The Design And Implementation Of Kinect-based Motion Capture System

Motion capture technology plays an important role in motion analysis and the production of special effects,CG animation,games and the other fields.The traditional motion capture systems still had lots of problems,though.For example,they were usually expensive,complicated to use,had poor realtime performance and were highly environmentally sensitive.These problems prevented them from being widely used.Nevertheless,with the development of innovative technology,Microsoft created the Kinect,a marker-less motion capture device which featured low producing cost and leading skeletal tracking algorithm,and had been put into widespread use.Nevertheless,some common issues still exist in Kinect-based motion capture systems,such as the motion data jitter and lack of the skeletal data.Another prominent problem is that the motion data can not be re-used.The major purpose of this research is to optimize the action data captured by Kinect,redirect the optimized data to the 3D character model so as to drive it to simulate the action of a real person,record and save the action script.On this basis,the Kinect-based motion capture system has been designed and implementedFirstly,the thesis focuses on the analytic study of the action capture and re-direction technology in terms of its present development in both domestic and abroad.A new overall design scheme of Kinect-based motion capture system is proposed following the analysis of the advantages and disadvantages of various motion capture systems and the system requirements.According to the MVC design pattern,the new system is divided into data acquisition layer,data interactive middleware,data processing layer and UI interface layer,in which the data processing layer is the key component composed of the skeleton data optimization module,the action redirection module and the action recording module.Sencondly,in terms of current issues of Kinect-based motion capture system,an optimized single joint repair algorithm is proposed to solve the existing problems;A missing-joint complement algorithm is implemented to deal with the continuous multi-joint loss problem;The skeletal motion smoothing algorithm is used to solve the joint jitter problem;The retargeting method based on the forward kinematics is used to solve the model driven problem;And the research of DAE model structure provides theoretical basis for recording and saving action script.Thirdly,the detailed system functions are designed and developed,including the user interface,the Kinect data acquisition module,the skeletal data optimization module,the action redirection module,the visualization module and the motion recording module.At latst,the thesis records the tests and analysis of all the proposals stated above in order to verify the feasibility and validity of the Kinect-based motion capture system.The results proves that the research work has reached the expected objectives.