Face Tracking with Occlusion

The Camshift algorithm fails to track face easily while it is occluded, so a new face tracking method is proposed in this paper. This method combines the Camshift algorithm and the GM(1,1) model with optimized background values. By using moving vector information, this method can effectively track face even occluded by other static objects. The GM(1,1) prediction model will reduce the searching region of the Camshift algorithm and enhance real-time performance. Furthermore this model is not only suitable for modeling of low increase exponential sequence but also suitable for high increase exponential sequence, so it adapts to the characteristic of human´s free motion. With occlusion, this method can improve accuracy of human face tracking and enhance robustness of the tracking algorithm by replacing the real values with the prediction values containing prediction errors.