基于证据理论的多源信息目标识别  

Mutli-sensor Target Identification based on Evidence Theory

作　　者：刘准钆[1] 程咏梅[1] 潘泉[1] 苗壮[1] 

机构地区：[1]西北工业大学自动化学院,西安710072

出　　处：《火力与指挥控制》2010年第10期43-46,共4页Fire Control & Command Control

基　　金：国家自然科学基金(60634030);航空科学基金(2006ZC53037);航空科学基金(2007ZC53037);高等学校博士学科专项科研基金资助项目(20060699032)

摘　　要：目标识别系统中多源传感器信息高度冲突时,利用DS规则直接融合得出的结果不合理,针对该问题,提出了期望证据融合算法。多个证据融合时,根据两证据间的矛盾信息大小确定其相互支持度,将证据支持度矩阵模最大特征值对应的特征向量归一化后得到各证据的权重系数,进而求出期望证据,并用DS规则迭代融合。通过数字仿真对多种方法进行了比较分析,表明期望证据法在传感器信息高度冲突时依然可以得到较为理想的融合结果。When the information of multi-sensors highly conflict in target identification system,the information fusion result will be unreasonable by DS combination rule.In order to solve the problem,evidence expectation approach is proposed.When several pieces of evidences are combined,the support degree for one another can be calculated according to the conflict information.Normal eigenvector for the maximal eigenvalue of evidence support degree matrix is considered to be weight coefficient.Then,expectation evidence can be gained,and it is combined by DS rule multiple times.Finally,several ways are compared and analyzed through numeric simulation,and the result suggests evidence expectation approach can still get ideal fusion result when the information of multi sensors highly conflict.

关 键 词：证据理论 矛盾因子 期望证据 目标识别 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]