The Research of Depth Perception Method Based on Sparse Random Grid

Abstract

In this paper, we propose a high-resolution depth sensing method based on structured light. In 3D contour scanning, passive binocular stereo vision is difficult to obtain enough 3D information for objects with inconspicuous surface features. To solve this problem, based on the binocular stereo vision principle and the structured light projection method in active vision, a method of obtaining sparse depth based on random mesh is proposed. Four templates are projected onto the surface of the object, which are random meshes and three templates with phase difference. In addition, the relative phase image is calculated according to the three-step phase-shifting mode. Finally, the depth map calculated in the conventional structured light method. In this paper, we demonstrate the effectiveness algorithm to show that our depth sensing are more accurate and resolution than the existing methods in the experiments.