Study on the Self-Explosion Defect Detection Methods in Glass Insulators Based on Modified YOLOv7

With the increasing utilization rate of glass insulators in transmission lines, in order to adapt to the current mode of unmanned aerial vehicle (UAV) -assisted inspection, a modified YOLOv7 glass insulator self-explosion defect detection method is proposed to meet the demand for glass insulator defect detection in the process of transmission line inspection. ST attention mechanism was introduced, and its module was added to the 11th layer of YOLOv7 network model to strengthen the detection effect of small targets and incomplete insulators. The coordinate loss function was modified, the formula of SIOU loss function was defined, and the position of detection box was corrected. Some glass insulator images in the data set CPLID and the glass insulator images taken by the UAV are used to train the modified network. The detection results show that the detection accuracy and speed are significantly modified.

I. Introduction

With the continuous expansion of the scale of power grid, the safety performance of line insulators is increasingly required. Due to the superior performance of glass insulators, the proportion of glass insulators used in the transmission network is increasing, and its quality and safety testing is particularly important. At present, manual inspection of transmission lines assisted by unmanned aerial vehicle (UAV) technology has become more and more normal^[1]. A large number of images of transmission line insulators can be obtained by UAV, and it is inevitable that false detection and omission will be caused by completely relying on manual identification. The use of computer for image detection can greatly save manpower and material resources, and effectively improve the detection efficiency.