文献类型：期刊文献

英文题名：3D Face Recognition Using Spherical Vector Norms Map

作者：Wang, Xue-Qiao[1,2];Yuan, Jia-Zheng[1,3];Li, Qing[1,2]

第一作者：王希庆

通讯作者：Wang, XQ[1];Wang, XQ[2]

机构：[1]Beijing Union Univ, Beijing Key Lab Informat Serv Engn, Beijing 100101, Peoples R China;[2]Beijing Union Univ, Comp Technol Inst, Beijing 100101, Peoples R China;[3]Capital Normal Univ, Beijing Adv Innovat Ctr Imaging Technol, Beijing 100048, Peoples R China

第一机构：北京联合大学北京市信息服务工程重点实验室

通讯机构：[1]corresponding author), Beijing Union Univ, Beijing Key Lab Informat Serv Engn, Beijing 100101, Peoples R China;[2]corresponding author), Beijing Union Univ, Comp Technol Inst, Beijing 100101, Peoples R China.|[1141734]北京联合大学智慧城市学院;[11417]北京联合大学;[11417103]北京联合大学北京市信息服务工程重点实验室;

年份：2017

卷号：33

期号：5

起止页码：1141-1161

外文期刊名：JOURNAL OF INFORMATION SCIENCE AND ENGINEERING

收录：;EI(收录号：20182905562712);Scopus(收录号：2-s2.0-85049875700);WOS:【SCI-EXPANDED(收录号:WOS:000410465400003)】；

基金：This paper was supported by the Foundation of Beijing Union University (No. Zk10201703).

语种：英文

外文关键词：spherical vector norms map; histograms of oriented gradients; 3D face recognition; linear discriminant analysis; face recognition grand challenge database

摘要：In this paper, we introduce a novel, automatic method for 3D face recognition. A new feature called a spherical vector norms map of a 3D face is created using the normal vector of each point. This feature contains more detailed information than the original depth image in regions such as the eyes and nose. For certain flat areas of 3D face, such as the forehead and cheeks, this map could increase the distinguishability of different points. In addition, this feature is robust to facial expression due to an adjustment that is made in the mouth region. Then, the facial representations, which are based on Histograms of Oriented Gradients, are extracted from the spherical vector norms map and the original depth image. A new partitioning strategy is proposed to produce the histogram of eight patches of a given image, in which all of the pixels are binned based on the magnitude and direction of their gradients. In this study, SVNs map and depth image are represented compactly with two histograms of oriented gradients; this approach is completed by Linear Discriminant Analysis and a Nearest Neighbor classifier.