Objects detecting and tracking with a new particle filter
Wei Sun, Long Fei Chen, Long Ren, Baolong Guo, Xianxiang Wu · 2012
Robust real-time tracking of non-rigid objects is a challenging task. Particle filtering has proven successfully for non-linear and non-Gaussian estimation problems. The article presents the integration of mean shift vector of the moving object into particle filtering. Color distributions are applied as they are robust to partial occlusion, which are rotation and scale invariant and computationally efficient. An initialization mechanism of particles based on Gaussian distribution is introduced, and a dynamic transfer matrix is used because moving objects may disappear and reappear. Comparisons with the mean shift tracker and a combination between the mean shift tracker and particle filtering show the advantages and limitations of the new approach.