| Target tracking is a challenging task in the field of computer vision. The subject involves a lot of advanced technologies, including artificial intelligence, image processing, pattern recognition and automatic control. The technology has broad application prospects in many fields, including the defense guidance, intelligent transportation, intelligent security, medical diagnostics, visual navigation and so on. For a variety of factors, object tracking accuracy, real-time and robustness can not be guaranteed.Existing tracking algorithms can usually be used in a specific environment. Some key object tracking technologies are studied and explored in this paper.The major research work of this paper is as follows:(1)Common target tracking algorithms including tracking method based on moving target detection, Mean Shift method, Kalman Prediction method, Particle Filter target tracking algorithms are studied, and the merits and shortcomings of the common algorithms are analyzed.(2)A robust multi-feature model fusion method is studied and designed. A feature model quality evaluation function is introduced to predict the corresponding weight and integrate the various features of the models. The accuracy and robustness of target tracking is improved to some extend by using this method of the adaptive multi-feature model integration.(3) In order to improve the time efficiency, particle number and particle radius selecting method is proposed in tracking process. This method can effectively improve the tracking in real-time.(4)A target tracking algorithm in video scene is proposed in this paper. The credibility of the model dynamic updates feature model likelihood and weight dynamically.Based on likelihood the algorithm updates the number of particles and proposal density. The proposed algorithm improves the tracking precision and real-time with more robustness, contrasting to traditional tracking algorithm,and the algorithm can adapt to the larger range of the video scene. |
PDF Full Text Request