A discounting approach to evidence conflict management
Liang-zhou Chen, Wenkang Shi, Yong Sheng Deng · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 2005
By combining several belief functions from distinct information sources, data fusion aims at obtaining a single Basic Probability Assignment (BPA) function. The classical Dempster's combining rule is the most popular rule of combinations, but it is a poor solution for the management of the evidence conflict at the normalization step. When deal with high conflict information it can even involve counter-intuitive results. A discount method to combine conflicting evidence based on evidence distance is presented; and the discount coefficient of the evidence in the system is also given. Numerical examples showed that the proposed method can provide reasonable results with good convergence efficiency.