基于余类零空间与最近距离的人脸识别算法

摘要

提出了一种新的人脸识别算法,即基于余类零空间与最近距离的人脸识别算法.通过构建不同类别的人脸图像的余类零空间与子空间,可以将不同类别的人脸最大化地区别出来.本算法的主要思想在于:测试图像与所属类别图像的子空间之间的距离最小,而与所属类别的图像的余类零空间距离最大.本算法基于ORL数据集与AR数据集进行了测试.从这些人脸数据集上的测试结果可以看出,本文提出的算法在PCA降维方法的基础上,比一些常见的算法所使用的判别方式更有效,如最近邻分类器(NN)所使用的最近距离判别方式、最近空间分类器(NS)所使用的最近空间距离判别方式、最近最远子空间分类器(NFS)所使用的最近最远空间距离判别方式等.%This paper presents a new scheme for face recognition, namely face recognition using null-space combined with nearest space distance classifier. By constructing the null-spaces and the subspaces of different types of human face images, different types of human face images are distinguished at the maximum degree. This idea considers that a test image has the shortest distance from its own class-specific subspace and has the farthest distance from its own class-specific null-space. The proposed classifier is evaluated on ORL database and AR database. Experiments on these databases demonstrate that the proposed scheme is more effective than some discriminants used by common classifiers, such as nearest distance used by nearest neighbor classifier, nearest space distance used by nearest space classifier and nearest-farthest subspace distance used by nearest-farthest subspace classifier.