Abstract: :According to the daily inspection requirements of power grid, a method of transmission channel fire identification based on multi-feature fusion is proposed to solve the problem of frequent transmission channel fire trip.The images of transmission line wildfires are enhanced by gamma transform and Sobel operator, and multistage segmentation algorithm consisting of superpixel segmentation, region growth method and morphological processing are used to obtain the images of flame regions. The color and texture features are extracted from segmented flame images and fused using a serial feature fusion approach and a BP neural network classifier is trained to achieve recognition of transmission channel wildfires. Experimental results demonstrate an overall accuracy rate of 92.7% for the proposed method, providing valuable guidance for power operation and maintenance personnel in monitoring fire incidents within transmission channels.

Key words: feature fusion, hill fire, classifier, Sobel operator