Insulator Defects Detection and Classification Method Based on YOLOV5

With the expansion of high-voltage transmission line construction, UAVs(Unmanned Aerial Vehicles) have been widely adopted in line inspection. The issue of self-explosion in glass insulators has attracted significant attention. In this study, we propose a method for self-explosion defect recognition based on YOLOv5 and incorporate various data augmentation techniques to enhance adaptability to complex environmental factors. Experimental results demonstrate the method's remarkable performance in defect detection, achieving an average precision (mAP) of 95.8% that satisfies practical detection requirements. Furthermore, for the detected self-explosion defects, the glass insulators are further classified into upper, middle, and lower defects based on the explosion location. Experimental findings indicate that this method effectively classifies self-explosion defects in glass insulators, holding significant implications for analyzing self-explosion patterns.