Research On Multi-target Tracking Method Based On Data Association

In the modern battlefield,the enhancement of target maneuvering,the development of stealth technology and the application of all kinds of interference technology make the battlefield environment more and more complex.Traditional multi-target tracking technology cannot deal with these situations effectively.To solve the above problems,this paper studies the method of track initiation and track maintenance in dense clutter environment based on data association as well as tracking of maneuvering targets.Finally,a complete multi-target tracking framework based on dynamic probabilistic data association is proposed.Detailed research work are as follows.1.This paper introduces the commonly used track initiation method,then an improved one-step delay track initiation algorithm is proposed,which improves the accuracy of track initiation.Then the probabilistic data association method,the joint probabilistic data association method and the multiple hypothesis data association method are studied and compared.Finally,a joint probabilistic data association method based on one-step delay track initiation is proposed.Which reduces the reliance on the prior information of the target and meets the requirements of the actual target tracking situation.2.To solve the problem of maneuvering target tracking in clutter environment,this paper studies the data association method based on multiple target.Aiming at overcoming the shortcomings of constant velocity model and constant acceleration model,an interacting multiple model algorithm based on tangential acceleration and normal acceleration model is introduced which can tracking the maneuvering target more accurately.Then this paper introduces how to combine interacting multiple model algorithm and data association algorithm.The interacting multiple model probabilistic data association method and the interacting multiple model joint probabilistic data association method are introduced which can track the maneuvering target in the clutter environment.3.This paper proposes a dynamic probabilistic data association method.It divides the whole tracking process into two stages: measurement matching phase and pipeline processing phase.By considering track initiation,track maintenance and track termination,it constitutes a complete multi-target tracking process.Finally the target recognition is achieved by the difference of model probabilities in the interacting multiple model algorithm.