Abstract: In order to improve the total yield of strawberries, reasonable monitoring and control of strawberry diseases is an effective means, a strawberry disease identification algorithm based on improved YOLOv5 is proposed. The detection algorithm uses CSPDarknet as the backbone feature extraction network, which can effectively improve the performance and training efficiency of the model. The EIOU loss function and Kmeans clustering algorithm are used to improve the convergence speed of the model. At the same time, CBAM attention mechanism is added to the model to improve the detection accuracy, and finally the CBAMYOLOv5l algorithm based on improved YOLOv5 is constructed. The experimental results show that the improved model improves the detection accuracy and still ensures efficient detection speed compared to the original model. In addition, the trained CBAMYOLOv5l target detection algorithm achieves an overall average accuracy of 96.52% under the validation set, with an average detection time of 27.52 ms. Compared with YOLOv4, YOLOv4Tiny, Faster_RCNN and other target detection algorithms, CBAMYOLOv5l algorithm has greater advantages in accuracy. It has good robustness and realtime performance in the actual strawberry orchard environment, and it can meet the needs of strawberry disease identification accuracy and reliably prompt the health status of strawberries, so as to timely achieve precise pesticide application and other protection measures.

Key words: strawberry, YOLOv5, machine vision, deep learning, disease recognition

摘要： 为提高草莓的总产量，合理监控和防治草莓病害是有效的手段，提出一种基于改进YOLOv5的草莓病害识别算法。该检测算法以CSPDarknet作为主干特征提取网络，能够有效提高模型的性能和训练效率，并使用EIOU Loss损失函数与Kmeans聚类算法，来提高模型的收敛速度。同时，在模型中增加CBAM注意力机制来提高检测精度，最终构建基于改进YOLOv5的CBAMYOLOv5l算法。试验结果表明，改进后的模型较之原始模型，在检测精度上有所提升且依然能保证高效的检测速度。另外，经过训练的CBAMYOLOv5l目标检测算法在验证集下的总体平均精度达到96.52%，平均检测时间为27.52 ms，对比YOLOv4、YOLOv4Tiny、Faster_RCNN等目标检测算法，该检测算法在精度上具有更大的优势，在实际的草莓果园环境中具有良好的鲁棒性与实时性，可以满足草莓病害识别精度的需求，能够可靠地提示草莓健康状态，从而及时地实现精准施药等保护措施。

关键词: 草莓, YOLOv5, 机器视觉, 深度学习, 病害识别