An Effective Particle Filter Tracking Varying Numbers of Multi-object

In the paper, we proposed an approach based on Bayesian framework to track varying number of objects using fixed camera. The approach is performed at both detection level and tracking level. At the detection level, a background-building algorithm is used to extract the spatial and color distribution of objects in a complex circumstance. At the tracking level, we used particle filter to track and label objects; to analyze the occurrences and probabilities of events such as continuation, birth and death, we update the correspondence matrix by matching features of object. We experiment the proposed approach on cars in highway video sequences, and verify the effectiveness and reliability of the method.