Research on Defect Detection Method of Substation Equipment Based on Cascade Network

Abstract

In order to accurately identify and locate the typical visual defects of substation equipment, a defect detection model of power equipment based on cascaded network structure is proposed. Firstly, aiming at the weak representation ability of small target defects in the original image, a super-resolution representation method based on generative confrontation network and double attention mechanism is proposed. Secondly, the content loss function and anti-loss function are used to optimize the generated network, and the discriminant network is used to guide the calculation of the representation process. Finally, the optimal super-resolution representation generation network and YOLOv5s network are cascaded to form a substation equipment defect detection model. The method is tested on the sample set taken in practical engineering application, and the mAP index is over 92%. The experimental results show that this method can be well ap-plied to the substation equipment defect detection task.