Tracking moving vehicles using particle filtering and level set

The paper presents a new framework for tracking moving vehicles in image sequences. Firstly, the Gaussian particle filtering algorithm was used to track the objects, and the external rectangle contour could be obtained. Secondly, based on the centroid of the external contour, the level set function without re-initialization is introduced to GAC model, curve evolution was performed to get the precise contour of the objects, and the extract results were fed back to the tracking frame; lastly the likelihood function was improved by updating the reference template dynamically. The experiment results show that the proposed method got the better tracking results in tracking moving vehicles for actual traffic environment.