基于平均偏离度的证据组合方法

Dempster-Shafer证据理论广泛应用于信息融合中, 但是在证据高冲突情况下基于经典D-S证据组合规则的融合结果存在反直观的问题。针对这一问题, 提出一种基于平均偏离度的证据组合方法。首先引入证据距离函数获得各证据体的相互支持度, 并将支持度归一化为证据的信任度。对所有的证据进行信任度加权平均, 获得一个参考证据。然后利用该参考证据对各个原始证据进行偏离度的判定及修正。最后利用Dempster-Shafer规则完成证据的组合。实验结果表明, 新方法提高了融合结果的可靠性和合理性, 可以有效地处理高冲突证据。

New evidence fusion method based on average deviation

Dempster-Shafer(D-S) method has been used widely in all kinds of data fusion system, but it has difficulty in dea-ling with combining evidences with high degree of conflict. In order to solve the problem, this paper proposed a new evidence fusion method based on the degree of average deviation strategy. Firstly, it obtained the mutual support degree of every evidence by evidence distance, and normalized the support degree to the credibility degree of evidence. It obtained a referenced evidence by averaging all evidences according to support degree, and then used the referenced evidence for deviation verification and modification of the original evidences. Lastly, it adopted the Dempster's rule of combination to combine evidences after disposing. The results of numerical examples show that this new method improves the reliability and rationality of the evidence combination results. Besides, this method can effectively deal with conflict evidence.