基于VC++的人脸识别系统的设计与实现
|
摘要
人脸识别技术作为一种极具有发展潜力的生物特征识别技术,在信用卡验证、视频会议、医学、档案管理、人机交互、公安系统中的罪犯识别等领域有着广泛的应用。由于人脸识别具有直接、方便、非侵犯性和用户友好等诸多优点,使其成为当前模式识别和人工智能领域非常活跃的一个研究方向。最近几十年,国内外许多研究学者都对人脸识别进行了大量的研究工作,并提出了很多实用有效的人脸识别算法。典型的人脸识别系统包括:人脸图像检测、特征提取、图像匹配和识别三部分。
本文主要对静止图像的人脸识别进行了深入研究,主要涉及到如下几方面:
1.详细的介绍了一维和二维的Gabor小波变换的方法,并通过计算验证了通过对Gabor滤波器参数的选择来表示人脸图像。深入研究了利用二维Gabor小波变换进行人脸特征提取的理论方法,在特征提取方面对传统的基于Gabor滤波器的人脸特征提取方法进行改进,提出了一种基于人脸有效区域的Gabor特征抽取算法。该方法首先将人脸图像经遮罩模板掩模,获取有效人脸区域,在有效区域内进行象素的Gabor特征抽取,对于有效区域外没有任何价值的像素区域,我们将其舍弃,这样就大大降低了人脸识别的时间和空间复杂度。经多次验证表明,该方法能够有效降低人脸特征向量维数,同时,具有与传统Gabor特征抽取算法同等的鲁棒性。
2.通过分析,提出了基于最佳Gabor特征的人脸识别的方法。该方法结合了人脸有效区域特征提取算法,利用遮罩模板提取眼睛、眉毛、鼻子和嘴唇所包含的人脸识别相关的主要信息,同时选取关键点并进行Gabor下采样,再利用主成分分析(PCA)方法对下采样后的Gabor特征进一步降维,然后再采用线性鉴别分析(LDA)方法进行压缩和特征选择。
3.本文所设计的人脸识别系统是在Windows XP系统的Visual C++6.0开发平台下实现的,并较为详细地介绍了系统主要部分的功能。Visual C++6.0提供的高度可视化的应用程序开发工具和丰富的Microsoft基本类库(MFC库),可使应用程序开发变得更加简单。
Face recognition technology as a kind of extremely has the potential of development in biometric technology, video conference, credit card verification, medical, file management, human-computer interaction, public security system of criminal identification field in a wide range of applications. Due to face recognition has direct, convenient, non-invasive and user friendly, and many other advantages, make it become the current pattern recognition and the field of artificial intelligence a very active research direction. In recent decades, many domestic research of face recognition of scholars have done a lot of research work, and puts forward a lot of practical and efficient face recognition algorithm. The typical face recognition system includes: face image detection, feature extraction, image matching and identification of three parts.
This thesis mainly for still images of face recognition is studied, the main involves the following aspects:
1.This thesis describes in detail a one-dimensional and two-dimensional Gabor wavelet transform method, and proved by calculation of Gabor filters through the selection of parameters to show facial image. A deep research using 2D Gabor wavelet transform the face of feature extraction method, on feature extraction theory of the traditional Gabor filter based on the face and improvement of the method of feature extraction is proposed based on the face of the valid area Gabor feature extraction algorithm. This method firstly will face image via masks template masking, obtain valid face region within the valid area, in the pixels Gabor feature extraction, outside the territory for effective without any value pixel region, we abandon it, thus greatly reduces the face recognition of time and space complexity. After repeated experiments show that this method can effectively reduce the face feature vector dimension, meanwhile, has Gabor feature extraction algorithm with traditional same robustness.
2.This thesis put forward through the analysis, based on the best Gabor characteristics of face recognition method. This method combines the third chapter of face effective mentioned feature extraction algorithm, using regional masks template extraction, eyebrow, nose and eyes lips contains the main face recognition related information, and at the same time selecting Gabor point and sampling, and then under the principal component analysis (PCA) under the method of sampling Gabor characteristics after further dimension reduction, and then using linear differential analysis (LDA) methods are compressed and feature selection.
3.This design of face recognition system is based on WindowsXP system in the Visual C++ 6.0 development of lans implementation, and introduces the system detail major parts of the function. Visual C++ 6.0 provide height visualization of application development tools and rich Microsoft Foundation Classes library (MFC library), can make the application development become more simple.
本文主要对静止图像的人脸识别进行了深入研究,主要涉及到如下几方面:
1.详细的介绍了一维和二维的Gabor小波变换的方法,并通过计算验证了通过对Gabor滤波器参数的选择来表示人脸图像。深入研究了利用二维Gabor小波变换进行人脸特征提取的理论方法,在特征提取方面对传统的基于Gabor滤波器的人脸特征提取方法进行改进,提出了一种基于人脸有效区域的Gabor特征抽取算法。该方法首先将人脸图像经遮罩模板掩模,获取有效人脸区域,在有效区域内进行象素的Gabor特征抽取,对于有效区域外没有任何价值的像素区域,我们将其舍弃,这样就大大降低了人脸识别的时间和空间复杂度。经多次验证表明,该方法能够有效降低人脸特征向量维数,同时,具有与传统Gabor特征抽取算法同等的鲁棒性。
2.通过分析,提出了基于最佳Gabor特征的人脸识别的方法。该方法结合了人脸有效区域特征提取算法,利用遮罩模板提取眼睛、眉毛、鼻子和嘴唇所包含的人脸识别相关的主要信息,同时选取关键点并进行Gabor下采样,再利用主成分分析(PCA)方法对下采样后的Gabor特征进一步降维,然后再采用线性鉴别分析(LDA)方法进行压缩和特征选择。
3.本文所设计的人脸识别系统是在Windows XP系统的Visual C++6.0开发平台下实现的,并较为详细地介绍了系统主要部分的功能。Visual C++6.0提供的高度可视化的应用程序开发工具和丰富的Microsoft基本类库(MFC库),可使应用程序开发变得更加简单。
Face recognition technology as a kind of extremely has the potential of development in biometric technology, video conference, credit card verification, medical, file management, human-computer interaction, public security system of criminal identification field in a wide range of applications. Due to face recognition has direct, convenient, non-invasive and user friendly, and many other advantages, make it become the current pattern recognition and the field of artificial intelligence a very active research direction. In recent decades, many domestic research of face recognition of scholars have done a lot of research work, and puts forward a lot of practical and efficient face recognition algorithm. The typical face recognition system includes: face image detection, feature extraction, image matching and identification of three parts.
This thesis mainly for still images of face recognition is studied, the main involves the following aspects:
1.This thesis describes in detail a one-dimensional and two-dimensional Gabor wavelet transform method, and proved by calculation of Gabor filters through the selection of parameters to show facial image. A deep research using 2D Gabor wavelet transform the face of feature extraction method, on feature extraction theory of the traditional Gabor filter based on the face and improvement of the method of feature extraction is proposed based on the face of the valid area Gabor feature extraction algorithm. This method firstly will face image via masks template masking, obtain valid face region within the valid area, in the pixels Gabor feature extraction, outside the territory for effective without any value pixel region, we abandon it, thus greatly reduces the face recognition of time and space complexity. After repeated experiments show that this method can effectively reduce the face feature vector dimension, meanwhile, has Gabor feature extraction algorithm with traditional same robustness.
2.This thesis put forward through the analysis, based on the best Gabor characteristics of face recognition method. This method combines the third chapter of face effective mentioned feature extraction algorithm, using regional masks template extraction, eyebrow, nose and eyes lips contains the main face recognition related information, and at the same time selecting Gabor point and sampling, and then under the principal component analysis (PCA) under the method of sampling Gabor characteristics after further dimension reduction, and then using linear differential analysis (LDA) methods are compressed and feature selection.
3.This design of face recognition system is based on WindowsXP system in the Visual C++ 6.0 development of lans implementation, and introduces the system detail major parts of the function. Visual C++ 6.0 provide height visualization of application development tools and rich Microsoft Foundation Classes library (MFC library), can make the application development become more simple.
引文
[1]章毓晋.基于子空间的人脸识别[M].北京:清华大学出版社, 2009,1-2
[2]周桐.基于PCA的人脸识别系统的设计与实现[D] .哈尔滨:哈尔滨工业大学,2007,13-17
[3] L.D.Harmon, S.C. Kuo, P. F. Ramig and U. Raudkivi.Identification of Human Face Profiles by Computer Pattern Recognition. IEEE. 1978,10(5):301～312
[4] Goldstein,A.Jay,Harmon,D.Leon and B.Ann.Identification of Human Faces. IEEE. 1971,59(5):748-760
[5] Kaya,Y. Kobayashi. A Basic Study on Human Face Recognition. Frontiers of Pattern Recognition.1972:265-289
[6]姚劲.基于Gabor小波变换的人脸识别研究[D].昆明:昆明理工大学,2008
[7] Weston J,Watkins C.Multi-class support vector machine[R].Royal Holloway College,Tech Rep:CSD-TR-98-04,1998
[8]李刚,高政.人脸识别理论研究进展[J],计算机与现代化,2003年第3期
[9] Guo G.,Li S.Z.,Chan K.L.Support vector machines for face recognition[J],International Journal of Computer Vision,2001,19:631-638
[10]阙晓辉.基于面部特征的人脸识别技术研究[D].武汉:武汉理工大学,2007,4-5
[11]程雪红.人脸检测与识别技术研究[D].西安:西安电子科技大学,2006,6-7
[12]李健,廖秋筠.人脸识别的方法研究[J] .微计算机信息,2006,22(4-2):254-256
[13] Turk M,Pentland,A.Eigenface for Recogniton[J].J Cog Neurosci,1991,23(3):7186
[14]马驰.人脸识别算法研究和实现[D].北京:北京交通大学,2007
[15] Hong Z.Algebraic feature extraction from image for recognition[J].Pattern ecognition,1991,24(3):211-219
[16] C.J.C.Burges.A Tutorial on Support Vector Machines for Pattern Recognition.Knowledge Discovery and Data Mining.1998,2(2):15～18
[17]夏良正.数字图像处理[M].南京:东南大学出版社,2005,153-154
[18]马俊容.单训练样本条件下人脸识别技术研究[D],长沙:湖南大学,2009,11-12
[19]李琳莉.人脸识别的图像处理算法[D],呼和浩特:内蒙古大学,2007,24-25
[20]杨盈昀,谢婷婷,施美楠.基于肤色的人脸检测算法研究[J].北京广播学院学报(自然科学版), 9(4),2002
[21]周志明,王以治,黄文芝,王宁宁.基于小波和支持向量机的人脸识别技术[J],计算机工程与应用,2004,12, 52-54
[22]赖剑煌,阮邦志,冯国灿.频谱脸:一种基于小波变换和Fourier变换的人像识别方法[J],中国图像图形学报,1999, 4(10), 811-816
[23]阮秋琦,阮宇智.数字图像处理(第二版)[M],北京,电子工业出版社,2003
[24]彼得鲁(英) .数字图像处理疑难解析[M」,北京:机械工业出版社,2005
[25]谢凤英,赵丹培.Visual C++数字图像处理[M].北京:电子工业出版社,2008
[26]龚理鸣.人脸识别实验平台设计及人脸识别算法实现[D],北京,北京邮电大学,2006,7-8
[27] Gabor D.Theory of connunication.Journal of the Institute of Electrical Engineers,1946,93(26):429-457
[28] B.S.Manjunath.Gabor Wavelet Transform and Application to Problems in Early Vision[J],1058-6393/92 IEEE:636-648
[29] Lades M,Vorbruggen J C,Buhmann J,et al.Distortion invariant object recognition in the dynamic link architecture.IEEE Transactions on Computers,1993,42(3):300-311
[30]李云峰.基于Gabor小波变换的人脸识别[D],大连:大连理工大学,2005,48-53
[31] Lee T S.Image representation using 2D Gabor wavelets.IEEE Transactions on Pattern Analysis and Machine Intelligence,1996,18(10):959-971
[32]周羽.红外图像人脸识别研究[D],大连:大连理工大学,2007,37-38
[33]李外云.基于ARM架构的嵌入式人脸识别技术研究[D].华东师范大学,2008,75-78
[34]程正兴.小波分析算法与应用[M],西安:西安交通大学出版社,1998
[35] Liu D.H.,Lam K.M.and Shen L.Optimal sampling of Gabor features for face recognition.Patern Recognition Leters. 2004,25:267-276
[36]胡文静,自动人脸识别技术研究及其在人员身份认证系统中的实现[D],华东师范大学,2006,89-90
[37]万峰,杜明辉.人脸识别中一种新的Gabor特征抽取算法[J]华南理工大学学报,32:5-8,2004
[38] S.Mallat ,“A Theory for Multiresolution Signal Decomposition : the Wavelet Representation-on,”IEEE Trans.PAM.1989,vol.11,674-693
[39]李永在.基于小波分析和主成分分析的人脸识别的研究[D] .山东:山东大学,2007,24-25
[40]朱宝生.基于小波变换的人脸统计特征识别系统研究[D] .南京:南京理工大学,2006,7-8
[41]李世雄.小波变换及其应用[M].北京:高等教育出版社,1997
[42] R.Lee , J. Liu , Face Recognition: Elastic Relation Encoding and Structural Matching.Proceeding of the IEEE International Conference on System , Man and Cybemetics ,1999, (2):172-177
[43]孙海燕.基于小波分析小域特征融合的人脸识别算法研究[D],北京:燕山大学,2009,33-36
[44]姚劲.基于Gabor小波变换的人脸识别研究[D],昆明:昆明理工大学,2008,62-65
[45]克鲁格林斯基(美).Visual C++ 6.0技术内幕[M],北京:清华大学出版社,1999
[46]张宏林.精通Visual C++数字图像处理典型算法及实现(第2版)[M].北京:人民邮电出版社,2006
[47]张宏林. Visual C++数字图像模式识别技术及工程实践[M],人民邮电出版社,2003
[2]周桐.基于PCA的人脸识别系统的设计与实现[D] .哈尔滨:哈尔滨工业大学,2007,13-17
[3] L.D.Harmon, S.C. Kuo, P. F. Ramig and U. Raudkivi.Identification of Human Face Profiles by Computer Pattern Recognition. IEEE. 1978,10(5):301～312
[4] Goldstein,A.Jay,Harmon,D.Leon and B.Ann.Identification of Human Faces. IEEE. 1971,59(5):748-760
[5] Kaya,Y. Kobayashi. A Basic Study on Human Face Recognition. Frontiers of Pattern Recognition.1972:265-289
[6]姚劲.基于Gabor小波变换的人脸识别研究[D].昆明:昆明理工大学,2008
[7] Weston J,Watkins C.Multi-class support vector machine[R].Royal Holloway College,Tech Rep:CSD-TR-98-04,1998
[8]李刚,高政.人脸识别理论研究进展[J],计算机与现代化,2003年第3期
[9] Guo G.,Li S.Z.,Chan K.L.Support vector machines for face recognition[J],International Journal of Computer Vision,2001,19:631-638
[10]阙晓辉.基于面部特征的人脸识别技术研究[D].武汉:武汉理工大学,2007,4-5
[11]程雪红.人脸检测与识别技术研究[D].西安:西安电子科技大学,2006,6-7
[12]李健,廖秋筠.人脸识别的方法研究[J] .微计算机信息,2006,22(4-2):254-256
[13] Turk M,Pentland,A.Eigenface for Recogniton[J].J Cog Neurosci,1991,23(3):7186
[14]马驰.人脸识别算法研究和实现[D].北京:北京交通大学,2007
[15] Hong Z.Algebraic feature extraction from image for recognition[J].Pattern ecognition,1991,24(3):211-219
[16] C.J.C.Burges.A Tutorial on Support Vector Machines for Pattern Recognition.Knowledge Discovery and Data Mining.1998,2(2):15～18
[17]夏良正.数字图像处理[M].南京:东南大学出版社,2005,153-154
[18]马俊容.单训练样本条件下人脸识别技术研究[D],长沙:湖南大学,2009,11-12
[19]李琳莉.人脸识别的图像处理算法[D],呼和浩特:内蒙古大学,2007,24-25
[20]杨盈昀,谢婷婷,施美楠.基于肤色的人脸检测算法研究[J].北京广播学院学报(自然科学版), 9(4),2002
[21]周志明,王以治,黄文芝,王宁宁.基于小波和支持向量机的人脸识别技术[J],计算机工程与应用,2004,12, 52-54
[22]赖剑煌,阮邦志,冯国灿.频谱脸:一种基于小波变换和Fourier变换的人像识别方法[J],中国图像图形学报,1999, 4(10), 811-816
[23]阮秋琦,阮宇智.数字图像处理(第二版)[M],北京,电子工业出版社,2003
[24]彼得鲁(英) .数字图像处理疑难解析[M」,北京:机械工业出版社,2005
[25]谢凤英,赵丹培.Visual C++数字图像处理[M].北京:电子工业出版社,2008
[26]龚理鸣.人脸识别实验平台设计及人脸识别算法实现[D],北京,北京邮电大学,2006,7-8
[27] Gabor D.Theory of connunication.Journal of the Institute of Electrical Engineers,1946,93(26):429-457
[28] B.S.Manjunath.Gabor Wavelet Transform and Application to Problems in Early Vision[J],1058-6393/92 IEEE:636-648
[29] Lades M,Vorbruggen J C,Buhmann J,et al.Distortion invariant object recognition in the dynamic link architecture.IEEE Transactions on Computers,1993,42(3):300-311
[30]李云峰.基于Gabor小波变换的人脸识别[D],大连:大连理工大学,2005,48-53
[31] Lee T S.Image representation using 2D Gabor wavelets.IEEE Transactions on Pattern Analysis and Machine Intelligence,1996,18(10):959-971
[32]周羽.红外图像人脸识别研究[D],大连:大连理工大学,2007,37-38
[33]李外云.基于ARM架构的嵌入式人脸识别技术研究[D].华东师范大学,2008,75-78
[34]程正兴.小波分析算法与应用[M],西安:西安交通大学出版社,1998
[35] Liu D.H.,Lam K.M.and Shen L.Optimal sampling of Gabor features for face recognition.Patern Recognition Leters. 2004,25:267-276
[36]胡文静,自动人脸识别技术研究及其在人员身份认证系统中的实现[D],华东师范大学,2006,89-90
[37]万峰,杜明辉.人脸识别中一种新的Gabor特征抽取算法[J]华南理工大学学报,32:5-8,2004
[38] S.Mallat ,“A Theory for Multiresolution Signal Decomposition : the Wavelet Representation-on,”IEEE Trans.PAM.1989,vol.11,674-693
[39]李永在.基于小波分析和主成分分析的人脸识别的研究[D] .山东:山东大学,2007,24-25
[40]朱宝生.基于小波变换的人脸统计特征识别系统研究[D] .南京:南京理工大学,2006,7-8
[41]李世雄.小波变换及其应用[M].北京:高等教育出版社,1997
[42] R.Lee , J. Liu , Face Recognition: Elastic Relation Encoding and Structural Matching.Proceeding of the IEEE International Conference on System , Man and Cybemetics ,1999, (2):172-177
[43]孙海燕.基于小波分析小域特征融合的人脸识别算法研究[D],北京:燕山大学,2009,33-36
[44]姚劲.基于Gabor小波变换的人脸识别研究[D],昆明:昆明理工大学,2008,62-65
[45]克鲁格林斯基(美).Visual C++ 6.0技术内幕[M],北京:清华大学出版社,1999
[46]张宏林.精通Visual C++数字图像处理典型算法及实现(第2版)[M].北京:人民邮电出版社,2006
[47]张宏林. Visual C++数字图像模式识别技术及工程实践[M],人民邮电出版社,2003