Mean Shift And Particle Filter-based Video Object Tracking Algorithm

Target tracking is a hot issue in vision field, which is widely used in video surveillance, bio-medicine, inertial guidance, navigation system and so on. While the target is moving, the disadvantage factors have a serious influence on the target tracking, such as the deformation, the interference of the complex background and various types of noise, shelter and light. Under the real complex condition, improving the accuracy, robustness and real-time performance of the tracking algorithm is important in the field of target tracking. This paper focuses on the application of Mean Shift algorithm and Particle Filter algorithm in the target tracking. The main studies are showed as follows:To the condition of changing target's scale, sheltered target and changing light, an improved Mean Shift algorithm is introduced. Firstly, the adaptive bandwidth Mean Shift algorithm is adopted to track the target effectively whose size is changing. Secondly, an adaptive filter Mean Shift algorithm based on LMS filter and Kalman prediction algorithm is put forward, which enhances the robustness of the algorithm under the condition of the deformation target and the sheltered target. Thirdly, CamShift algorithm based on the HSV color model is introduced. The motion prediction is adopted to enhance the adaptability of the track algorithm under the condition of changing light and sheltered target.Correlative tracking strategy and multi-feature fusion is adopted to improve Particle Filter. With some correlative tracking strategy, the correlative tracking method based on Particle Filter is accomplished to verify the robustness of the Particle Filter. In addition, an observation model is built to combine with color features and texture features as well as the scale variables is added in dynamic system model. Then a multi-feature fusion based on Particle Filter is introduced to improve the robustness of the tracking algorithm.At last, the search strategy is improved based on the combination of Mean Shift and Particle Filter. By the clustering effect of Mean Shift which muster the particle to the real target area, the Mean Shift algorithm is embedded into the Particle Filter algorithm for target tracking. Moreover the Bhattachyarya coefficient is used to measure the situation of target tracking, furthermore, the performance of tracking algorithm is improved by combining with the search strategy and the improved Mean Shift and Particle Filter combination algorithm.