Stereo matching algorithm based on binocular vision

Abstract

Binocular stereo vision technology has been widely used in target detection, unmanned driving and other aspects due to its advantages of low cost and wide range of application. It is a research hotspot in the field of computer vision. Binocular stereo vision is mainly to obtain 3D depth information of the scene by analyzing two or more views of the scene. Stereo matching is the core part of the scene, and the performance of the matching algorithm directly affects the effect of 3D reconstruction. This paper mainly analyzes and summarizes the local stereo matching algorithm, global stereo matching algorithm and semi-global stereo matching algorithm by studying stereo matching algorithms at home and abroad, expounds the characteristics and performance of various matching algorithms, and makes a comparative analysis of various algorithms.