Regional fuzzy binocular stereo matching algorithm based on global correlation coding for 3D measurement of rail surface

Abstract In view of the uncontrollable imaging quality of the electromagnetic railgun’s rail, based on the existing 3D measurement system, which combines binocular vision with active dot matrix illumination, a regional fuzzy binocular stereo matching algorithm based on global correlation coding is presented. Considering the lack of imaging details and textures and the problem of point target matching, the proposed stereo matching algorithm can get more complete and reliable matching results. Compared with the existing matching algorithms (such as SSD, SIFT, DP, BP, GC, etc.), the proposed matching algorithm can ensure the overall matching accuracy of left and right views, and has higher matching integrity. The implementation of the algorithm is that, the image spatial correlation of target image is firstly decoupled by coding, and then the L1 correlation constraint is calculated for the matched images. The combination of cyclic displacement matrix and random matrix is adopted here encoding the target image, which can make the coding result more robust. Meanwhile, the 3D measurement experiment of rail surface is done under dim light condition. The experimental results show that the matching integrity can reach 99.44%. Through the experiment, the 3D depth map of the target scene is obtained.