摘要
针对Dempster理论在合成冲突较大证据时出现与实际结果相悖的问题,首先根据各证据之间的相似度系数判断证据之间是否存在冲突,对存在冲突的3条以上证据,计算证据之间的兰氏距离并构建相似矩阵;然后,根据相似矩阵的特征值确定证据的重要性系数,并通过重要性系数对BPA函数进行转化;最后,用Dempster方法进行合成。仿真结果表明:该改进冲突证据合成方法可以有效处理冲突证据,避免了一票否决的问题,比其他几种典型的合成方法有更快的识别速度,同时降低了识别结果的不确定性。
To solve the problem of the unrealistic results arising in using Dempster＇s method for the combination of highly conflicting evidences, an improved evidence combination method is presented based on Lance Williams distance. First of all, the similarities between evidences are used as an ap- proach to judging whether conflicts exist. If there are more than 3 evidences, Lance Williams distance between evidences is calculated and a similarity matrix is established. The corresponding maximum characteristic value of similarity matrix is used to obtain the evidencesr important coefficients. Then, BPA functions are transformed by means of the important coefficients. Finally the Dempster＇s method is used for the combination. The simulation results show that this method can deal with conflicting evidences effectively, calculate faster than other improved algorithms, reduce more uncertainty in recognizing results, meanwhile retain the advantages of Dempster＇s method.
作者
姜礼平
李冬雷
季傲
JIANG Li-ping LI Dong-lei JI Ao(College of Science, Naval Univ. of Engineering, Wuhan 430033, Chin)
出处
《海军工程大学学报》
CAS
北大核心
2016年第6期1-4,22,共5页
Journal of Naval University of Engineering
基金
国家自然科学基金资助项目(6107419)
全军军事学研究生课题资助项目(2014JY451)
关键词
冲突证据
证据理论
兰氏距离
特征值
conflicting evidence
evidence theory
Lance Williams distance
characteristic value
