Research on insulator defect detection algorithm based on improved YOLOv5s
Insulator defect detection is a key step in the development of smart grids.At present,insula-tor defect detection based on computer vision has been widely used in intelligent inspections.This paper se-lects the YOLOv5s model as the basic network to improve detection accuracy while ensuring network operation speed.Firstly,the CBAM attention module is added to the backbone feature extraction network to enhance the feature extraction capability of the model;secondly,the BiFPN structure is used in the neck structure to fuse multi-scale features to reduce feature loss and improve the feature fusion capability of the model;finally EIoU Loss is used as the loss function of the network regression loss,which solves the problem of being sensitive to insulators of various scales in aerial images and improves the convergence speed of the network.Verified by experimental results,the improved network model has a mAP value of 94.13%and a Recall value of 84.94%when the detection speed does not change much,which are 5.71%and 14.57%higher than the YOLOv5s network model.At the same time,the model size is reduced to 13.5 MB.Compared with other improved mod-els,the precision and recall rate have been significantly improved,which can better meet the needs of practi-cal applications.
万方数据
维普
