基于局部分割的快速人脸识别  

Fast Face Recognition Based on Local Segmentation

作　　者：尉宇[1] 罗浩[1] 

机构地区：[1]武汉科技大学信息科学与工程学院,武汉430081

出　　处：《科学技术与工程》2017年第31期133-138,共6页Science Technology and Engineering

基　　金：国家自然科学基金(61673304)资助

摘　　要：针对人脸识别系统易受光照,表情与遮挡等因素的影响,提出一种基于局部分割的快速人脸识别算法。首先,建立高斯肤色模型,并融合几何特征快速实现人脸粗定位;然后,利用局部初次匹配与全局特征的方法排除背景环境对人脸检测的干扰以及减少匹配过程中的计算量,同时运用LBP算子对2D-Fr FT的幅度与相位特征的互补信息进行编码;最后,采用最近邻分类器进行分类识别。本文的方法在公共人脸图像数据库上进行仿真实验。结果表明,该算法简单、鲁棒性高、在速度与准确性方面具有良好的性能。In view of the influence of face recognition system,such as illumination,facial expression and occlusion,etc,this paper proposes a novel fractional Fourier features method based on local segmentation for fast face recognition. Firstly,human face is separated from input image by constructing Gaussian skin color model and employing geometrical relationships to rapidly achieve the purpose of detecting potential face region. Then local initial matching and global feature methods are adopted to eliminate the impact of surroundings and improve the matching speed,meanwhile,complementary information of 2 D-Fr FT is taken full advantage of performing feature extraction and local LBP operator encodes its texture features. Eventually,a nearest neighbor classifier is utilized for classification. The proposed scheme is examined on public available database. The experimental results show that our proposed method is not only simple and robust,but good performance in recognition speed and rate.

关 键 词：人脸识别 颜色空间 分数阶傅里叶变换(FrFT) 局部二值模式 模板匹配 二维主成分分析 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]