Abstract: A second-order feature selection method based on complete 2DPCA is proposed for face recognition. The method based on complete 2DPCA uses immune algorithm and five-fold cross-validation method which is based on the nearest neighbor classifier to complete second-order selection of facial features in order to overcome the demerits of PCA and some improved PCA methods. Experiments based on ORL face library and Yale face library show that the method can achieve good recognition effect.

Key words: face recognition, complete Two-dimensional Principle Component Analysis(2DPCA), immune algorithm, nearest neighbor classifier

摘要： 提出一种基于完全二维主元分析(2DPCA)的二次特征选择方法用于人脸识别。该方法针对PCA及其改进方法的不足，结合完全2DPCA方法，用免疫算法和基于最近邻分类器的5阶交叉验证方法完成对人脸特征二次选择。基于ORL人脸数据库和Yale人脸数据库的实验结果表明，该方法识别效果较好。

关键词: 人脸识别, 完全二维主元分析, 免疫算法, 最近邻分类器

CLC Number: 

• TP301.6