A Trajectory-Based Point Tracker Using Chaos Evolutionary Programming

A trajectory-based point tracker using chaos evolutionary programming (CEP) algorithm is proposed in this thesis. While motion constraints such as rigidity and small motion which are imposed by previous approaches are liberated, the proposed CEP is proved to be effective for establishing point correspondence between two consecutive frames sampled at a fixed interval. The whole point trajectory within the sample interval is then reconstructed by polynomial interpolation. Our experimental results demonstrate that the proposed point tracker can accurately locate target under different kinds of situations like object deformation, occlusion, and sudden motion as well.