对目标进行识别时, 应用Dempster证据组合规则融合冲突证据会产生不合理的结论。针对这个问题, 提出一种基于加权马氏距离的证据理论改进方法。使用证据理论前对证据进行预处理, 引入加权的马氏距离来度量不同证据被其他证据支持的程度。利用平均证据代替冲突证据, 将支持度作为证据的权值, 再应用Dempster证据组合规则得出识别结果。通过仿真实验, 将该方法与现有方法进行了对比分析, 结果表明该方法较其他方法能更有效地融合高度冲突的证据, 提高了目标识别的准确性。

In target recognition, the use of Dempsters rule for conflict evidence fusion may result in unreasonable conclusion. To solve the problem, we proposed a weighted Mahalanobis distance based method for improving the evidence theory. Before the application of evidence theory, the evidences were preprocessed and the support degree of each evidence was calculated out by introducing weighted Mahalanobis distance. The conflict evidence was superseded by average evidence and the weight of evidence was obtained by support degree. Then the evidences were fused by evidence theory combination rule to recognize the target. In simulations, the proposed method was compared with the existing combination methods, and the result shows that this method is more effective in combining conflict evidence than the others.