Abstract
Aiming at the current deployment of visual cameras for transmission lines, the fusion technology of visual images for transmission lines is studied. A convolutional neural network fusion algorithm based on super pixel and guided filtering is proposed. Firstly, the source image is decomposed at the basic level and the detail level, and the images on two scales are segmented into hyper-pixels. The distance metric is used to achieve clustering of hyper-pixels. Based on the hyper-pixel clustering block, a minimum spanning tree of hyper-pixels is formed to obtain the connected optimized guide image, and the original image fusion effect is optimized using guide filtering. At the same time, considering and designing the initial image imaging jitter and online collaborative computing deployment that exist in the fusion, to better meet the requirements of high computing speed and low computing resources for transmission line scenarios.