Study On Knowledge-based Multi-target Tracking

With the development of science and technology, it is increasingly difficult to trackin modern complex warfare environment. Therefore, the multi-target trackingtechnology faces more severe challenges. It is required to remain stable and accuratetracking for multi-targets in the complex environment, such as dense clutter, low SNR,target strong mobility. But the traditional multi-target tracking techniques, which arebased on kinematic parameters, can not work well in the complex environment. In orderto adapt to the complex tracking environment, it is need to more priori information andapply to the tracking system to improve the tracking performance. The prioriinformation---amplitude information of measurements and road information was studiedto improve multi-target tracking algorithms in the paper, so as to improve themulti-target tracking performance in the complex environment.The paper introduced the theory of multi-target tracking at first, outlining thecommon target models and the multi-target algorithm of data association and trackingfilter, and analyzing the characteristics of algorithms. These are to lay the theoreticalfoundation for the subsequent research of the multi-target tracking algorithms based onknowledge.Then, for maneuvering multi-target, in order to reduce the impact of clutter, etc., onthe tracking performance, the paper presented the C-IMMJPDA-AI (ComprehensiveIMMJPDA based on amplitude information) algorithm and the C-IMMNNCJPDA-AI(Comprehensive IMM Near Neighbor Cheap JPDA based on amplitude information)algorithm. The amplitude of target returns is stronger than those coming from falsealarms. Thus, it is can improve the tracking performance of maneuvering multi-target byusing amplitude information. Simulation results show that, compared with thetraditional method, C-IMMJPDA-AI and C-IMMNNCJPDA-AI improve the reliabilityand validity of the data association in the low-observable conditions, thereby reducingthe rate of losing track and improving the multi-target tracking performance. At last, thepaper compared the performance of C-IMMJPDA-AI algorithm andC-IMMNNCJPDA-AI algorithm in detail, and given the suitable environment scenarios of the two algorithms.Finally, in order to reduce the impact of strong maneuvering on the trackingperformance of the road-constrained ground targets, an Adaptive Road-ConstrainedVariable Structure Interacting Multiple Model (ARCVS-IMM) algorithm was proposedin the paper. In the algorithm, a road information model was set up and threedetermination conditions were put forward. According to the road information, and thethree determination conditions, adaptively adjust the models. The simulation resultsverify that, compared with the IMM algorithm, the algorithm improves the trackingperformance of road-constrained ground targets and reduces the computation by usingroad information.