Multiple Non-rigid Objects Tracking by Modified Kernel Particle Filter

This paper proposed a new method for multi-target tracking in video sequences by combining two trackers, sum-of-squared differences (SSD) and kernel particle filter (KPF). In our work, the idea of Object Likelihood Value of pixel is proposed. Instead of using direct propagation resample result from the previous sample set, a weighted SSD displacement is used for reinitializing and resample before next KPF iteration. Experiment results on soccer athletes tracking show that the combination of SSD with KPF tracker forms a simple and powerful multiple non-rigid object tracking system.