Recognition of a Moving Object in a Stereo Environment Using a Content Based Image Database

We have developed a two-camera system that is capable to detect a moving object in the workspace. After noise removal and preprocessing the disparity map and the 3D model is produced. This detected moving object and its most important features are forwarded to a content based retrieval system that finds the most similar stored objects in the database.
The found object is surrounded by the smallest rectangle or the convex hull. To do this the feature points had to be found by the Susan and Harris type corner detection algorithms.
As we have used two cameras, the 3D description of the objects are also given by the disparity map. Considering the literature various CBIR systems have been developed, ours is a special one that can fruitfully cooperate with the stereo system.