一种考虑分类器不确定性的多通道D-S融合方法

A D-S based multi channel information fusion method using classifiers uncertainty measurement

不同的多个分类器能利用被识别对象的互补信息 ,从而提高系统的综合识别能力。然而 ,各分类器的分类结果中存在不确定性信息 ,不确定性值的大小反映了分类结果的优劣 ,这个值可以通过距离的方式来计算。本文从几何角度出发 ,提出了分类器结果的不确定性测量方法 ,在利用 D- S方法合成多个分类器的输出时 ,将这些不确定性信息也融入计算之中。数值分析结果表明 。

Classifiers of different types complement one another in classification performance, and the fusion of multiple different classifiers can improve the accuracy of pattern recognition system. There exists uncertainty in the outputs of classifiers, which reflects the performance of the classification results. A geometrically inspired measurement of uncertain is proposed in this paper, and this measurement is considered when combining multiple classifiers using D S evidence theory. Numerical analysis results show that this method is effective and reasonable.