Directional-edge-based object tracking employing on-line learning and regeneration of multiple candidate locations

An object tracking algorithm employing on-line learning and regeneration of multiple candidate locations has been developed. By introducing a directional-edge-based feature representation of images, being inspired by the biological principle, the system is robust against illumination variation. In order to further enhance the performance, an on-line learning technique and a statistical multiple candidate locations approach have been developed. As a result, the system is also robust against object size variation, partial occlusion, and object deformation. The performance of this algorithm has been verified by experiments performed under varying disturbing circumstances.