一种改进型冲突证据合成算法研究.docVIP

一种改进型冲突证据合成算法研究 刘娟，胡泽，葛亮 西南石油大学电气信息学院，四川，成都，610500 摘要：证据理论在多传感器数据融合中，已经成为一种基本的数据融合算法。针对证据理论在处理冲突证据时，可能会得到与客观事实相悖的结果，提出一种基于证据距离，调整证据权值的合成规则。根据证据间距离的不同，得到每个证据的支持度，然后根据支持度分配不同的权值，该权值反映冲突证据对最终融合结果的影响。为验证该模型的可靠性，与其他方法进行对比，证明了新的合成规则能够有效处理冲突证据，得到合理的融合结果。实验表明，该方法与其他方法相比，具有较快的收敛速度，高度的可靠性，可以快速有效地识别目标。 关键词：证据理论；冲突证据；支持度；识别目标 中图分类号：TP301.6 文献标识码：A Research of a Modified Combination Rule of conflict Evidence LIU Juan, HU Zhe, GE Liang School of Electronic and Information Engineering, Southwest Petroleum University, Chengdu 610500，ChinaAbstract: Evidence theory has become a basic data fusion algorithm in the multi-sensor data fusion. D-S evidence method would involve invalid result while evidences conflict with each other. Considering these problems, on the basis of Support Degree of evidences, a new evidence combination rule was presented in this paper. According to the different evidence distance，Support Degree was introduced. Then, the weight was distributed differently on the base of different Support Degree. These weights reflect the influence on the final fusion results. In order to validate the reliability of the new evidence combination rule，this paper compared with other rules and the results show that this new evidence combination rule can deal with conflict evidence effectively and get reasonable fusion results. Experimental results show that the new rule has fast convergent speed and high reliability. Compared with the existing combination rules, the newly proposed rule is much fast in recognizing target. Key words: Evidence Theory; Conflict Evidence; Support Degree; Recognizing Target 1引言 Dempster-Shafer(D-S)证据理论由Dempster于1968年提出，之后Shafer 进一步将其完善。证据理论通过引入信任函数、似然函数[]，为不确定数据的表达和融合提供了比较灵活的算法。但是，当证据之间存在冲突或不一致时，传统的D-S证据合成规则就不能使用或者得出与事实相悖的结果，所以对冲突证据合成的研究一直是学术界研究的焦点。 经过比较近20年来证据理论的研究成果，发现其改进算法可分为两大类：一类对Dempster合成规则进行修改，通过将冲突信息在各子集上进行重新分配，达到优化乘性策略所得结果的目的[]。此类算法通常具有某些局限性，例如：Yager规则[]对证据的扰动过于敏感，具有一票否决的缺点。Lefevre规则[]处理高冲突证据风险巨大，此时合成规则不适用。张山鹰合成规则[]算法复杂，不容易得到融合结果。一类则修改模型而不改变D-