An efficient image stitching method of wind-turbine blades for damage measurement in online repair robots

When there is a large damage on the wind-turbine blade, a single image taken at close range cannot cover the entire damage area, thus affecting a series of online repair operations of the repair robot. Therefore, an efficient image stitching method of wind-turbine blades for damage measurement in online repair robots is proposed. Firstly, the nonlinear scale space is constructed by AKAZE algorithm for feature point detection, and then PCA and SIFT-like descriptor are combined to form PCA-SIFT-like algorithm to describe the obtained feature points. Secondly, Brute-force matching is combined with KNN to form the Brute-force-KNN algorithm to quickly screened out the initial correct matching set. Then the wrong matches are further eliminated by the RANSAC algorithm. Thirdly, the multi-resolution fusion algorithm is adopted to eliminate stitching artifacts resulting from image registration, thereby achieving seamless stitching of wind-turbine blade images. Finally, aiming at the problem of engineering application of wind-turbine blade image stitching in the scene of robot high-altitude repair work, an experimental platform based on the wind-turbine blade repair robot is built, and image stitching software is developed. On this basis, the online stitching experiment of wind-turbine blade images was carried out. Experimental results show that the proposed stitching algorithm can meet the online operation requirements of wind-turbine blades for online repair robots. Therefore, the proposed image stitching method for wind-turbine blades is feasible and effective in engineering applications, which can provide technical support for robot automatic measurement and repair of blade damage, and has good engineering application value.