一种分级的DS证据合成策略  

Multi-Level DS Evidence Combination Strategy

作　　者：何兵[1] 胡红丽[1] 

机构地区：[1]北京航空航天大学计算机学院虚拟现实与可视化新技术研究室,北京100083

出　　处：《计算机工程与应用》2004年第10期88-90,98,共4页Computer Engineering and Applications

基　　金：国家863高技术研究发展计划:分布式虚拟综合环境(编号:2001AA115130)

摘　　要：该文提出了一种分级的DS证据合成策略,该方法利用证据间的一致性信息将证据分级(体现为可信度),并通过增加低级别证据不确定性的思想修正证据,最后采用DS证据理论对修正后的证据加以合成。由于奇异证据一般而言与其他证据间的一致性较差,依据该文的方法计算得到的可信度低,故增加低级别证据的不确定性可以有效地降低奇异证据对最终合成结果的影响,从而提高合成判决结果的可靠性。仿真实验表明,该方法在解决不一致证据间的合成问题方面体现出良好的适应能力。A Multi-Level DS evidence combination method is proposed to solve the problem of combining evidences with varying degree of conflict.Such a method comprehensively utilizes the advantages of DS theory and additive strategy.First the confidence to an evidence is decided by analyzing the consistence among this evidence and other evidences.Second,the unknown property of those evidences with low confidence will be increased.At the end,those modified evidences are combined by DS evidence theory to achieve the decision.It is because generally the queer evidences own low consistence with other evidences,so they have low confidence degree.Therefore by increasing the low confidence evidences' unknown ability distribute,we can effectively control queer evidence's affects to the final combination result.According to the simulation results,such a strategy could improve the reliability of combination result as well as the decision.

关 键 词：信息融合 证据理论 人工智能 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]