Super Pixel-Level Contamination State Detection Model of Composite Insulator Specimens Under Different Lighting Conditions Based on Hyperspectral Technology and Model Transfer

Accurate and real-time monitoring of the surface contamination status of insulators is of great significance for preventing pollution flashover accidents and maintaining system safety. This article introduces hyperspectral imaging technology to detect the contamination status of composite insulators. To address the impact of different outdoor lighting conditions on the detection of composite insulator contamination status, a composite insulator contamination status evaluation model is established based on the hyperspectral data collected indoors. Then, gradient spectral angle matching (GSAM) method is used to extract insulator images from the hyperspectral images collected under different outdoor lighting conditions, and the insulator image is segmented into multiple super-pixel regions using the energy-driven sampling method. The super-pixel average spectral data were used instead of the pixel spectral data, and the contamination state was evaluated after the series processing of the two model transfer methods of correlation alignment method and piecewise direct standardization (PDS). The results show that the use of super-pixel segmentation method reduces the amount of data to be measured by 98% while ensuring the visualization of insulator pollution distribution, and the model transfer method improves the classification accuracy of the model from 50% to 90%, effectively reducing the impact of changes in lighting conditions on hyperspectral detection.