基于pignistic概率距离的冲突证据合成方法  

New method to combine conflict evidence based on pignistic probability distance

作　　者：王栋[1] 李齐[1] 蒋雯[2] 徐晓滨[3] 邓勇[1] 

机构地区：[1]上海交通大学电子信息学院,上海200240 [2]西北工业大学电信学院,陕西西安710062 [3]杭州电子科技大学自动化学院,浙江杭州310018

出　　处：《红外与激光工程》2009年第1期149-154,共6页Infrared and Laser Engineering

基　　金：863高技术发展计划(2006AA02Z320);国家自然基金资助项目(30700154);目标特征与环境光学国防重点实验室基金资助项目(51476040105JW0301);航空科学基金资助项目(20070511001);浙江省自然基金资助项目(Y107458)

摘　　要：Dempster Shafer证据理论在数据融合中有着广泛的应用。但当证据之间高度冲突时,应用DS证据合成公式往往会得出错误的结果。在现有的证据合成改进方法中,均使用冲突系数k来度量证据之间冲突的程度。最新的研究表明:影响证据冲突的因素不仅仅是冲突系数k,还和证据的pignistic概率距离等因素有关。在原有的冲突系数k之上,引入pignistic概率距离,提出了一种新的冲突证据合成方法。仿真结果表明:当证据之间的冲突小时,合成结果近似于D-S证据合成公式,当证据之间的冲突较大时,这种方法在公式模型上比其他方法完善,冲突证据的合成结果更加合理。Dempster Shafer evidence theory is widely used in data fusion. However, illogical results may be obtained by classical Dempster Shafer combination rule when collected evidence highly conflicts each other. Many methods are proposed to solve this problem. In these methods, the conflict coefficient k is used to measure the conflicting degree between the evidence. However, researches proclaim that not only the conflict coefficient k, but also other relating factors have an effect on conflict degree between evidence. A new combination method based on the classical conflict coefficient k and pignistic probability distance was proposed. A numerical example was used to show that： while the conflict between the evidence is small, the result is close to D-S evidence combination rule; while it is large, the method is more perfect in the rule model due to the more reasonable results.

关 键 词：信息融合 证据理论 组合规则 证据冲突 目标识别 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]