A Novel Algorithm for Object Tracking with Particle Filtering and GVF-Snake

A novel adaptive object tracking algorithm with particle filtering and improved GVF-snake is proposed in this paper. Firstly, the initial contours of objects are obtained by using the background differencing method, and through the powerful searching ability of GVF-snake, the snake can be converged to the real contours of objects; by using the increasing or decreasing contour point algorithm according to the distances among them, the proposed algorithm can reach the purpose of tracking moving and deformable objects adaptively. Secondly, through the combination of the particle filtering and improved GVF-snake, an energetic particle filtering (EPF) object tracking algorithm is proposed; the tracking tactic for partially occluded objects is also proposed to overcome the effects such as shadows and occlusions in a complex environment. Experiments show that the proposed algorithm can fully utilize the contour information of partially occluded objects in the entire tracking process, and obtain better tracking results even though the tracked object is occluded for long time.