Three Dimentional Motion Analysis Of Human Body Based On Voxel Data

In the real world, most of the computer vision tasks are related to human activities, such as intelligent human-machine interfaces, virtual reality, advanced user interface, intelligent environment, entertainment, intelligent video surveillance systems, motion analysis, medication, education and so on. Currently, 3D human motion analysis is a hot while difficult problem in the field of computer vision. In the process of motion capture, the more the number of cameras is, the more precise 3D human figures are, the more the amount of data of each frame is. In addition, people obtain data of 3D human motion through motion capture more and more frequently, which is used as true value of the original in experiment.Therefore, after a lot of international related research, experiments were carried out both in multi-camera system and in motion capture system. In the multi-camera system, we processed the video-acquisition data to obtain the parameters of 3D human body skeleton. In the motion capture system, we analyzed 3D human motion data got by motion capture and we got the 3D human body skeleton. Both made good results. Our innovations are as follows.1. Give a standard skeleton model in line with human body topology. Experimental results prove that the skeleton model can represent the diversity and stability of human motion well.2. Propose an algorithm to find skeleton motion from voxel data in representation of human motions and gestures. Experimental results show that 3D human skeleton model can display human movement gesture efficiently.3. Present a probability evolutionary algorithm (PEA) to process parameters of skeleton model. The optimization function can maximize the match of the motions between the huaman skeleton and the voxel.4. Propose a method of 3D thinnning algorithm to the problem of human skeleton computation for human pose initialization. Experiments prove that the algorithm can give out the human skeleton parameters automatically from voxel data.5. Give an algorithm to compute the pose of 3D human skeleton from the data of motion capture system. The experimental results show that the method is robust and effective for wide range of human motion.