An efficient AdaBoost tracking algorithm based on the particle framework

Object tracking is a classic problem in the computer vision field. Its main purpose is to get the locations, the motion parameters, the trajectories and other information of the objects from the videos. An Adaboost tracking algorithm based on the particle framework is proposed in this paper. Due to the existence of high noises in the training samples, the difficulty of the AdaBoost training and the possibility of the unsuccessful tracking would be increased to some extent. So we use the random distributed particles to avoid the tracking falling into the local optimum. And by combining some weak classifiers with weights, the strong classifier of the Adaboost is used to update the training template on line. The experiments results show that our algorithm improves the tracking accuracy significantly compared to the original AdaBoost tracking.