证件照片的一种混合人脸识别方法  

A Mixed Face Recognition Method Based on ID Card

作　　者：华强胜[1] 陈志刚[1] 

机构地区：[1]中南大学信息科学与工程学院,长沙410083

出　　处：《计算机工程》2003年第10期65-67,共3页Computer Engineering

基　　金：国家留学回国人员基金(教外司留[2000]479);教育部骨干教师资助计划(教技司[2000]143)

摘　　要：在仔细分析证件照片中人脸特点的基础上，提出了一种把人脸的几何特征矢量匹配和人脸的分块加权匹配相结合的思想。该方法针对一般人脸识别方法不能有效消除人脸表情影响的特点，首先对人脸进行快速准确的眼睛定位、图像摆正以及标准化处理，然后一方面抽取能够避免人脸表情影响的几何特征向量，另一方面对标准人脸进行分块加权匹配，最后进行综合识别。对JAFFE人脸库的试验结果表明，该方法识别准确率高，能够有效地消除人脸表情在识别中的影响，结果令人满意。Based on the analysis of human face in ID card, this paper proposes a new face recognition method, which integrates the geometric eigenvector matching with the weighted face subdivision matching. The current recognition methods cannot effectively smooth out the disturbance caused by all kinds of face expressions, so the presented method firstly makes a rapid and correct location of the eyes, balances and normalizes the image. Then on one hand, it extracts the geometric eigenvectors, which can avoid the expression disturbances; on the other hand, it makes a weighted face subdivision matching on the normalized faces. Finally it outputs the final result based on the two previous matchings. The experimental results on the JAFFE face database show that the given method not only has a high recognition rate, but also can effectively smooth out the face expression disturbances, so it gets a satisfactory result.

关 键 词：人脸表情影响 人脸标准化 人脸几何特征 人脸分块匹配 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]