Summary: Particle filter is a sequential Monte Carlo method in a recursive Bayesian filtering framework. The efficiency and accuracy of the particle filter depends on two key factors: how many particles are used and how these particles are re-allocated. We estimate the number of required particles by Kullback-Leibler distance (KLD), which is called KLD-sampling. Meanwhile, we employ a hybrid dynamic model to generate diversified particles, which suits object's agile motion. Besides, we employ the mean shift analysis as a local mode seeking mechanism to make each particle more informative?. We demonstrate this proposed algorithm performance in tracking the ball from sports video clips.. Details are given in paper ICME'07.
