Motion Tracking Based on Vision on Augmented Reality

One of the key techniques for augmented reality is registration. In order to overlay virtual objects on the real word at the right positions and orientations, the accuracy of registration is of most importance. In most of the current augmented reality systems, the real objects are static,thus little attention is paid to dynamic object tracking. An algorithm is proposed for estimating 3D motion parameters of a real object using optical flow of markers. According to the motion equation of 3D rigid body and the pinhole model of the camera, the relative position and orientation between the camera and the dynamic object can be obtained. The algorithm has a simple structure,requires only a single CCD camera and a few markers with different colors. It effectively extends the application areas of augmented reality systems.