基于证据加权融合的改进D-S目标识别算法  

Improved D-S Target Recognition Algorithm Based on Weighted Evidence Fusion

作　　者：郭伟震 张歆茗 方敏[1] 朱仁萍[1] 周莉[1] 

机构地区：[1]鲁东大学信息与电气工程学院,山东烟台264039 [2]思高方达金融服务(上海)有限公司,上海200122

出　　处：《鲁东大学学报（自然科学版）》2016年第1期31-37,共7页Journal of Ludong University:Natural Science Edition

基　　金：国家自然基金(61273152);国家自然科学基金青年项目(61304052)

摘　　要：为有效融合高冲突证据,在总结已有D-S改进算法(CID-S)的基础上提出了新的基于证据加权融合的改进D-S目标识别算法(NID-S).该算法充分考虑了待融合的两证据源与其他证据源间的关联关系,根据相似度矩阵非负对称的性质求取各证据源的相似度、可信度等特征,进而利用证据源的可信度进行加权平均求和,得到不同证据源对各焦元的支持度,并将其作为一组中心值;然后构造相似度函数,并计算不同证据源中的各焦元信度赋值与其对应焦元的中心值的关联程度,从而确定参与产生矛盾信息的焦元的可信度;最后,在局部冲突信息再分配的基础上,给出改进的D-S组合规则.并与D-S算法、已有改进D-S算法进行仿真分析、对比,结果表明:无论是在相对好的探测环境还是强干扰的环境下,本文所提出的改进D-S算法的目标识别结果均优于D-S及其已有改进算法,且是一种实时性好、抗干扰能力较强,适用范围广泛的目标识别算法.In order to effectively fuse the high conflict evidences, a new improved D-S （NID-S） target recog- nition algorithm based on weighted evidence fusion is proposed on the basis of summarizing the existing ira-proved D-S （CID-S） target recognition algorithm. The relationship between the two pieces of evidence source to be fused and other pieces of evidence source is fully considered by this algorithm, and according to the non- negative properties of the similarity matrix, the similarity and the credibility of each piece of evidence can be respectively obtained. Further the support degree of each focal element of different evidence sources, which is taken as a set of central values, is obtained by using the sum of weighted average of the credibility degree a- bout evidence source. Then the similarity function is constructed and the correlation degree between the belief assignment of each focal element and the central value of the corresponding focal element of different evidence sources is calculated, thus the credibility of the focal element involved in the production of contradictory infor- mation is determined. Finally, an improved D-S combination rule of evidence theory is given based on the re- distribution of local conflict information. And the simulation result of the new improved D-S algorithm is ana- lyzed and compared with the ones of D-S algorithm and the existing improved D-S algorithm. The results show that the new improved D-S algorithm is better than the D-S and the existing improved D-S algorithms whether in a relatively good environment or strong interference scenario. And it has the advantage of good real-time per- formance, strong anti-interference ability and wide range of application.

关 键 词：D-S证据组合规则 冲突信息 可信度 目标识别算法 

分 类 号：TN95[电子电信—信号与信息处理]