基于证据理论的多传感器信息融合改进方法  

An Improved Approach to Multi-Sensors Information Fusion Based on Evidence Theory

作　　者：朱江乐[1] 章卫国[1] 张竞凯[1] 池程芝[1] 

机构地区：[1]西北工业大学自动化学院,西安710072

出　　处：《计算机测量与控制》2013年第2期547-549,共3页Computer Measurement &Control

基　　金：航空科学基金(20100753009)

摘　　要：证据理论是信息融合的一种重要方法,被广泛应用于不确定推理中,但在融合高冲突信息时,常会得出不合常理的结论;针对这一问题,提出一种改进融合方法:在原有局部冲突分配规则的基础上,引入冲突分配系数,该系数由待融合的2个证据中同一焦元的基本概率指派值在辨识空间所占比例和相似度构成,以此进一步细分融合证据间的冲突,使融合结果更趋合理;算例验证显示该方法能够有效融合高冲突信息,对比其他方法,可以加快收敛速度,提高准确性,降低决策风险。Evidence theory is an important method in information fusion and widely used in uncertain reasoning. But it often draws counter--intuitive conclusion after fusing high--conflict information. In order to solve such disadvantage, an improved fusion method of evidence theory is proposed. Based on proportional conflict redistribution rule, a conflict distribution coefficient is introduced. This coefficient consists of the BPA value proportion and similarity of the same focal element between two evidences. Comparing with other methods in the same test, this new method is appropriate for high--conflict information fusion and can accelerate the convergence.

关 键 词：证据理论 多传感器 信息融合 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]