Research on the classification and detection of insulator pollution levels based on the improved algorithm of NSST-PCNN image fusion

Aiming at the problem of pollution accumulation on the surface of insulators of transmission lines, an innovative image fusion improvement algorithm based on non-downsampled shear wave transform (NSST) and pulse-coupled neural network (PCNN) is proposed, taking into account the complementary advantages of infrared images and visible light images. On this basis, a convolutional neural network algorithm is utilized to automatically detect and classify the degree of insulator surface contamination. First, the insulators in the substation were photographed using a dual-lens imaging device to obtain infrared images and visible light images with different levels of contamination; then pre-processing operations such as image denoising and image alignment were performed, and image fusion was realized using the improved NSST-PCNN algorithm. Finally, comparison experiments between insulator pollution degree classification algorithms for multi-source image fusion datasets; and between single-image source datasets and multisource image fusion datasets were conducted. The experimental results show that the improved NSST-PCNN fusion algorithm proposed in this paper can effectively improve the accuracy of insulator pollution level classification and provide new technical support for non-contact insulator pollution level detection.