Research On Knowledge-based Target Tracking

With the development of science and technology.The technology of multiple targets is facing more and more requirements.in some scenarios, especially like dense clutter, uneven clutter distribution, low detection rate, low SNR and so on, the traditional tracking algorithm performance was not as good as expected, so the knowledge-based multi-target tracking system is presented. knowledge-based system solve the problem of traditional tracking system that only kinematic parameters is used and that is very important for multi-target tracking performance. The main work of this paper firstly introduce the scene of uneven distribution clutter and low SNR, then discuss useage of amplitude information, clutter distribution information to improve tracking performance. secondly introduce the scene of targets that are close to each other and similar in kinematic parameters, this paper present classified information of the target to improve the multi-target tracking performance.This paper firstly discuss some basic theories of multi-target tracking technology, including multi-target tracking model,tracks management, data association and filtering method.Then briefly descripte the reason of chosing these method adopted in this paper.To avoid undesirable defects under scenario that low SNR and the uneven distribution of clutter in traditional tracking system. This paper combined the information of clutter distribution with amplitude information. Firstly, according to information of clutter distribution of different regions,present different measurements filter threshold and strategies.When the success rate of track initiation is ensured the method presented greatly reduces the error rate of track initiation.For tracking targets, this paper combine information of clutter distribution with amplitude information, then proposed aapproximate method of calculating the probability density function of angular distribution and then to correct data association algorithm in dense clutter regions. The simulation results show that the corrected algorithm CIMMDJPDA-AICLUTTER can improve the tracking performance under scenarios of uneven clutter distribution and low SNR.For the case that we are able to get the measurements classification of sensor.This paper use the classification information to assist in the tracking system.This paper describes how to joint target classification information to JPDA.And the deffects of CIMMJPDA-C algorithm are also analyzed.For the propblem of gate overlap this paper propose corrected CIMMJPDA-C and that is CIMMJPDA-MC algorithm.When the clutter is strong or SNR is low this paper propose CIMMJPDA-AIMC that combine amptitude.The simulation results show that the corrected algorithm can get a better performance when target tracking gates overlaped or clutter is strong.