Research On Multi-feature Fusion Algorithm Based On Ds Evidence Theory

Because of the demands from military technology,automation and intelligence,information fusion has been broadly concerned in academia and industry.DS evidence theory has been a major mathematical method applied in the information fusion systems and offers a suitable framework to combine the uncertain information.This dissertation thoroughly studies the information fusion algorithm based on DS evidence theory mainly by the following parts:Firstly,we expound the theoretical foundation of DS evidence theory and the principles of DS combination rule through the theoretical deduction of the synthetic formula.This chapter is the theoretical basis of the paper.Secondly,the reason of three common paradoxes in DS evidence theory is discussed and the improved fusion algorithms to solve the paradoxes are mainly introduced.In order to evaluate the advantages and disadvantages of the improved fusion algorithms,we summarize an evaluation model by reviewing the related literatures which shows the properties of an improved fusion algorithm should be satisfied.Thirdly,the fusion methods of high conflict evidence are proposed.In consideration of the relation of the focal elements among a body of evidence,we take the average basic probability assignment value of focus as a reference to modify the original evidences;as to the relation between the bodies of evidence,we use the distance measurement method and the similarity function method to obtain the weight coefficients of the bodies of evidence.The above two fusion algorithms are simulated and verified by using the improved algorithm evaluation criterion and the results compared with the other improved fusion algorithms indicate that the two evidence fusion models have the property of anti conflict and strong robustness in dealing with the high conflict evidence fusion systems.Finally,based on the study of paradoxes,we apply the improved information fusion algorithm to emitter signals identification.By extracting the multi-dimensional entropy features of 9 kinds of signals,we use the improved information fusion algorithm to match and recognize modulation signals in condition of the SNR of signals are known and unknown respectively.The results show that this algorithm has strong applicability,and it still can get good recognition results when lacking of priori information.