基于改进YOLOv8n的真实场景下草莓病害检测方法

doi:10.13733/j.jcam.issn.20955553.2024.12.039

摘要/Abstract

摘要：

针对实际种植环境下草莓病害目标检测中，存在背景复杂、检测精度低等主要问题，提出一种改进YOLOv8n的草莓病害检测算法YOLOv8n-SD。搜集并处理真实场景下草莓叶、花、果的常见病害图像以构建试验数据集。在YOLOv8n模型的基础上对其进行优化改进，利用多尺度并行计算与补丁感知注意力对主卷积模块进行重构，提出C2f-PPA模块，有效融合多尺度特征信息，提高模型的特征捕获能力。引入ADown模块，减少下采样过程中的信息损失，提高模型的推理速度和鲁棒性。提出一种任务对齐的共享动态检测头（Task-aligned Dynamic Head，TDyH），增强定位分支和分类分支之间的信息交互，降低模型参数的同时，提高检测精度和准确性。根据试验结果，改进后的YOLOv8n-SD模型的检测精度达到83.7%，相较于原YOLOv8n提高3.3%，mAP@0.5与mAP@0.5∶0.95分别达到76.9%和59.9%，分别提升1.6%和2.3%。改进后的算法能精确识别草莓生长各阶段的常见病害，并满足边缘设备的轻量化需求和实时检测需求。

关键词: 草莓病害, 目标检测, YOLOv8n, ADown, 轻量化, 实时检测

Abstract:

Addressing the primary issues of complex backgrounds and low detection accuracy in strawberry disease target detection under practical farming conditions, an improved YOLOv8n strawberry disease detection algorithm as YOLOV8N-SD is proposed. Images of common strawberry leaf, flower, and fruit diseases under real-world scenarios are collected and processed to construct an experimental dataset. Optimizations and improvements are made to the YOLOv8n model. The primary convolutional module is reconstructed by using multi-scale parallel computing and patch-perceptive attention, introducing the C2f-PPA module. This effectively integrates multi-scale feature information, enhancing the model's feature capturing capability. Additionally, the ADown module is incorporated to reduce information loss during downsampling, thereby improving the model's inference speed and robustness. A Task-aligned Dynamic Head (TDyH) is proposed to strengthen information exchange between the localization and classification branches. This reduces model parameters while simultaneously enhancing detection precision and accuracy. According to experimental results, the improved YOLOv8n-SD model achieves a detection accuracy of 83.7%, representing a 3.3% increase over the original YOLOv8n. Its mAP@0.5 and mAP@0.5∶0.95 scores reach 76.9% and 59.9% respectively, marking improvements of 1.6% and 2.3% compared to the baseline. This enhanced algorithm not only accurately identifies common diseases at various stages of strawberry growth but also meets the lightweight and real-time detection requirements of edge devices.

Key words: strawberry diseases, target detection, YOLOv8, ADown, lightweight, real-time detection

中图分类号:

李嘉诚, 陈中举, 许浩然. 基于改进YOLOv8n的真实场景下草莓病害检测方法[J]. 中国农机化学报, 2024, 45(12): 267-274.

Li Jiacheng, Chen Zhongju, Xu Haoran. A method for strawberry disease detection in real scenarios based on improved YOLOv8n[J]. Journal of Chinese Agricultural Mechanization, 2024, 45(12): 267-274.

参考文献

［1］　丁新伦, 张洁, 吴祖建. 草莓上新发现的病毒研究现状与展望［J］. 园艺学报, 2024, 51(7): 1639-1648.
Ding Xinlun, Zhang Jie, Wu Zujian. Recent advances in new viruses infecting strawberry ［J］. Acta Horticulturae Sinica, 2024, 51(7): 1639-1648.
［2］　黄宁旺, 蒋琪玮, 王晶如, 等. 设施内不同遮阴处理对草莓生长和光合特性及产量品质的影响［J］. 中国果树, 2024(5): 68-74.
Huang Ningwang, Jiang Qiwei, Wang Jingru, et al. Effects of varying shading treatments under facility conditions on the growth, photosynthetic characteristics, yield and quality of strawberry ［J］. China Fruits, 2024(5): 68-74.
［3］　章璞, 乔波, 陈义明. 基于改进YOLOv8的多阶段草莓检测算法［J］. 中国农机化学报, 2024, 45(10): 274-280.
Zhang Pu, Qiao Bo, Chen Yiming. A multistage strawberry detection algorithm based on improved YOLOv8 ［J］. Journal of Chinese Agricultural Mechanization, 2024, 45(10): 274-280.
［4］　王瑞彬, 杨世忠, 高升. 融合残差网络与注意力机制的草莓检测［J］. 中国农机化学报, 2024, 45(1): 266-273.
Wang Ruibin, Yang Shizhong, Gao Sheng. Strawberry detection combining residual network with attention mechanism ［J］. Journal of Chinese Agricultural Mechanization, 2024, 45(1): 266-273.
［5］　黄铝文, 郑梁, 黄煜, 等. 基于多尺度卷积与通道域增强的草莓病害识别方法［J］. 江苏农业科学, 2023, 51(10): 202-210.
Huang Lüwen, Zheng Liang, Huang Yu, et al. Strawberry disease recognition method based on multiscale convolution and channel domain enhancement ［J］. Jiangsu Agricultural Sciences, 2023, 51(10): 202-210.
［6］　Targ S, Almeida D, Lyman K. Resnet in resnet: Generalizing residual architectures ［J］. arXiv preprint arXiv: 1603.08029, 2016.
［7］　胡晓波, 许桃胜, 黄伟, 等. 交互式双分支特征融合的草莓病害程度快速诊断方法［J］. 农业机械学报, 2023, 54(11): 225-235.
Hu Xiaobo, Xu Taosheng, Huang Wei, et al. Interactive bilateral feature fusion network for realtime strawberry disease diagnosis ［J］. Transactions of the Chinese Society for Agricultural Machinery, 2023, 54(11): 225-235.
［8］　乔珠峰, 赵秋菊, 郭建鑫, 等. 基于改进YOLOv5的草莓病害智能识别终端设计［J］. 中国农机化学报, 2024, 45(3): 205-211.
Qiao Zhufeng, Zhao Qiujiu, Guo Jianxin, et al. Design of strawberry disease intelligent identification terminal based on improved YOLOv5 ［J］. Journal of Chinese Agricultural Mechanization, 2024, 45(3): 205-211.
［9］　Wang Q, Wu B, Zhu P, et al. ECA-Net: Efficient channel attention for deep convolutional neural networks ［C］.Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2020: 11534-11542.
［10］　亢洁, 刘佳, 刘文波, 等. 基于多感受野与多尺度融合的草莓病害分割算法［J］. 江苏农业科学, 2024, 52(4): 206-215.
Kang Jie, Liu Jia, Liu Wenbo. Study on strawberry disease segmentation algorithm based on multi receptive field and multiscale fusion module ［J］. Jiangsu Agricultural Sciences, 2024, 52(4): 206-215.
［11］　Siddique N, Paheding S, Elkin C P, et al. U-Net and its variants for medical image segmentation: A review of theory and applications ［J］. IEEE Access, 2021, 9: 82031-82057.
［12］　邱畅, 田光兆, 赵嘉威, 等. 基于改进YOLOv5的草莓病害识别［J］. 中国农机化学报, 2024, 45(3): 198-204.
Qiu Chang, Tian Guangzhao, Zhao Jiawei, et al. Strawberry disease identification based on improved YOLOv5 ［J］. Journal of Chinese Agricultural Mechanization, 2024, 45(3): 198-204.
［13］　Woo S, Park J, Lee J Y, et al. Cbam: Convolutional block attention module ［C］. Proceedings of the European Conference on Computer Vision, 2018: 3-19.
［14］　Yang Z, Wang X, Li J. EIoU: An improved vehicle detection algorithm based on vehiclenet neural network ［C］. Journal of Physics: Conference Series. IOP Publishing, 2021, 1924(1): 012001.
［15］　Ahmed M, Seraj R, Islam S M S. The kmeans algorithm: A comprehensive survey and performance evaluation ［J］. Electronics, 2020, 9(8): 1295.
［16］　Xu S, Zheng S C, Xu W, et al. HCF-Net: Hierarchical context fusion network for infrared small object detection ［J］. arXiv preprint arXiv:240310778, 2024.
［17］　Wang C Y, Yeh I H, Liao H Y M. YOLOv9: Learning what you want to learn using programmable gradient information ［J］. arXiv preprint arXiv:240213616, 2024.
［18］　Feng C, Zhong Y, Gao Y, et al. Tood: Taskaligned onestage object detection ［C］. 2021 IEEE/CVF International Conference on Computer Vision. IEEE Computer Society, 2021: 3490-3499.
［19］　Zhu X, Hu H, Lin S, et al. Deformable convnets v2: More deformable, better results ［C］. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2019: 9308-9316.
［20］　Li C, Li L, Jiang H, et al. YOLOv6: A singlestage object detection framework for industrial applications ［J］.arXiv preprint arXiv: 2209.02976, 2022.
［21］　Woo S, Debnath S, Hu R, et al. Convnext v2: Codesigning and scaling convnets with masked autoencoders ［C］. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023: 16133-16142.
［22］　Qin D, Leichner C, Delakis M, et al. Mobilenetv4-universal models for the mobile ecosystem ［J］. arXiv preprint arXiv: 2404.10518, 2024.
［23］　Ding X, Zhang Y, Ge Y, et al. UniRepLKNet: A universal perception largekernel convnet for audio video point cloud timeseries and image recognition ［C］. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024: 5513-5524.
［24］　Tan M, Pang R, Le Q V. Efficientdet: Scalable and efficient object detection ［C］. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2020: 10781-10790.
［25］　Dai X, Chen Y, Xiao B, et al. Dynamic head: Unifying object detection heads with attentions ［C］. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2021: 7373-7382.

[1]	张传栋, 亓璐, 丁华立. 基于改进 YOLOv8n模型的多品种葡萄簇检测方法[J]. 中国农机化学报, 2024, 45(9): 220-226.
[2]	李娟 , 金志雄 . 基于轻量化 Transformer的农作物检测机器人路径规划[J]. 中国农机化学报, 2024, 45(9): 227-233.
[3]	李邦国, 王辉, 宋杨, 任志伟, 刘跃华, 徐乐程. 基于无人驾驶小麦收割机立体视觉感知系统[J]. 中国农机化学报, 2024, 45(9): 244-249.
[4]	杨巧梅, 崔婷婷, 袁永榜, 罗桦 . 轻量化 YOLO模型在农作物微小病虫害检测中的应用研究[J]. 中国农机化学报, 2024, 45(9): 265-270.
[5]	赵晓霞, 程曼, 袁洪波. 基于StrongSORT算法的羊只多目标跟踪方法[[J]. 中国农机化学报, 2024, 45(8): 180-188.
[6]	冯晓, 张辉, 刘运超, 张微, 李小红, 马中杰, . 基于改进YOLOv5s的硬核期葡萄簇检测[J]. 中国农机化学报, 2024, 45(8): 240-245.
[7]	周思捷, 刘天奇, 陈天华. 基于改进YOLOv5算法的水稻病害识别研究[J]. 中国农机化学报, 2024, 45(8): 246-253.
[8]	岳耀华, 张伟, 亓立强, . 基于改进YOLOv5的田间大豆花朵生长状态检测方法研究[J]. 中国农机化学报, 2024, 45(7): 188-193.
[9]	张延军, 赵建鑫, . 基于多路神经网络的黄花菜关节点检测算法研究[J]. 中国农机化学报, 2024, 45(7): 228-234.
[10]	吴利刚, 吕媛媛, 周倩, 陈乐, 张梁, 史建华. 基于改进YOLOv5的轻量级黄花成熟检测方法[J]. 中国农机化学报, 2024, 45(7): 235-242.
[11]	张晨一, 张晓乾, 任振辉. 基于改进YOLOv3的自然场景中未成熟青苹果识别研究[J]. 中国农机化学报, 2024, 45(7): 243-248.
[12]	熊志刚, 陈为真. 一种YOLOv5-DK的柠檬初期病虫害检测算法研究[J]. 中国农机化学报, 2024, 45(7): 249-254.
[13]	薛霞, 刘鹏, 周文. 基于改进YOLOv8算法的实时细粒度植物病害检测[J]. 中国农机化学报, 2024, 45(5): 188-194.
[14]	漆海霞, , , 李承杰, 黄桂珍. 基于轻量化YOLOv4的死淘鸡目标检测算法[J]. 中国农机化学报, 2024, 45(5): 195-201.
[15]	罗鑫, 李加强, 何超. 基于改进YOLOv4的澳洲坚果视觉监测方法[J]. 中国农机化学报, 2024, 45(5): 217-222.