单样本人脸识别中的虚拟样本扩展方法研究  

Extension Samples Methods of Face Recognition with One Training Image Per Person

作　　者：孙士明[1] 王芹芹[1] 纪友芳[1] 

机构地区：[1]中国石油大学计算机与通信工程学院,东营257061

出　　处：《微计算机应用》2010年第9期6-11,共6页Microcomputer Applications

摘　　要：在单训练样本人脸识别问题中,每个人都只有一幅人脸样本图像,因为没有充分具有代表性数量的训练样本,一些常用方法识别率明显下降,有的甚至不能使用。针对这一问题的解决方法是引进缩放、旋转、剪切、翻转和镜像等几何变换增加虚拟样本数量以提高传统识别方法的识别率。对增加后的样本使用传统的二维主成分分析方法做了验证,识别率得到了明显提高。At present there are many methods that could deal well with face recognition when there is sufficient number of representative training samples.However,few of them can work well when only one training sample per class is available.In order to deal with this question,geometry transformed which include scaling,rotating,shearing,flipping and mirroring etc is applied to the single example image to obtain its derived image.2DPCA is applied to extract features from derived face image and the recognition rate is improved significantly.

关 键 词：人脸识别 单样本 虚拟样本 二维主成分分析 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]