Registration of 3D Reconstruction Point Clouds Based on the Assistant Cameras Array System

Three-dimensional reconstruction plays an important role in the real practice. Since the three-dimensional reconstruction technique based on structured light is seriously restricted by the light conditions, the three-dimensional reconstruction technique based on passive optical binocular vision is widely applied with the superiority of unlimited by the environment conditions. However, the precision of the reconstruction is not satisfying. The point cloud registration technique of three-dimensional reconstruction can greatly improve the drawback, but the traditional ICP algorithm endures a low iteration speed owing to large iterations and the precision is still unsatisfying. This paper presents a point cloud registration technique algorithm of three-dimensional reconstruction based on assistant camera array system, which creatively takes advantage of run length encoding to do the corner recognition, and can acquire accurate registration results without iteration. Experimental results indicate a greater improvement in accuracy and speed compared with ICP algorithm.