Stereo vision reconstruction method based on common perpendicular constraint

Qiaoyu Xu,Huai Yao,Xin Qi,Dong Ye,Rensheng Che

DOI: https://doi.org/10.19650/j.cnki.cjsi.2008.11.036

2008-01-01

Abstract:To solve the reconstruction problem of stereo vision measurement, this paper presents a method of stereo vision reconstruction based on common perpendicular constraint. The proposed method consists of two processes-, firstly, by minimizing the distances between the space point and two projection rays, determining the constraint that the space point is on the common perpendicular of the two projection rays; then taking the middle of the common perpendicular section as the iteration initial value, using Levenberg-Marquardt algorithm, the optimum solution of the space point is obtained by minimizing the error between the imaged points and the space point projections. The method not only minimizes the distances between the space point and the two projection rays, but also eliminates the effect of the image noise and the distances between the space point and the two cameras. The method can quickly and effectively converge to the optimum solution of the space point. Both computer simulation and real measurement data are used to compare the proposed method with other common methods, the results show that this method is more precise than other methods, and the measurement accuracy reaches to 0.023 mm within 0.04 pixel image noise.