一种基于模糊积分的多分类器人脸识别方法

提出了一种基于模糊积分的多分类器人脸识别新方法,首先采用两种不同的K-L变换及Fisher线性鉴别分析方法对原始图象进行特征抽取和压缩,然后将压缩后的特征设计成不同的分类器,再利用模糊积分对这些分类器进行融合,达到主观期望和客观证据间的最佳匹配,从而提高了对目标的识别率,并将该方法应用到人脸识别中,实验结果表明,该方法提高了识别率,证明其有效性。

A Novel Face Recognition Method Based on Multiclassifier Fusion with Fuzzy Integral

A novel face recognition method based on multiclassifier fusion with fuzzy integral is presented. First of all,two different kinds of K-L transform methods and Fisher linear discriminant analysis are used repectively to extract the features from the original image and reduce the dimensionalities of those features. Then,different classifiers can be designed on the basis of compressed features. Finally multiclassifier fusion is employed with fuzzy integral method. This method is put into use for face recognition and the experimental results show that the recognition rate is improved,so the validity is proved.