The Research On Video Target Tracking Algorithm Based On Particle Filter

Video target tracking is the core issue in the field of computer vision. It has been widely applied for the military and civil purposes. With the progress of the relevant research in recent years, a variety of effective video target tracking algorithms have been proposed. However, the video target tracking has been confronted with many challenges in the practical application, such as the complicated background, illumination change,similar-object interference and target occlusion. Hence, it has been a challenging research topic to design a robust video target tracking algorithm.From the perspective of the filter improvement and target appearance model design,this paper will propose some improvement methods. The main content of this paper is presented as follows.Firstly, this paper will elaborate on the basic theories and algorithm process of particle filter. It will also analyze some problems existing in the ordinary particle filter.Secondly, aimed at the problem of particle degeneration, this paper has proposed that the firefly algorithm can be utilized to optimize the sampling process of particle filter. On the strength of the optimum-seeking feature, the firefly algorithm can make the particle lean toward the high-likelihood region before the weight update, thus effectively mitigating the particle degeneration. In order to address the issue of particle diversity loss,this paper has designed a new resampling method. The new method can not only improve the particle diversity but also enhance the resampling efficiency.Moreover, this paper has also analyzed the state transition model and observation model used in the video target tracking. The improved particle filter based on the firefly algorithm will be applied to the video target tracking. By solving the problem of the particle failing to completely cover the proximity of target location for the non-use of observation value, it has improved the robustness of the tracking algorithm under the complicated background. Finally, the proposed methods are tested by using standard video sequences, and the results verified the validity and superiority of the algorithm.In order to solve the problem of tracking instability in the event of target occlusion and illumination change for the simple video information, it has designed a trackingalgorithm based on the multi-feature adaptive fusion. The multi-feature information, such as color, texture and edge, can be utilized to describe the target. The democratic fusion strategy has been used to integrate a variety of information. It can also enable the tracking algorithm to adjust the multi-feature weights to the current tracking status adaptively, thus realizing the complementary advantage between information.Lastly, the standard video sequence has been utilized to test the improvement method.It is concluded that those algorithms are effective and excellent.