文章导读
总览 评价 明悦 1, , 阮秋琦 2,* ( 1、 北京交通大学信息科学研究所,北京 100044; 2、 北京交通大学信息科学研究所,北京,100044,中国; ) 摘要: 鲁棒的分区域边界球描述子用于辅助三维人脸识别。在我们的框架中,我们首先运用曲率信息和人脸的形
明悦1,, 阮秋琦2,*
(
1、北京交通大学信息科学研究所,北京 100044; 2、北京交通大学信息科学研究所,北京,100044,中国; )
摘要:
鲁棒的分区域边界球描述子用于辅助三维人脸识别。在我们的框架中,我们首先运用曲率信息和人脸的形状特性分割原始的三维点云数据。然后,提取的区域被投影到边界球带上用以反映人脸不同区域的显著的形状信息。接着,一个正交区域和全局回归技术用来提取辨别特征矢量。基于不同人脸数据集的实现结果显示平衡区域和全局的人脸特性可以取得高质量的性能。对比以前的传统方法,我们的框架具有一致的有效性、鲁棒性和通用性。
关键词:
分区域边界球描述子;三维人脸识别;正交区域和全局回归技术
MING Yue1,, RUAN Qiuqi2,*
(
1、Institute of Information Science, Beijing Jiaotong University, Beijing 100044; 2、Institute of Information Science, Beijing Jiaotong University, Beijing 100044, P.R. China; )
Abstract:
A robust regional bounding sphere representation (RRBSR) is introduced to facilitate 3D face recognition. In our framework, we first segment a group of regions on each 3D facial point cloud by curvature information and facial shape characteristics. Then, the extracted regions are projected on the bounding spherical bands in order to reflect the distinctive shape information of facial different regions. Next, an orthogonal regional and global regression (ORGR) is utilized to extract the discriminant feature vectors. Experimental results based on the different 3D face databases demonstrate that balancing regional and global facial characteristics allow for the high-qualified performance. Compared with the previous popular approaches, our framework has a consistently better performance in terms of effectiveness, robustness, and universality.
Tag:
点此返回栏目查看更多>>>参考论文