Aiming at the problems of large difference in accuracy and low detection accuracy of various types of apple leaf diseases, an improved YOLOv5 detection algorithm for accurate identification of apple leaf diseases (YOLOV5-CSEP) was proposed. Firstly, C3Ghost module was introduced to replace the C3 module of YOLOv5 backbone network to reduce the number of parameters and calculation amount of the model. Secondly, the hybrid attention module C-SAM was added to the backbone network to improve the feature extraction capability of the backbone network, and the CA attention module was added to the neck network to suppress the interference of complex background information. Finally, an enhanced path aggregation network (E-PANet) was introduced to fully integrate multi-scale features and improve the accuracy and robustness of the network to detect various types of apple leaf diseases. Experiments showed that all performance indexes of the improved algorithm were improved, the accuracy rate reached 93.2%, and the average accuracy (mAP@0.5) reached 87.9%. Compared with the original YOLOv5 algorithm, it was improved by 3.4% and 1.7% respectively, and the calculation amount was reduced by 11%.

针对多类苹果叶片病害准确率差异大、检测精度不高的问题，提出一种改进YOLOv5准确判别苹果叶片病害的检测算法(YOLOv5-CSEP)。首先，引入C3Ghost模块替换原YOLOv5主干网络C3模块，减少模型的参数量与计算量；其次，将混合注意力模块C-SAM加入主干网络中，提高主干网络的特征提取能力，在颈部网络中加入CA注意力模块，抑制复杂背景干扰关注目标信息；最后，引入增强型路径聚合网络(E-PANet)充分融合多尺度特征，提升网络对多类苹果叶片病害检测的准确性与鲁棒性。试验表明，改进后算法的各项性能指标均有提升，精确率达到93.2%，平均精度均值mAP@0.5达到87.9%，与原YOLOv5算法相比分别提高3.4%与1.7%，计算量减少11%。