基于信息融合的三维人脸识别
Three-Dimension Face Recognition Based on Information Fusion
廖广军 1曾玮 2刘屿 1李致富 3胡跃明1
作者信息
- 1. 华南理工大学 自动化科学与工程学院,广东 广州510640; 华南理工大学 精密电子制造装备教育部工程中心,广东 广州510640
- 2. 华南理工大学 自动化科学与工程学院,广东 广州510640
- 3. 华南理工大学 精密电子制造装备教育部工程中心,广东 广州510640; 华南理工大学 机械与汽车工程学院,广东 广州510640
- 折叠
摘要
为了克服二维人脸识别的不足,提出采用Kinect采集设备并基于信息融合进行三维人脸识别。在人脸检测方面,利用双眼位置定位人脸,并通过肤色模型和深度信息实现人脸过滤。在人脸识别方面,将动态时间规整算法应用于人脸轮廓快速提取,并提出了一种改进的局部区域二值化特征(LBP)匹配来实现人脸识别,将彩色图像信息和人脸生理特征信息的分析应用于区域LBP算法的权值确定。实验表明:文中提出的基于信息融合的三维人脸识别方法具有使用价值,识别方法的实时性和鲁棒性方面得到一定提升。
Abstract
In order to overcome the disadvantages of two-dimension face recognition,a three-dimension method using a Kinect instrument is proposed based on the information fusion.In this method,for the face detection,the eyes' location is used to locate the face,and the skin color model as well as the depth information is employed to realize the face filtering.For the face recognition,the DTW (Dynamic Time Warping)algorithm is applied to the rapid extraction of the face contour,and an improved LBP (Local Binary Pattern)matching algorithm is proposed, in which the weight value distribution is determined by analyzing the color image information and the face physiolo-gical characteristics.Experimental results show that the proposed method for face recognition is applicable and is of improved real-time capability and robustness.
关键词
人脸检测/三维人脸识别/信息融合/动态时间规整/区域二值化特征匹配Key words
face detection/three-dimension face recognition/information fusion/dynamic time warping/local binary pattern matching引用本文复制引用
基金项目
出版年
2013