基于肤色的人脸检测系统
|
摘要
人脸检测作为人脸信息处理领域中的一项关键技术,在安防、视频会议等实际应用领域有着广泛的应用。但是由于人脸的复杂性和实际应用中的较高要求,人脸检测仍是研究中的难点问题。
本文归纳和分析了当前人脸检测的典型算法,然后设计了一个基于肤色分割和人脸验证的人脸检测系统。重点研究了不同色彩空间下的肤色特点和人脸验证等内容,并运用编程工具实现了本文讨论的算法,验证了算法的准确性。
论文的研究工作主要有以下几个方面:
(1)对色彩空间和肤色模型进行了较深入的研究,运用程序提取人脸数据库目标像素点的各个分量值,并对其进行统计分析,获取最佳分割阈值。在肤色分割的过程中,运用了图像处理中常用的形态学去噪以及设定区域像素累积阈值的方法消除图像中的噪声。随后又利用了人脸的几何特征对二值化图中可能的肤色区域进行大小、长宽比例和肤色填充率的分析。最后运用唇色检测方法,对候选人脸区域进行验证,最终标定检测到的人脸区域。
(2)本文运用直方图方法进行了统计分析,利用Visual C++编程工具对该算法进行了实现,并且在人脸测试图像集上作了许多实验。结果表明本系统对复杂背景下的彩色人脸检测较为准确,但是由于实际应用中的复杂度、以及背景的不确定性、复杂、干扰,本系统还是存在一些漏检和误检的情况,适应性方面有所欠缺。总的来说,本系统具有一定的实用性。
As a key technology of the field of face information processing, face detection has been widely applied in production and life, such as safety protection, video meeting and so on. However, due to the complexity of the problem and high demand in practical application, this problem becomes a hot research presently.
This dissertation studys current typical algorithms on face detection,and then designs a color-based face segmentation and face detection verification system. This dissertation mainly focus on different color spaces and color characteristics of face verification, and use the program to inspect the validity of the arithmetic mentioned in this dissertation.
The main contents of the dissertation are as follows:
(1) In order to acquire the best threshold value, the color space and comparison of different complexion’s clustering are lucubrated, and program tools are used to acquire the every weight of object pixel point in different color space in homemade face set. In the process of color segmentation, the methods are designed to eliminate noise including the morphological method in image processing and the method of setting the region picture element accumulation threshold value. Then the size of the color region, aspect ratio and color fill rate analysis about object region is carried out by utilizing face’s geometrical characters, in order that candidate face areas are output as processing result. Finally, candidate areas can be determined by using a lip color detection, and face region can be detected efficiently.
(2) The whole system designs and debugs in visual C++ 6.0. A large number of experiments for homemade face testing set has been done. The results show that system’s correctness is remarkable. But due to the complexity of practical application and the uncertainty of the background, there are some missing faces and false faces. In a word, the system is practical to a certain regions.
本文归纳和分析了当前人脸检测的典型算法,然后设计了一个基于肤色分割和人脸验证的人脸检测系统。重点研究了不同色彩空间下的肤色特点和人脸验证等内容,并运用编程工具实现了本文讨论的算法,验证了算法的准确性。
论文的研究工作主要有以下几个方面:
(1)对色彩空间和肤色模型进行了较深入的研究,运用程序提取人脸数据库目标像素点的各个分量值,并对其进行统计分析,获取最佳分割阈值。在肤色分割的过程中,运用了图像处理中常用的形态学去噪以及设定区域像素累积阈值的方法消除图像中的噪声。随后又利用了人脸的几何特征对二值化图中可能的肤色区域进行大小、长宽比例和肤色填充率的分析。最后运用唇色检测方法,对候选人脸区域进行验证,最终标定检测到的人脸区域。
(2)本文运用直方图方法进行了统计分析,利用Visual C++编程工具对该算法进行了实现,并且在人脸测试图像集上作了许多实验。结果表明本系统对复杂背景下的彩色人脸检测较为准确,但是由于实际应用中的复杂度、以及背景的不确定性、复杂、干扰,本系统还是存在一些漏检和误检的情况,适应性方面有所欠缺。总的来说,本系统具有一定的实用性。
As a key technology of the field of face information processing, face detection has been widely applied in production and life, such as safety protection, video meeting and so on. However, due to the complexity of the problem and high demand in practical application, this problem becomes a hot research presently.
This dissertation studys current typical algorithms on face detection,and then designs a color-based face segmentation and face detection verification system. This dissertation mainly focus on different color spaces and color characteristics of face verification, and use the program to inspect the validity of the arithmetic mentioned in this dissertation.
The main contents of the dissertation are as follows:
(1) In order to acquire the best threshold value, the color space and comparison of different complexion’s clustering are lucubrated, and program tools are used to acquire the every weight of object pixel point in different color space in homemade face set. In the process of color segmentation, the methods are designed to eliminate noise including the morphological method in image processing and the method of setting the region picture element accumulation threshold value. Then the size of the color region, aspect ratio and color fill rate analysis about object region is carried out by utilizing face’s geometrical characters, in order that candidate face areas are output as processing result. Finally, candidate areas can be determined by using a lip color detection, and face region can be detected efficiently.
(2) The whole system designs and debugs in visual C++ 6.0. A large number of experiments for homemade face testing set has been done. The results show that system’s correctness is remarkable. But due to the complexity of practical application and the uncertainty of the background, there are some missing faces and false faces. In a word, the system is practical to a certain regions.
引文
[1] R.Chellappa,C.L.Wilson and S.Sirohey. Human and machine recognition of faces: A survey[J].Proceedings of the IEEE,1995,83 (5):705～741
[2]张翠平,苏光大.人脸识别技术综述.中国图象图形学报,2000,5(11):885-894
[3]艾海舟,肖习攀,徐光佑.人脸检测与检索[J],计算机学报,2002,25(5):449-548
[4]赵丽红,刘纪红,徐心和.人脸检测方法综述[J],计算机应用研究,2004,(9):l-4
[5]王科俊,姚向辉.人脸图像检测与识别方法综述[J].自动化技术与应用,2004,23(12):5-8
[6]梁路宏,艾海舟等.人脸检测研究综述.计算机学报,2002,25(5):449-458
[7]卢春雨.人脸自动识别若干问题研究与系统实现[博士学位论文].清华大学,1998
[8] http://www.cqvip.com/qk/97838a/2003003/8038878.html
[9]山世光,高文,陈熙林.基于纹理分布和变形模板的面部特征提取[J],软件学报,2001,12(4):570-577
[10]孙艳丰,梁永涛等.基于肤色模型和重心模板的人脸检测.北京工业大学学报,2006,32(5)
[11]周杰,卢春雨,张长水,李衍达.人脸自动识别方法综述,电子学报,2000,28(4):102-106
[12]梁路宏,艾海舟,何克忠.基于多模板匹配的单人脸检测.中国图象图形学报,1999,10(4)
[13] K.Sung and T.Poggio. Example-based learning for view based human face detection. IEEE Transactions on Pattern Analysis and Machine Intelligence ,1998, 20(l):39-51
[14] http:// www.cc.gatech.edu/Wtljx/Publication.html
[15] Ming-HsuanYang,Dvaid J. Kriegman and Narendra Ahuja. Detecting Faces in Images: A survey, IEEE Transactions on Pattern Analysis and Machine Intelligence,2002,24(l):34-58
[16] L.Stringa. Automatic person recognition by acoustic and geometric features[J],Technical Report 9205-04,IRST,1991
[17] R.Brunelli,T.Poggio. Face Recognition Through Geometrical Features[R], Proceedings of ECCV, 1992,792-800
[18] N.Intrator,D.Reisfeld,Y.Yeshurun. Extraction of Facial Features for Recognition Using Neural Networks[R], Proceedings of International Workshop on Automatic Face and Gesture Recognition,1995,260-265
[19] H.Zabrodsky ,S.Peleg. Symmetry as a Continuous Feature. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1995,17:1154-1166
[20] Akamatsu. The IEEE Third International Conference on Automatic Face and Gesture Recognition[R], 1998,4: 14-16
[21] Srihari, Govindaraju. Analysis of Textual Images Using the Hough Transform[J],Machine Vision Applications,1989,2:141-153
[22] G.Yang,T.S.Huang. Human Face Detection in Complex Background[J],Pattern Recognition, 1994,27(1):53-63
[23]梁路宏,艾海舟,肖习攀.基于模板匹配与支持矢量机的人脸检测[J],计算机学报.2002,1(25):22-29
[24] Z.KouzaniA,F.Sammut K. Commonsense knowledge based face detection. Proceedings Conference on Intelligent Engineering Systems, 1997,215-220
[25] B. Moghaddam, A.Pentland. Probabilistic visual learning for object representation. IEEE Transactions on Pattern Analysis and Machine Intelligence,1997,19(7):696-710
[26] M.Yang, N.Ahuja, D.Kriegman. Face detection using mixtures of linear subspaces. In:Proc Conference on Automatic Face and Gesture Recognition, Grenoble, France,2000,70-76
[27] Z.Ghahramani,E.Hinton. The EM algorithm for mixtures of factor analyzers .University of Toronto: Technical Report CRG-TR-96-1,1996.
[28] A.Yuille,P.Hallinan,D.Cohen. Feature Extraction from Faces Using c Templates. International Journal of Computer Vision,1992,8(2):99-111
[29] K-KSung. Learning and Example Selection for Object and Patten Detection[J]PhD thesis,Massaehusetts Inst of Technology,1996.
[30] Platt J C. Sequential minimal optimization: A fast algorithm for training support vector machines. MSR: Technical Report MSR-TR-98-14,1998
[31] E.Osuna, R.Freund, F.Girosi. Training support vector machines: Application to face detection. Proceedings Computer Vision and Pattern Recognition, Puerto Rico,1997.130-136
[32]梁路宏,艾海舟等.基于多关联模板匹配与人工神经网确认的人脸检测,电子学报,2001,29(6):
[33] Rafael C.Gonzalez,Richard E.Woods著,阮秋琦,阮宇智等译,冈萨雷斯数字图像处理(第二版)
[34]郑南宁.计算机视觉与模式识别.北京:国防工业出版社,1998
[35]杨枝灵,王开.Visua1C++数字图像获取处理及实践应用.北京:人民邮电出版社,2003
[36]陶霖密,彭振云,徐光枯.人体的肤色特征[J].软件学报,2001,5.
[37]沈常宇,许潘园.肤色建模和肤色分割的人脸定位研究[J].光电工程, 2007年09期
[38]张宏林.数字图像模式识别技术及工程实践.北京:人民邮电出版社,2003
[39]陆继祥.基子24位彩色人脸图像嘴唇的分割与提取,计算机工程2003,2:147-148
[2]张翠平,苏光大.人脸识别技术综述.中国图象图形学报,2000,5(11):885-894
[3]艾海舟,肖习攀,徐光佑.人脸检测与检索[J],计算机学报,2002,25(5):449-548
[4]赵丽红,刘纪红,徐心和.人脸检测方法综述[J],计算机应用研究,2004,(9):l-4
[5]王科俊,姚向辉.人脸图像检测与识别方法综述[J].自动化技术与应用,2004,23(12):5-8
[6]梁路宏,艾海舟等.人脸检测研究综述.计算机学报,2002,25(5):449-458
[7]卢春雨.人脸自动识别若干问题研究与系统实现[博士学位论文].清华大学,1998
[8] http://www.cqvip.com/qk/97838a/2003003/8038878.html
[9]山世光,高文,陈熙林.基于纹理分布和变形模板的面部特征提取[J],软件学报,2001,12(4):570-577
[10]孙艳丰,梁永涛等.基于肤色模型和重心模板的人脸检测.北京工业大学学报,2006,32(5)
[11]周杰,卢春雨,张长水,李衍达.人脸自动识别方法综述,电子学报,2000,28(4):102-106
[12]梁路宏,艾海舟,何克忠.基于多模板匹配的单人脸检测.中国图象图形学报,1999,10(4)
[13] K.Sung and T.Poggio. Example-based learning for view based human face detection. IEEE Transactions on Pattern Analysis and Machine Intelligence ,1998, 20(l):39-51
[14] http:// www.cc.gatech.edu/Wtljx/Publication.html
[15] Ming-HsuanYang,Dvaid J. Kriegman and Narendra Ahuja. Detecting Faces in Images: A survey, IEEE Transactions on Pattern Analysis and Machine Intelligence,2002,24(l):34-58
[16] L.Stringa. Automatic person recognition by acoustic and geometric features[J],Technical Report 9205-04,IRST,1991
[17] R.Brunelli,T.Poggio. Face Recognition Through Geometrical Features[R], Proceedings of ECCV, 1992,792-800
[18] N.Intrator,D.Reisfeld,Y.Yeshurun. Extraction of Facial Features for Recognition Using Neural Networks[R], Proceedings of International Workshop on Automatic Face and Gesture Recognition,1995,260-265
[19] H.Zabrodsky ,S.Peleg. Symmetry as a Continuous Feature. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1995,17:1154-1166
[20] Akamatsu. The IEEE Third International Conference on Automatic Face and Gesture Recognition[R], 1998,4: 14-16
[21] Srihari, Govindaraju. Analysis of Textual Images Using the Hough Transform[J],Machine Vision Applications,1989,2:141-153
[22] G.Yang,T.S.Huang. Human Face Detection in Complex Background[J],Pattern Recognition, 1994,27(1):53-63
[23]梁路宏,艾海舟,肖习攀.基于模板匹配与支持矢量机的人脸检测[J],计算机学报.2002,1(25):22-29
[24] Z.KouzaniA,F.Sammut K. Commonsense knowledge based face detection. Proceedings Conference on Intelligent Engineering Systems, 1997,215-220
[25] B. Moghaddam, A.Pentland. Probabilistic visual learning for object representation. IEEE Transactions on Pattern Analysis and Machine Intelligence,1997,19(7):696-710
[26] M.Yang, N.Ahuja, D.Kriegman. Face detection using mixtures of linear subspaces. In:Proc Conference on Automatic Face and Gesture Recognition, Grenoble, France,2000,70-76
[27] Z.Ghahramani,E.Hinton. The EM algorithm for mixtures of factor analyzers .University of Toronto: Technical Report CRG-TR-96-1,1996.
[28] A.Yuille,P.Hallinan,D.Cohen. Feature Extraction from Faces Using c Templates. International Journal of Computer Vision,1992,8(2):99-111
[29] K-KSung. Learning and Example Selection for Object and Patten Detection[J]PhD thesis,Massaehusetts Inst of Technology,1996.
[30] Platt J C. Sequential minimal optimization: A fast algorithm for training support vector machines. MSR: Technical Report MSR-TR-98-14,1998
[31] E.Osuna, R.Freund, F.Girosi. Training support vector machines: Application to face detection. Proceedings Computer Vision and Pattern Recognition, Puerto Rico,1997.130-136
[32]梁路宏,艾海舟等.基于多关联模板匹配与人工神经网确认的人脸检测,电子学报,2001,29(6):
[33] Rafael C.Gonzalez,Richard E.Woods著,阮秋琦,阮宇智等译,冈萨雷斯数字图像处理(第二版)
[34]郑南宁.计算机视觉与模式识别.北京:国防工业出版社,1998
[35]杨枝灵,王开.Visua1C++数字图像获取处理及实践应用.北京:人民邮电出版社,2003
[36]陶霖密,彭振云,徐光枯.人体的肤色特征[J].软件学报,2001,5.
[37]沈常宇,许潘园.肤色建模和肤色分割的人脸定位研究[J].光电工程, 2007年09期
[38]张宏林.数字图像模式识别技术及工程实践.北京:人民邮电出版社,2003
[39]陆继祥.基子24位彩色人脸图像嘴唇的分割与提取,计算机工程2003,2:147-148