Research of infrared thermal on-line detection technology of zero resistance insulator

The zero resistance insulator greatly threats the safe and steady operation of the power system. A new method of detecting the zero resistance insulator on-line based on the infrared thermal image is presented. The recognition algorithm for zero resistance insulator with image processing is designed. The soft thresholding de-noising method is designed and Bayesian estimation is applied in the de-noising of infrared thermal images. The improved bi-threshold OTSU segmenting method is applied to wipe out the disturbed objects and keep only the plate area of insulators. Two-factor analysis of variance is used to assess the features of the infrared thermal images. Additional momentum and self-adaptive gradient descent algorithm are applied to improve back-propagation (BP) neural network in this research. Testing results show that these methods can achieve good effect.