轻量化 YOLO模型在农作物微小病虫害检测中的应用研究

doi:10.13733/j.jcam.issn.2095-5553.2024.09.040

摘要/Abstract

摘要：

针对农作物病虫害检测中早期微小病虫害变化目标识别准确率不高的问题，提出一种轻量化农作物微小病虫害检测算法 YOLO-MobileNet-CBAM。该算法采用 MobileNetV3轻量级卷积模块替换 YOLOv5s的主干提取网络来降低参数运算量，并引入 CBAM注意力机制从通道和空间两个维度对重要的特征提取进行强化，有效增强小目标的聚焦准确度。卷积模块中用 H-SiLU代替原模型的 SiLU激活函数提高训练速度，避免梯度消失问题。通过将 SIoU函数代替原模型中的 GIoU函数计算预测框回归损失，将形状损失计算在内，进一步提高小目标定位精度。通过特征金字塔输出 4个不同尺度的检测头识别大面积病害、微小病害及虫害目标，增加微小目标的检测精度。结果表明，YOLO-MobileNet-CBAM在微小病虫害目标检测任务中精确率达 92. 38%，召回率达 90. 24%，平均精度大于 90%。实现模型轻量化，同时有效提高检测精确度，为手持式终端检测应用提供技术支持。

关键词: 农作物, 微小病虫害检测, 轻量化模型, YOLO-MobileNet-CBAM

Abstract:

In response to the problem of insufficient accuracy in early small change target recognition in crop pest detection，a lightweight plant pest detection algorithm YOLO-MobileNet-CBAM is proposed. This algorithm replaces the backbone extraction network of YOLOv5s with a lightweight convolutional module of MobileNetV3 to reduce parameter computation，and introduces CBAM attention mechanism to strengthen important feature extraction from both channel and spatial dimensions，effectively enhancing the detection accuracy of small targets. It improves training speed and avoids gradient vanishing problems by replacing the original model′s SiLU activation function with H-SiLU in the convolutional module. The prediction box regression loss function utilizes the SIoU function instead of the GIoU function in the original model，accounting for shape loss to further improve accuracy of small target localization. Finally，four detection heads with varying scales are output via the feature pyramid to identify large.scale diseases，small diseases and pest targets，thereby enhancing the detection accuracy of small targets. The results show that YOLO-MobileNet-CBAM achieves an accuracy rate of 92. 38%，a recall rate of 90. 24%，and an average accuracy of over 90% in detecting small pests and diseases. It achieves lightweight model design while effectively improving detection accuracy，and provides technical support for handheld terminal detection applications.

Key words: crops；small diseases and pests；lightweight model；YOLO-MobileNet-CBAM ,

中图分类号:

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.

Yang Qiaomei, Cui Tingting, Yuan Yongbang, Luo Hua. Research on the application of lightweight YOLO model in detection of small crop diseases and pests [J]. Journal of Chinese Agricultural Mechanization, 2024, 45(9): 265-270.

参考文献

［ 1］联合国粮农组织：全球每年四成农作物因虫害造成的损失达 4 466亿元［ EB/OL］. https：//k. sina. com. cn/article_7517400647_1c0126e47059013hai. html， 2021-06-03.

［ 2］谢春燕，吴达科，王朝勇，等 .基于图像和光谱信息融合的病虫害叶片检测系统［ J］.农业机械学报， 2013，44（S1）：269-272.

Xie Chunyan，Wu Dake，Wang Chaoyong，et al. Insect pest leaf detection system based on information fusion of image and spectrum［J］. Transactions of the Chinese Society for Agricultural Machinery， 2013， 44（S1）： 269-272.

［ 3］田有文，程怡，吴琼，等 .农产品病虫害高光谱成像无损检测的研究进展［ J］.激光与红外， 2013，43（12）： 1329-1335.

［ 4］赵越，赵辉，姜永成，等 .基于深度学习的马铃薯叶片病害检测方法［ J］.中国农机化学报， 2022，43（10）： 183-189.

Zhao Yue，Zhao Hui，Jiang Yongcheng，et al. Detection method of patato leaf diseases based on deep learning［J］. Journal of Chinese Agricultural Mechanization，2022，43（10）：183-189.

［ 5］王江晴，冀星，莫海芳，等 .基于轻量化 VGG的植物病虫害识别［ J］.中国农机化学报， 2022，43（4）：25-31.

Wang Jiangqing， Ji Xing， Mo Haifang， et al. Plant disease detection based on lightweight VGG［J］. Journal of Chinese Agricultural Mechanization， 2022， 43（4）： 25-31.

［ 6］赵法川，徐晓辉，宋涛，等 .融合多头注意力的轻量级作物病虫害识别方法［ J］.华南农业大学学报， 2023（6）： 1-16.

Zhao Fachuan， Xu Xiaohui， Song Tao， et al. A lightweight crop pest identification method based on multi.head attention［J］. Journal of South China Agricultural University，2023（6）：1-16.

［ 7］曹跃腾，朱学岩，赵燕东，等 .基于改进 ResNet的植物叶片病虫害识别［ J］.中国农机化学报， 2021，42（12）： 175 -181.

Cao Yueteng， Zhu Xueyan， Zhao Yandong， et al. Recognition of plant leaf diseases，insect，and pests based on improved ResNet［J］. Journal of Chinese Agricultural Mechanization，2021，42（12）：175-181.

［ 8］杨堃，范习健，薄维昊，等 .基于视觉加强注意力模型的植物病虫害检测［ J］.南京林业大学学报（自然科学版）， 2023，47（3）：11-18.

［ 9］梁万杰，曹静，孙传亮，等 .基于深度学习的农作物病害识别系统研发［J］.中国农机化学报， 2023，44（9）：169-175.

Liang Wanjie，Cao Jing，Sun Chuanliang，et al. Research and development of crop diseases intelligent recognition system based on deep learning［J］. Journal of Chinese Agricultural Mechanization，2023，44（9）：169-175.

［10］胡凯，骆润玫，刘泽乾，等 .基于改进 SSD的广佛手病虫害检测方法［ J］.南京农业大学学报， 2023，46（4）： 813-821.

Hu Kai， Luo Runmei，Liu zeqian， et al. Detection of bergamot diseases and pests based on improved SSD［J］. Journal of Nanjing Agricultural University，2023，46（4）： 813-821.

［11］徐会杰，黄仪龙，刘曼 .基于改进 YOLOv3模型的玉米叶片病虫害检测与识别研究［ J］.南京农业大学学报， 2022， 45（6）：1276-1285.

Xu Huijie，Huang Yilong，Liu Man，et al. Research on pest detection and identification of corn leaf based on improved YOLOv3 model［J］. Journal of Nanjing Agricultural University，2022，45（6）：1276-1285.
［12］Redmon J，Divvala S，Girshick R，et al. You only look once：Unified，real.time object detection［C］. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition，2016：779-788.
［13］LiY G，Wei H，Han Z，et al. Deep learning.based safety helmet detection in engineering management based on convolutional neural networks［J］. Advances in Civil Engineering，2020（6）：1-10.
［14］Woo S，Park J，Lee J Y，et al. CBAM：Convolutional block attention module［C］. Proceedings of the European Conference on Computer Vision（ECCV），2018：3-19.
［15］Howard A， Sandler M， Chu G， et al. Searching for mobilenetv3［C］. Proceedings of the IEEE/CVF International Conference on Computer Vision， 2019： 1314-1324.
［16］Lin T Y，Dollár P，Girshick R，et al. Feature pyramid networks for object detection［C］. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition，2017：2117-2125.
［17］Liu S，Qi L，Qin H，et al. Path aggregation network for instance segmentation［C］. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition， 2018：8759-8768.
［18］Rezatofighi H，Tsoi N，Gwak J Y，et al. Generalized intersection over union：A metric and a loss for bounding box regression［C］. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition， 2019：658-666.
［19］Loss G Z S I U. More powerful learning for bounding box regression［J］. arxiv preprint arxiv：2205. 12740，2022.
［20］Ren S， He K， Girshick R， et al. Faster R-CNN： Towards real.time object detection with region proposal networks［J］. IEEE Transactions on Pattern Analysis and Machine Intelligence，2016，39（6）：1137-1149.