Adversarial Reconstruction for Outdoors Insulator Anomaly Detection and Recognition in High-Speed Railway Traction Substation

As an indispensable part of the high-speed railway traction substation,the outdoors insulator ensures the normal operation of transmission network and defends the stability of transmission line. In order to avoid the serious transmission fault and decrease the economic loss, finding a effective way to find the anomaly regions of insulator is a meaningful issue. Therefore, a method is presented for outdoors insulator anomaly detection and recognition in high-speed railway traction substation based on adversarial reconstruction in this paper. First, the R3Det is employed to fix the position of insulator from the input image of traction substation. At the second stage, the insulator image which located from the first stage is entered into our designed generative adversarial networks to reconstruct input image. In the end, the structural similarity algorithm is adopted to compare the difference between the input image and the generated image to find the anomalous region, then template matching is used for recognition of the anomalous area. Test results on Heishan traction substation show the effective performance of our designed method.