Algorithm for Three-Dimensional Reconstruction of Nonrigid Objects Using a Depth Camera

摘要

—An algorithm for 3D reconstruction of objects with nonrigid shape using an RGB-D depth camera is proposed. The algorithm can be used in medicine, agriculture, robotics, virtual reality, and human–computer interaction. The proposed algorithm makes it possible to accurately reconstruct a 3D object with one depth camera without restricting camera movement and without using a priori information about an object shape. The reconstruction process consists of the following steps: input of information using an RGB-D camera, registration with a modified iterative closest point algorithm, and dynamic construction of a dense 3D model of objects. The efficiency of the proposed algorithm is evaluated using experimental data and is compared with the modern methods of registration. The results show that the proposed algorithm can accurately reconstruct 3D nonrigid objects on complex scenes with one depth camera.