| 基于网格纵横局部二值模式的三维人脸识别 |
| |
| 引用本文： | 汤兰兰,达飞鹏. 基于网格纵横局部二值模式的三维人脸识别[J]. 仪器仪表学报, 2016, 37(6): 1413-1420 |
| |
| 作者姓名： | 汤兰兰 达飞鹏 |
| |
| 作者单位： | 东南大学四牌楼校区自动化学院,东南大学四牌楼校区自动化学院 |
| |
| 基金项目： | 国家自然科学基金项目（面上项目,重点项目,重大项目）;教育部博士点基金 |
| |
| 摘 要： | 提出一种基于网格纵横局部二值模式的三维人脸识别方法。为了充分体现人脸表面的细节差异,有效表示出由表情引起的局部形状变化,本文首先在人脸网格表面的半刚性区域内检测关键点集,并且由关键点确定中心面片,由中心面片及其周围有序环确定关键点邻域;其次将网格上中心面片及其周围有序环看成一个整体,从纵向(相邻环上对应面片之间)和横向(同一环上相邻标号面片之间)分别提取网格纵向局部二值模式描述符和网格横向局部二值模式描述符;然后对这两者进行特征融合得到网格纵横局部二值模式描述符;最后利用LC-KSVD2字典学习算法在Bosphorus数据库和FRGC v2.0数据库上完成识别实验。在Bosphorus数据库上各表情的平均Rank-1识别率为97.6%,在FRGC v2.0数据库上的Rank-1识别率为97.9%,该实验结果充分表明本文所提算法具有较高的识别精度,并且对表情变化具有一定的鲁棒性。
|
| 关 键 词： | 3D人脸识别 关键点检测 网格纵向局部二值模式 网格横向局部二值模式 特征融合 LC-KSVD2算法 |
| 收稿时间： | 2016-03-24 |
| 修稿时间： | 2016-05-03 |
A 3D Face Recognition method based on the Local Binary Pattern from Vertical and Horizontal on the mesh |
| |
| Tang Lanlan,Gai Shaoyan,Da Feipeng and Deng Xing. A 3D Face Recognition method based on the Local Binary Pattern from Vertical and Horizontal on the mesh[J]. Chinese Journal of Scientific Instrument, 2016, 37(6): 1413-1420 |
| |
| Authors: | Tang Lanlan Gai Shaoyan Da Feipeng Deng Xing |
| |
| Affiliation: | 1. School of Automation, Southeast University, Nanjing 210096, China; 2. Key Laboratory of Measurement and Control of CSE, Ministry of Education, Nanjing 210096, China,1. School of Automation, Southeast University, Nanjing 210096, China; 2. Key Laboratory of Measurement and Control of CSE, Ministry of Education, Nanjing 210096, China,1. School of Automation, Southeast University, Nanjing 210096, China; 2. Key Laboratory of Measurement and Control of CSE, Ministry of Education, Nanjing 210096, China and 1. School of Automation, Southeast University, Nanjing 210096, China; 2. Key Laboratory of Measurement and Control of CSE, Ministry of Education, Nanjing 210096, China |
| |
| Abstract: | A 3D face recognition method based on the local binary pattern from vertical and horizontal on the mesh is proposed in this paper to reflect the differences in the details on the face mesh fully and describe the local shape changes due to expression variations efficiently. First, the keypoints in the semi-rigid region of the face are detected, and the neighborhood of a keypoint is referred to the region constituted by the central facet which is determined by the keypoint and its surrounding concentric ordered rings. Then, the central facet and its surrounding ordered ring facets are taken as a whole, and the local binary pattern from vertical on the mesh(mesh-VLBP) descriptor and the local binary pattern from horizontal on the mesh(mesh-HLBP) descriptor are proposed, respectively, where vertical refers to the relation between the corresponding facets on two adjacent rings and horizontal refers to the relation between the facets with adjacent labels on one ring. Next, the feature-level fusion of the two descriptors is performed to obtain the Local Binary Pattern from vertical and horizontal on the mesh(mesh-VHLBP) descriptor. Finally, the recognition experiments are conducted using Label consist-KSVD2 learning algorithm on the Bosphorus database and the FRGC v2.0 database. The average Rank-1 recognition rate of expression variations on the Bosphorus database is 97.6% and the Rank-1 recognition rate on the FRGC v2.0 database is 97.9%, which strongly demonstrate that the proposed method is efficient and robust to expression variations. |
| |
| Keywords: | 3D face recognition keypoints detection mesh-VLBP mesh-HLBP feature-level fusion LC-KSVD2 algorithm |
| 本文献已被 CNKI 等数据库收录！ |
| 点击此处可从《仪器仪表学报》浏览原始摘要信息 |
|
点击此处可从《仪器仪表学报》下载免费的PDF全文 |
|
