Research Of Maneuvering Target Tracking Algorithm Based On Nonlinear Filtering

The maneuvering target tracking has promising foreground, and has widely application in civil and military fields. Many experts in-country or abroad have paid a lot of attention on this topic. They have done deeply research on it and acquired a lot of important achievements. The problem of the maneuvering target tracking is to estimate target movement precisely as targets is maneuvering. Since the variety in the tracking environment and unknown maneuverability of the targets, more complicated requirements have been demanded in many applications to the problem. Therefore, based on the achievement of the former's researchers, this article has been done is amore deeply and more systemic manner.Firstly, the paper introduces the constituent elements and main content of the target tracking, and summarizes the process of foundation of target tracking model. Some important non-linear filtering algorithms are researched and implemented. Secondly, the maneuvering target tracking and adaptive filtering algorithm, which are based on current statistical model, are studied. The interactive multiple models algorithm is described, the non-linear filtering algorithm are used in the two algorithms shown above. At last, according to the superiority of the two algorithms, a modified interactive multiple models algorithm based on current statistical model is presented. It can adjust adaptively the extreme value of accelerator, and can decrease the difficulty of the extreme value's selection.Sequentially, combining with a new achievement, a maneuvering target tracking algorithm for the unknown target number which based on Rao-Blackwellized particle filter is researched. This algorithm can track the targets effectively and can estimate the number of tracks at the same time. However, because of the effect of the parameterβ, when there are a few particles in the process of filtering, tracking results may have a long time delay. To solve this problem, the paper introduces an improved algorithm. We decrease the value ofβ, and connect the tracks which may be the same track. Using this approach, the estimation number of targets become more precisely without the increase of particles at any time stamp.