Study Of Track Correlation Algorithm In Multi-radar And Multi-target Tracking

As one of the key techniques in the high level of generality, multi-sensorinformation fusion technology has been widely used in the field of target tracking. Inthe multi-sensor multi-target tracking system, the distributed structure has a series ofadvantages such as low cost, high reliability, and strong ability to generate, so it isadopted as optimization program. In the distributed multi-sensor multi-target trackingsystem, the effect of track correlation and the performance of the multi-target trackingalgorithmare the key to influence the target tracking accuracy. Somerelevant improvedoptimization algorithms are proposed in this paper in order to improve the effect oftrack correlation and the performance of the multi-target tracking. The main researchcontents of this paper are as follows:Firstly, the optimal allocation algorithm based on deviation of the standarddeviation of track correlation for three sensors is proposed, in order to improve theaccuracy of optimal track correlation algorithm for distributed three sensor systemunder complex detective environment. Further, the stability of the above improvedalgorithm is not high under worse environment, more solutions of weighted fusionalgorithm is studied, a new weighted track correlation algorithm based on likelihoodfunction is proposed, which makes a number of satisfied solutions from the allocationweighted and fused. Reasonable weights not only optimize the quality of the fusedtrack, but also improve the multi-target tracking accuracy to some extent, and alsoeffectively avoid the insufficient of low track correlation accuracy, which caused byrelying on only a set of optimal track correlation results to track correlation whendetective environment is complex. More, the D-S (Dempster-Shafer) evidencecombination rule is used to introduce the multi-source information into the above twokinds of improved algorithms, and the improved weighted correlation algorithms fusedwith multi-source information are proposed in this paper. Last, make the multi-targettracking effect of these relevant algorithms compared and analyzed, and summarize theperformance features and applicable conditions of different algorithms.Secondly, in view of the traditional optimal allocation algorithm spent a long time,low dimensional track correlation algorithm based on the optimal fusion algorithm of two-two track is proposed. Further, consider that the correlation effect of abovealgorithm becomes worse under complex detective environment, an improved lowdimensional track correlation algorithm is proposed, which uses the solutioncomponents of some good solutions from the two tracks to associate with the track ofthe third local node, then weight the unbalanced allocation. Compared with the originaloptimal allocation algorithm of track correlation for three local nodes, the aboveimproved low dimensional track correlation algorithms have the advantages of lowertime cost and better real-time performance. In order to alleviate the heavycomputational burden of the optimal allocation algorithm of multi-dimensionalcorrelation, this paper extends the above improved low dimensional track correlationalgorithm to four local nodes track correlation system, and get better effects ofmulti-target tracking. More, to optimize the track correlation effect of low dimensionalcorrelation algorithm in the poor detection environment, these improved lowdimensional correlation algorithms fused with multi-feature information are given inthis paper. Last, make the multi-target tracking effect of these relevant algorithmscompared and analyzed, and summarize the merits and demerits, get the applicableconditions of different algorithms.Thirdly, both the track correlation effect and multi-target tracking algorithm havea direct impact on multi-target tracking accuracy, in order to improve the targettracking effect of distributed multi-target tracking algorithm, a new joint probabilisticdata correlation algorithm based on the optimal assignment algorithm for local nodes isproposed in this paper. A number of good solutions formed by the optimal assignmentproblem of single sensor point-track correlation are used in the new algorithm todetermine the number of measurement and clutter which have entered the target gate,and then the joint probabilistic data correlation algorithm is used to calculate targetstate. Theoretical analysis and simulation experiment results show that, not only theadvantage of strong anti-jamming capability of the joint probabilistic data correlationalgorithm is inherited, but also the problems that the insufficient of easily to get errorcorrelations for the pure point-track optimal assignment algorithm when the detectioncondition is poor, and the defect of easily to appear track offset for the pure jointprobabilistic data correlation algorithm when the detection condition is not bad are avoided. More, this paper extends the above ideas to the two kinds of improvedthree-dimensional algorithms and three kinds of low dimensional correlationalgorithms, which are mentioned in the above of this paper. Last, these differentalgorithms are compared and analyzed, the results show that, when optimize the trackcorrelation algorithm and multi-target tracking algorithm at the same time, the stabilityand the ability of anti-interference of the algorithm can be improved in a large extent,finally, the accuracy of the tracking effect for distributed multi-sensor multi-targettracking system is improved.