摘要
针对复杂环境下人脸识别难度大的问题,提出了一种熵权法融合局部Gabor特征方法。计算类熵加权向量;计算局部归一化输入图像的Borda计数矩阵,从而消除低值Gabor jet比较矩阵;通过将分数层类熵加权Gabor特征与LGBP和LGXP融合解决了完成人脸的识别。在FERET、AR和FRGC 2.0人脸数据库上的实验结果表明,该方法对轻微姿态变化具有显著鲁棒性,并且对人眼检测中高达3像素的误差具有鲁棒性,相比其他几种人脸识别方法,该方法取得了更好的识别效果。
For the big challenge difficulty of face recognition under the complex environments, a fusion method based on entropy weighted method and local Gabor features is proposed. Class entropy weighting vectors are calculated. Borda count matrix is calculated to remove low-value Gabor jet comparison matrix. Layered class entropy weighted Gabor feature is fused with LGBP and LGXP respectively so as to finish face recognition. Experimental results on FERET, AR, and FRGC 2.0 databases show that proposed method shows significant robustness to slight pose variations and errors of up to3 pixels in eye detection. It has better recognition efficiency than several other recognition methods.
出处
《计算机工程与应用》
CSCD
北大核心
2016年第5期134-140,共7页
Computer Engineering and Applications
基金
国家自然科学基金(No.F020704)
广东省教育科学"十二五"规则课题(No.2012JK304)
广东省科技计划项目(No.2013B070206046)
关键词
GABOR特征
人脸识别
鲁棒性
局部归一化
熵权法
Gabor feature
face identification
robust
local normalization
entropy weighted method
