Research And Application On Data Fusion Based On Basic Probability Assignment Function Adjusting

With the development of science and technology and the needs of modern industry, data fusion as a new interdisciplinary received extensive attention and rapid development in recent years. Data fusion of multiple sensors can be integrated to provide complementary and redundant information, access the object in a more comprehensive and accurate information, quick decisions and judgments. For complex state type of the sensors in the industrial, and the measured parameters of the problem difficult to obtain accurately and reliably, this paper present a data fusion based on the basic probability assignment adjusting.Firstly, this paper expounds the basic principles of data fusion, and the concept, and the key technology of data fusion, summarize and classify fusion algorithm, and expound the data fusion algorithm, neural network theory methods and evidence in detail. In the same time this paper research the structure of neural network design in depth and analyze the advantages and disadvantages of evidence theory, as following:that data fusion is difficult to obtain the basic probability assignment problem is given which uses neural network technologies to obtain the basic probability assignment method; A data fusion method based on basic probability assignment adjusting was proposed, building the neural network structure of target recognition; The data fusion method is applied to anti-money laundering field, giving the steps identified transactions, building transactions in distributed environment recognition system.The simulation results show the feasibility and effectiveness. And the method is self-learning ability. The method overcomes the existing experience to acquire the basic probability assignment strong subjective shortcomings.