3D Human Motion Capture And Body Model Measurement Us-ing Kinect

3D human body measurement is one of the hotspots in the cross-field of comput-er vision and computer graphics. It has a wide range of applications in health care, security, electronic business and entertainment. Currently, mature 3D human body measurement systems based on LiDAR or multi-view cameras are often expensive and complicated, which limits its growing applications in our daily life. Because of the low price and user-friendly design of depth camera, researchers are paying more and more attention to 3D human body measurement algorithms based on depth camera. How-ever, state-of-art methods based on depth camera usually cannot handle the dressed human body with large scale of motion.Nowadays, academia has seen a rapid development of motion capture technology, which can provide 4D human body reconstruction in both temporal and spatial do-main. Based on multi-layer human motion filter, this thesis introduces a novel method of measuring human body models with large scale of motion. Besides that, this thesis provides an efficient method to eliminate the effect of clothes deformation. The exper-iments of our system show that our human body measurement algorithm gives more accurate result than state-of-art methods.Also, in this thesis, we formulate the problem of depth image denoising as param-eterization of quadratic surface, and recover accurate and smooth local structure from original depth image. While handling the 3D correspondence, we introduce a novel spectral technique to keep the global consistency of two sets of 3D point clouds, which efficiently improves the result of non-rigid registration.