3D Point Cloud Matching without Projection Constraint based on Gray Gradient Optimization

Abstract

Stereo vision measurement is the main method of 3D reconstruction. However, it relies on projection methods, such as structured light projection, grating projection, phase shift projection, to achieve point matching in different shooting positions, which is not conducive to portable application. In this paper, the idea of 3D reconstruction without projection constraint was proposed, which can be realized by the process of mark points recognition and matching, feature edge extraction, matching region division, and point cloud matching. In order to improve the accuracy of point cloud matching, the matching algorithm based on gray gradient optimization was adopted. To verify its effectiveness, the checkerboard with $5*7$ squares was used in the 3D reconstruction experiment, the proposed method was compared with the interpolation method. The results show that our method outperformed the interpolation method in flatness ($\lt 0.06 {\mathrm{mm}}$). Moreover, it can provide a reference for stereo vision measurement without projection constraints.