Research On Maneuvering Object Tracking Algorithm Based On Performance Feedback

Multi-source information fusion is a procedure to perform detection, correlation,and estimation using information and data from multiple sources, in order to obtainmore precise estimation of object position and identity. It is widely used in militaryand civil applications. Target tracking is one of the important research contents ofmulti-source fusion technology, and also is the current hot research areas. With theemergence of some new types of maneuvering target, such as hypersonic target andsmall slow target, etc., the traditional target tracking algorithm is faced with enormouschallenges.According to the uncertainty of target motion mode and the sources ofmeasurement when tracking maneuvering target, and data association problems in theprocess of the multiple target tracking, a hybrid grid multiple model algorithm basedon current statistical model is proposed and the closed-loop structure of multiplemaneuvering target tracking is built in this paper. the main work of this thesis isdescribed as follows:Firstly, the research background and significance of the thesis is briefly described.The current progress of target tracking algorithms are summarized. The related theoryand methods of target tracking technology is introduced.Secondly, due to decline problem of tracking accuracy caused by accelerationestimation inaccuracy in hybrid grid multiple model algorithm, a new HGMMalgorithm based on current statistics model is proposed, since the current accelerationmean in CS model can reflect true target motion model more accurately. Gridpartition is performed by estimation of the acceleration mean in CS model, andpossible model set of the target is generated online. Then interactive multiple modelalgorithm is adopted to track target. The proposed algorithm performance is testedand compared in general and strong maneuvering scenarios, and simulation resultsdemonstrate that the proposed algorithm not only improves the accuracy formaneuvering target tracking, but also can be applied to most of the tracking scenarios.Thirdly, due to the problem of low object tracking accuracy and even lost trackwhen tracking multiple maneuvering targets, CSHGMM mentioned above is used tosolve multiple maneuvering target tracking problem, and a multiple object tracking algorithm based on performance feedback is proposed. Since the covarianceinformation is fed back to new information and filter gain of the next time, it is usedto adjust the tracking gate on one hand, and is used as a metric to select correlationalgorithms on the other hand to build multi-level feedback structure in multiplemaneuvering target tracking. Simulation results demonstrate that the proposedalgorithm can improve tracking accuracy effectively and timeliness.Finally, the main work and further research of the thesis is summarized.