The Research Of 3D Human Motion Capture Based On Reference Points

Technology of Three-dimensional Human Motion Capture (HMC, Human Motion Capture) has wide range of applications in animation, film special effects, computer games, video surveillance, virtual reality, intelligent human-computer interaction, clinical rehabilitation, the robot moves design, sports training and other areas and in recent years the field of computer vision is to become one of the most active issue.This paper present a thesis of Human Motion Capture based on reference points tracking. This thesis solved the problems of tracking human and capturing three-dimensional human motion in videos and investigates on tracking the articulated motion with reference points and evaluating the three-dimensional motion of the figure in the form of articulated skeleton chains knowing the camera parameter in monocular videos.First of all the paper complete the calibration of matrix of intrinsic and extrinsic parameters based on the classic pinhole camera model, and then elaborated the basic principles of particle filter. The paper actualize particle filter tracking algorithm based on weighted color histogram and applied it on two-dimensional human joint motion capture system based on reference points. The tracking algorithm proposed above takes"joint + region color"as it's match features which make the tracking simple and matching accurately. After having the two-dimensional data of joints, the paper realizes the algorithm of the three-dimensional human capture based on restrict of human body structure and elaborated derivation of the algorithm.At last the paper does the error analysis for the algorithm proposed in the paper. The experiment proved that the algorithm can analysis three-dimensional motion of human body quantify and can be used to provide the standards for other three-dimensional reconstruction system without markers.