Study On Face Tracking Based On Sequence Monte Carlo Filtering

Human face can provide much valuable information in the study of safe monitoring,perceptual human-computer interface and compute vision.Therefore many scholars have been attracted to study the human face detection and tracking, and many algorithms have also been proposed.Although the multitudinous results have been harvested, yet the following questions have not been solved effectively: timeliness, robustness and cover processing. In this paper we make some researches on those problems above and propose a set based on the Monte Carlo filter human face tracking algorithm, which fully considers the non-rigid characteristic of face and the complex environment of tracking face. The main contents of the paper are summarized as follows:A skin color and the spatial information based real-time face detection algorithm is presented. After analysis the characters of the human face, we draw a conclusion that the performance both in fast and accuracy can be obtained through suitable character selection and using essential technology such as matching method and the movement prediction. Experiments results demonstrate the algorithm is effective under complex situation such as translation, incline, out-of-plane rotation of face.Combined the merit of the above matching algorithm, a sequence Monte carlo filter based face tracking algorithm under complex environment is provide. Monte Carlo filter can deal those nonlinear and nongaussian estimation problems,and what's more, it doesn't need the measurement equation which is hard to build in face tracking. Experiments results demonstrate the algorithm is effective under complex situation such as changing background and partial occlusion. Finally, the validity and the feasibility of the algorithm are analyzed, and the further research direction is proposed.