摘要
为增强三维人脸识别系统对表情变化的鲁棒性,提出一种融合深度数据与面部刚性区域的人脸识别系统。首先根据人脸几何特征定位鼻尖点,以该点为中心切割出有效的面部区域,把各种姿态的人脸统一矫正到正面姿态;然后以深度图作为整体特征采用二维主成分分析(2DPCA)算法进行匹配,以面部刚性区域作为局部特征采用改进的迭代最近点(ICP)算法进行匹配;最后将所提取的整体特征匹配结果和局部特征匹配结果进行融合。在CASIA 3D人脸库上的实验结果表明,构造的系统识别率高于使用单一特征的系统,且对表情变化有较好的鲁棒性。
In order to enhance the robustness of 3D face recognition to expression change, a face recognition system is proposed by fusing the depth data and facial rigid region in this paper. First, the tip point of the nose is located according to the facial geometric features, the effective facial region is cut around the center of the tip point, and the faces with dif- ferent poses are transformed to the normalized front pose. Then the depth image as global feature is matched using the 2D principal component analysis （2DPCA） algorithm; and the facial rigid region as local feature is matched using the modi- fied iterative closest point （ICP） algorithm. Finally, the extracted matching results of the global and local features are fused. The experiment results on the CASIA 3D face database show that the proposed system has higher recognition rate than the system with single feature, and also has better robustness to facial expression change.
出处
《仪器仪表学报》
EI
CAS
CSCD
北大核心
2014年第2期299-304,共6页
Chinese Journal of Scientific Instrument
基金
中央高校基本科研专项(CDJXS12160004)资助项目
关键词
人脸识别
局部特征
整体特征
特征融合
刚性区域
face recognition
local feature
global feature
feature fusion
rigid region
