基于信息总不确定度的冲突证据组合修正方法

现有的冲突证据组合修正方法仅从证据距离、模糊度等描述信息不确定性的一个或几个方面对证据体的基本概率分配函数进行修正，对证据的关联性考虑不够充分。该文提出基于信息总不确定度的冲突证据组合修正新方法。该文在笛卡尔乘积的基础上定义提出组合总不确定度的概念，并给出根据融合前各证据体总不确定度预测融合后证据体组合总不确定度值域的方法。对冲突证据，利用各证据体总不确定度与组合总不确定度的比值，求出对证据基本概率分配函数的修正权重，再根据Dempster规则进行加权平均组合。信息融合的算例分析结果表明，与现有方法相比，该方法融合结果的总不确定度更小，更有利于融合结果的后续决策分析与数据应用。

Combination of Conflicting Evidence by Using the Total Uncertainty Degree of Information

The common way of conflicting evidence combination is to modify the basic probability mass assignment of evidence bodies by a certain indicator which can reflect or describe the information uncertainty of the conflicting evidence. In existing conflicting evidence combination methods, indicators such as the distance of evidence and ambiguity are used. However, these indicators reflect only one or several aspects of the characteristics of the conflicting information uncertainty. A novel method of conflicting evidence combination is proposed based on the total uncertainty degree of information. The concept of combined total uncertainty of information is defined based on Cartesian product. An approach of predicting the range of fused informations combined total uncertainty degree by the total uncertainty degree of each body of evidence before information fusion is also presented. Weights for each evidence body are obtained according to the total uncertainty degree of each evidence body and the combined total uncertainty on their Cartesian product. Then, the bodies of conflicting evidence are combined by the weighted average according to Dempsters rule. Results of numerical examples of information fusion show that, compared with the existing approaches, the total uncertainty degree of the combined information obtained by the proposed method is smaller, which means the combined information is more helpful to subsquent decision analysis and data applications.