Research On Multi-target Tracking Method In Clutter Environment

With the continuous development and improvement of radar technology and equipment,radar target detection and tracking systems have an important position in my country's national defense industry.Accurately discovering all targets in the surveillance area and predicting their trajectories in real time is one of the key research directions in the field of radar data processing.Aiming at the problem of stable tracking of multiple maneuvering targets in a clutter environment,this thesis improves the algorithm in two aspects: data association and tracking filtering.The main research contents are as follows:1.First,the basic framework of multi-target tracking is introduced.The basic principles of traditional joint probabilistic data association and joint integrated probabilistic data association algorithms are introduced respectively.After that,the commonly used target motion models and tracking filtering methods: Kalman filtering and Smooth Variable Structure Filter are introduced.Through simulation experiments,the robustness of the smooth variable structure filter is verified.2.Aiming at the combinatorial explosion problem that classical data association algorithms are prone to appear in multi-target and clutter-intensive environments,this thesis proposes a joint integrated probabilistic data association method based on association hypothesis network.This method first generates a target list through a heuristic clustering method,then builds an association hypothesis network,splits the likelihood matrix,and uses some joint events to calculate the association probability.Aiming at the problem of repeated subtrees in the process of splitting the confirmation matrix in traditional data association algorithms,this algorithm merges them to avoid partial redundant association weight calculations and reduce unnecessary calculations.Finally,it is verified by simulation experiments that compared with the classic Bayesian data association algorithm,this method improves the tracking accuracy of the system,has better real-time performance,and can quickly and correctly perform data association in a clutter environment.3.In the target tracking system,due to the uncertain parameters,the filtering algorithm of the Kalman system is prone to divergence when the target is maneuvering.In order to improve the robustness of the tracking method under maneuvering conditions,this thesis proposes a maneuvering multi-target tracking algorithm based on an improved smooth variable structure filter.The traditional smooth variable structure filtering method only modifies the position information.In addition to gain correction on the position state,the improved method proposed in this thesis also performs gain correction on the state information indirectly related to the measurement,which further improves the tracking accuracy of the algorithm.When the sensor only provides the position status,i.e.the number of measurements processed is less than the number of statuses,the improved method proposed in this thesis has better robustness.After that,the joint integrated probabilistic data association algorithm based on the association hypothesis network proposed in this thesis is combined with the improved smooth variable structure filtering algorithm to obtain a new multi-target tracking method.Through simulation experiments,it is verified that the algorithm can achieve stable and accurate tracking of multiple maneuvering targets in a clutter environment,and has good robustness.4.Finally,the multi-target tracking method and the traditional method proposed in this thesis are respectively processed on the real echo data collected by the radar,and the performance of the results is analyzed.Experimental results show that compared with traditional algorithms,the multi-target tracking algorithm proposed in this thesis has higher correlation accuracy,higher tracking accuracy,better results,and has obvious advantages in the presence of clutter,interference,and track crossing.