Improved Evidence Fusion Algorithm And Its Application

Information fusion technology is broadly used in military and civilian, and hasreceived extensive attention of scholars at home and abroad. While the evidence fusionalgorithm, that including the classical D-S and DSmT evidence combination rule, as aneffective tool to solve the problem of multi-source imperfect information fusion in theinformation fusion methods, has a good application prospects, but the disadvantages ofcombination rule limits the further application of the algorithm. Therefore, this papermainly discusses the improved algorithm of evidence combination rule and itsapplication, proposes the improved fusion algorithms for different detectionenvironment, and then gives the improved multi-sensor maneuvering target trackingalgorithms. The main contents of the study are summarized as follows:Analysis of disadvantages of the existing D-S and DSmT evidence fusion rule andits improved algorithm, the improved algorithms that based on2-D information fusionand3-D information fusion are proposed, which from the two aspects oflow-dimensional and high-dimensional.The2-D information fusion algorithms based on DSmT and D-S evidencecombination rule are proposed, by analyzing the angle of global and local of thecredibility of focal element involved in the redistribution of conflict information, andthrough data analysis and simulation experiment in the background of multi-sensortarget identification system are implemented to compare with the performance ofcorresbonding algorithms, the results show that the proposed algorithm based onimproved2-D information of evidence fusion can effectively accelerate the convergencerate of the main focal element, improve the accuracy of information fusion results indifferent extent, and reduce the uncertainty of information fusion system.To improve the accuracy of multi-target recognition and tracking by making fulluse of the imperfect information, the improved3-D information fusion algorithms basedon D-S and DSmT evidence combination rule are proposed. The characteristics ofinconsistent factor formed by3-D information fusion through using D-S combinationrule of evidence theory is analyzed, and two different conflicting informationassignment algorithms are proposed. Through analyzing the advantages anddisadvantages of two kinds of algorithms, an adaptive evidence fusion algorithm which can adapt the changing of evidence interference is proposed. In the improved3-DDSmT algorithm, firstly,according to the characteristics of contradictory informationformed by3-D information DSmT fusion rule, three kinds of different local conflictsredistribution assignment algorithms are proposed;secondly, the improved2-Dinformation DSmT algorithm that fusing multiple feature information is proposed,which from the perspective of rational utilization of target feature information. Inaddition, according to the advantages of two kinds of improved algorithm, the improved3-D information DSmT algorithm of fusing multiple feature information is proposed.Through theoretical analysis and simulation results verify the feasibility andeffectiveness of the3-D improved algorithm proposed in this paper.To resolve the shortage of large amount of calculation and the relatively poorreal-time performance of joint probabilistic data association(JPDA) algorithm and itsimproved algorithm in multi-sensor multi-maneuvering target tracking, a newdistributed interactive multiple models(IMM) generalized probability dataassociation(GPDA) algorithm is proposed. Considering the calculated fusion weights ofmulti-sensor on the same goal model in the existing general algorithm,only use thecorresponding likelihood function of target location information to calculate, putforward to use the trace size of the covariance matrix of the model and combined withthe definition of membership function in fuzzy mathematics, to determine the weight ofeach local nodes on the same target correspondence model of track fusion, and then givethe fuzzy methods to fuse model state estimation in fusion centers,that based ondistributed multi-sensor multiple models generalized probability data associationalgorithm. In addition, on the basis of the improved algorithm, a new interactingmultiple models of generalized probability data association algorithm is given, whichusing part of model trace size of the covariance matrix to update the probability ofinteraction model,with the condition of the model estimation error correlation. Finally,analyzing the advantages and disadvantages of the existing multiple feature informationfusion algorithm, the fusing multi-source information of distributed interactivegeneralized probability data association algorithm of grey-distance that based on D-Sevidence combination rule is proposed. Through simulation experiments, the proposedall kinds of new algorithms are compared with the existing correlation algorithm, and analyzed the performance and application characteristics of the new algorithms.A new structure method of three-dimensional assignment model is proposed, that isapplicable to distributed systems of three-dimensional track associated cost matrix andcentralized system of three-dimensional measurement associated cost matrix.This paperstudies the fuzzy optimal assignment(OA) algorithm and fuzzy generalized probabilisticdata association(GPDA) algorithm in multi-sensor multi-target tracking, while itsapplication to interact with multiple maneuvering target tracking, fuzzy multiplemaneuvering target tracking algorithms are respectively given under the structure ofdistributed and centralized. On the basis of the improved algorithm, the introduction ofgrey-distance algorithm based on D-S evidence combination rule to fuse multi-sourceinformation, and then fuzzy association of multi-source information algorithms aregiven. Through simulation experiments, compares and analyses the differentperformance characteristics and application environment of the improved maneuveringtarget tracking algorithms.