Robust proposal distribution for adaptive visual tracking in a particle filtering frame work

Different techniques are available in the literature for target tracking in video sequences. We focus our attention mainly on particle filter because of its power and versatility. We propose a metric named resampling force which measures the effectiveness of resampling stage. Resampling force represents how much the best old particles are regenerated after resampling. Furthermore, we extend the basic particle filter by developing a stage named compensation which acts before resampling stage. It controls the concentration of particles around the local maximal. Experimental results over a set of real-world video sequences demonstrate better performance of our method compared to basic particle filter.