基于四元数分数阶伪Zernike矩的人脸验证算法  

Face verification algorithm based on quaternion fractional pseudo-Zernike moment

作　　者：俞铭 褚红健 李佑文 Yu Ming;Chu Hongjian;Li Youwen(Guodian Nanjing Automation Co.,Ltd.,Nanjing 211100,China;Nanjing SAC Rail Traffic Engineering Co.,Ltd.,Nanjing 210032,China)

机构地区：[1]国电南京自动化股份有限公司,江苏南京211100 [2]南京国电南自轨道交通工程有限公司,江苏南京210032

出　　处：《江苏科技信息》2021年第24期49-51,共3页Jiangsu Science and Technology Information

摘　　要：为了提高人脸验证的效率和准确率,文章提出一种基于四元数分数阶伪Zernike矩的人脸验证方法,首先将彩色人脸图像按RGB三通道分成IR,IG,IB三幅图像;然后采用四元数分数阶伪Zernike矩变换快速计算方法将三通道上的图像变换成一个四元数分数阶伪Zernike矩特征矩阵,获得彩色人脸图像的特征矩阵;最后使用K-NN算法将所述彩色人脸图像的识别特征矩阵与预先建立的人脸图像特征库中的识别特征矩阵进行匹配,完成人脸验证。在公开数据集上的实验结果表明,提出的四元数分数阶伪Zernike矩提取的彩色人脸特征具有较高的鲁棒性,提出的快速计算算法有效地提高了人脸验证模型的验证效率。In order to improve the efficiency and accuracy of face verification,a face verification method based on quaternion fractional pseudo-Zernike moment is proposed.First,color face images are divided into IR,IG and IB images according to RGB channels.Then the image on the three channels is transformed into a quaternion fractional pseudomatch the recognition feature matrix of the color face image with the recognition feature matrix in the pre-established face image feature database to complete face verification.The experimental results on open datasets show that the proposed computation algorithm effectively improves the verification efficiency of the face verification model.

关 键 词：人脸验证 四元数伪Zernike矩 快速计算 

分 类 号：TP277[自动化与计算机技术—检测技术与自动化装置]