Modeling-Based Decoding of the Point Pattern for Active Stereo Vision

Abstract

Active stereo vision (ASV) computes the parallax and depth information from the coded structured light patterns. Thus, it could overcome the difficulties of measuring objects without textures and colors. Pattern decoding is used to obtain the indexed points with pixel coordinates from the acquired pattern image for the subsequent stereo matching. However, decoding of the structured light patterns at locations of color crosstalk, specular reflection and occlusion remains challenging. In this paper, we propose a pattern modeling method to decode the structured light point pattern robustly. The mapping model between the points in the designed pattern and the points in the segmented pattern is computed based on the natural neighbor interpolation. Based on the modeled points, the missing points in the segmented pattern are restored and the noise points in the segmented pattern are removed. The robustness is achieved in the sense of hundred percent point segmentation completeness. Due to the hundred percent completeness, the points in the corresponding blocks are matched directly according to their indexes. Experimental results verified the effectiveness of the proposed active stereo vision approach and also proved that the proposed pattern modeling method could solve the segmentation problems caused by specular reflection, color crosstalk and occlusion effectively. Note to Practitioners—The proposed approach tries to solve the segmentation problems caused by color crosstalk, specular reflection and occlusions for exiting point-pattern based 3D surface imaging techniques. The proposed approach utilizes the natural neighbor interpolation technique to compute the mapping model between the designed points and the segmented points robustly. Based on the mapping model, the missing points could be restored and the noise points could be removed. As a result, the robust segmentation results with all the projected points could be obtained. The segmentation problems caused by specular reflection, color crosstalk and occlusions are thus solved successfully. Although the proposed approach is developed based on our designed point pattern, it could also be easily expanded to other structured light point patterns. Currently, the proposed approach only works for the point patterns and much more work is still required before it could be applied to other types of structured light patterns, such as the line patterns or the phase patterns.