一种快速分层递阶DSmT近似推理融合方法(B)  

A Fast Approximate Reasoning Method in Hierarchical DSmT(B)

作　　者：李新德[1] 杨伟东[1] 吴雪建[1] Jean Dezert 

机构地区：[1]东南大学自动化学院复杂工程测量与控制教育部重点实验室,江苏南京210096 [2]ONERA(The French Aerospace Lab),29 Av.de la Division Leclerc

出　　处：《电子学报》2011年第A03期31-36,共6页Acta Electronica Sinica

基　　金：国家自然科学基金(青年基金)(No.60804063);江苏省自然科学基金(No.BK2010403);图像信息处理与智能控制教育部重点实验室开放基金(No200902);航空科学基金(No.20100169001);东南大学优秀青年教师教学;科研资助计划;东南大学创新基金(No.3208000501)

摘　　要：针对Dezert-Smarandache Theory(DSmT),随着鉴别框架中焦元数目的增多,其组合推理运算成指数增长,已成为制约该理论广泛应用与发展的瓶颈问题.为了解决这个难题,本文在进一步深入研究仅单子焦元赋值几个关键问题的基础上,主要针对超幂集空间中部分单子和冲突焦元具有信度赋值的情况,通过比例分配原则,把冲突焦元的信度赋值分配到相应的单子焦元上,然后根据仅单子焦元情形下的近似推理方法进行处理,即利用二叉树分组技术对单子焦元进行刚性分组,实现细粒度超幂集空间向粗粒度超幂集空间映射.最后通过从计算效率、信息损失和相似度的角度分别比较新、老方法,比较结果充分地验证了新方法的优越性.With the increment of focal element numbers in discernment framework,the computation amount in DSmT will exponentially go up,which has been the bottleneck problem to block the wide application and development of DSmT.To solve this problem,in this paper,on the base of further studying several key issues of hierarchical DSmT only with singletons,an extension is done to be fit for the case that there are singletons and conflicting focal elements with assignments in hyper-power set simultaneously.The conflicting focal elements are discoupled in terms of the proportion of assignments taken by elements involved in it.This is very helpful for further processing information according to the method only with singletons.That is,these discoupled focal elements in hyper-power set are forced to group through bintree to realize the mapping from the refined hyper-power set to the coarsened one.Finally,by comparing new method with old one from different views,i.e.computation effecieny,loss of information and similarity,the superiority of new one over old one is testified well.

关 键 词：近似推理 信息融合 分层递阶 Dezert-Smarandache THEORY 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]