为快速、准确识别田间环境下的三七叶片病害，提出1种基于改进YOLOv5s算法的三七叶片病害识别方法.以三七的六种典型病害圆斑病、灰霉病、黄锈病、白粉病、炭疽病和病毒病为对象，通过引入SE-Attention（spatial excitation attention）注意力机制，增强了YOLOv5s模型对病害部位关键信息的捕捉能力，提高模型对不同通道特征信息的敏感性.此外，模型采用加权双向特征金字塔网络（bidirectional feature pyramid network， BiFPN），提升特征融合的有效性，实现高层次的特征表达，从而提高模型的泛化能力.试验结果表明，改进后的YOLOv5s模型准确率和召回率分别为86.8%和85.3%，mAP_0.5和mAP_0.5：0.95分别为87.4%和76.4%，相较于原始YOLOv5s模型，准确率、召回率、mAP_0.5和mAP_0.5：0.95分别提升10.4、13.0、11.0和18.6个百分点，最后通过对比不同深度学习目标检测模型对三七病害进行检测，结果表明本试验所提出的算法可以在保证精度与速度的前提下，实现田间三七病害的智能化检测，为实际应用场景下三七病害自动检测提供理论依据.

In order to quickly and accurately identify Panax notoginseng leaf diseases in the field environment， this study proposes a detection method based on an improved YOLOv5s algorithm. The approach focuses on six common diseases affecting Panax notoginseng： cercospora leaf spot， gray mold， yellow rust， powdery mildew， anthracnose and viral disease. By incorporating the SE－Attention （Spatial Excitation Attention） mechanism， the model enhances its ability to capture critical information from diseased regions， thereby improving sensitivity to feature channels. Furthermore， the model utilizes a weighted Bidirectional Feature Pyramid Network（BiFPN） to improve the effectiveness of feature fusion， achieving higher－level feature representation and enhancing the model’s generalization capability. The experimental results show that the improved YOLOv5s model achieves an accuracy of 86.8% and a recall rate of 85.3%， with mAP_0.5 and mAP_0.5∶0.95 are 87.4% and 76.4% respectively. Compared with the original YOLOv5s model， the improved version demonstrates gains of 10.4， 13.0， 11.0 and 18.6 percentage point in accuracy， recall rate， mAP_0.5， and mAP_0.5∶0.95， respectively. Additionally， comparative tests with other deep learning－based object detection models confirm that the improved model effectively detects Panax notoginseng diseases. These results demonstrate that the proposed algorithm offers an optimal balance between detection accuracy and computational efficiency， enabling real－time disease identification in Panax notoginseng under field conditions. This provides a theoretical foundation for the automated detection of leaf diseases in practical agricultural applications.