Electrical Insulator Defects Detection Method Based on YOLOv5
Zhiqiang Feng, Li Guo, Darong Huang, Runze Li
Abstract
For electrical transmission lines, insulator inspection is an important indicator for power system safety operation. Manual visual inspection activities are usually performed in insulator statue recognition and maintenance, but it is time-consuming, unsafe, and low-efficient. As the development of image processing and machine learning, automatic insulator defect detection has been drawn more attention in electrical equipment inspection in recent years. This paper proposes an automatic insulator detection method using YOLOv5 object detection model. By comparing performance with 4 different versions of YOLOv5, experimental results show that YOLOv5x model with K-means clustering can achieve highest accuracy at 86.8%, and MAP is 95.5%. In addition, this model can efficiently identify and locate the insulator defects across transmission lines, so as to avoid unsafe manual detection and improve the detection efficiency.