Contour Processing&System Platform Implementation For Motion Capture

In recent years, video-based human motion capture has been widely applied to entertainment industry such as animation, games, movies and public life such as virtual reality, human behavior intelligent monitoring for the rapid development of the computer vision and artificial intelligence, which makes the research of video-based human motion capture become a hot area. It’s very important to research on it and protect cultural heritage such as national dance is one of our goal in this paper.The existing motion capture system can be summarized as mechanical, electromagnetic, acoustic and optical, but there are some shortcomings of the existence of motion capture systems such as expensive motion capture equipment, dependence on environment and performers’ unnatural movement, of which, the unnatural of movement is fatal shortcoming for the motion capture. Video-based human motion capture systems studied in this paper can overcome these shortcomings.This paper first introduces the general situation of human motion capture, mainly including the research background, methods, present situation and existing results. This paper, combined with the goal of cultural heritage protection, mainly research on video-model-based human motion capture system without mark, we focus on part of the problems of it, basically including the visual bottom processing which contains human contour extraction and the human edge detection, projection of3D model edge to the planar and platform implementation of human motion capture system.In this paper, background subtraction based on the Gaussian mixture model is applied to extract contour, in which we improve the background update model, deal with the affect of shadow and increase the continuity by recovering missing detection. In order to overcome the dependence of environment for background subtraction method, we presented a new way to extract human silhouette based on HOG. False edge is removed by using the method of SUSAN. Demo platform is built based on OpenCV and OpenGL, showing out every step and corresponding result of the human motion capture system, which includes image and video processing, dealing with BVH file, multithreaded programming and Windows message handling.