PCA算法提供了一个高维和低维间的线性变换矩阵,这个变换矩阵可以通过求取协方差矩阵的特征向量获得。特征值较大的特征向量反映人脸最大差异性;根据脸部固定结构特点构造人脸平均模板,利用模板匹配来检测图像中的人脸,计算待测图像与特征空间的距离进一步判别是否是数据库中人脸。实验表明,PCA算法在视频监控系统的人脸识别中可以很好地实现人脸特征提取和检测。
The PCA algorithm provides a linear transformation matrix from high -dimensional matrix to low-dimensional matrix. And this linear trans- formation matrix can be obtain by solved the proper vector of the covariance matrix, and the bigger characteristic value of the proper vector can reflect the biggest diversity of person's face. The person's face average template can be constructed according to the face fixed unique feature. And this template can be used to matching the face in the image. The result of the experiment indicates that the PCA algorithm is successful which is used in realizing the person face feature extraction and examination in the person face recognition system.