Monocular 3D Motion Capture and its Application to Augmented Reality

• Main Authors: Ting-Wen Chen, 陳鼎文
• Other Authors: Huei-Yung Lin
• Format: Others
• Language: zh-TW
• Published: 2009
• Online Access: http://ndltd.ncl.edu.tw/handle/03382088522960350910

碩士 === 國立中正大學 === 電機工程所 === 97 === Image-based 3D motion capture is a challenging research topic in computer vision. Since the video sequence only provide 2D information of the scene, many researchers used stereo vision or shape-from-silhouette techniques to obtain the depth for 3D human motion estimation. In this thesis, only a single camera with fixed position is used for data acquisition. It is not required for the users to wear any portable sensing devices such as gyro or markers. We present a method to match the silhouettes obtained from the video sequence and the image projection from a computer generated virtual 3D human model. By adjusting the 3D pose of the virtual 3D model iteratively based on the kinematic constraints of human motion and skin color detection, the human pose and the associate 3D motion parameters can be uniquely identified and the self-occlusion problem can be refined. Finally, we integrate the 3D motion capture system to the augmented reality application. The parameters obtained from the motion capture system are used for the interaction between the user and the environment through a head mounted display.