摘要
为了实现铁路实名制检票时旅客的人脸图像与其身份证上人脸图像的自动比对,提出1种基于面部不变特征的人脸身份认证算法。以人的面部不变特征为前提,采用改进的尺度不变特征变换算法,提取现场采集的旅客的人脸图像及其身份证上的人脸图像的关键点,将靠近关键点的区域划分为部分重叠的子区域,然后以图像的词包模型差为基元构建人脸差特征空间,对训练图像的类别信息进行建模;对支持向量机(SVM)分类器训练分类的过程进行优化,训练优化的SVM分类器;最后,使用人脸差特征空间和训练好的SVM分类器进行加权投票,确认身份证上的人脸图像与现场采集的人脸图像是否为同一个人,实现旅客身份的认证。实验结果表明,在图像采集的尺度、角度和光照等不可控的情况下,该算法能够达到较高的认证速度和准确率。
To achieve the face automatic verification between the face images collected at the scene and the ID cards＇ of passengers, the algorithm using facial invariant features is proposed. Based on facial invariant features, the improved scale invariant feature transform algorithm is adopted to extract the key points of face images. The region near the key points is divided into partially overlapping sub-regions. Then the fa- cial difference space based on the bag of visual words model is constructed to model the category information of training samples. Next, the training process of Support Vector Machine （SVM） classifiers is optimized and the classifiers are trained. Finally, using facial difference space and the trained SVM classifiers by the method of weighted voting, it decides whether it is the same person of the collected face image and the image of ID card＇s, and it achieves the face verification of the passengers. Experiments show that the algorithm can accomplish face verification effectively, even in the real scene with low image quality, uncontrollable light and angle variations. It enhances the certification speed and achieves the higher accuracy.
出处
《中国铁道科学》
EI
CAS
CSCD
北大核心
2015年第1期133-139,共7页
China Railway Science
基金
国家自然科学基金资助项目(201313342056)
铁道科学技术研究发展中心科研项目(1119DZ4303)
关键词
火车票实名制
自动检票
人脸认证
尺度不变特征变换
词包模型
人脸差特征空间
Real-name system for ticket
Automatic ticket checking
Face verification
Scale invariant fea ture
Bag-of-visual words model
Facial difference space
