Adaptive visual tracking of moving objects modeled with unknown parameterized shape contour

The paper presents a new approach to establish a visual tracking system which can automatically detect the shape contours of moving objects, extract their shape parameters, and continuously track locations of these moving objects in the shape parameter space. The system consists of a robust shape detection algorithm based on the randomized Hough transform (RHT) method and an adaptive object tracking algorithm. The paper demonstrates that, through the combination of a motion detector and the RHT method, the resultant shape detection algorithm is computationally efficient and is robust to noise and object occlusion. On the other hand, the tracking algorithm tracks the shape contours and updates the shape parameters corresponding to the translational, rotational, and scaling motion of moving objects. The tracking algorithm uses a decomposition of the parameter space into lower dimensional subspaces for computational efficiency. Finally the paper provides several simulation and experiment results to validate the new approach.