Maneuvering Detection Based Multi-Model Fusion Algorithm Research For Moving Target Tracking

With the development of science and technology of industry, moving targets have the characteristics of high maneuver performance and low detection degrees, which put forward higher requirement of maneuver judgment, accurate state estimation and real-time application to target tracking. At present, most researches on target tracking are based on IMM algorithm. However, IMM algorithm has some shortcomings of filtering estimation, model set optimization and algorithm structure. Therefore, aiming at moving targets maneuvering characteristic, this paper investigates interacting multiple model fusion algorithm based on maneuvering detection for moving target tracking.Firstly, according to the problem of KF gain matrices with massive calculation, we utilize the SS-KF parallel filters in filtering estimation, and then propose the steady state KF IMM algorithm based on maneuvering detection. SS-KF algorithm has low complexity that makes it qualified for project use, so SS-IMM algorithm reduces its complexity obviously. However, the process noise of SS-IMM algorithm has asignificant effect on tracking performance. Consequently, we introduce maneuvering detection strategy and combine it with SS-IMM algorithm under different process noise. The new algorithm can effectively reduce the computation time and improve the tracking performance.Secondly, from the model set optimization and algorithm structure adjustment, we present a new algorithm named SIMM-EV algorithm. It combines the extended viterbi algorithm with scalar-weight matrix method, which can improve the shortcomings of IMM algorithm. SIMM-EV algorithm can not only solve the problem of the interaction of invalid model, but also integrate new function of mending the structure defects of IMM. Therefore, SIMM-EV algorithm effectively improves the maneuvering target tracking precision, especially speed tracking.At last, on the basis of MD-SS-IMM and SIMM-EV algorithm, we design a new practical approach named MD-SSMM-EV for maneuvering target tracking. It optimizes the tracking performance through the algorithm fusion way, and brings in maneuvering detection strategy for judging target states, then implements algorithm switching according the detection results. In allusion to the three proposed algorithms, this paper firstly conducts simulation experiments of relevant algorithms. Then, based on vehicle tracking experiments, we acquire the real tracking trajectory and measured states by differential GPS equipment and conventional GPS equipment, respectively. Experimental results demonstrate that the proposed algorithms in this paper outperform the conventional IMM algorithm both in tracking accuracy and computational complexity greatly, which can be competitive alternatives in the real-time application.