Study On The Algorithm For Multi-sensor Multi-target Tracking System Based On Information Fusion

Information fusion technology has been widely used in military and civil area. It is wellknown that multi-sensor multi-target tracking is an important application area of informationfusion technology. And multi-sensor multi-target tracking based on information fusion hasimproved tracking performance by fusing the valid measurements from all the sensors.Moreover, researches have shown that tracking performance of multi-sensor system issuperior to that of the single-sensor tracking system.This paper mainly discusses several commonly used centralized multi-radar multi-targettracking algorithms, centralized multi-passive-sensor multi-target tracking algorithms andmulti-radar multi-target tracking algorithms with fusing multi-feature information. The mainwork of this paper is as follows:Firstly, this paper analyses several common sub-optimal data association algorithms,including probabilistic data association (PDA) algorithm, joint probabilistic data association(JPDA) algorithm and generalized probabilistic data association (GPDA) algorithm. Thispaper further discusses advantages and disadvantages of the algorithms above by carrying outcorresponding computer simulations.Secondly, in view of the problem of long time spent of traditional optimal assignmentalgorithm of data association, this paper put forward a new optimal assignment algorithm ofdata association (OA). The new algorithm takes dynamic information to construct the modelof optimal assignment algorithm and calculation of new algorithm is easy. On basis of thenew algorithm, this paper proposes an improved OA (IOA) algorithm of data association andcarries out corresponding simulation analysis. IOA algorithm makes use of marginalassociation probability to replace the original statistic probability and improves themulti-target tracking performance in poor detection scenario. In addition, on basis of making afull consideration about the advantage and disadvantage of OA and GPDA algorithm, byeffectively fusing OA and GPDA algorithm, this paper gets a new algorithm (OA-GPDA),GPDA algorithm based on OA algorithm, which takes full advantage of the two algorithms.Thirdly, based on discussing the feature-aided multi-radar multi-target trackingalgorithms, including sub-optimal and optimal data association algorithms, this paper introduces multi-feature information into PDA algorithm, JPDA algorithm, GPDA algorithm,OA algorithm and IOA algorithm. By carrying out computer simulations, this paper analysesand compares the tracking performance between tracking algorithms based on state estimateand tracking algorithms based on multi-feature information.Finally, in view of the problem of heavy calculation in three-dimensional (3-D)assignment algorithm of data association of multi-passive-sensor system, this paper developsa new data association algorithm of multi-passive-sensor multi-target tracking system. Ingeneral detection scenario, the new algorithm not only further advances the multi-targettracking performance, but also decreases the time spent. However, in the scenario with hightarget density, the new algorithm becomes worse. Therefore, another new algorithm fusingcourse information is proposed which overcomes the disadvantage. Moreover, considering theadvantage of GPDA algorithm, on basis of new3-D assignment algorithm of data association,this paper put forward an improved3-D assignment algorithm of data association whichimproves association probability between measurements and targets, and proposes fusingalgorithm which takes full advantage of GPDA algorithm and the new3-D assignmentalgorithm of data association. In addition, this paper further summarizes and analyses theperformance and application characteristics of the corresponding algorithms.