Integrated Data Fusion Using Dempster-Shafer Theory

This paper proposes an integrated data fusion approach. The approach is based on the Dempster-Shafer evidence theory, and includes four main aspects: the construction of basic probability assignment, a novel reliability coefficient function converting similarity to initial weight factors, an improved fusion approach by reassigning reliability coefficient, and the "Discount Rule." Utilizing the integrated approach, conflicting data are fused more accurately and effectively than using the single fusion method. Experimental results show that the belief assignment results of the proposed approach are in accordance with the practical situation.