Visually guided 3-D animation system using real-time marker-less motion-capture with multiple-camera fusion

A realtime marker-less motion-capture system is described that can easily and seamlessly map objects in the real world into a virtual environment. In general, virtual environment applications, such as man-machine seamless interaction, require the system to estimate the motion parameters for natural objects such as human bodies in realtime. To achieve this, the developed motion-capture system achieves multiple-camera fusion and reconstructs the parameters for complete humanbody motion using real-time inverse kinematics. Implementation of this system demonstrated its ability to work in realtime and online on a pc-cluster.