Study On Key Issues In Generalized Evidence Theory

Evidence theory is one of the most widely used methods in the field of information fusion,which has obvious advantages in the representation and modeling of uncertain information.However,the classical combination rule in the theory of evidence can not effectively deal with conflict evidence,especially the conflict of evidence caused by incomplete frame of discernment,and seriously restrict the further promotion and application of evidence theory.As a generalization of classical compatible evidence theory,generalized evidence theory defines the generalized basic probability assignment function,generalized belief function and plausibility function,gives a new evidence combination rules and conflict management model,to solve the basic problems of conflict recognition,conflict management and conflict evidence fusion.When the framework is complete,the generalized evidence theory is degenerated from the open world to the classical closed world D-S evidence theory.This paper mainly discusses the key problems of the generalized basic probability assignment and the form of combination rules in generalized evidence theory,in order to promote further deepen research on the theory and application of generalized evidence theory in the field of information fusion.This paper mainly completed the following research work:Firstly,chapter-I systematically summarizes the development background and research status of information fusion,evidence theory,introduces the basic concept and application framework of evidence theory.Some problems of the evidence theory are discussed and analyzed.Correspondingly,the mainstream solutions to these problems have been summed up.Then,in chapter-II,the theoretical framework of generalized evidence theory and the related concepts and definitions are introduced in detail.The forms,characteristics and properties of generalized combination rule are analyzed in detail.The conflict processing model of generalized evidence theory is briefly introduced.Secondly,the paper studies the generation scheme of the basic probability assignment,which is the key problem in the theory of evidence and the theory of generalized evidence.This paper presents a method for generating the basic probability assignment based on the sample difference degree in open world.This paper describes in detail how to use the method to model the characteristics of the data which is bound to the open world environment in the practical engineering application.The rationality and validity of the proposed method is demonstrated by a practical example.Then,the paper studies the generalized combination rule in the generalized evidence theory and analyzes the unreasonable part.On the basis of the generalized combination rule,a modified combination rule is proposed,fully compatible with the original generalized evidence theory,and combined with the mathematical properties of commutative and associativity.The new rule has the ability of dealing with information fusion problem of effective modeling and processing of open world.Through numerical examples,examples of geometric analysis and comparison of the two kinds of evidence combination rule,the experimental results show that the improved combination rules of evidence in the distribution of conflict is more reasonable;and the effectiveness of the new combination rule is verified by the application of target recognition.Finally,the main work of this paper is summarized,and the future research work and the problems worthy of further research are prospected.