A visual tracking system using an RFID-tag

We propose a visual tracking system that identifies an object in front of the camera using an RFID-tag, which is put onto each object. Identifying the object, the system downloads the CAD model of the object via the Internet, and uses the model for visual tracking. The visual tracking consists of two procedures: (1) the estimation of the initial pose and position relative to the camera, and (2) the tracking of the motion between two consecutive images. In the former procedure, the system detects the shape component from the initial image, and estimates the pose and position with the shape of the component. In the latter procedure, the motion is estimated using both the CAD model and the difference of the two images. The proposed system shows the framework that enables a vision system to handle enormous number of objects.