Research On Particle Filter Tracking Based On Improved Mean-Shift Algorithm

Video object tracking is the critical technology of computer vision and the hotspot of the certain research domain. As a very complicated problem, object tracking problem relates to many aspects of computer vision research. In this paper, the object detection and tracking of the moving objects and relative problems in the video are mainly studied. The objective of this paper is to optimize the selection of the initial target and increase the precision of the tracking contrail by the improved algorithm of object tracking, which is called the algorithm of the particle filter based on improved Mean-Shift. The main research achievements are as followings:Firstly, the status quo and research methods of object tracking were analyzed and summarized. The common algorithms of target detection are introduced, and the method of mixed-Gauss background model based on color-background and color- gradient is adopted.Secondly, the method of traditional Mean-Shift was improved. The selection of initial object was achieved based on improved Mean-Shift. Simulation results testified the effect of improved artificial potential field. Simulation results testified the effect of improved artificial potential field.Thirdly, based on analyzing Particle filter theory, the application Particle filter algorithm during object tracking was determined. According to the retrogression of single Particle filter algorithm, the Particle filter algorithm would be combined with Mean-Shift algorithm. The diversity is maintained and the time is reduced due to litter particle.Finally, the experiments based on object tracking were done, and compares between traditional method and improved method were obtained. A series of experiment results and data analysis show the method's validity and practicability.