Fusion Algorithm Of Conflict Evidence Based On DS Theory

In recent years,with the increasing demand for information fusion,fusion technology has not only made great progress in information processing,but also been successfully studied and applied in many fields such as military field,fault diagnosis and target identification.Among them,D-S evidence theory has the advantage of expressing and dealing with uncertainties well without prior information,so it can model problems and process data more optimally in the process of fusion,which improves the accuracy of fusion and makes decision results more accurate.However,if there are conflict evidences,the evidence fusion using D-S evidence theory can not achieve good results or even the results are contrary to common sense,so we need to improve the evidence theory.The current research focuses on revising evidence sources and modifying combination rules.However,compared with two methods,it is found that pretreatment of evidence sources will not destroy the good quality of Dempster combination rules,which is more advantageous than revising combination rules.In this paper,the conflict evidences are processed and synthesized from two aspects of evidence source pretreatment and evidence fusion.The main research contents are as follows:Firstly,the uncertainty of fusion is analyzed,and an inverse DP probability conversion method is proposed to reduce the uncertainty in the case of evidence conflict and compound focal elements.Based on DP synthesis rule,the uncertainty is gradually reduced by dividing the potential,and the basic probability assignment function is transformed into probability function to fuse.Secondly,a weighted fusion algorithm based on confidence distance is proposed to solve the problem of low credibility in the process of conflict evidence fusion.Using confidence distance measure to measure evidence,the evidence are transformed into distance matrix form.The credibility is obtained by matrix correlation calculation,and then the reliability is allocated by being weighted to modify the evidence source.In the end,the basic evidence is fused.Finally,aiming at the measurement of evidence conflict degree,a method of conflict evidence synthesis based on exponential divergence is proposed.The index cross-entropy is used to measure the conflict evidence,and the conflict coefficient between the evidences is constructed into a distance matrix,and the data fusion is carried out by weighted fusion.A large number of simulation and comparative studies verify the effectiveness and reliability of the proposed algorithm.