Stereo matching method of distribution box hole image based on improved AD-Census algorithm

Abstract

In mountain tunnels’ mechanical and electrical engineering, it is necessary to carry out cable distribution plates. To ensure the accuracy of the cable distribution plate, it is essential to measure the distance between the distribution box hole and tunnel mouth and the size of the distribution box hole. Stereo matching is a crucial step in using binocular vision for measurement, and the quality of the matching effect directly affects the measurement accuracy of the following steps. This study focuses on the measurement of tunnel distribution box hole size and studies the stereo-matching algorithm of distribution box hole images. AD-Census, a semi-global stereo-matching algorithm, is selected as the main stereo-matching algorithm in this paper, and the stereo-matching experiment is carried out. According to the effect of stereo matching, the AD-Census algorithm is improved, and the MSD-Census algorithm fused with edge information is proposed to obtain better disparity information, which lays a foundation for the subsequent measurement of distribution box holes.