Appraisal of an enhanced Particale Filter for object tracking

An enhanced particle Filter (PF) is introduced for object tracking. In this work, a new likelihood model is proposed. It depends on multiple of likelihood functions: position likelihood; gray level intensity likelihood; and similarity likelihood. Also, it combines information about the tracked object to get a robust and an accurate tracking performance. The proposed enhanced PF is implemented and evaluated. Its results are compared with a single likelihood function PF tracker, as well as, a correlation tracker and an edge tracker. The experimental results demonstrate the superior performance of the proposed tracker in terms of accuracy, robustness and occlusion compared with other methods.