A robust and real-time algorithm for human face tracking using improved particle filtering

In view of the problem that face tracker based on particle filtering using only histogram cue is frequently disturbed by background, a particle swarm optimization particle filtering(PSOPF) face tracking algorithm is proposed. An AdaBoost classifier is used to initialize the target tracking and update the template. To solve the problem of degeneration, the distribution of particles is optimized by PSO. Experimental results show that the proposed algorithm can track the human face steadily and be robust to the rotation of face, illumination changes, background interference and partial occlusion. The demand for general real-time performance(30 fps) can also be satisfied.