An Automatic Defect Detection Method for Gas Insulated Switchgear

Abstract

According to the characteristics of gas insulated switchgear (GIS) equipment fault detection and internal structure, a GIS pipe inspection robot with visual equipment was designed. Based on inspection images, a novel method of defect detection was proposed for small target detection inside the GIS. Firstly, image enhancement is applied for image acquired in low light based on Retinex-Net. Then, small target detection algorithm is implemented using improved YOLOv3 for foreign object detection inside GIS. In order to improve detection speed, the number of parameters of YOLOv3 is reduced by model pruning operation, which holding the accuracy of the model. The experimental results demonstrate that the mAP (Mean Average Precision) reaches 0.86, and the proposed algorithm is capable of real-time detection as faster than 15 frames per second (fps).