Research On Target Tracking And Data Fusion Algorithm

With the rapid development of electronic technology, target tracking and data fusion technology have been widely used in many areas.Target tracking is an important part of radar data processing,its core is filtering algorithm, so the research of filtering algorithm is very important for stable tracking. The essence of data fusion is integrating information from multiple sensors to extend the temporal and spatial coverage and improve the accuracy, reliability, and information utilization of the system through the fusion algorithms.The target tracking algorithms and data fusion algorithms are researched in this dissertation.Among them, data fusion is the fusion with the estimated level and closely related with the target tracking algorithm.At first, the basic linear Kalman filtering(KF) algorithms is researched in this dissertation.The extended Kalman filtering(EKF) and the unscented Kalman filtering(UKF) are researched when the observation equation and state equation are based on non-linear mathematical model.The two nonlinear algorithms, namely,EKF and UKF are analyzed through the simulation.Simulation results verify that UKF has better tracking accuracy in strong nonlinear environment, and EKF has the same tracking accuracy in weak nonlinear environment.Secondly, the data association algorithms involved in target tracking are researched, including probabilistic data association(PDA) algorithm, joint probabilistic data association(JPDA) algorithm and fuzzy C-Means(FCM) clustering association algorithm. Among them, PDA algorithm is mainly applied to single-target tracking, and the JPDA algorithm and FCM algorithm is mainly applied to multi-target tracking. The simulation results show PDA algorithm is effective for data association with single-target in the clutter environment. Under the target cross-motion scenes, JPDA algorithm and FCM algorithm are compared to verify that FCM has higher convergence rate, smaller loss rate and shorter running time.Finally, according to the structure of the fusion system, the centralized and the distributed fusion algorithms are researched. Among the centralized fusion algorithms, composite measurement fusion, the sequential filtering fusion, the augmented filtering fusion are researched. The simulation results show the effectiveness of the three algorithms and verify that the sequential filtering fusion is optimal with sync data and less sensors scenes, the composite measurement fusion is optimal with higher real-time requirements and more sensors scenes. Among the distributed fusion algorithms, the weighted covariance fusion, the adaptive tracking fusion, and the information matrix fusion are researched. The simulation results show the effectiveness of the three algorithms and verify that the weighted covariance fusion has the same fusion performance which is better than the adaptive tracking fusion and the advantage of the adaptive tracking fusion advantage is the least time-consuming.