基于对称性特征分块主成分分析的人脸识别方法

为了进一步提高人脸识别的精度,考虑在分块主成分分析算法中引入对称性思想。首先对图像进行分块并分别求其奇偶对称脸,然后利用主成分分析算法提取图像的主要鉴别特征。该方法充分考虑了光照等多种因素对识别率的影响,利用人脸图像的对称性增加了样本数量,以有效提高识别率。在ORL人脸库上的实验显示,在每类训练样本数为7、提取特征数为20的情况下,基于对称性特征的分块主成分分析方法的人脸识别率为95%,说明该方法是有效的。

Method of Face Recognition Based on Symmetry Block Principal Component Analysis

In order to improve the performance of face recognition,this paper includes the idea of symmetry in block principal component analysis algorithm.Firstly block the images,and then get the even symmetry faces and the odd symmetry faces,applying the principal component analysis algorithm for extracting the image of the main characters.This method considers light and other factors which affect the recognition rate.The use of the symmetry in face images increases the sample number to effectively improve the recognition rate.Finally,experiment on the ORL face shows that,when the training number is 7 and the extracted Eigen number is 20,the face recognition rate using our method is 95%.This verifies the effectiveness of this method.