The Research Of Multi-Source Information Fusion Method

With the rapid development of computers and network technology, information fusion technology has been rapid developing.In recent years, multi-source information fusion technology has been attracting great concern of military and civilian applications, and then its application became more and more widely. Currently, the country’s information technology is one of the major tasks of social construction, and multi-source information fusion technology of modern information technology is an integral part of a technology. This paper’s research is based on the background of multi-source information fusion monitoring and management for security issue.Firstly, this paper researches relevant knowledge of multi-source information fusion theory and fusion methods, especially for the D-S evidence theory fusion method and BP neural network fusion method with more detailed analysis. D-S evidence theory of evidence combination rules and its improved methods are not good solutions to the conflict of evidences; BP neural network fusion method has the weaknesses that it must be retrained when the samples changed in order to better fusion result. So, these two fusion algorithms are integrated to an framework so that these two fusion methods can make up each other, and then this paper proposes a fusion algorithm based on the context of the right value of the composite. Experiment verifies the correctness and validity of the algorithm of this paper proposes, and Compared to the D-S evidence theory fusion algorithm, this can algorithm effectively resolve conflicts of evidence fusion.This paper proposes a multi-sensor event estimation model based on the characteristics of spatial and temporal when the fire broke, the model uses the ratio of the monitoring and threshold as a basis to judge, so as to facilitate to compare the possibility that the fire broke out between different types of sensors. And then this paper proposes based on clustering fusion algorithm according to the model. The algorithm is based on clustering computing, selects the sensor detecting the incident of maximum of probability as cluster center and clustering, and then gets fusion result after set option and the computing is short. Finally, experiment verifies the feasibility and effectiveness of the algorithm.