Research On Technology Of Feature-Based Video Motion Capture

In recent years, motion capture technology has been used frequently in the domain of compute animation, and the influence of it has become larger, so it has been receiving increasing attention from lots of researchers in many fields. But these existing motion capture systems suffered from many constraints, include requiring the actors to wear special clothing or to attach visual or magnetic markers to their bodies. Besides these, the price of motion capture devices is too expensive. These constraints restrict this technology to further application.To solve this problem and remove these restraints, we want to extract human motion information from video directly, and make use of these information to generate motion capture data file. It is one of the hottest research issues in computer vision filed in recent few years. Now, it has a comprehensive application in the analysis of athlete performance, human-computer interface, virtual reality and video animation etc.The research of this thesis will implement a simple motion capture system, or a substitute of existing motion capture systems. It doesn't need any auxiliary devices. Through it, we can implement human motion measurement, record and recovery. In this paper, we will integrate motion capture and human motion analysis technology together. The key issues, such as human motion tracking and 3D motion recovery, are deeply studied. To verify our thoughts, we develop a video based human motion capture prototype system at last, it can extract human motion information from a video clip with fixed background, and generate a motion capture data file to implement the process of human motion capture. The experimental results demonstrate the feasibility and practicability of our approaches.