摘要
三维人脸数据的获取会受到成本以及可访问性的影响。通过对深度相机(如Xtion pro live)获取人脸数据过程的研究可知,它能够很容易获得彩色和深度结合(RGB-D)图。针对RGB-D图,使用局部和整体混合识别,利用局部二值的平均信息熵模式(LBEP),快速提取RGB-D图的直方图信息和特征向量,根据不同区域在表情不同情况下的变化程度,对不同区域的识别效果赋予不同的权值,进行加权运算。实验结果表明,相比现有的二维和三维人脸识别算法,改进的LBEP算法识别率有明显的提升。
The extraction of 3D face images data was affected by costs and accessibility.According to the study of acquiring face data process by depth camera（Xtion pro live）,RGB-D diagrams can be easily obtained.As for the RGB-D image,the local and global hybrid identification was used to extract the histogram information and feature vector based on the local binary average entropy pattern（LBEP）.According to different changes in some regions,the rate of recognition was given different weights and then they were calculated.Experimental results show that compared with 2D face recognition and 3D face recognition algorithm,improved LBEP algorithm apparently increases the discrimination ratio.
出处
《计算机工程与设计》
北大核心
2016年第12期3366-3370,共5页
Computer Engineering and Design
基金
山西省自然科学基金项目(2015011045)
