Three-dimensional reconstruction with single-shot structured light dot pattern and analytic solutions

Abstract Structured light three-dimensional reconstruction is one of the most important techniques in computer vision. However, most existing methods require multiple projections of the designed patterns to achieve the closed form solution, which makes them fail to measure the dynamic objects. In this paper, we propose a two-view based single-shot structured light 3D reconstruction approach that calculates the 3D coordinates of the object with analytical solutions. The designed structured light dot pattern is projected onto the object, distorted and imaged in two cameras. For each point in the structured light dot pattern, there is a ray of light that goes through the optical center and the physically imaged point on the camera’s image plane. Two rays could be determined by two cameras and their intersection solves the unknown point on the object in the 3D world coordinate system. To segment the dot pattern robustly and efficiently, an iterative dot segmentation method is proposed. Experimental results verified the effectiveness of the proposed approach in reconstructing objects without bright colors, rich textures, large discontinuities and obvious occlusions.