Abstract:

In order to meet the real-time requirement of shape matching in target tracking process, this paper proposes a method of regional shape extraction using a cluster of concentric discrete circles (CCDC). The main idea is to locate the object’s shape center on the center of CCDC, map the shape on the CCDC, and calculate the eigenvalue of each circle and then describe the shape with an eigenvector. The advantage of this method lies in that it is fast in computation and less affected by the object shape and size. Besides, it preserves the properties of translational, scaling and rotational invariance, with a computing speed of 10 times of the moment  invariants and 300 times of the Zernike moments. It can well meet the requirements of real-time in tracking process.