基于人脸特征的身份认证系统的研究与设计
|
摘要
基于生物特征的身份认证技术在社会生活中具有越来越重要的地位和作用,已逐渐成为信息安全的重要研究领域。在多种生物特征中,基于人面部特征的识别和认证因为具有无侵害性、成本低、隐蔽性、不需要被测者特殊配合等优点,得到了广泛的关注和重视,具有广阔的应用前景。
本文通过对几种常见的身份认证技术的分析,总结了现有的身份认证系统存在的问题和不足,在此基础上将生物特征之一的人脸特征应用于身份认证技术中,设计并实现了一个基于人脸特征的身份认证原型系统,对系统的功能、框架结构和涉及的关键技术等进行了研究和分析,并对所设计系统的性能进行了测试。结果表明,该系统能够较为准确地完成身份认证任务,且对一定范围内的人脸姿态、光照变化等具有较强的健壮性。
研究的关键技术包括人脸和眼睛的检测与定位、图像预处理、人脸特征提取和分类器设计等。本文利用AdaBoost学习算法构建了人脸检测和眼睛检测分类器,实现了人脸和眼睛的检测与定位,实验表明,该算法能够实时、准确地检测并定位出人脸和眼睛。针对人脸图像存在的姿态、光照变化等问题,结合人眼定位时的眼睛坐标和图像预处理技术提出了人脸图像的规范化处理方法,有效地减少了姿态、光照变化对系统性能带来的影响。特征提取方法的优劣直接影响到系统的认证性能。由于Gabor小波核函数具有较强的空间位置和方向选择性,能够捕捉对应于空间和频率的局部结构信息,表达人脸最有用的局部特征,因此本文采用人脸图像的Gabor滤波响应来表示人脸特征。为解决Gabor特征维数过高的问题,结合主分量分析和线性鉴别分析方法各自的优势提出了两种特征提取方案,并在简单分类器和SVM分类器上进行了实验。实验表明,GP特征提取方案在SVM分类器上获得了较好的实验效果。
The identity verification technology based on biometric features played an increasingly important role in our society; it has been a key research area in information security. Among the measured biometric features, facial feature identification and verification are gaining popularity and diverse applications for the reason that they are considered to be non-invasive, low cost, and natural biometric technologies.
In this thesis, several common identity verification technologies are analyzed and the problems and shortages of the present identity verification system are summarized. Base on this, face features are applied to identity verification technology and a face-based identity verification prototype system is designed and implemented. The function module, frame, and key technologies of the system are researched and analyzed, and the performance of the system is tested. The results show that the system can achieve identity verification task precisely, and robustness to pose and lighting variance.
The key technologies in research include the detection and location of face and eyes, the image pre-processing, the extraction of facial features and the design of classifier. In this thesis, AdaBoost learning algorithm is applied to construct the classifier to realize face detection and eye location. The result shows that the algorithm can detect and locate face and eyes timely and precisely. Face image pre-processing method is put forward, which reduces the affect of pose and lighting variance on the system.
That whether the method of feature extraction is good or not can directly affects the verification ability of the system. Due to the better spatial and orientation selection of the Gabor wave kernel function which can capture local structure informations corresponding to spatial and frequency to express the most useful local features of faces, the thesis use the Gabor response of face image to represent face features. In order to deduce the high dimensions of the Gabor features, the basis of researching the advantages of both the principal component analysis and linear discriminant analysis, two feature extraction methods are put forward and examined on simple classifier and SVM classifier. The results show that GP feature extraction method on the SVM classifier obtains the better result.
本文通过对几种常见的身份认证技术的分析,总结了现有的身份认证系统存在的问题和不足,在此基础上将生物特征之一的人脸特征应用于身份认证技术中,设计并实现了一个基于人脸特征的身份认证原型系统,对系统的功能、框架结构和涉及的关键技术等进行了研究和分析,并对所设计系统的性能进行了测试。结果表明,该系统能够较为准确地完成身份认证任务,且对一定范围内的人脸姿态、光照变化等具有较强的健壮性。
研究的关键技术包括人脸和眼睛的检测与定位、图像预处理、人脸特征提取和分类器设计等。本文利用AdaBoost学习算法构建了人脸检测和眼睛检测分类器,实现了人脸和眼睛的检测与定位,实验表明,该算法能够实时、准确地检测并定位出人脸和眼睛。针对人脸图像存在的姿态、光照变化等问题,结合人眼定位时的眼睛坐标和图像预处理技术提出了人脸图像的规范化处理方法,有效地减少了姿态、光照变化对系统性能带来的影响。特征提取方法的优劣直接影响到系统的认证性能。由于Gabor小波核函数具有较强的空间位置和方向选择性,能够捕捉对应于空间和频率的局部结构信息,表达人脸最有用的局部特征,因此本文采用人脸图像的Gabor滤波响应来表示人脸特征。为解决Gabor特征维数过高的问题,结合主分量分析和线性鉴别分析方法各自的优势提出了两种特征提取方案,并在简单分类器和SVM分类器上进行了实验。实验表明,GP特征提取方案在SVM分类器上获得了较好的实验效果。
The identity verification technology based on biometric features played an increasingly important role in our society; it has been a key research area in information security. Among the measured biometric features, facial feature identification and verification are gaining popularity and diverse applications for the reason that they are considered to be non-invasive, low cost, and natural biometric technologies.
In this thesis, several common identity verification technologies are analyzed and the problems and shortages of the present identity verification system are summarized. Base on this, face features are applied to identity verification technology and a face-based identity verification prototype system is designed and implemented. The function module, frame, and key technologies of the system are researched and analyzed, and the performance of the system is tested. The results show that the system can achieve identity verification task precisely, and robustness to pose and lighting variance.
The key technologies in research include the detection and location of face and eyes, the image pre-processing, the extraction of facial features and the design of classifier. In this thesis, AdaBoost learning algorithm is applied to construct the classifier to realize face detection and eye location. The result shows that the algorithm can detect and locate face and eyes timely and precisely. Face image pre-processing method is put forward, which reduces the affect of pose and lighting variance on the system.
That whether the method of feature extraction is good or not can directly affects the verification ability of the system. Due to the better spatial and orientation selection of the Gabor wave kernel function which can capture local structure informations corresponding to spatial and frequency to express the most useful local features of faces, the thesis use the Gabor response of face image to represent face features. In order to deduce the high dimensions of the Gabor features, the basis of researching the advantages of both the principal component analysis and linear discriminant analysis, two feature extraction methods are put forward and examined on simple classifier and SVM classifier. The results show that GP feature extraction method on the SVM classifier obtains the better result.
引文
[1]International Biometric Group,The Biometrics Market and Industry Report 2007-2012.http://www.biometricgroup.com/reports/public/market.report.html.
[2]山世光.人脸识别中的若干关键问题的研究[D],中国科学院博士学位论文,2004.
[3]Zhao W,Chellappa R,Phillps P Jet al.Face Recognition:A Literature Survey[J].ACM Computing Surveys(CSUR),2003,35(4):399-458.
[4]Yang M H,Kriegman D J,Ahuja N.Detecting Faces in Images:A Survey[J].IEEE Transactions on Pattern Analysis and Machine Intelligence.2002,24(1):34-58.
[5]P.J.Philips,P.J.Grother,R.J.Micheals et al.Face Recognition Vendor Test 2002:Evaluation Report,Technical Report[R],National Institute of Standards and Technology,2003.
[6]Ding X Q,Fang C H.Discussions on Some Problems in Face Recognition[R].In:Li S Z,Lai J H,Tan T N et al,eds.Sinobiometrics2004.Berlin:Springer-Verlag,2004,3338:47-56
[7]Mu xiang Zhang.Adaptive protocol for Entity Authentication and Key Agreement in Mobile Networks[C].ICISC 2003,LNCS 2971,2004:166-183.
[8]Zhang N,Shi Q,Merabti M.Anonymous public-key certificates for anonymous and fair document exchange[J].IEEE Proceedings:Communications.2000,6(147):345-350.
[9]Sung K,Poggio T.Example-based learning for view based human face detection[J].IEEE Transaction on Pattern Analysis and Machine Intelligence,1998,20(1):39-51.
[10]艾英山,张德贤.人脸识别方法的综述与展望[J],计算机与数字工程,2005,33(10):24-27.
[11]苏建波,徐波.应用模式识别技术导论.人脸识别与语音识别[M].上海:上海交通大学出版社,2001:125-132.
[12]Yang M H.Recent Advances in Face Detection[C].IEEE 17th International Conference on Pattern Recognition 2004 Tutorial,Cambridge,United Kingdom,2004.
[13]Garcia C,Tziritas G.Face detection using quantized skin color regions merging and wavelet packet analysis[J].IEEE Transaction on Multmedia,1999,21(3):264-277.
[14]Yang G Z,Huang T S.Human face detection in a complex background[J].Pattern Recognition,1994,27(1):53-63.
[15]Moghaddam B,Pentland A.Probabilistic visual learning for object representation[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,1997,19(7):696-710.
[16]Moghaddam B,Pentland A.Beyong linear eigenspaces:Bayesian matching for face recognition[A].In:Wechsler H,Philips P J,Bruce V,et al,eds.Face Recognition:From Theory to Applications[C],Berlin:Springer-Verlag,1998:230-243.
[17]Rowley H A,Baluja S,Kanade T.Neural network-based face detection[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,1998,20(1):23-38.
[18]Osuna E,Freund R,Girosi F.Training support vector machines:an application to face detection[C].Proceedings of IEEE Computer Society Conference on Computer Vision and Pattern Recognition,Los Alamitos,CA,1997:130-136
[19]Viola P,Jones M.Rapid Object Detection Using a Boosted Cascade of Simple Features[C].IEEE Conference on Computer Vision and Pattern Recognition,2001,1:511-518.
[20]Freund Y,Schapire R E.A decision theoretic generalization of online learning and an application to boosting[J].Journal of Computer and System Science,1997,55(1):119-123.
[21]Schapire R E.The Strength of Weak Learnability[J].Machine learnling,1990,5(2):197-227.
[22]Tang X S,Ou Z Y,Su T M.Robust Precise Eye Location by Adaboost and SVM Techniques[A].In:Wang J,Liao X F,Yi Zh,eds.Advances in Neural Networks-ISNN 2005.Berlin:Springer-Verlag,2005,3497:93-98.
[23]R.O.Duda,P.E.Hart,D.G.Stork著,李宏东,姚天祥等译.模式分类[M].北京:机械工业出版社,2003:46-53.
[24]Sobottka K,Pitas I.Face localization and facial feature extraction based on shape and color information[C].International comference on Image Processing,1996,3(9):483-486.
[25]刘小军.人脸识别技术研究[D].中国科学院博士学位论文,2001.
[26]C.Liu and H.Wechsler.Gabor feature based classification using the enhanced fisher linear discriminant model for face recognition[J].IEEE Transaction on Image Processing,2002,4(11):467-476.
[27]Lades M,Vorbruggen J C,Buhmann J,et al.Distortion invariant object recognition in the dynamic link architecture[J].IEEE Transactions on Computers,1993,42(3):300-311.
[28]Lee T S.Image representation using 2D Gabor wavelets[J].IEEE Transaction on Pattern Analysis and Machine Intelligence,1996,18(10):959-971.
[29]Rao R,Ballard D.An active vision architecture based on iconic representations[J],Artificial Intelligence,1995,(78):461-505.
[30]Schiele B,Crowley J L.Recognition without correspondence using multidimensional receptive field histograms[J],Int.J.Comput.Vis.,2000,1(36):31-52.
[31]Webb A R.Statistical Pattern Recognition[M],Second Edition.New York:Wiley,2002.
[32]Turk M,Pantland A P.Eigenfaces for Recognition[J].Jounal of Cognitive Neuroscience.1991,3(1):71-86.
[33]Belhumeur P N,Hespanha J P,Kriegman D J.Eigenfaces vs Fisherfaces:Recognition Using Class Specific Linear Projection[J].IEEE Transaction on Pattern Analysis and Machine Intelligence,1997,19(7):711-720.
[34]Philips P J,Martin A,Wilson L C et al.An introduction to evaluating biometric systems[J],Computer,2000(33):56-33.
[35]杨非.面向确认任务的人脸识别技术研究[D].中国科学院硕士学位论文.2006.
[36] Barras C, Gauvain L J. Feature and Score Normalization for Speaker Verification of Cellular Data[C].Proceedings of IEEE International Conference on Acoustics, Speech and Signal Processing(ICASSP),2003,4(2):49-52.
[37] Burges C J C. A Tutorial on Support Vector Machines for Pattern Recognition[J], Data Mining and Knowledge Discovery, 1998,2(2): 121 -167.
[38] Auckenthaler R, Carey M, Lloyd-Thomas H. Score Normalization for Text-Independent Speaker Verification Sytems[J]. Digital Signal Processing (DSP) Journal,2000(10):42-54.
[39] M. Ben, R. Blouet, F. Bimbot. A Monte-Carlo Method For Score Normalization in Automatic Speaker Verification Using Kullbace-Leibler Distances[C]. In Proc. Int.Conf.Acoustics, Speech and Signal Processing (ICASSP),2002(l):689-692.
[40] D.Garcia-Romero, J.Gonzalez-Rodriguez, J.Fierrez-Aguilar et al. U-Norm Likelihood Normalisation in PIN-Based Speaker Verification Systems[C]. In Springer LNCS-2688, 4~(th) Int'l.Conf. Audio-and Video-Based Biometric Person Authentication(AVBPA 2003), Guidlford,2003:208-213
[41] Norman Poh, Samy Bengio. Improving Single Modal and Multimodal Biometric Authentication Using F-Ratio Client-Dependent Normalisation[R]. IDIAP RR 04-52,2004.
[42] N.Poh, S.Bengio. Why Do Multi-Stream, Multi-Band and Multi-Modal Approaches Work on Biometric User Authentication Tasks? [C]. In IEEE Int'l Conf. Acoustics, Speech, and Signal Processing (ICASSP),2004(V):893-896.
[43] Vapnik V Statistical Learning Theory[M]. New York:Wiley,1998.
[44] Boser B E, Guyon I M, Vapnik V. A Training Algorithm for Optimal Margin Classifiers[A]. Proceedings of the Fifth Annual Workshop on Computational Learning Theory[C]. Pittsburgh,PA,USA: [s.n.],1992:144-152.
[45] Cortes C, Vapnik V. Support vector networks[J]. Machine Learning,1995,20(3):273-297.
[46] Vapnik V. The Nature of Statistical Learning Theory[M]. Springer, Second edition, 1999.
[47] Maria Trias. Face Verification based on Support Vector Machines[D]. USA:California University,2005.
[48] Greory Matthew Wagner, M.S.,B.S. Face Authentication with Pose Adjustment Using Support Vector Machines with a Hausdorff-based Image Kernel[D]. USA:Texas Tech University,2007.
[49] www.csie.ntu.edu.tw/cjlin/libsvm.
[2]山世光.人脸识别中的若干关键问题的研究[D],中国科学院博士学位论文,2004.
[3]Zhao W,Chellappa R,Phillps P Jet al.Face Recognition:A Literature Survey[J].ACM Computing Surveys(CSUR),2003,35(4):399-458.
[4]Yang M H,Kriegman D J,Ahuja N.Detecting Faces in Images:A Survey[J].IEEE Transactions on Pattern Analysis and Machine Intelligence.2002,24(1):34-58.
[5]P.J.Philips,P.J.Grother,R.J.Micheals et al.Face Recognition Vendor Test 2002:Evaluation Report,Technical Report[R],National Institute of Standards and Technology,2003.
[6]Ding X Q,Fang C H.Discussions on Some Problems in Face Recognition[R].In:Li S Z,Lai J H,Tan T N et al,eds.Sinobiometrics2004.Berlin:Springer-Verlag,2004,3338:47-56
[7]Mu xiang Zhang.Adaptive protocol for Entity Authentication and Key Agreement in Mobile Networks[C].ICISC 2003,LNCS 2971,2004:166-183.
[8]Zhang N,Shi Q,Merabti M.Anonymous public-key certificates for anonymous and fair document exchange[J].IEEE Proceedings:Communications.2000,6(147):345-350.
[9]Sung K,Poggio T.Example-based learning for view based human face detection[J].IEEE Transaction on Pattern Analysis and Machine Intelligence,1998,20(1):39-51.
[10]艾英山,张德贤.人脸识别方法的综述与展望[J],计算机与数字工程,2005,33(10):24-27.
[11]苏建波,徐波.应用模式识别技术导论.人脸识别与语音识别[M].上海:上海交通大学出版社,2001:125-132.
[12]Yang M H.Recent Advances in Face Detection[C].IEEE 17th International Conference on Pattern Recognition 2004 Tutorial,Cambridge,United Kingdom,2004.
[13]Garcia C,Tziritas G.Face detection using quantized skin color regions merging and wavelet packet analysis[J].IEEE Transaction on Multmedia,1999,21(3):264-277.
[14]Yang G Z,Huang T S.Human face detection in a complex background[J].Pattern Recognition,1994,27(1):53-63.
[15]Moghaddam B,Pentland A.Probabilistic visual learning for object representation[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,1997,19(7):696-710.
[16]Moghaddam B,Pentland A.Beyong linear eigenspaces:Bayesian matching for face recognition[A].In:Wechsler H,Philips P J,Bruce V,et al,eds.Face Recognition:From Theory to Applications[C],Berlin:Springer-Verlag,1998:230-243.
[17]Rowley H A,Baluja S,Kanade T.Neural network-based face detection[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,1998,20(1):23-38.
[18]Osuna E,Freund R,Girosi F.Training support vector machines:an application to face detection[C].Proceedings of IEEE Computer Society Conference on Computer Vision and Pattern Recognition,Los Alamitos,CA,1997:130-136
[19]Viola P,Jones M.Rapid Object Detection Using a Boosted Cascade of Simple Features[C].IEEE Conference on Computer Vision and Pattern Recognition,2001,1:511-518.
[20]Freund Y,Schapire R E.A decision theoretic generalization of online learning and an application to boosting[J].Journal of Computer and System Science,1997,55(1):119-123.
[21]Schapire R E.The Strength of Weak Learnability[J].Machine learnling,1990,5(2):197-227.
[22]Tang X S,Ou Z Y,Su T M.Robust Precise Eye Location by Adaboost and SVM Techniques[A].In:Wang J,Liao X F,Yi Zh,eds.Advances in Neural Networks-ISNN 2005.Berlin:Springer-Verlag,2005,3497:93-98.
[23]R.O.Duda,P.E.Hart,D.G.Stork著,李宏东,姚天祥等译.模式分类[M].北京:机械工业出版社,2003:46-53.
[24]Sobottka K,Pitas I.Face localization and facial feature extraction based on shape and color information[C].International comference on Image Processing,1996,3(9):483-486.
[25]刘小军.人脸识别技术研究[D].中国科学院博士学位论文,2001.
[26]C.Liu and H.Wechsler.Gabor feature based classification using the enhanced fisher linear discriminant model for face recognition[J].IEEE Transaction on Image Processing,2002,4(11):467-476.
[27]Lades M,Vorbruggen J C,Buhmann J,et al.Distortion invariant object recognition in the dynamic link architecture[J].IEEE Transactions on Computers,1993,42(3):300-311.
[28]Lee T S.Image representation using 2D Gabor wavelets[J].IEEE Transaction on Pattern Analysis and Machine Intelligence,1996,18(10):959-971.
[29]Rao R,Ballard D.An active vision architecture based on iconic representations[J],Artificial Intelligence,1995,(78):461-505.
[30]Schiele B,Crowley J L.Recognition without correspondence using multidimensional receptive field histograms[J],Int.J.Comput.Vis.,2000,1(36):31-52.
[31]Webb A R.Statistical Pattern Recognition[M],Second Edition.New York:Wiley,2002.
[32]Turk M,Pantland A P.Eigenfaces for Recognition[J].Jounal of Cognitive Neuroscience.1991,3(1):71-86.
[33]Belhumeur P N,Hespanha J P,Kriegman D J.Eigenfaces vs Fisherfaces:Recognition Using Class Specific Linear Projection[J].IEEE Transaction on Pattern Analysis and Machine Intelligence,1997,19(7):711-720.
[34]Philips P J,Martin A,Wilson L C et al.An introduction to evaluating biometric systems[J],Computer,2000(33):56-33.
[35]杨非.面向确认任务的人脸识别技术研究[D].中国科学院硕士学位论文.2006.
[36] Barras C, Gauvain L J. Feature and Score Normalization for Speaker Verification of Cellular Data[C].Proceedings of IEEE International Conference on Acoustics, Speech and Signal Processing(ICASSP),2003,4(2):49-52.
[37] Burges C J C. A Tutorial on Support Vector Machines for Pattern Recognition[J], Data Mining and Knowledge Discovery, 1998,2(2): 121 -167.
[38] Auckenthaler R, Carey M, Lloyd-Thomas H. Score Normalization for Text-Independent Speaker Verification Sytems[J]. Digital Signal Processing (DSP) Journal,2000(10):42-54.
[39] M. Ben, R. Blouet, F. Bimbot. A Monte-Carlo Method For Score Normalization in Automatic Speaker Verification Using Kullbace-Leibler Distances[C]. In Proc. Int.Conf.Acoustics, Speech and Signal Processing (ICASSP),2002(l):689-692.
[40] D.Garcia-Romero, J.Gonzalez-Rodriguez, J.Fierrez-Aguilar et al. U-Norm Likelihood Normalisation in PIN-Based Speaker Verification Systems[C]. In Springer LNCS-2688, 4~(th) Int'l.Conf. Audio-and Video-Based Biometric Person Authentication(AVBPA 2003), Guidlford,2003:208-213
[41] Norman Poh, Samy Bengio. Improving Single Modal and Multimodal Biometric Authentication Using F-Ratio Client-Dependent Normalisation[R]. IDIAP RR 04-52,2004.
[42] N.Poh, S.Bengio. Why Do Multi-Stream, Multi-Band and Multi-Modal Approaches Work on Biometric User Authentication Tasks? [C]. In IEEE Int'l Conf. Acoustics, Speech, and Signal Processing (ICASSP),2004(V):893-896.
[43] Vapnik V Statistical Learning Theory[M]. New York:Wiley,1998.
[44] Boser B E, Guyon I M, Vapnik V. A Training Algorithm for Optimal Margin Classifiers[A]. Proceedings of the Fifth Annual Workshop on Computational Learning Theory[C]. Pittsburgh,PA,USA: [s.n.],1992:144-152.
[45] Cortes C, Vapnik V. Support vector networks[J]. Machine Learning,1995,20(3):273-297.
[46] Vapnik V. The Nature of Statistical Learning Theory[M]. Springer, Second edition, 1999.
[47] Maria Trias. Face Verification based on Support Vector Machines[D]. USA:California University,2005.
[48] Greory Matthew Wagner, M.S.,B.S. Face Authentication with Pose Adjustment Using Support Vector Machines with a Hausdorff-based Image Kernel[D]. USA:Texas Tech University,2007.
[49] www.csie.ntu.edu.tw/cjlin/libsvm.