Improved anti-occlusion tracking algorithm based on face detection

Abstract

An improved anti-occlusion tracking algorithm based on face detection is elaborated in this paper. The Kalman filter and meanshift algorithm along with the coefficient of Bhattacharyya being used as a parameter for occlusion discovery consists of the framework of the base of this tracking system. Only occlusion takes place, will the face detection be involved in addition to the basic algorithm in the algorithm in even frames while the same pattern is implemented in odd frames as the none-occluded situation. Meanshift module is used without the interference of occlusion and face detection takes effects on the contrary in the way that linear prediction is provided by Kalman filter in both situation. The algorithm in this paper, testified from the experimental results, can effectively offset the instability brought by occlusion and recovery the trajectory of the objective instantaneously.