基于线性和非线性算法的高效人脸识别系统

doi:10.16180/j.cnki.issn1007-7820.2019.07.016

摘要/Abstract

摘要：

文中提出了一种基于外观的线性和非线性人脸识别方法,所用的线性算法有主成分分析(PCA)和线性判别分析(LDA)。两种非线性方法分别是核主成分分析(KPCA)及核费希尔分析(KFA),线性降维投影方法基于二阶相依性编码模式信息,非线性方法用于处理三个或更多像素之间的关系。首先通过Gabor对图片进行预处理,然后采用线性、非线性分析进行降维。通过马哈利诺比斯-余弦(Mahcos)度量用于定义两幅图像通过相应的降维技术后的相似性度量。实验表明,当与Gabor小波一同使用时,LDA和KFA的性能最高,分别为CMC和ROC结果的93.33%。通过对AT&T数据库400幅图像的综合分析,发现线性和非线性算法的性能受图像分类数目、图像预处理及识别测试集的人脸图像数目的影响。

关键词: 主成分分析, 线性判别分析, 核主成分分析, 核费希尔分析, 人脸识别, 降维技术

Abstract:

This paper presented a linear and non-linear face recognition method based on appearance. The linear algorithms used include principal component analysis (PCA) and linear discriminant analysis (LDA). Two non-linear methods were kernel principal component analysis (KPCA) and nuclear fisher analysis (KFA).Linear dimensionality reduction projection method was based on second-order dependency coding mode information, and non-linear method was used to deal with the relationship between three or more pixels. In the last stage,the Mahcos measure was used to define the similarity measure of two images after the corresponding dimension reduction technology.Experiments show that when used together with Gabor wavelet,LDA and KFA had the highest performance, 93.33% of CMC and ROC results, respectively. Through the comprehensive analysis of 400 images in AT&T database, it was found that the performance of linear and non-linear algorithms was affected by the number of image classification, image preprocessing and the number of face images in recognition test set.

Key words: principal component analysis, linear discriminant analysis, kernel principal component analysis, kernel fisher analysis, face recognition, dimension reduction technology

中图分类号:

TP319

刘芬. 基于线性和非线性算法的高效人脸识别系统[J]. 电子科技, 2019, 32(7): 82-86.

LIU Fen. An Efficient Face Recognition System Based on Linear and Nonlinear Algorithms[J]. Electronic Science and Technology, 2019, 32(7): 82-86.

图/表 3

参考文献 16

[1]	Olivaresmercado J, Toscanomedina K, Sanchezperez G , et al. Face recognition system based on MOTIF features[J]. Journal of Modern Optics, 2018,65(2):1-9.
[2]	Flobak Å, Vazquez M, Lægreid A , et al. CImbinator: A Web based tool for drug synergy analysis in small and large scale datasets[J]. Bioinformatics, 2017,33(15):2410-2412.
[3]	Yue W, Qiang J . Facial landmark detection: a literature survey[J].International Journal of Computer Vision, 2018(5):1-28.
[4]	Abbad A, Douini Y, Abbad K , et al. Post-processing of dimensionality reduction methods for face recognition[J]. Pattern Recognition & Image Analysis, 2017,27(2):266-275.
[5]	Yang J, Wang H, Ding H , et al. Nonlinear dimensionality reduction methods for synthetic biology biobricks’ visualization[J]. BMC Bioinformatics, 2017,18(1):47-49.
[6]	Luo A, An F, Zhang X , et al. A hardware-efficient recognition accelerator using haar-like feature and SVM classifier[J]. IEEE Access, 2019,39(9):1-19.
[7]	张持健, 刘雪, 张贺 , 等. 基于PCA和非线性SVC的小数据人脸识别[J]. 无线电通信技术, 2019,45(1):77-81.
	Zhang Chiijian, Liu Xue, Zhang He , et al. Small data face recognition based on PCA and nonlinear SVC[J]. Radio Communication Technology, 2019,45(1):77-81.
[8]	李昆仑, 李尚然, 王琳 , 等. 核协同近邻表示的人脸识别算法[J]. 小型微型计算机系统, 2018,39(10):194-199.
	Li Kunlun, Li Shangrun, Wang Lin , et al. Face recognition algorithms based on kernel co-neighbor representation[J]. Minicomputer System, 2018,39(10):194-199.
[9]	Ghasemzadeh A, Azad S S, Esmaeili E . Breast cancer detection based on Gabor-wavelet transform and machine learning methods[J].International Journal of Machine Learning & Cybernetics, 2018(5):1-10.
[10]	Harit A, Joshi J C C,Gupta K K. . Facial emotions recognition using gabor transform and facial animation parameters with neural networks [C]. Indianapolis:Materials Science & Engineering Conference Series, 2018.
[11]	Glaze C M ,Filipowicz A L S,Kable J W,et al .A bias-variance trade-off governs individual differences in on-line learning in an unpredictable environment[J].Nature Human Behaviour, 2018(3):1-7.
[12]	Beiji Z, Mohammed N, Chengzhang Z , et al. Overall Gabor Classifier (OGC) with kernel partial least square discrimination[J]. Journal of Computational & Theoretical Nanoscience, 2017,14(8):3727-3736.
[13]	徐梦珂, 许道云, 魏明俊 . 基于2D-KPCA的拉普拉斯特征映射人脸识别[J]. 计算机应用研究, 2017,34(7):2212-2215.
	Xu Mengke, Xu Daoyun, Wei Mingjun . Face recognition based on laplace feature mapping of 2D-KPCA[J]. Computer Applied Research, 2017,34(7):2212-2215.
[14]	邓道举, 李秀梅 . 基于KPCA和投影字典对学习的人脸识别算法[J]. 计算机系统应用, 2018,27(5):145-150.
	Deng Daoju, Li Xiumei . Face recognition algorithms based on KPCA and projection dictionary for learning[J]. Computer System Applications, 2018,27(5):145-150.
[15]	李懿, 刘晓东 . 一种新的模糊稀疏表示人脸识别算法[J]. 大连理工大学学报, 2017,57(2):189-194.
	Li Yi, Liu Xiaodong . A new fuzzy sparse representation face recognition algorithm[J]. Journal of Dalian University of Technology, 2017,57(2):189-194.
[16]	Ragab A, Carnavalet X D, Yacout S , et al. Face recognition using multi-class logical analysis of data[J]. Pattern Recognition & Image Analysis, 2017,27(2):276-288.

	人脸识别性能指标
	CMC		ROC			EPC
	不包含Gabor滤波器的一级识别率(使用评估集/%)	包含Gabor滤波器的一级识别率(使用评估集/%)	不包含Gabor滤波器的验证率(使用评估集/%)	包含Gabor滤波器的验证率(使用评估集/%)	包含Gabor滤波器的验证率(使用测试集/%)	包含Gabor滤波器的总错误率的一半(使用测试集/%)	不包含Gabor滤波器的总错误率的一半(使用测试集/%)
PCA	66.07	74.15	66.75	92.49	63.12	1.61	4.72
LDA	86.03	93.31	76.42	93.33	91.87	1.56	4.09
KPCA	49.18	79.98	51.42	92.49	92.46	2.68	8.80
KFA	85.65	93.32	60.69	93.32	93.33	1.85	6.66