证据理论的模糊折扣方法  

A Fuzzy Discounting Method for Evidence Theory

作　　者：柯小路[1] 马荔瑶[1] 朱儒[1] 王永[1] 

机构地区：[1]中国科学技术大学自动化系,合肥230027

出　　处：《小型微型计算机系统》2015年第2期285-289,共5页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(60974103)资助

摘　　要：证据理论是常用的一种决策级信息融合方法,能有效处理不确定性信息.然而该方法在融合冲突证据时却可能会得到与常理相悖的结论.针对这一问题,对已有解决方法进行了总结,在此基础上提出一种新的折扣方法,综合考虑了证据之间的相互冲突情况和各证据自身的区分能力.鉴于冲突、区分能力等概念的定义具有模糊性,引入模糊集合予以表示,定义了它们的模糊隶属函数,并设计了相应的模糊推理规则以获取折扣值.算例表明,新方法对一致性证据影响甚微,而对冲突证据能够进行有效处理,得到合理的融合结果.Evidence theory is widely applied to the fields of decision-level information fusion. It＇s an effective tool to manage messages with uncertainty. However, the result generated by the Dempster＇s rule may be counterintuitive when two bodies of evidence are con- flicting. We draw a conclusion about the extant methods on this problem, and then propose a new discounting method to overcome this defect. Conflict is taken into account,as well as the discernibility of each evidence. Considering that it＇s hard to define a threshold to distinguish ＇Big＇ from ＇Small＇ either for conflict or for discernibility,fuzzy sets are therefore introduced and fuzzy membership func- tions are defined for them, respectively. A set of fuzzy rules are also designed to reasoning discounting factors. Numerical examples show that the new method has little influence on consonant evidence while it can effectively discount conflicting evidence, and that rea- sonable results would be generated.

关 键 词：证据理论 模糊理论 证据折扣 组合规则 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]