An improved fusion algorithm of evidence theory
Tian Xue-yi, Yibing Li · 2010
In this paper, aiming at the problem that the DS evidence theory can't deal with the evidence conflict and the small disturbance of the basic probability distribution function can cause drastic result changes, an improved fusion algorithm of evidence theory is presented. Firstly, it uses the evidence distance to obtain the corresponding of evidence conflict, then obtains the evidence credibility by using the entropy of evidence conflict and makes use of the evidence credibility to redistribute the basic probability distribution function. Finally, it gets the fusion result by the ordinary combination formula. The simulation results prove that the algorithm can not only fuse the evidence with serious conflict, but also has better recognition result when the basic probability distribution function brings small disturbance.