模糊支持向量机在人脸识别中的应用
|
摘要
支持向量机作为一种机器学习方法,较好地解决了非线性、高维数等实际问题,是机器学习领域的研究热点,为人脸识别提供了一条有效的路径。
鉴于样本通常具有模糊特性且分布有稀疏的差别,在研究了现有的一些模糊支持向量机方法的基础上,本文提出了基于模糊K近邻的模糊支持向量机方法。该方法首先针对每一类样本,计算出样本均值,得到样本类中心点；然后计算出样本与中心点的距离,根据距离计算出样本的初始隶属度；计算每个样本的K个近邻点,按照模糊K近邻方法计算样本的隶属度,将初始隶属度和模糊K近邻隶属度以一定比例融合,得出样本的最终隶属度值。结合稻米图像检测问题,验证了该方法的有效性。
针对人脸图像的特征提取,本文在重点分析了主成分分析方法和二维主成分分析方法的基础上,提出了用二维特征求解样本隶属度,用主成分特征进行支持向量机分类的方法。该方法结合了二维主成分特征在选取少量分量时人脸重构图像稳定的优点和主成分特征重构图像局部特征清晰的优点。为了与二维主成分特征分类结果进行比较,还通过引入矩阵内积,给出了针对二维特征的三类核函数。实验表明,本文提出的利用两种特征进行分类的方法在人脸识别中具有较高的精度。
As a machine learning method, support vector machine can solve the non-linear, high dimension and other practical problem, it is the focus in the field of machine learning, and provides a valid path for face recognition.
Because the sample usually has a lot of vague information, and the sparseness of the samples distribution are different, through studying some of the existing fuzzy support vector machine method, a fuzzy support vector machine method based on fuzzy K neighbors is presented in this paper. In this method, the sample mean is calculated,and the center of each calss is got; then the distance between the sample and the center is calculated, according to the distance sample's initial membership is got;by finding K neighbors for each sample point, the sample membership degree is calculated according to the fuzzy K nearest neighbor method,then the paper integrates initial membership degree with fuzzy K neighbors membership by a certain percentage that obtaines the final membership value of the sample.Combine with rice image detection, the validity of this method is verified.
According to analyze the main component analysis and the two-dimensional principal component analysis regarding face image feature extraction, using two-dimensional features for the calculation of the membership and the principal component for support vector classification is proposed in this paper. The method combines the stabilization of two-dimensional principal component in reconstructing face image and the obviousness of the principal component to the reconstructed image local characteristics. In order to contrast with the sort results of two-dimensional characteristics, through introduction of matrix inner product, three types of two-dimensional characteristics kernel function are given. Experiments show that the method of this paper has a high classification accuracy for face recognition.
鉴于样本通常具有模糊特性且分布有稀疏的差别,在研究了现有的一些模糊支持向量机方法的基础上,本文提出了基于模糊K近邻的模糊支持向量机方法。该方法首先针对每一类样本,计算出样本均值,得到样本类中心点；然后计算出样本与中心点的距离,根据距离计算出样本的初始隶属度；计算每个样本的K个近邻点,按照模糊K近邻方法计算样本的隶属度,将初始隶属度和模糊K近邻隶属度以一定比例融合,得出样本的最终隶属度值。结合稻米图像检测问题,验证了该方法的有效性。
针对人脸图像的特征提取,本文在重点分析了主成分分析方法和二维主成分分析方法的基础上,提出了用二维特征求解样本隶属度,用主成分特征进行支持向量机分类的方法。该方法结合了二维主成分特征在选取少量分量时人脸重构图像稳定的优点和主成分特征重构图像局部特征清晰的优点。为了与二维主成分特征分类结果进行比较,还通过引入矩阵内积,给出了针对二维特征的三类核函数。实验表明,本文提出的利用两种特征进行分类的方法在人脸识别中具有较高的精度。
As a machine learning method, support vector machine can solve the non-linear, high dimension and other practical problem, it is the focus in the field of machine learning, and provides a valid path for face recognition.
Because the sample usually has a lot of vague information, and the sparseness of the samples distribution are different, through studying some of the existing fuzzy support vector machine method, a fuzzy support vector machine method based on fuzzy K neighbors is presented in this paper. In this method, the sample mean is calculated,and the center of each calss is got; then the distance between the sample and the center is calculated, according to the distance sample's initial membership is got;by finding K neighbors for each sample point, the sample membership degree is calculated according to the fuzzy K nearest neighbor method,then the paper integrates initial membership degree with fuzzy K neighbors membership by a certain percentage that obtaines the final membership value of the sample.Combine with rice image detection, the validity of this method is verified.
According to analyze the main component analysis and the two-dimensional principal component analysis regarding face image feature extraction, using two-dimensional features for the calculation of the membership and the principal component for support vector classification is proposed in this paper. The method combines the stabilization of two-dimensional principal component in reconstructing face image and the obviousness of the principal component to the reconstructed image local characteristics. In order to contrast with the sort results of two-dimensional characteristics, through introduction of matrix inner product, three types of two-dimensional characteristics kernel function are given. Experiments show that the method of this paper has a high classification accuracy for face recognition.
引文
[1]C. Cortes, V. N. Vapnik. Support vector networks. Machine Learning,1995,20(3):273-297.
[2]Vapnik V. Statistical Learning Theory. New York:John Wiley&Sons,1998.
[3]徐建华,张学工,李衍达.支持向量机的新发展[J].控制与决策,2004,19(5)：481-484.
[4]王国胜,钟义信.支持向量机的理论基础—统计学习理论[J].计算机工程与应用,2001,19(5)：19-20.
[5]Inoue. T, Abe S. Fuzzy support vector machines for pattern classification. In:Proceedings of International,Joint.Gonference on, Neural Networks, x..2001,2:1449-1454.
[6]Lin C F, Wang S D. Fuzzy support vector-machines. IEEE. Transactions on Neural Networks,2002,13(2):464-471.
[7]Suykens J, Vandewalle J. Least squares support vector machines classifiers. Neural Processing Letters,1999,9(3):293-300.
[8]袁玉波,严杰,徐成贤.多项式光滑的支持向量机[J].计算机学报,2005,28(1)：9-17.
[9]刘畅,孙德山.模糊支持向量机隶属度的确定方法[J].计算机工程与应用,2008,44(11)：41-46
[10]吴青,刘三阳,杜喆.基于边界向量提取的模糊支持向量机方法[J].模式识别与人工智能,2008,21(3)：332-337
[11]王琳,闫德勤,梁宏霞.基于熵和蚁群算法的模糊支持向量机新方法[J].计算机应用,2009,29(7)：1890-1893
[12]唐静远,师奕兵.采用模糊支持向量机的模拟电路故障诊断新方法[J].电子测量与仪器学报,2009,23(6)：7-12
[13]杨海军,太雷.基于模糊支持向量机的上市公司财务困境预测[J].管理科学学报,2009,12(3)：102-110
[14]Meir Perez.A Hybrid Fuzzy-SVM classifier, applied to gene expression profiling for automated leukaemia diagnosis. In:IEEE 25th Convention of Digital Object Identifier, 2008:041-045
[15]B. E. Boser, I. M. Guyon, V. N. Vapnik. A training algorithm for optimal margin classifiers. The 5th Annual ACM Workshop on COLT. Pittsburgh:ACM Press,1992.
[16]吕晓丽,李雷,曹未丰.基于二叉树的SVM多分类算法[J].信息技术,2008,32(4)：1-3.
[17]张永,迟忠先,米滢.一类直接构造的模糊多类支持向量分类器[J].计算机工程与应用,2008,44(8),12-15.
[18]唐发明,王仲东,陈绵云.一种新的二叉树多类支持向量机算法[J].计算机工程与应用,2005,41(7)：24-26.
[19]Gold C, Sollicbb P. Model selection for support vector machine classification. Neurocomputing,2003,55:221-249.
[20]袁小芳,王耀南.基于混沌优化算法的支持向量机参数选取方法[J].控制与决策,2006,21(1)：111-113.
[21]Joachims T. Makong Large-scale support vector machine learning paractical. In:B. Scholkopf, C. J. C. Burges, A. J. Smola, eds. Advances in Kernel Methods-Support Vector Learning, Cambridge, MA, MIT Press.1999.
[22]Platt J. Sequential nanimal optimization:A fast algorithm for training support vector machines:[Technical Report MSR-TR-98-14]. Mecrosoft Research.1998.
[23]Mangasarian 0 L, Musicant D R. Successive overrelaxation for support vector machines. IEEE Transactions on Neural Networks,1999,10:1032-1037.
[24]Mangasarian O L,Musicant D R. Lagrangian wupport vector machines. Journal of Machine Learning Research,2001,1:161-177.
[25]W. W. Bledsoe. Man-machine Facial Recognition. Techonical Report. PRI:22, Panoramic Res. Inc, Palo Alto, CA,1966.
[26]Xiaobao Li, Nicholas Roeder, Face Reangnition Extraction From Front-View Images, Pattern Recognition,1995,28(8):1168-1179.
[27]Brunelli R, Poggio T, Face Recognition:Features Versus Templates. IEEE Trans. on Pattern Analysis and Machine Intelligence,1993,15(10):1042-1052.
[28]洪子泉,杨静宇.基于奇异值特征和统计模型的人像识别算法[J].计算机研究与发展,1994,31(3)：60-65.
[29]杨绍华.一种基于小波变换和PCA的人脸识别改进方法[J].吉林工程技术师范学院学报,2008,24(9)：62-65.
[30]林劫,李建平.基于Gabor小波和模型自适应的鲁棒人脸识别方法[J].计算机应用研究,2008,25(9)：2868-2871.
[31]於东军,杨静宇.基于神经网络的人脸特征提取及识别[J].计算机工程与应用,2002,38(11)：33-47.
[32]孙亚.基于粒子群BP神经网络人脸识别算法[J].计算机仿真,2008,25(8)：201-204.
[33]边肇祺,张学工.模式识别(第2版)[M].北京：清华大学出版社,2000.
[34]张翔,肖小玲等.基于样本之间紧密度的模糊支持向量机方法[J].软件学报,2006,17(5)：951-958
[35]李天真,周柏清.基于计算机视觉技术的稻米检测研究[J].粮食与食品工业,2005,12(4)：50-55
[36]方长云,谢黎虹.计算机视觉技术在稻米品质检测中的应用[J].中国稻米,2008,6：1-4
[37]中华人民共和国国家标准优质稻谷GB/T17891-1999[S].北京：中国标准出版社,1999
[38]Cootes T F. Active shape models-their training and application[J]. Computer Vision and Image Understanding,1995,61(1):38-59
[39]Cootes T F, Taylor C J. Active shape models-'smart snakes'[C]//Proceedings of British Machine Vision Conference, Leeds,1992:266-275
[40]Cootes T F, Taylor C J. Multi-resolution search with active shape models[C]//Proceedings of International Conference on Pattern Recognition, Manchester,1994,1:610-612
[41]Cootes T F, Taylor C J. Statistical models of appearance for computer vision[OL]. [2007-02-01].http://www.isbe.mainac.uk
[42]Matthews I and Baker S. Active appearance models revisited. International Journal of Computer Vision,2004,60(2):135-164
[43]Faggian N, Paplinski A, and Chin T J. Face recognition from video using active appearance model segmentation. IEEE International Conference on Pattern Recogniton, HongKong, China, Aug.2006,1:287-290.
[44]SORIANOM, HUOV NENS. Using the skin lotus to cope with changing illunination conditions in color-based face tracking[C]//IEEE Nordic Signal Processing Symposium. Kolmarden, Sweden:IEEE,2000:383-386
[45]T. Kanade. Computer recognition of human faces based on Stuttgart:Birkhauser,1997.
[46]Yuille. Feature extraction from faces using deformable templates. International Journal of computer vision,1992,8(2):99-111
[47]Nastar. Flexible Images:Matching and recogniton and learned deformations. Computer vision and Image Understanding.1997,65(2):179-191
[48]Yongguo Liu, Gang Chen. Face Recognition Based on Independent Component Analysis and Fuzzy Support Vector Machine. Intelligent Control and Automation,2006:9889-9892
[49]吴建华,李娜等.基于CGA和ICA的人脸特征提取方法研究[J].计算机应用,2007,27(8)：2038-2040
[50]武妍,吴桂云.基于ICA/SVM的混合学习方法及在人脸检测中的应用[J].计算机工程与应用,2004,40(36)：87-101
[51]姚雪梅等.基于ICA和SVM的局部人脸识别方法[J].图形图像,2005：15-27
[52]郑宇杰,杨静宇等.基于独立成分分析和模糊支持向量机的人脸识别方法[J].系统仿真学报,2005,17(7)：1768-1770
[53]郑宇杰,於东军等.一种基于ICA和LDA组合的人脸识别新方法[J].计算机科学,2006,33(4)：194-197
[54]Turk M and Pentland A. Face Processing:Models for recognition. Proc. Intelligent Robots and Computer Vision, SPIE,1989,1,192:22-32
[55]徐艳,陈孝威.基于差分图像和PCA主元分析的人脸检测[J].计算机应用与软件,2008,25(8)：229-273
[56]Jian Yang, David Zhang, Jing-Yu Yang. Tow dimensional PCA:A new approach to appearance-based face representation and recognition. IEEE Trans. Pattern Anal. Machine Intell.2004,26(1):131-137
[57]Kressel UHG. Pairwise classification and support vector machines[M]. Cambridge:MIT Press,1999
[58]Vapnik V. The Nature of statistical learning theory [M]. New York:Springer Verlag,1995
[59]Kuncheva L. Combining classifiers by clustering, selection and decision templates[R]. Technical report(2000), University of Wales, UK
[2]Vapnik V. Statistical Learning Theory. New York:John Wiley&Sons,1998.
[3]徐建华,张学工,李衍达.支持向量机的新发展[J].控制与决策,2004,19(5)：481-484.
[4]王国胜,钟义信.支持向量机的理论基础—统计学习理论[J].计算机工程与应用,2001,19(5)：19-20.
[5]Inoue. T, Abe S. Fuzzy support vector machines for pattern classification. In:Proceedings of International,Joint.Gonference on, Neural Networks, x..2001,2:1449-1454.
[6]Lin C F, Wang S D. Fuzzy support vector-machines. IEEE. Transactions on Neural Networks,2002,13(2):464-471.
[7]Suykens J, Vandewalle J. Least squares support vector machines classifiers. Neural Processing Letters,1999,9(3):293-300.
[8]袁玉波,严杰,徐成贤.多项式光滑的支持向量机[J].计算机学报,2005,28(1)：9-17.
[9]刘畅,孙德山.模糊支持向量机隶属度的确定方法[J].计算机工程与应用,2008,44(11)：41-46
[10]吴青,刘三阳,杜喆.基于边界向量提取的模糊支持向量机方法[J].模式识别与人工智能,2008,21(3)：332-337
[11]王琳,闫德勤,梁宏霞.基于熵和蚁群算法的模糊支持向量机新方法[J].计算机应用,2009,29(7)：1890-1893
[12]唐静远,师奕兵.采用模糊支持向量机的模拟电路故障诊断新方法[J].电子测量与仪器学报,2009,23(6)：7-12
[13]杨海军,太雷.基于模糊支持向量机的上市公司财务困境预测[J].管理科学学报,2009,12(3)：102-110
[14]Meir Perez.A Hybrid Fuzzy-SVM classifier, applied to gene expression profiling for automated leukaemia diagnosis. In:IEEE 25th Convention of Digital Object Identifier, 2008:041-045
[15]B. E. Boser, I. M. Guyon, V. N. Vapnik. A training algorithm for optimal margin classifiers. The 5th Annual ACM Workshop on COLT. Pittsburgh:ACM Press,1992.
[16]吕晓丽,李雷,曹未丰.基于二叉树的SVM多分类算法[J].信息技术,2008,32(4)：1-3.
[17]张永,迟忠先,米滢.一类直接构造的模糊多类支持向量分类器[J].计算机工程与应用,2008,44(8),12-15.
[18]唐发明,王仲东,陈绵云.一种新的二叉树多类支持向量机算法[J].计算机工程与应用,2005,41(7)：24-26.
[19]Gold C, Sollicbb P. Model selection for support vector machine classification. Neurocomputing,2003,55:221-249.
[20]袁小芳,王耀南.基于混沌优化算法的支持向量机参数选取方法[J].控制与决策,2006,21(1)：111-113.
[21]Joachims T. Makong Large-scale support vector machine learning paractical. In:B. Scholkopf, C. J. C. Burges, A. J. Smola, eds. Advances in Kernel Methods-Support Vector Learning, Cambridge, MA, MIT Press.1999.
[22]Platt J. Sequential nanimal optimization:A fast algorithm for training support vector machines:[Technical Report MSR-TR-98-14]. Mecrosoft Research.1998.
[23]Mangasarian 0 L, Musicant D R. Successive overrelaxation for support vector machines. IEEE Transactions on Neural Networks,1999,10:1032-1037.
[24]Mangasarian O L,Musicant D R. Lagrangian wupport vector machines. Journal of Machine Learning Research,2001,1:161-177.
[25]W. W. Bledsoe. Man-machine Facial Recognition. Techonical Report. PRI:22, Panoramic Res. Inc, Palo Alto, CA,1966.
[26]Xiaobao Li, Nicholas Roeder, Face Reangnition Extraction From Front-View Images, Pattern Recognition,1995,28(8):1168-1179.
[27]Brunelli R, Poggio T, Face Recognition:Features Versus Templates. IEEE Trans. on Pattern Analysis and Machine Intelligence,1993,15(10):1042-1052.
[28]洪子泉,杨静宇.基于奇异值特征和统计模型的人像识别算法[J].计算机研究与发展,1994,31(3)：60-65.
[29]杨绍华.一种基于小波变换和PCA的人脸识别改进方法[J].吉林工程技术师范学院学报,2008,24(9)：62-65.
[30]林劫,李建平.基于Gabor小波和模型自适应的鲁棒人脸识别方法[J].计算机应用研究,2008,25(9)：2868-2871.
[31]於东军,杨静宇.基于神经网络的人脸特征提取及识别[J].计算机工程与应用,2002,38(11)：33-47.
[32]孙亚.基于粒子群BP神经网络人脸识别算法[J].计算机仿真,2008,25(8)：201-204.
[33]边肇祺,张学工.模式识别(第2版)[M].北京：清华大学出版社,2000.
[34]张翔,肖小玲等.基于样本之间紧密度的模糊支持向量机方法[J].软件学报,2006,17(5)：951-958
[35]李天真,周柏清.基于计算机视觉技术的稻米检测研究[J].粮食与食品工业,2005,12(4)：50-55
[36]方长云,谢黎虹.计算机视觉技术在稻米品质检测中的应用[J].中国稻米,2008,6：1-4
[37]中华人民共和国国家标准优质稻谷GB/T17891-1999[S].北京：中国标准出版社,1999
[38]Cootes T F. Active shape models-their training and application[J]. Computer Vision and Image Understanding,1995,61(1):38-59
[39]Cootes T F, Taylor C J. Active shape models-'smart snakes'[C]//Proceedings of British Machine Vision Conference, Leeds,1992:266-275
[40]Cootes T F, Taylor C J. Multi-resolution search with active shape models[C]//Proceedings of International Conference on Pattern Recognition, Manchester,1994,1:610-612
[41]Cootes T F, Taylor C J. Statistical models of appearance for computer vision[OL]. [2007-02-01].http://www.isbe.mainac.uk
[42]Matthews I and Baker S. Active appearance models revisited. International Journal of Computer Vision,2004,60(2):135-164
[43]Faggian N, Paplinski A, and Chin T J. Face recognition from video using active appearance model segmentation. IEEE International Conference on Pattern Recogniton, HongKong, China, Aug.2006,1:287-290.
[44]SORIANOM, HUOV NENS. Using the skin lotus to cope with changing illunination conditions in color-based face tracking[C]//IEEE Nordic Signal Processing Symposium. Kolmarden, Sweden:IEEE,2000:383-386
[45]T. Kanade. Computer recognition of human faces based on Stuttgart:Birkhauser,1997.
[46]Yuille. Feature extraction from faces using deformable templates. International Journal of computer vision,1992,8(2):99-111
[47]Nastar. Flexible Images:Matching and recogniton and learned deformations. Computer vision and Image Understanding.1997,65(2):179-191
[48]Yongguo Liu, Gang Chen. Face Recognition Based on Independent Component Analysis and Fuzzy Support Vector Machine. Intelligent Control and Automation,2006:9889-9892
[49]吴建华,李娜等.基于CGA和ICA的人脸特征提取方法研究[J].计算机应用,2007,27(8)：2038-2040
[50]武妍,吴桂云.基于ICA/SVM的混合学习方法及在人脸检测中的应用[J].计算机工程与应用,2004,40(36)：87-101
[51]姚雪梅等.基于ICA和SVM的局部人脸识别方法[J].图形图像,2005：15-27
[52]郑宇杰,杨静宇等.基于独立成分分析和模糊支持向量机的人脸识别方法[J].系统仿真学报,2005,17(7)：1768-1770
[53]郑宇杰,於东军等.一种基于ICA和LDA组合的人脸识别新方法[J].计算机科学,2006,33(4)：194-197
[54]Turk M and Pentland A. Face Processing:Models for recognition. Proc. Intelligent Robots and Computer Vision, SPIE,1989,1,192:22-32
[55]徐艳,陈孝威.基于差分图像和PCA主元分析的人脸检测[J].计算机应用与软件,2008,25(8)：229-273
[56]Jian Yang, David Zhang, Jing-Yu Yang. Tow dimensional PCA:A new approach to appearance-based face representation and recognition. IEEE Trans. Pattern Anal. Machine Intell.2004,26(1):131-137
[57]Kressel UHG. Pairwise classification and support vector machines[M]. Cambridge:MIT Press,1999
[58]Vapnik V. The Nature of statistical learning theory [M]. New York:Springer Verlag,1995
[59]Kuncheva L. Combining classifiers by clustering, selection and decision templates[R]. Technical report(2000), University of Wales, UK