Abstract
In this thesis, a real-time Human In Virtual Environment (HIVE) system is developed that can track and analyze human body motion in a real environment and then simulate a synthetic actor in virtual environment. Here we propose a model-based method to analyze the real human motion, and then synthesize a virtual actor performing the same action. This system consists of four phases: segmentation, 3D human model generation, human motion analysis and virtual actor emulation. In the segmentation phase, we generate the background model first, and then subtract the image sequence from the background model to extract the foreground object. In 3D human model generation phase, we analyze the foreground image and estimate the shape parameters of the 3D human model such as the length and radius of limbs. In human motion analysis phase, we change the motion parameters of the 3D human model and then project the 3D human model onto 2D image plane to match the 2D model with the foreground human body silhouette, so that we can estimate the motion parameters. Finally, we create a 3D virtual actor by computer graphic OpenGL.