一种应用PCA的人脸图像集识别方法

A Face Images Set Recognition Method Applying PCA

提出一种适用于同类别的图像子集的类别判断方法。同类别的图像子集经过PCA主特征提取后,选择较大的p个特征值对应的线性无关的特征向量,组成特征矩阵,则同类别的图像子集可以转化成一个特征矩阵,图像子集与特征矩阵一一对应,进而整个图像库能够用矩阵集合来表示。定义一种矩阵间的距离及最小二乘距离,通过计算待测图像子集对应特征矩阵与图像库中不同类别对应特征矩阵之间的距离或最小二乘距离,判断待测试图像子集所属类别。

Proposes a classification judgment method applied to images subset belonged to the same classification. After extracting main features by PCA in images subset of the same classification, the p larger linearly independent feature vectors corresponding to features of the same amount have been chosen, composing feature matrix. Then images subset of the same classification can be transformed to a feature matrix, shaping a one-to-one correspondence between the images subset and feature matrix. In turn, the entire images library can be expressed by matrix sets. Defines a distance and least squares between matrices. By computing the distance or the least squares between a feature matrix corresponding to one images subset under test and feature matrices corresponding to images set of the entire images library, the classification of images subset under test can be judged.