Object Tracking and Shoslif Tree Based Classification Using Shape and Color Features

This paper proposes a video-surveillance system that is able to track and automatically classify objects detected in a scene. A common object representation based on corners is used for tracking and for classification. Recognition module is based on a self organizing decision tree (SHOSLIF) that is automatically generated by the system during a learning phase. Both color and shape related features are used for object classification. Presented results confirm the validity of the proposed approach.