A method of 3D human-motion capture and reconstruction based on depth information
Wei Quan, Shan Jiang, Han Cheng, Yu Zhang, Bai Lijuan, Haimei Zhao · 2016
During the past decades, motion capture has become a very important role in computer animation. However, real-time, anti-interference and robust motion tracking of video sequences is still very challenging. Kinect is a popular motion sensor which can obtain human motion data. But when taking into practice, there are no adequate implementations on how to extract skeleton data from the depth image, bind skeleton model with motion data. In this paper, we propose a method for human motion capture based on the characteristic of Kinect2, made an improvement based on the traditional method. Finally, we develop a 3D human-motion capture and reconstruction system. A detailed procedure of implementation is demonstrated. Experiments have shown that motions of virtual character are controlled by the real person completely and accurately, and the system has advantages of real-time, low-cost and high accuracy.