Super‐resolution reconstruction method of infrared images of composite insulators with abnormal heating based on improved SRGAN

Abstract

Abstract Abnormal heating of composite insulators on ultra‐high voltage (UHV) transmission lines is widespread under high humidity and heat in southern China. The infrared imaging technology can be used of quickly and effectively detect insulators with abnormal heating and determine their defect types. However, it may cause misjudgement when the infrared images are not clear. A super‐resolution reconstruction method of infrared images of composite insulators with abnormal heating based on improved super‐resolution generative adversarial networks (SRGAN) is proposed in this paper, and the dense residual network in SRGAN is perfected by introducing the residual channel attention (ResCA). This method overcomes the problems that the resolution of images reconstructed by traditional methods is not significantly improved, and images are too smooth, and the edge details are easily lost. Compared with the traditional methods, the five reference‐free image quality evaluation indexes of the image reconstructed by the improved SRGAN are increased by −0.73% to 63.84%. Further the effectiveness and superiority of the improved SRGAN are verified by comparing the changes in the distribution of grey values of infrared images before and after reconstruction. Finally, the improved SRGAN is used to perform super‐resolution reconstruction of the low‐resolution infrared image examples.