Visual tracking system using an ID-tag and the network

We propose a visual tracking system that downloads object models from the network. A visual ID-tag is attached onto each object, and the system identifies the object by recognizing the tag in the image. In order to realize robust recognition, we design a circular tag made of a retroreflector, and mount a lighting system that illuminates objects in sync with the camera. Observing the change of pixel values through several frames, the proposed system extracts the tags from the image. The vision system estimates the initial pose and position of the object and downloads its CAD data based on the extracted ID-tag. Then, the system visually tracks the object based on the 3D CAD model.