Research On Multi-sourse Information Fusion Method Based On D-S Evidence Theory

Multi-source information fusion in the military and civilian areas has shown a broad application prospect.However,due to the complexity of multi-source information and many uncertain factors in the system,the problem of uncertain information processing gradually highlights,becoming a common problem faced by information fusion system applied in most fields.As an important information fusion method,D-S evidence theory has its advantages in the representation and reasoning of uncertain information.However,it is also faced with the problems of the construction of basic probability assignment function of evidence and the tremendous amount of computation in the fusion process in the application of the actual fusion systems.This dissertation studies these two aspects with a view to D-S evidence theory can be applied better in multi-source information fusion system,in order to improve the performance of information fusion.In view of generating the basic probability assignment function,a method of construction based on the fuzzy membership degree is proposed.This method firstly gets the intersection point of observation value and membership function,and then the traditional ?-cut method is improved according to the intrinsic characteristics of membership degree distribution.Finally,the structure of the focal element is determined,and the membership value is transformed into the basic probability assignment function.The simulation results show that the proposed method counteracts the drawbacks of the traditional ?-cut method,and the generated basic probability assignment function has more accurate information description ability and good arithmetic performance.The example application proves the validity and feasibility of the method.In order to solve the problem of the calculation of evidence fusion process,an approximate calculation method based on improved energy function is proposed.This method firstly improves the traditional energy function by adding a correction factor,and on this basis,determines the focal element control rules,and then takes into account the complete set of relations between the focal points to correct the evidence,which can reduce the computational complexity and ensure the approximate accuracy.Through the simulation analysis,the validity and rationality of the method are verified.Finally,a fault identification information fusion system was built based on the cold rolling data.The simulation results show that the method proposed in this dissertation has a high correct recognition rate for fault identification system in cold tandem rolling process,the basic probability assignment function generated by the proposed method can describe the information of the system very well and has good operation characteristics,and the proposed approximate calculation method can reduce the computational complexity of evidence fusion process while preserving the essence of information as much as possible.