Dynamic Planning Stereo Matching Study Based on Improved SGBM Algorithm
Yimou Li, Guo-Ping Fu, Yang Gao, Lei Wang, Wei Wang · 2024
This paper describes an improved stereo matching technique aimed at achieving high-precision non-contact measurements of stationary objects. The method is based on the SGBM (Semi-Global Block Matching) algorithm, which utilizes a binocular camera to calibrate the camera to the object and stereo-correct the image. The SGBM algorithm, based on the Census transform, is employed to match the stereo-corrected binocular images, generate parallax maps, and determine the 3D coordinate points. By utilizing parallax data, the depth information of the plane is calculated, and a dense point cloud of the image is created. The relative error of the binocular vision system developed in this paper ranges from 0.54% to 1.30%, confirming the algorithm's excellent accuracy and stability, and providing precise data for applications that require high accuracy.