An improved color-based tracking by particle filter

Color has proven an efficient visual feature for tracking non-linear objects in real time using Particle Filter (PF). However, the points generated by the Monte Carlo(MC)random sampling often form the possible gaps and clusters in sample set which affect the tracking accuracy and speed in PF. To solve these problems, we propose an improved method for tracking color objects with the points generated by Quasi-Monte Carlo(QMC) sampling to replace the MC points, which could break the correlation of the original ones, and use the robust color feature to model the object to improve the tracking accuracy and efficiency. Simulation shows that the proposed method is superior to the traditional one, and improves the tracking accuracy and speed.