C-4c5c-80B6-F28383C3B503-F004.jpg)\n\n 图4 \n\n[![Image 16](https:\u002F\u002Fwww.smartag.net.cn\u002Ffileup\u002F2096-8094\u002FFIGURE\u002F2026-8-2\u002FImages\u002F2096-8094-2026-8-2-158\u002Fthumbnail\u002F78D45F7B-E79C-4c5c-80B6-F28383C3B503-F005.jpg)](https:\u002F\u002Fwww.smartag.net.cn\u002FCN\u002F10.12133\u002Fj.smartag.SA202511008#figureClass13755)\n\n**图5**\n\nC3k2_EMA模块\n\n![Image 17](https:\u002F\u002Fwww.smartag.net.cn\u002Ffileup\u002F2096-8094\u002FFIGURE\u002F2026-8-2\u002FImages\u002F2096-8094-2026-8-2-158\u002F78D45F7B-E79C-4c5c-80B6-F28383C3B503-F005.jpg)\n\n 图5 \n\n[![Image 18](https:\u002F\u002Fwww.smartag.net.cn\u002Fimages\u002Ftable-icon.gif)](https:\u002F\u002Fwww.smartag.net.cn\u002FCN\u002F10.12133\u002Fj.smartag.SA202511008#inline_content13756)\n\n**表2**\n\nCD-YOLO模型训练参数\n\n| 训练参数 | 值 |\n| :---: | :---: |\n| 训练周期 | 100 |\n| 初始学习率 | 0.001 |\n| 优化器 | Adam |\n| 每批次图像数量 | 24 |\n| 动量 | 0.937 |\n| 优化器权重衰减系数 | 0.000 5 |\n| 输入尺寸 | 640×640 |\n\n 表2 \n\n[![Image 19](https:\u002F\u002Fwww.smartag.net.cn\u002Fimages\u002Ftable-icon.gif)](https:\u002F\u002Fwww.smartag.net.cn\u002FCN\u002F10.12133\u002Fj.smartag.SA202511008#inline_content13757)\n\n**表3**\n\n胡萝卜苗检测研究消融试验结果\n\n| 试验编号 | 改进方式 | _P_\u002F% | _R_\u002F% | mAP 0.5\u002F% | 计算量\u002FGFLOPs | 模型大小\u002FMB | 单张图片处理时间\u002Fms |\n| :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: |\n| DWConv | C3k2_EMA | DynamicHead |\n| 1 | × | × | × | 78.2 | 74.9 | 81.6 | 21.5 | 19.2 | 14.3 |\n| 2 | √ | × | × | 79.8 | 73.6 | 81.4 | 14.8 | 13.8 | 11.5 |\n| 3 | × | √ | × | 78.7 | 77.1 | 82.9 | 21.3 | 19.2 | 11.7 |\n| 4 | × | × | √ | 79.9 | 76.7 | 83.7 | 21.5 | 19.8 | 12.3 |\n| 5 | √ | √ | × | 79.7 | 74.5 | 81.7 | 14.7 | 13.9 | 10.6 |\n| 6 | √ | × | √ | 79.0 | 74.9 | 81.8 | 15.0 | 14.4 | 9.8 |\n| 7 | × | √ | √ | 80.2 | 75.1 | 83.2 | 21.6 | 19.8 | 11.6 |\n| 8 | √ | √ | √ | 81.2 | 76.4 | 84.0 | 15.3 | 14.4 | 9.6 |\n\n 表3 \n\n[![Image 20](https:\u002F\u002Fwww.smartag.net.cn\u002Ffileup\u002F2096-8094\u002FFIGURE\u002F2026-8-2\u002FImages\u002F2096-8094-2026-8-2-158\u002Fthumbnail\u002F78D45F7B-E79C-4c5c-80B6-F28383C3B503-F006.jpg)](https:\u002F\u002Fwww.smartag.net.cn\u002FCN\u002F10.12133\u002Fj.smartag.SA202511008#figureClass13758)\n\n**图6**\n\nC3k2与C3k2_EMA模块特征图对比\n\n![Image 21](https:\u002F\u002Fwww.smartag.net.cn\u002Ffileup\u002F2096-8094\u002FFIGURE\u002F2026-8-2\u002FImages\u002F2096-8094-2026-8-2-158\u002F78D45F7B-E79C-4c5c-80B6-F28383C3B503-F006.jpg)\n\n 图6 \n\n[![Image 22](https:\u002F\u002Fwww.smartag.net.cn\u002Ffileup\u002F2096-8094\u002FFIGURE\u002F2026-8-2\u002FImages\u002F2096-8094-2026-8-2-158\u002Fthumbnail\u002F78D45F7B-E79C-4c5c-80B6-F28383C3B503-F007.jpg)](https:\u002F\u002Fwww.smartag.net.cn\u002FCN\u002F10.12133\u002Fj.smartag.SA202511008#figureClass13759)\n\n**图7**\n\n不同模型在胡萝卜苗测试集上的效果对比\n\n![Image 23](https:\u002F\u002Fwww.smartag.net.cn\u002Ffileup\u002F2096-8094\u002FFIGURE\u002F2026-8-2\u002FImages\u002F2096-8094-2026-8-2-158\u002F78D45F7B-E79C-4c5c-80B6-F28383C3B503-F007.jpg)\n\n 图7 \n\n[![Image 24](https:\u002F\u002Fwww.smartag.net.cn\u002Ffileup\u002F2096-8094\u002FFIGURE\u002F2026-8-2\u002FImages\u002F2096-8094-2026-8-2-158\u002Fthumbnail\u002F78D45F7B-E79C-4c5c-80B6-F28383C3B503-F008.jpg)](https:\u002F\u002Fwww.smartag.net.cn\u002FCN\u002F10.12133\u002Fj.smartag.SA202511008#figureClass13760)\n\n**图8**\n\nYOLOv11s与CD-YOLO在遮挡条件下胡萝卜苗检测效果对比 注：图中蓝色框为检测框，红色框为漏检目标。\n\n![Image 25](https:\u002F\u002Fwww.smartag.net.cn\u002Ffileup\u002F2096-8094\u002FFIGURE\u002F2026-8-2\u002FImages\u002F2096-8094-2026-8-2-158\u002F78D45F7B-E79C-4c5c-80B6-F28383C3B503-F008.jpg)\n\n 图8 \n\n[![Image 26](https:\u002F\u002Fwww.smartag.net.cn\u002Fimages\u002Ftable-icon.gif)](https:\u002F\u002Fwww.smartag.net.cn\u002FCN\u002F10.12133\u002Fj.smartag.SA202511008#inline_content13761)\n\n**表4**\n\n不同轻量化主干网络检测胡萝卜苗效果对比\n\n| 主干网络 | _P_\u002F% | _R_\u002F% | mAP 0.5\u002F% | 计算量\u002FGFLOPs | 参数量 | 模型大小\u002FMB | 单张图片处理时间\u002Fms |\n| :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: |\n| MobileNetv3 | 76.2 | 73.0 | 79.9 | 13.2 | 7.21×10 6 | 16.9 | 14.6 |\n| ShuffleNetv2 | 73.0 | 72.3 | 77.8 | 10.4 | 5.34×10 6 | 11.0 | 12.2 |\n| EfficientVit | 76.0 | 74.5 | 79.7 | 14.6 | 7.39×10 6 | 15.7 | 11.4 |\n| CD-YOLO | 81.2 | 76.4 | 84.0 | 15.3 | 7.06×10 6 | 14.4 | 9.6 |\n\n 表4 \n\n[![Image 27](https:\u002F\u002Fwww.smartag.net.cn\u002Fimages\u002Ftable-icon.gif)](https:\u002F\u002Fwww.smartag.net.cn\u002FCN\u002F10.12133\u002Fj.smartag.SA202511008#inline_content13762)\n\n**表5**\n\n胡萝卜苗检测研究不同模型对比效果\n\n| 模型 | _P_\u002F% | _R_\u002F% | mAP 0.5\u002F% | 计算量\u002FGFLOPs | 模型大小\u002FMB | 单张图片处理时间\u002Fms |\n| :---: | :---: | :---: | :---: | :---: | :---: | :---: |\n| SSD | 63.5 | 81.4 | 72.2 | 63.5 | 90.5 | 36.6 |\n| Faster R-CNN | 69.1 | 84.5 | 77.7 | 142.6 | 108.1 | 113.3 |\n| YOLOv5s | 80.2 | 73.9 | 80.5 | 16.2 | 15.1 | 14.2 |\n| YOLOv8s | 79.1 | 74.8 | 81.6 | 49.1 | 22.5 | 9.8 |\n| YOLOv11s | 78.2 | 74.9 | 81.6 | 21.5 | 19.2 | 14.3 |\n| YOLOv8s-P2 | 79.3 | 74.4 | 81.5 | 55.3 | 24.8 | 10.7 |\n| DWG-YOLOv8 | 78.2 | 73.2 | 82.0 | 17.4 | 14.7 | 10.5 |\n| HAD-YOLO | 77.9 | 74.3 | 81.1 | 35.4 | 12.5 | 19.3 |\n| CD-YOLO | 81.2 | 76.4 | 84.0 | 15.3 | 14.4 | 9.6 |\n\n 表5 \n\n[![Image 28](https:\u002F\u002Fwww.smartag.net.cn\u002Fimages\u002Ftable-icon.gif)](https:\u002F\u002Fwww.smartag.net.cn\u002FCN\u002F10.12133\u002Fj.smartag.SA202511008#inline_content13763)\n\n**表6**\n\nYOLOv11s与CD-YOLO在不同测试子集上效果对比\n\n| 测试子集 | 模型 | 图像数量 | _P_\u002F% | _R_\u002F% | mAP 0.5\u002F% |\n| :---: | :---: | :---: | :---: | :---: | :---: |\n| 常规子集 | YOLOv11s | 357 | 78.7 | 76.3 | 77.6 |\n| CD-YOLO | 357 | 79.8 | 78.1 | 79.4 |\n| 遮挡子集 | YOLOv11s | 122 | 75.9 | 71.1 | 75.5 |\n| CD-YOLO | 122 | 79.6 | 75.1 | 80.6 |\n| 整体测试集 | YOLOv11s | 479 | 78.9 | 74.7 | 76.9 |\n| CD-YOLO | 479 | 79.7 | 76.2 | 79.9 |\n\n 表6 \n\n[![Image 29](https:\u002F\u002Fwww.smartag.net.cn\u002Fimages\u002Ftable-icon.gif)](https:\u002F\u002Fwww.smartag.net.cn\u002FCN\u002F10.12133\u002Fj.smartag.SA202511008#inline_content13764)\n\n**表7**\n\nYOLOv11s与CD-YOLO遮挡条件下胡萝卜苗检测效果对比\n\n| 模型 | 目标总数\u002F个 | 正确识别数\u002F个 | 漏识数\u002F个 | 识别率\u002F% | 漏识率\u002F% |\n| :---: | :---: | :---: | :---: | :---: | :---: |\n| YOLOv11s | 1 652 | 1 337 | 315 | 80.9 | 19.1 |\n| CD-YOLO | 1 652 | 1 431 | 221 | 86.6 | 13.4 |\n\n 表7 \n\n[![Image 30](https:\u002F\u002Fwww.smartag.net.cn\u002Ffileup\u002F2096-8094\u002FFIGURE\u002F2026-8-2\u002FImages\u002F2096-8094-2026-8-2-158\u002Fthumbnail\u002F78D45F7B-E79C-4c5c-80B6-F28383C3B503-F009.jpg)](https:\u002F\u002Fwww.smartag.net.cn\u002FCN\u002F10.12133\u002Fj.smartag.SA202511008#figureClass13765)\n\n**图9**\n\nYOLOv11s与CD-YOLO模型热力图可视化对比\n\n![Image 31](https:\u002F\u002Fwww.smartag.net.cn\u002Ffileup\u002F2096-8094\u002FFIGURE\u002F2026-8-2\u002FImages\u002F2096-8094-2026-8-2-158\u002F78D45F7B-E79C-4c5c-80B6-F28383C3B503-F009.jpg)\n\n 图9 \n\n[![Image 32](https:\u002F\u002Fwww.smartag.net.cn\u002Fimages\u002Ftable-icon.gif)](https:\u002F\u002Fwww.smartag.net.cn\u002FCN\u002F10.12133\u002Fj.smartag.SA202511008#inline_content13766)\n\n**表8**\n\nYOLOv11s与CD-YOLO模型在不同设备上检测帧率对比 ( 帧\u002Fs)\n\n| 模型 | 工作站 | 边缘算力设备 | TensorRT |\n| :---: | :---: | :---: | :---: |\n| YOLOv11s | 89.6 | 10.5 | 29.7 |\n| CD-YOLO | 104.2 | 12.9 | 32.5 |\n\n 表8 \n\n[![Image 33](https:\u002F\u002Fwww.smartag.net.cn\u002Fimages\u002Ftable-icon.gif)](https:\u002F\u002Fwww.smartag.net.cn\u002FCN\u002F10.12133\u002Fj.smartag.SA202511008#inline_content13767)\n\n**表9**\n\nYOLOv1","智慧农业(中英文) 2026 8(2): 158-174","2026-07-27T00:00:00Z","论文",10,false,69,{"impact":17,"substance":18,"depth":19,"authority":20,"freshness":21,"relevant":22,"comment":23},15,20,18,14,2,1,"核心期刊论文，提出改进YOLOv11s的胡萝卜苗检测方法，具技术新颖性，但时效性较低。",[25],{"name":10,"url":6},[27,28,29,30],"智慧农业","YOLOv11","胡萝卜苗","目标检测",{"code":32,"slug":33,"name":34,"path":35,"parent_code":36},"B1","vegetables","蔬菜","\u002Ftopic\u002Fhorticulture\u002Fvegetables","B",[38],{"code":39,"slug":40,"name":41,"path":42,"parent_code":43},"I2","agricultural-ai","农业AI与大模型","\u002Ftopic\u002Fdigital-agriculture\u002Fagricultural-ai","I",{"status":45,"locked":14,"version":46,"method":47,"evidence":48},"classified","2026-10-22.2","rules",{"candidates":49,"agricultural_evidence":52},[50,57,63],{"code":39,"score":51,"supported":52,"title":53,"abstract":55,"tags":56},12,true,[54],"YOLO",[54],[30],{"code":32,"score":58,"supported":52,"title":59,"abstract":61,"tags":62},11,[60],"萝卜",[60],[],{"code":64,"score":22,"supported":14,"title":65,"abstract":66,"tags":67},"I1",[],[],[27],[69,70],"智慧农业 目标检测 胡萝卜苗 YOLOv11","智慧农业 目标检测","智慧农业目标检测胡萝卜苗YOLOv11-143",9,"10.12133\u002Fj.smartag.sa202511008","2026-08-01T06:01:55.365250Z",{"total":76,"page":22,"page_size":76,"items":77},6,[78,145,210,314,374,432],{"id":79,"title":80,"url":81,"summary":82,"summary_zh":83,"content":8,"source_name":84,"source_url":81,"published_at":85,"category":12,"cover_url":8,"hotness":13,"is_selected":14,"score":86,"score_detail":87,"sources":91,"tags":93,"primary_topic":98,"secondary_topics":99,"classification":101,"search_phrases":121,"slug":124,"view_count":125,"doi":126,"paper":127,"created_at":144},5387,"LIGHTWEIGHT FEATURE FUSED ATTENTION DRIVEN MULTI LEVEL HYBRID MOBILENET V2 FRAMEWORK FOR EARLY TOMATO LEAF DISEASE DETECTION","https:\u002F\u002Fdoi.org\u002F10.62441\u002Factainnovations.vi.1918","It is crucial to find and treat tomato crop diseases early and accurately to improve crop efficiency and make sure there is no scarcity of crop. However, computerized analysis is provocative due to lighting differences, compound backgrounds, and graphic likenesses among infection regions. Research work proposed a lightweight Attention Driven Multi Level Hybrid MobileNet-V2 algorithm that fuses the handcrafted texture structures with Deep Learning (DL) features, then image preprocessing is incorporated using Gaussian, Bilateral, and Guided filters to minimize noise by preserving the disease lesions. LAB color model along with Contrast Limited Adaptive Histogram Equalization (CLAHE) are used to improve the contrast. Otsu thresholding is involved to effectively separate the leaf segment from diseased parts. Local Binary Pattern (LBP) and Gray Level Co-occurrence Matrix (GLCM) descriptors are used to obtain fine-grained features, while high level features are gained by pretrained MobileNet-V2 model. Features fusion is fed to the fully connected layers of deep learning model for training and validation purpose. Results have revealed that the suggested model attains better classification performance as compared to existing models, making it an effective explanation for real time tomato leaf disease detection for smart agriculture applications.","早期准确发现并治疗番茄作物病害，对于提高作物效率、确保作物不短缺至关重要。然而，由于光照差异、复杂背景以及感染区域之间的图形相似性，计算机化分析具有挑战性。研究提出了一种轻量级注意力驱动多层混合MobileNet-V2算法，该算法将手工纹理结构与深度学习（DL）特征相融合，然后结合高斯、双边和引导滤波器进行图像预处理，以在保留病害病变的同时最小化噪声。采用LAB颜色模型和对比度受限自适应直方图均衡化（CLAHE）来提高对比度。引入Otsu阈值分割以有效分离叶片片段与病害部分。使用局部二值模式（LBP）和灰度共生矩阵（GLCM）描述符获取细粒度特征，同时通过预训练的MobileNet-V2模型获取高层特征。特征融合后输入深度学习模型的全连接层进行训练和验证。结果表明，与现有模型相比，所提出的模型取得了更好的分类性能，使其成为智慧农业应用中实时番茄叶片病害检测的有效解决方案。","Acta Innovations","2026-10-18T00:00:00Z",64,{"impact":51,"substance":19,"depth":88,"authority":13,"freshness":89,"relevant":22,"comment":90},16,8,"提出融合手工纹理特征与MobileNet-V2的轻量番茄叶病检测方法，有一定方法新颖性，但属单篇论文、应用范围有限，未达每日精选门槛。",[92],{"name":84,"url":81},[27,94,95,96,97],"农业人工智能","深度学习","番茄病害","图像识别",{"code":32,"slug":33,"name":34,"path":35,"parent_code":36},[100],{"code":39,"slug":40,"name":41,"path":42,"parent_code":43},{"status":45,"locked":14,"version":46,"method":47,"evidence":102},{"candidates":103,"agricultural_evidence":52},[104,110,115],{"code":32,"score":17,"supported":52,"title":105,"abstract":107,"tags":109},[106],"tomato",[108,106],"番茄",[96],{"code":39,"score":72,"supported":52,"title":111,"abstract":112,"tags":114},[],[95,113],"deep learning",[94,97,95],{"code":64,"score":116,"supported":52,"title":117,"abstract":118,"tags":120},7,[],[27,119],"smart agriculture",[27],[122,123],"MobileNet-V2 番茄叶片病害","番茄叶病 深度学习 检测","MobileNet-V2番茄叶片病害-5387",0,"10.62441\u002Factainnovations.vi.1918",{"doi":126,"openalex_id":128,"authors":129,"venue":84,"cited_by_count":125,"oa_url":135,"card":136,"direction":142,"ingested_from":143},"W7222977034",[130,132],{"name":131,"orcid":8},"Mrs. Vijetalaxmi A Hegde",{"name":133,"orcid":134},"Abhay Ramakant Kasetwar","https:\u002F\u002Forcid.org\u002F0000-0002-5800-8764","https:\u002F\u002Factainnovations.com\u002Findex.php\u002Fpub\u002Farticle\u002Fdownload\u002F1918\u002F981",{"tldr":137,"method":138,"finding":139,"direction":140,"opportunity":141},"提出轻量注意力多级混合MobileNet-V2框架，融合手工纹理与深度学习特征实现番茄叶片病害早期检","高斯\u002F双边\u002F引导滤波去噪，LAB+CLAHE增强，Otsu分割，LBP与GLCM","所提模型分类性能优于现有模型，可有效用于智慧农业实时番茄叶片病害检测。","农业人工智能与决策模型","可探索多作物跨域泛化、田间复杂光照下轻量模型部署与可解释性，及病害早期分级检测。","智慧农业 \u002F 农业物联网","openalex","2026-10-20T23:30:10.424511Z",{"id":146,"title":147,"url":148,"summary":149,"summary_zh":150,"content":8,"source_name":151,"source_url":148,"published_at":152,"category":12,"cover_url":8,"hotness":13,"is_selected":14,"score":153,"score_detail":154,"sources":157,"tags":159,"primary_topic":161,"secondary_topics":162,"classification":164,"search_phrases":181,"slug":184,"view_count":22,"doi":185,"paper":186,"created_at":209},4494,"Tomato ripeness classification with weight-based and texture-adaptive pooling in CNNs","https:\u002F\u002Fdoi.org\u002F10.64960\u002Feasr.2026.266223","This research introduces a weight-based mixed pooling and adaptive fuzzy pooling mechanism based on feature map texture analysis for attention mechanism Squeeze-and-Excitation Networks (SE) and Convolutional Neural Networks (CNNs) in tomato ripeness classification. Traditional CNNs and SE networks rely on deterministic pooling strategies that can be suboptimal for capturing diverse visual features such as texture and color gradients in fruit. A novel fuzzy pooling system is proposed that dynamically adjusts the fusion ratio of min and max pooling based on real-time texture analysis of feature maps via a Laplacian filter. This mechanism was integrated into Darknet53, ResNet34, and ResNet101 architectures, both directly and within the SE attention block. The method was evaluated on a five-class tomato ripeness dataset using 10-fold cross-validation against traditional and fixed-weight pooling methods. The fuzzy and mixed pooling approach demonstrated superior performance on certain models which include ESEnet: Darknet53 + SE (Fuzzy pool) achieved an average F1 score across folds of 0.8429, outperforming Darknet53 + baseline SE (Average pool) at 0.8350; Extended Net: ResNet34 + Fuzzy pool achieved an average F1 score across folds of 0.8305, outperforming baseline ResNet34 at 0.8130 and ResNet101 + Fuzzy pool achieved an average F1 score across folds of 0.8284, outperforming baseline ResNet101 at 0.8267. This improvement was realized with minimal increase in computational overhead. Specifically, measurements of model parameters (M) and FLOPs (GMACs) showed only a slight increase across all architectures after integrating SEnet into the system, and zero increase after integrating the Laplacian filter and fuzzy logic. The study concludes that logic-based adaptive pooling is a robust strategy that improves feature representation in CNNs, achieving a balance between classification accuracy and computational cost. These findings offer practical value for developing more reliable automated harvesting systems and provide a generalizable enhancement for a wide range of computer vision classification tasks.","本研究提出了一种基于特征图纹理分析的权重混合池化与自适应模糊池化机制，用于番茄成熟度分类中的注意力机制挤压激励网络（Squeeze-and-Excitation Networks, SE）和卷积神经网络（Convolutional Neural Networks, CNNs）。传统的CNN和SE网络依赖确定性池化策略，在捕捉果实纹理和颜色梯度等多样化视觉特征时可能并非最优。本文提出了一种新颖的模糊池化系统，通过拉普拉斯滤波器对特征图进行实时纹理分析，动态调整最小池化和最大池化的融合比例。该机制被集成到Darknet53、ResNet34和ResNet101架构中，既可直接集成，也可嵌入SE注意力模块内。该方法在五类番茄成熟度数据集上采用10折交叉验证，与传统池化及固定权重池化方法进行了对比评估。模糊池化和混合池化方法在部分模型上表现出优越性能，包括：ESEnet：Darknet53 + SE（模糊池化）在各折上的平均F1得分为0.8429，优于Darknet53 + 基线SE（平均池化）的0.8350；扩展网络：ResNet34 + 模糊池化在各折上的平均F1得分为0.8305，优于基线ResNet34的0.8130；ResNet101 + 模糊池化在各折上的平均F1得分为0.8284，优于基线ResNet101的0.8267。这一改进仅带来了极小的计算开销增加。具体而言，模型参数量（M）和浮点运算次数（GMACs）的测量结果表明，将SEnet集成到系统后所有架构仅有轻微增加，而集成拉普拉斯滤波器和模糊逻辑后则无任何增加。研究结论认为，基于逻辑的自适应池化是一种鲁棒策略，能够改善CNN中的特征表示，在分类精度和计算成本之间实现平衡。这些发现为开发更可靠的自动化收获系统提供了实用价值，并为广泛的计算机视觉分类任务提供了一种可推广的增强方法。","Engineering and Applied Science Research","2026-10-07T00:00:00Z",66,{"impact":89,"substance":18,"depth":155,"authority":51,"freshness":72,"relevant":22,"comment":156},17,"提出基于纹理自适应模糊池化的CNN改进方法，在番茄成熟度分类上取得小幅但稳定的F1提升且算力开销几乎不增，对自动采摘与农业视觉识别有参考价值，但属细分方法类论文，影响面有限。",[158],{"name":151,"url":148},[27,94,108,97,160],"自动采摘",{"code":32,"slug":33,"name":34,"path":35,"parent_code":36},[163],{"code":39,"slug":40,"name":41,"path":42,"parent_code":43},{"status":45,"locked":14,"version":46,"method":47,"evidence":165},{"candidates":166,"agricultural_evidence":52},[167,171,177],{"code":32,"score":17,"supported":52,"title":168,"abstract":169,"tags":170},[106],[108,106],[108],{"code":39,"score":89,"supported":52,"title":172,"abstract":173,"tags":176},[],[174,175],"计算机视觉","computer vision",[94,97],{"code":64,"score":22,"supported":14,"title":178,"abstract":179,"tags":180},[],[],[27],[182,183],"番茄 成熟度 分类 卷积神经网络","Darknet53 ResNet 番茄 成熟度","番茄成熟度分类卷积神经网络-4494","10.64960\u002Feasr.2026.266223",{"doi":185,"openalex_id":187,"authors":188,"venue":151,"cited_by_count":125,"oa_url":203,"card":204,"direction":142,"ingested_from":143},"W7220840158",[189,191,193,195,198,200],{"name":190,"orcid":8},"Rut Yatigul",{"name":192,"orcid":8},"Wisanu Jitviriya",{"name":194,"orcid":8},"Eiji Hayashi",{"name":196,"orcid":197},"Aran Blattler","https:\u002F\u002Forcid.org\u002F0000-0002-8684-6239",{"name":199,"orcid":8},"Watcharin Tangsuksant",{"name":201,"orcid":202},"Teppakorn Sittiwanchai","https:\u002F\u002Forcid.org\u002F0000-0001-9653-8750","https:\u002F\u002Fph01.tci-thaijo.org\u002Findex.php\u002Feasr\u002Farticle\u002Fdownload\u002F266223\u002F176569",{"tldr":205,"method":206,"finding":207,"direction":140,"opportunity":208},"提出基于纹理自适应模糊池化机制，提升CNN番茄成熟度分类性能。","用Laplacian滤波分析特征图纹理，动态调整最大最小池化融合比，集成到Dar","模糊池化在多个模型上F1提升，且计算开销几乎不增加。","可探索自适应池化在其它果蔬成熟度或田间复杂光照下的泛化能力。","2026-10-09T23:30:09.993787Z",{"id":211,"title":212,"url":213,"summary":214,"summary_zh":215,"content":8,"source_name":216,"source_url":213,"published_at":217,"category":12,"cover_url":8,"hotness":13,"is_selected":14,"score":218,"score_detail":219,"sources":221,"tags":223,"primary_topic":227,"secondary_topics":228,"classification":235,"search_phrases":292,"slug":295,"view_count":22,"doi":296,"paper":297,"created_at":313},4038,"Use of artificial intelligence (AI) in vegetable crops: A review","https:\u002F\u002Fdoi.org\u002F10.33545\u002F2664844x.2026.v8.i10a.1995","Artificial Intelligence (AI) is emerging technology in vegetable production, enabling data-driven decision making and precision agriculture practices. AI refers to the intelligence exhibited by machines that can perceive their environment, process large datasets, and make decisions to optimize agricultural productivity. With the integration of technologies such as machine learning, sensors, drones, satellites, Geographic Information Systems (GIS), and Global Positioning Systems (GPS), AI enables efficient monitoring and management of crop production systems In vegetable cultivation, AI technologies assist farmers in several critical operations including soil monitoring, climate assessment, pest and disease diagnosis, irrigation scheduling, and yield prediction. Sensors such as wind vane, cup count anemometer, rain gauge, lux meter, soil moisture sensor, and crop canopy sensors collect real-time environmental and crop data. These data was processed through data loggers and machine learning algorithms to provide accurate recommendations for crop management. Drone-based imaging systems and multispectral cameras further support precision farming by identifying weeds, nutrient deficiencies and disease symptoms at early stages. The use of AI in vegetables also highlights image-based predictive agriculture systems where captured images are processed through stages such as. image cleaning, feature extraction, and model training to detect pests and diseases. Several case studies demonstrate the use of AI in vegetable crops such as tomato, capsicum, and chilli for monitoring canopy growth, irrigation management. Additionally, robotic technologies in harvesting address labor shortages in agriculture. Crop yield prediction of AI in vegetables increases by 40% sensors and algorithms reduces labor cost by 30% moisture sensor increases water use efficiency by 40% spraying done reduces use of pesticide by 60% robotic arms increases harvesting accuracy by 40% artificial neural network data integration saves 25% cost in disease monitoring.","人工智能（AI）正在成为蔬菜生产领域的新兴技术，推动数据驱动决策与精准农业实践。AI是指机器所表现出的智能，能够感知环境、处理大规模数据集并做出决策以优化农业生产力。通过整合机器学习、传感器、无人机、卫星、地理信息系统（GIS）和全球定位系统（GPS）等技术，AI实现了对作物生产系统的高效监测与管理。在蔬菜种植中，AI技术可协助农民完成多项关键作业，包括土壤监测、气候评估、病虫害诊断、灌溉调度和产量预测。风向标、杯式风速计、雨量计、照度计、土壤水分传感器和作物冠层传感器等设备可采集实时的环境与作物数据。这些数据通过数据记录器和机器学习算法进行处理，为作物管理提供精准建议。基于无人机的成像系统与多光谱相机通过早期识别杂草、养分缺乏和病害症状，进一步支持精准农业。AI在蔬菜中的应用还突出了基于图像的预测性农业系统，该系统将采集的图像经过图像清洗、特征提取和模型训练等阶段处理，以检测病虫害。多个案例研究展示了AI在番茄、辣椒和尖椒等蔬菜作物中用于监测冠层生长和灌溉管理的应用。此外，收获机器人技术可缓解农业劳动力短缺问题。AI在蔬菜中的作物产量预测通过传感器和算法提高了40%，劳动力成本降低了30%，土壤水分传感器使水分利用效率提高了40%，喷雾作业使农药使用量减少了60%，机械臂使收获准确率提高了40%，人工神经网络数据集成使病害监测成本节省了25%。","International Journal of Agriculture and Food Science","2026-10-01T00:00:00Z",75,{"impact":19,"substance":18,"depth":88,"authority":51,"freshness":72,"relevant":22,"comment":220},"系统综述AI在蔬菜生产中的应用，含产量、成本、水肥药效等量化数据，对智慧农业主题有聚合价值。",[222],{"name":216,"url":213},[27,94,224,225,226],"病虫害识别","精准农业","蔬菜种植",{"code":32,"slug":33,"name":34,"path":35,"parent_code":36},[229,230],{"code":39,"slug":40,"name":41,"path":42,"parent_code":43},{"code":231,"slug":232,"name":233,"path":234,"parent_code":43},"I4","gis","GIS与空间分析","\u002Ftopic\u002Fdigital-agriculture\u002Fgis",{"status":45,"locked":14,"version":46,"method":47,"evidence":236},{"candidates":237,"agricultural_evidence":52},[238,246,252,259,264,271,278,285],{"code":39,"score":239,"supported":52,"title":240,"abstract":242,"tags":245},19,[241],"artificial intelligence",[243,244,241],"机器学习","machine learning",[94,224],{"code":32,"score":19,"supported":52,"title":247,"abstract":249,"tags":251},[248],"vegetable",[34,108,250,248,106],"辣椒",[226],{"code":231,"score":72,"supported":52,"title":253,"abstract":254,"tags":258},[],[255,256,257],"地理信息","GIS","geographic information",[],{"code":64,"score":89,"supported":52,"title":260,"abstract":261,"tags":263},[],[225,262],"precision agriculture",[27,225],{"code":265,"score":76,"supported":52,"title":266,"abstract":267,"tags":270},"F4",[],[268,269],"农药","pesticide",[],{"code":272,"score":76,"supported":52,"title":273,"abstract":274,"tags":277},"H2",[],[275,276],"土壤","soil",[],{"code":279,"score":76,"supported":52,"title":280,"abstract":281,"tags":284},"H3",[],[282,283],"灌溉","irrigation",[],{"code":286,"score":287,"supported":52,"title":288,"abstract":289,"tags":291},"I3",3,[],[290],"multispectral",[],[293,294],"AI 蔬菜 精准农业","蔬菜 病虫害 图像识别","AI蔬菜精准农业-4038","10.33545\u002F2664844x.2026.v8.i10a.1995",{"doi":296,"openalex_id":298,"authors":299,"venue":216,"cited_by_count":125,"oa_url":8,"card":308,"direction":140,"ingested_from":143},"W7215128283",[300,302,304,306],{"name":301,"orcid":8},"Hrishikesh Nirmala Balkrishna Thube",{"name":303,"orcid":8},"Dr. SK Kshirsagar",{"name":305,"orcid":8},"MG Agale",{"name":307,"orcid":8},"GS Shinde",{"tldr":309,"method":310,"finding":311,"direction":142,"opportunity":312},"综述人工智能在蔬菜作物生产中的应用，涵盖监测、诊断、灌溉、产量预测与采收机器人。","综述机器学习、传感器、无人机、卫星、GIS\u002FGPS及图像处理在蔬菜生产中的应用案","AI使蔬菜产量预测提高40%、节水40%、减农药60%、采收准确率提高40%。","可针对小农户蔬菜场景开发低成本多源数据融合模型，并验证AI长期经济与生态效益。","2026-10-02T23:30:40.356457Z",{"id":315,"title":316,"url":317,"summary":318,"summary_zh":319,"content":8,"source_name":320,"source_url":317,"published_at":321,"category":12,"cover_url":8,"hotness":13,"is_selected":14,"score":15,"score_detail":322,"sources":324,"tags":326,"primary_topic":330,"secondary_topics":331,"classification":333,"search_phrases":354,"slug":357,"view_count":287,"doi":358,"paper":359,"created_at":373},3284,"Computer vision-based size estimation module for selective harvesting of lettuce","https:\u002F\u002Fdoi.org\u002F10.14719\u002Fpst.14546","Selective harvesting of leafy vegetables remains a major challenge in agricultural automation, particularly for lettuce that meet predefined size or maturity criteria. Vision-based perception systems play a critical role in enabling selective harvesting by identifying plants that meet predefined size or maturity criteria. This study presents the development and validation of computer vision-based size estimation software designed to support selective harvesting decisions in leafy vegetable production. A C #-based image processing application was developed to perform red, green and blue (RGB) colour segmentation, contour detection and pixel-to-metric size estimation using images acquired from a standard Universal Serial Bus (USB) webcam. The system converts pixel measurements into real-world dimensions through calibrated scaling factors that account for variations in camera-to-plant distance. Field experiments were conducted using three commercial lettuce cultivars Romaine, Iceberg and Curly. Vision-based canopy width measurements showed strong agreement with manual measurements (N = 60) and exhibited very strong correlations with plant weight (Romaine type: r = 0.9803; Iceberg type: r = 0.9935; Curly type: r = 0.9707; p \u003C 0.001). Linear regression models achieved high predictive performance (R² = 0.942–0.987), while error metrics indicated low prediction errors (Mean absolute error (MAE): 5.62–8.64 g; root mean square error (RMSE): 7.30–11.06 g). Bland-Altman analysis further confirmed strong agreement between vision-based estimates and manual measurements. This study primarily validates a lightweight vision-based measurement algorithm intended for future integration into robotic harvesting systems. The results demonstrate that simple RGB-based image processing can provide reliable size estimation for selective harvesting decisions in leafy vegetable production systems.","叶菜类蔬菜的选择性收获仍是农业自动化领域的一项重大挑战，尤其是对符合预定大小或成熟度标准的生菜而言。基于视觉的感知系统通过识别符合预定大小或成熟度标准的植株，在实现选择性收获中发挥着关键作用。本研究介绍了一款基于计算机视觉的大小估算软件的开发与验证，该软件旨在为叶菜类蔬菜生产中的选择性收获决策提供支持。研究开发了一款基于C#的图像处理应用程序，利用标准通用串行总线（USB）网络摄像头采集的图像，执行红绿蓝（RGB）颜色分割、轮廓检测以及像素到实际尺寸的估算。该系统通过经过校准的缩放因子将像素测量值转换为实际尺寸，并考虑了摄像头与植株之间距离的变化。田间试验采用三个商业化生菜品种——罗马生菜、结球生菜和皱叶生菜。基于视觉的冠幅测量结果与人工测量结果高度一致（N = 60），并与植株重量呈现极强的相关性（罗马型：r = 0.9803；结球型：r = 0.9935；皱叶型：r = 0.9707；p \u003C 0.001）。线性回归模型取得了较高的预测性能（R² = 0.942–0.987），误差指标显示预测误差较低（平均绝对误差（MAE）：5.62–8.64 g；均方根误差（RMSE）：7.30–11.06 g）。Bland-Altman分析进一步证实了基于视觉的估算值与人工测量值之间的高度一致性。本研究主要验证了一种轻量级视觉测量算法，旨在未来集成到机器人收获系统中。结果表明，简单的基于RGB的图像处理能够为叶菜类蔬菜生产系统中的选择性收获决策提供可靠的大小估算。","Plant Science Today","2026-09-22T00:00:00Z",{"impact":51,"substance":18,"depth":88,"authority":51,"freshness":72,"relevant":22,"comment":323},"该研究开发了基于RGB图像的轻量级生菜尺寸估计模块，与人工测量高度一致，为机器人选择性采收提供了可靠感知方案，具有较高的技术参考价值。",[325],{"name":320,"url":317},[27,94,327,328,329],"生菜","机器视觉","选择性采收",{"code":39,"slug":40,"name":41,"path":42,"parent_code":43},[332],{"code":32,"slug":33,"name":34,"path":35,"parent_code":36},{"status":45,"locked":14,"version":46,"method":47,"evidence":334},{"candidates":335,"agricultural_evidence":52},[336,340,344,350],{"code":39,"score":88,"supported":52,"title":337,"abstract":338,"tags":339},[175],[174,175],[94,328],{"code":32,"score":13,"supported":52,"title":341,"abstract":342,"tags":343},[],[34,327,248],[327],{"code":345,"score":287,"supported":52,"title":346,"abstract":347,"tags":349},"F6",[],[348],"robotic harvesting",[],{"code":64,"score":22,"supported":14,"title":351,"abstract":352,"tags":353},[],[],[27],[355,356],"生菜 选择性采收 机器视觉","计算机视觉 生菜 尺寸估计","生菜选择性采收机器视觉-3284","10.14719\u002Fpst.14546",{"doi":358,"openalex_id":360,"authors":361,"venue":320,"cited_by_count":125,"oa_url":317,"card":368,"direction":140,"ingested_from":143},"W7214000589",[362,365],{"name":363,"orcid":364},"Erhan Kahya","https:\u002F\u002Forcid.org\u002F0000-0001-7768-9190",{"name":366,"orcid":367},"Fatma Funda Özdüven","https:\u002F\u002Forcid.org\u002F0000-0003-4286-8943",{"tldr":369,"method":370,"finding":371,"direction":140,"opportunity":372},"开发基于RGB图像处理的生菜尺寸估计软件，用于选择性采收决策。","C#图像处理，USB摄像头采集，RGB分割与轮廓检测，标定像素-实际尺寸转换。","视觉冠幅与重量强相关（r=0.97-0.99），回归模型R²达0.94-0.99，误差低。","可集成到机器人采收系统，并探索多品种、多光照条件下的鲁棒性及实时边缘部署。","2026-09-23T23:30:38.203393Z",{"id":375,"title":376,"url":377,"summary":378,"summary_zh":8,"content":8,"source_name":379,"source_url":8,"published_at":321,"category":12,"cover_url":8,"hotness":13,"is_selected":14,"score":380,"score_detail":381,"sources":384,"tags":386,"primary_topic":389,"secondary_topics":390,"classification":397,"search_phrases":418,"slug":421,"view_count":422,"doi":8,"paper":423,"created_at":431},3249,"设施番茄采摘机器人识别定位与采摘方法研究——江苏大学雷志龙等 基于改进YOLO v5-HSV融合算法识别准确率95.01%","http:\u002F\u002Fwww.qikanzj.com\u002Fhek\u002Fnyjxxb\u002Fmulu\u002F455758.html","智慧农业(中英文)期刊发表江苏大学雷志龙、刘畅、王权团队研究：针对设施单果番茄采摘需求，设计一款设施番茄智能采摘平台，主要由升降机构、采摘机构、识别与定位系统等部分组成。平台整体结构由低压一体化伺服滚珠丝杠副升降机构、六轴协作机械臂和力控末端执行器组成。基于改进YOLO v5-HSV融合算法来识别检测，通过对H分量进行图像阈值分割，提高对成熟目标果实识别的准确率，有效排除未成熟番茄和枝叶背景的干扰；通过眼在手外的标定方法，使用ZED双目相机进行定位。搭建的设施番茄采摘机样机平台在现场采摘试验中识别准确率达到95.01%，采摘成功率为87.96%，单果平均采摘时间为14.56s。","智慧农业(中英文)·江苏大学",82,{"impact":19,"substance":382,"depth":19,"authority":20,"freshness":13,"relevant":22,"comment":383},22,"核心期刊论文，方法有改进、数据完整，识别准确率与采摘成功率等指标明确，对设施农业智能装备研发有参考价值。",[385],{"name":379,"url":377},[27,94,387,328,388],"采摘机器人","设施番茄",{"code":32,"slug":33,"name":34,"path":35,"parent_code":36},[391,392],{"code":39,"slug":40,"name":41,"path":42,"parent_code":43},{"code":345,"slug":393,"name":394,"path":395,"parent_code":396},"agricultural-robots","机器人与无人农场","\u002Ftopic\u002Finputs-equipment\u002Fagricultural-robots","F",{"status":45,"locked":14,"version":46,"method":47,"evidence":398},{"candidates":399,"agricultural_evidence":52},[400,405,409,413],{"code":39,"score":401,"supported":52,"title":402,"abstract":403,"tags":404},13,[54],[54],[94,328],{"code":32,"score":51,"supported":52,"title":406,"abstract":407,"tags":408},[108],[108],[388],{"code":345,"score":72,"supported":52,"title":410,"abstract":411,"tags":412},[387],[],[387],{"code":64,"score":414,"supported":52,"title":415,"abstract":416,"tags":417},4,[],[27],[27],[419,420],"江苏大学 设施番茄 采摘机器人","YOLO v5 番茄 识别定位","江苏大学设施番茄采摘机器人-3249",5,{"doi":8,"openalex_id":8,"authors":424,"venue":8,"cited_by_count":125,"oa_url":8,"card":425,"direction":140,"ingested_from":430},[],{"tldr":426,"method":427,"finding":428,"direction":140,"opportunity":429},"设计设施番茄智能采摘平台，融合改进YOLO v5与HSV实现识别定位与采摘。","改进YOLO v5-HSV融合算法、ZED双目相机眼在手外标定、六轴机械臂力控末","识别准确率95.01%，采摘成功率87.96%，单果平均采摘时间14.56秒。","可探索多果簇、遮挡与弱光环境下识别定位鲁棒性，并优化采摘效率与末端力控。","agent","2026-09-23T00:04:33.401768Z",{"id":433,"title":434,"url":435,"summary":436,"summary_zh":437,"content":8,"source_name":438,"source_url":435,"published_at":439,"category":12,"cover_url":8,"hotness":13,"is_selected":14,"score":440,"score_detail":441,"sources":444,"tags":446,"primary_topic":450,"secondary_topics":451,"classification":457,"search_phrases":484,"slug":487,"view_count":21,"doi":488,"paper":489,"created_at":510},2805,"Synergistic effects of spectral preprocessing and machine learning algorithms for nitrogen estimation in tomato using hyperspectral spectral data","https:\u002F\u002Fdoi.org\u002F10.1038\u002Fs41598-026-71908-1","Abstract Accurate estimation of leaf nitrogen content is essential for optimizing fertilization management and improving crop productivity. This study proposes a nondestructive hyperspectral imaging framework combined with advanced machine learning algorithms to classify nitrogen levels in tomato leaves (Solanum lycopersicum L., Royal variety). A total of 300 hyperspectral samples were collected from plants subjected to three nitrogen treatments (N-30%, N-60%, and N-90%) under controlled conditions. Reflectance data (400–1000 nm) were calibrated and preprocessed using Standard Normal Variate (SNV), Multiplicative Scatter Correction (MSC), and Savitzky–Golay (SG) filtering. Both unsupervised and supervised algorithms were systematically evaluated. Clustering results demonstrated that preprocessing substantially influenced class separability. The MSC + GMM combination yielded the best unsupervised results, with the lowest Davies–Bouldin index (0.443) and the highest silhouette coefficient (0.975), indicating improved cluster compactness and separation. In supervised learning, Neural Networks (NN) consistently outperformed the other models, achieving 100% accuracy and an ROC–AUC of 1.00 under SNV, MSC, and SG preprocessing. Gradient Boosting (GB) and Support Vector Machine (SVM) also demonstrated strong predictive capability, whereas conventional models, including LDA, LR, KNN, and NB, showed moderate improvements following preprocessing. Statistical analysis confirmed the significant effect of spectral preprocessing on both clustering and classification outcomes ( p \u003C 0.05). Overall, integrating hyperspectral imaging with appropriate preprocessing and nonlinear machine learning models provided strong classification performance under the controlled experimental conditions. However, the exceptionally high performance observed for some models should be interpreted cautiously because of the relatively small dataset and experimentally controlled nitrogen treatments. Moreover, the absence of an independent external validation dataset limits the assessment of generalizability across growing conditions, cultivars, and field environments. Therefore, independent validation using larger and more diverse datasets is required before broader application of the proposed framework in practical precision agriculture.","摘要 准确估算叶片氮含量对于优化施肥管理和提高作物生产力至关重要。本研究提出了一种无损高光谱成像框架，结合先进的机器学习算法对番茄叶片（Solanum lycopersicum L.，Royal品种）的氮水平进行分类。在受控条件下，从接受三种氮处理（N-30%、N-60%和N-90%）的植株中共采集了300个高光谱样本。反射率数据（400–1000 nm）经标准正态变量变换（SNV）、多元散射校正（MSC）和Savitzky–Golay（SG）滤波进行校准和预处理。系统评估了无监督和有监督算法。聚类结果表明，预处理对类别可分性产生了显著影响。MSC + GMM组合取得了最佳无监督结果，具有最低的Davies–Bouldin指数（0.443）和最高的轮廓系数（0.975），表明聚类紧密度和分离度得到改善。在有监督学习中，神经网络（NN）始终优于其他模型，在SNV、MSC和SG预处理下均达到100%的准确率和1.00的ROC–AUC。梯度提升（GB）和支持向量机（SVM）也表现出较强的预测能力，而传统模型（包括LDA、LR、KNN和NB）在预处理后表现出中等程度的改善。统计分析证实了光谱预处理对聚类和分类结果均有显著影响（p \u003C 0.05）。总体而言，将高光谱成像与适当的预处理及非线性机器学习模型相结合，在受控实验条件下提供了较强的分类性能。然而，由于数据集相对较小且氮处理为实验受控条件，某些模型所表现出的极高性能应谨慎解读。此外，缺乏独立的外部验证数据集限制了对跨生长条件、品种和田间环境泛化能力的评估。因此，在将该框架广泛应用于实际精准农业之前，需要使用更大规模、更多样化的数据集进行独立验证。","Scientific Reports","2026-09-17T00:00:00Z",77,{"impact":17,"substance":442,"depth":155,"authority":20,"freshness":13,"relevant":22,"comment":443},21,"高光谱结合机器学习实现番茄叶片氮素无损估测，方法系统、结论明确，但样本量小且缺乏外部验证，属细分领域方法学进展。",[445],{"name":438,"url":435},[27,94,447,448,449],"精准施肥","番茄种植","高光谱遥感",{"code":32,"slug":33,"name":34,"path":35,"parent_code":36},[452,453],{"code":39,"slug":40,"name":41,"path":42,"parent_code":43},{"code":286,"slug":454,"name":455,"path":456,"parent_code":43},"remote-sensing","遥感与农情监测","\u002Ftopic\u002Fdigital-agriculture\u002Fremote-sensing",{"status":45,"locked":14,"version":46,"method":47,"evidence":458},{"candidates":459,"agricultural_evidence":52},[460,464,468,473,480],{"code":32,"score":17,"supported":52,"title":461,"abstract":462,"tags":463},[106],[108,106],[448],{"code":39,"score":17,"supported":52,"title":465,"abstract":466,"tags":467},[244],[243,244],[94],{"code":286,"score":51,"supported":52,"title":469,"abstract":471,"tags":472},[470],"hyperspectral",[470],[449],{"code":474,"score":116,"supported":52,"title":475,"abstract":476,"tags":479},"F3",[],[477,478],"施肥","fertilization",[447],{"code":64,"score":116,"supported":52,"title":481,"abstract":482,"tags":483},[],[225,262],[27],[485,486],"农业人工智能 高光谱遥感 智慧农业 番茄种植","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番茄种植-2805","10.1038\u002Fs41598-026-71908-1",{"doi":488,"openalex_id":490,"authors":491,"venue":438,"cited_by_count":125,"oa_url":435,"card":504,"direction":140,"ingested_from":143},"W7213472002",[492,495,498,501],{"name":493,"orcid":494},"Mohammad Vahedi Torshizi","https:\u002F\u002Forcid.org\u002F0000-0003-3648-1515",{"name":496,"orcid":497},"Sajad Sabzi","https:\u002F\u002Forcid.org\u002F0000-0003-2439-5329",{"name":499,"orcid":500},"Mohsen Azadbakht","https:\u002F\u002Forcid.org\u002F0000-0002-5726-9321",{"name":502,"orcid":503},"Razieh Pourdarbani","https:\u002F\u002Forcid.org\u002F0000-0003-0766-8305",{"tldr":505,"method":506,"finding":507,"direction":508,"opportunity":509},"用高光谱成像结合预处理与机器学习，对番茄叶片氮素水平进行分类。","300个高光谱样本，SNV、MSC、SG预处理，聚类与多种监督分类算法对比。","MSC+GMM聚类最优，神经网络在三种预处理下均达100%准确率。","农业遥感与作物表型","样本少且无外部验证，可扩展多品种、多环境田间数据并做独立验证以提升泛化性。","2026-09-17T23:30:59.361085Z"]