Human Motion Capture Based On Binocular Vision

In recent years, the human motion capture technology is widely used in various fields, which has been caused more and more attention by scholars. Therefore, the study of motion capture technology has been become a hot topic at present. However, traditional motion capture methods not only need to use expensive equipment, but also the sensors that attached to the body influence the free movement of human. Thus, a novel human motion capture method based on binocular stereo vision is proposed in this paper. The basic idea of the proposed method is to use the theory of computer vision to restore the3D human motion skeleton information directly from the2D video sequence, which can save the cost of motion capture and compensate the defect of traditional methods. This paper has done the following three research work:(1) Firstly, the fundamental theory of binocular stereo vision technology and motion capture technology are illustrated in detail, which describes the basic theories and framework of binocular stereo vision measurement system, in the following the motion capture and related technologies are elaborated;(2) Secondly, the whole technical route of human motion reconstruction based on multi-view video is discussed, which the key technologies include multi-camera calibration, markers extraction, markers tracking and3D reconstruction. In the procedure of tracking markers, we propose a novel multi-view tracking algorithm. This algorithm is made up of two steps, one is binocular stereo vision tracking which combines the Extended Kalman Filter with the smooth of markers’trajectory, and another is multi-view data fusion.(3) Finally, the implementation of human motion capture system based on binocular stereo vision is analyzed, which contains image acquisition, binocular camera calibration, binocular rectification and stereo matching. In the end, we realize human motion capture system from binocular video by combining these technologies with the related technologies of motion reconstruction.Through the analysis of experimental data errors show that the error of camera calibration is small. Results of binocular rectification and stereo matching also can meet the requirements of experiment.3D skeleton motion capture of binocular video in the Matlab environment is also desirable. Therefore, the effectiveness and feasibility of the proposed algorithm is proved in this paper.