基于AdaBoost算法的自动人脸检测与识别
|
摘要
AdaBoost(Adaptive Boosting)算法是1995年Freund和Schapire提出的一种快速人脸检测算法,是人脸检测领域里程碑式的进步,这种算法根据弱学习的反馈,适应性地调整假设的错误率,在保证检测速度的前提下,检测正确率得到了很大的提高。
本文在深入学习和研究AdaBoost算法的基础上,收集了大量的人脸样本和非人脸样本,训练出了共有1152个弱分类器组成的16级强分类器的级联分类器。
为了克服Adaboost检测算法因分类器运算量巨大而导致的检测速度较慢的缺点,在检测过程中设置了分类器前端优化。优化包括两个方面:首先,在图像进入检测算法之前,利用运算量较小的RGB肤色模型进行肤色检测,找到图像中最大的人脸区域,从而预测检测窗口放大尺度;其次,在检测算法的前端利用肤色积分图筛选肤色像素比例较小的子窗口,减小分类器的负担。
静态图像的人脸检测算法对每幅图像都进行全图检测,不考虑帧间信息,而动态序列图像中帧与帧之间是有一定联系的,所以动态序列图像人脸检测过程中使用了人脸跟踪来预测下一帧人脸的位置和区域,大大提高检测速度,满足了处理动态序列图像的实时性要求。
最后,使用基于几何特征的方法识别了检测到的人脸。
The AdaBoost (Adaptive Boosting) algorithm, introduced in 1995 by Freund and Schapire, is an algorithm to detect faces very quickly, which is a landmark advance in the field of face detection. This algorithm adjusts adaptively the errors of the weak hypotheses by the feedback of weaklearn, which advances detection accuracy rates greatly and doesn't reduce detection efficiency.
This research, on the basis of studying the AdaBoost algorithm thoroughly, has collected lots of human face samples and non-face samples and trained a cascade classifier of 16 strong classifiers which has been made of by 1152 weak classifiers.
In order to conquer the defect of lower detection speed when the AdaBoost algorithm has tremendous computations, the research has embedded improvements in front of classifier in the process of face detection. The improvements has tow parts: firstly, before the image inputted the AdaBoost algorithm, applying the skin color model RGB which needs fewer computations to detect the skin color pixels and finding the maximum face area in the image; secondly, in front of the AdaBoost algorithm using skin color integral image to abandon the sub-windows which having lower proportion of skin color pixels, which decrease the burden of classifier greatly.
Because some information exist between tow frames in dynamic sequence images, but the face detection algorithm for static image has not used the information, we have used face track to forecast the position and area of the face in the next frame. This method has advanced the speed of face detection, and satisfied the timely demand of dealing with dynamic sequence images.
Lastly, the research has used the method based on the geometry feature to identify the face detected.
本文在深入学习和研究AdaBoost算法的基础上,收集了大量的人脸样本和非人脸样本,训练出了共有1152个弱分类器组成的16级强分类器的级联分类器。
为了克服Adaboost检测算法因分类器运算量巨大而导致的检测速度较慢的缺点,在检测过程中设置了分类器前端优化。优化包括两个方面:首先,在图像进入检测算法之前,利用运算量较小的RGB肤色模型进行肤色检测,找到图像中最大的人脸区域,从而预测检测窗口放大尺度;其次,在检测算法的前端利用肤色积分图筛选肤色像素比例较小的子窗口,减小分类器的负担。
静态图像的人脸检测算法对每幅图像都进行全图检测,不考虑帧间信息,而动态序列图像中帧与帧之间是有一定联系的,所以动态序列图像人脸检测过程中使用了人脸跟踪来预测下一帧人脸的位置和区域,大大提高检测速度,满足了处理动态序列图像的实时性要求。
最后,使用基于几何特征的方法识别了检测到的人脸。
The AdaBoost (Adaptive Boosting) algorithm, introduced in 1995 by Freund and Schapire, is an algorithm to detect faces very quickly, which is a landmark advance in the field of face detection. This algorithm adjusts adaptively the errors of the weak hypotheses by the feedback of weaklearn, which advances detection accuracy rates greatly and doesn't reduce detection efficiency.
This research, on the basis of studying the AdaBoost algorithm thoroughly, has collected lots of human face samples and non-face samples and trained a cascade classifier of 16 strong classifiers which has been made of by 1152 weak classifiers.
In order to conquer the defect of lower detection speed when the AdaBoost algorithm has tremendous computations, the research has embedded improvements in front of classifier in the process of face detection. The improvements has tow parts: firstly, before the image inputted the AdaBoost algorithm, applying the skin color model RGB which needs fewer computations to detect the skin color pixels and finding the maximum face area in the image; secondly, in front of the AdaBoost algorithm using skin color integral image to abandon the sub-windows which having lower proportion of skin color pixels, which decrease the burden of classifier greatly.
Because some information exist between tow frames in dynamic sequence images, but the face detection algorithm for static image has not used the information, we have used face track to forecast the position and area of the face in the next frame. This method has advanced the speed of face detection, and satisfied the timely demand of dealing with dynamic sequence images.
Lastly, the research has used the method based on the geometry feature to identify the face detected.
引文
[1]Zhao W.,Chellappa R.,et al.Face recognition:a literature survey.ACM Computing Surveys,2003,35(4):399-458.
[2]马希荣,王志良.远程教育中和谐人机情感交互模型的研究[J].计算机科学,2005,32(9):182-183.
[3]高秀梅,杨静宇,金忠等.基于核的Foley-Sammon鉴别分析与人脸识别[J].计算机辅助设计与图形学学报,2004,16(7):962-967.
[4]陈粟,倪林.一种特征脸分析和小波变换相结合的人脸识别方法[J].计算机应用,2004,24(10):75-78.
[5]丁宾,高新波,姬红兵.基于离散余弦变换的人脸画像识别方法[J].计算机工程,2004,30(20):151-153.
[6]陈才扣,杨静宇,杨健.一种融合PCA和KFDA的人脸识别方法[J].控制与决策,2004,19(10):1147-1151.
[7]廖红文,冯国灿等.压缩域上人脸识别的研究[J].中山大学学报(自然科学版),2004,43(5):16-19.
[8]高西奇,周洪祥,何振亚.基于小波变换的主元分析人脸图像识别[J].东南大学学报,1996,26(2):137-141.
[9]沈获帆,滕晓龙,刘重庆.基于Gabor小波和支持向量机的人脸识别方法[J].红外与激光工程,2004,33(6):600-602.
[10]彭进业,王大凯等.基于小波分解系数的贝叶斯人脸识别方法[J].光子学报,2001,30(10):1263-1269.
[11]马希荣,王嵘.一种基于最近邻决策的点集分类方法的确定与实现[J].计算机科学,2007,34(12):183-185.
[12]荆晓远,金忠,杨静宇.基于带通滤波和遗传算法的人脸图像预处理及识别[J].中国图像图形学报,1998,3(10):840-844.
[13]史泽林,李德强,黄莎白.基于模糊准则的小波特征选择在人脸识别中的应用[J].信息与控制,2005,34(1):50-54.
[14]Manjunath B.S.,Chellappa R.,Malsburg C.A feature based approach to face recognition.Proc.IEEE Conf.On Computer Vision and Pattern Recognition,1992:373-378.
[15]Samaria F.,Young S.HMM based architecture for face identification.Image Vis.Comput.,1994,12:537-583.
[16]Brunelli R.,Poggio Y.HyperBF networks for gender classification.Proc.DARP Image Understanding Workshop,1992:311-314.
[17]Pentland A.,Moghaddrn B.,Starner T.View based and modular eigenspaces for face recognition.Proe.IEEE Conf.On Computer Vision and Pattern Recognition,1994:1-7.
[18]Huang J.,Hersele B.,Blanz V.Component based face recognition with 3D morphable models.International Conference on Audio- and Vdeo-Based Person Authentication,2003:1-5.
[19]Turk M.,Pentland A.Eigenfacesfor recognition.J.Cogn.Nenrosei.,1991,3:72-86.
[20]Bartlett M.S.,Lades H.M.,Sejnowski T.Independent component representation for face recognition.Proe.SPIE Sym.On Electronic Imagings:Science and Technology,1998:528-539.
[21]Moghaddam B.,Jebara T.,Pentland A.Bayesian face recognition.Pattern Recognition,2000,33:1171-1782.
[22]Belhumeur P N.,Hespanha J.P,Kriegam D.J.Eigenfaces vs.Fisherfaces:Recognition using class specific linear projection.IEEE Trans.On Pattern Analysis and Machine Intelligence,1997,19(7):711-720.
[23]Zhong J.,Yang J.,and Hu Z.Face recognition based on the uncorrelated discriminant transformation.Pattern Recognition,2001,34:1405-1416.
[24]Cheng Y,Liu K.,et al.A robust algebraic method for human and face recognition.Proe.11~th Int.Conf.On Patt.Reeog.,1992:221-224.
[25]Phillips P.J.Support vector machines applied to face recognition.Adv.Neural Inform.Process.Syst.,1998,11:803-809.
[26]Li Yongmin,Gong Shaogang.Recognising trajectories of facial identities using kernel discriminant analysis.Image and Vision Computing,2003,21:1077-1086.
[27]Wiskott L.Fellous J.M.Face recognition by elastic bunch graph matching.IEEE Trans.Patt.Anal.Mach.Intell.,1997,19:775-779.
[28]Liu C.,Wechsler H.Evolutionary pursuit and its application to face recognition.IEEE Trans.Patt.Anal.Math.Intell.,2000,22:570-582.
[29]Liu C.J.Gabor-based kernel PCA with fractional power polynomial models for face recognition.IEEE Trans.On Patt.Anal.and Mach.Intelt.,2004,26(5):572-581.
[30]田欣.基于不同色彩空间的肤色模型[J].西安科技学院学报,2001,21(4):369-371.
[31]Stem H.Efros B.Adaptive color space switching for face tracking in multi-colored lighting environments.Fifth IEEE International Conference,Automatic Face and Gesture Recognition.2002:236-241.
[32]Wang Yanjiang,Yuan Baozong.Human face tracking using Gaussian mixture models and fuzzy shape analysi.ICSP'04 7th International Conference,Signal Processing.2004,2:1322-1325.
[33]梁路宏,艾海舟,何克忠,张钹.基于仿射模板匹配的多角度单人脸定位[J].计算机学报,2000,23(6):641-646.
[34]梁路宏,艾海舟,何克忠,张钹.基于多关联模板匹配的人脸检测[J].软件学报,2001,12(1):94-102.
[35]王洪群,彭嘉雄,强赞霞.基于边缘和纹理特征相结合的快速人脸精确定位方法[J].计算机工程与应用,2004,40(7):27-32.
[36]Rowley H.A.Neural Network-Based Face Detection.IEEE Transactions on Pattern Analysis andMachine Intelligence,1998,20(1):23-38.
[37]E.Osuna R.Freund F.Girosi.Training support vector machines:An application to face detection.IEEE Conf,Computer Vision and Pattern Recognition.Hilton Head Island,South Carolina,2000:130-136.
[38]陈茂林,戚飞虎.自组织隐马尔可夫模型的人脸检测研究[J].计算机学报,2002,25(11):1165-1169.
[39]张九龙,张毅坤.人脸检测的贝叶斯特征判别方法[J].计算机工程与应用,2004,40(29):164-165.
[40]Schapire R E.The strength of weak learn ability.Machine Learning.1990.5(2):197-227
[41]Kearns M.Valiant L G.Learning Boolean Formulae or Factoring:[Technical Report TR-1488].Cambridge.MA:Havard University Aiken Computation Laboratory.1988
[42]Keams M,Valiant L G.Crytographic Limitation on Learning Boolean Formulae and Finite Automata.In:Proc of the 21 annual ACM Sytoposlum on Theory of Computing.New York NY:ACM press.1989:433-444
[43]Valiant L G.A theory of the learnable.Communications of the ACM.1984,27(11):1134- 1142
[44]Freund Y.Boosting a weak learning algorithm by majority.Information 1994 and computation,1995,141(2):256-285
[45]Keams M.I.Vazirani L G.Learning Boolean formulae or finite automata is as hard as storing:[Technical Report TR -14- 88]Harvard University Aiken Computation Laboratory,Aug.1988
[46]Kearns M,Valiant L G.Cryptographic limitations on learning Boolean formulae and finite automata.Journal of the Association for Computing Machinery,41(1):67-95
[47]Freund Y.Schapire R E.A decision theoretic generalization of online learning and an application to boosting.Journal of Computer and System Science,1997,55(1):119- 139
[48]Viola P.,Jones M.J.Robust Real-Time Face Detection.International Journal of Computer Vision 2004,57(2):137-154.
[49]Peer P,Kovac J,Solina F.2003 Human Skin Colour Clustering for Face Detection.In:Submined to Eurocon 2003 International Conference on Computer as a Tool,2003.
[50]涂承胜,刁力力,鲁明羽,陆玉昌.Boosting家族AdaBoost系列代表算法[J].计算机科学,2003 Vol.30 No.3
[51]涂承胜,陆玉昌.BoostingN论基础[J].计算机科学,2004,Vol.31 No.10
[52]Bo Wu,HaizhouAi,Chang Huang,Shihong Lao.Fast rotation invariant multi-view face detection based on real Adaboost.Sixth IEEE International Conference,Automatic Face and Gesture Recognition.2004:79-84.
[53]邓亚峰,苏光大,傅博.计算机工程[J].2006(11):222-224.
[54]陈启泉,邱文字,陈维斌.标准正面人脸图像的特征提取[J].华侨大学学报,2000,2l(4):413-418.
[2]马希荣,王志良.远程教育中和谐人机情感交互模型的研究[J].计算机科学,2005,32(9):182-183.
[3]高秀梅,杨静宇,金忠等.基于核的Foley-Sammon鉴别分析与人脸识别[J].计算机辅助设计与图形学学报,2004,16(7):962-967.
[4]陈粟,倪林.一种特征脸分析和小波变换相结合的人脸识别方法[J].计算机应用,2004,24(10):75-78.
[5]丁宾,高新波,姬红兵.基于离散余弦变换的人脸画像识别方法[J].计算机工程,2004,30(20):151-153.
[6]陈才扣,杨静宇,杨健.一种融合PCA和KFDA的人脸识别方法[J].控制与决策,2004,19(10):1147-1151.
[7]廖红文,冯国灿等.压缩域上人脸识别的研究[J].中山大学学报(自然科学版),2004,43(5):16-19.
[8]高西奇,周洪祥,何振亚.基于小波变换的主元分析人脸图像识别[J].东南大学学报,1996,26(2):137-141.
[9]沈获帆,滕晓龙,刘重庆.基于Gabor小波和支持向量机的人脸识别方法[J].红外与激光工程,2004,33(6):600-602.
[10]彭进业,王大凯等.基于小波分解系数的贝叶斯人脸识别方法[J].光子学报,2001,30(10):1263-1269.
[11]马希荣,王嵘.一种基于最近邻决策的点集分类方法的确定与实现[J].计算机科学,2007,34(12):183-185.
[12]荆晓远,金忠,杨静宇.基于带通滤波和遗传算法的人脸图像预处理及识别[J].中国图像图形学报,1998,3(10):840-844.
[13]史泽林,李德强,黄莎白.基于模糊准则的小波特征选择在人脸识别中的应用[J].信息与控制,2005,34(1):50-54.
[14]Manjunath B.S.,Chellappa R.,Malsburg C.A feature based approach to face recognition.Proc.IEEE Conf.On Computer Vision and Pattern Recognition,1992:373-378.
[15]Samaria F.,Young S.HMM based architecture for face identification.Image Vis.Comput.,1994,12:537-583.
[16]Brunelli R.,Poggio Y.HyperBF networks for gender classification.Proc.DARP Image Understanding Workshop,1992:311-314.
[17]Pentland A.,Moghaddrn B.,Starner T.View based and modular eigenspaces for face recognition.Proe.IEEE Conf.On Computer Vision and Pattern Recognition,1994:1-7.
[18]Huang J.,Hersele B.,Blanz V.Component based face recognition with 3D morphable models.International Conference on Audio- and Vdeo-Based Person Authentication,2003:1-5.
[19]Turk M.,Pentland A.Eigenfacesfor recognition.J.Cogn.Nenrosei.,1991,3:72-86.
[20]Bartlett M.S.,Lades H.M.,Sejnowski T.Independent component representation for face recognition.Proe.SPIE Sym.On Electronic Imagings:Science and Technology,1998:528-539.
[21]Moghaddam B.,Jebara T.,Pentland A.Bayesian face recognition.Pattern Recognition,2000,33:1171-1782.
[22]Belhumeur P N.,Hespanha J.P,Kriegam D.J.Eigenfaces vs.Fisherfaces:Recognition using class specific linear projection.IEEE Trans.On Pattern Analysis and Machine Intelligence,1997,19(7):711-720.
[23]Zhong J.,Yang J.,and Hu Z.Face recognition based on the uncorrelated discriminant transformation.Pattern Recognition,2001,34:1405-1416.
[24]Cheng Y,Liu K.,et al.A robust algebraic method for human and face recognition.Proe.11~th Int.Conf.On Patt.Reeog.,1992:221-224.
[25]Phillips P.J.Support vector machines applied to face recognition.Adv.Neural Inform.Process.Syst.,1998,11:803-809.
[26]Li Yongmin,Gong Shaogang.Recognising trajectories of facial identities using kernel discriminant analysis.Image and Vision Computing,2003,21:1077-1086.
[27]Wiskott L.Fellous J.M.Face recognition by elastic bunch graph matching.IEEE Trans.Patt.Anal.Mach.Intell.,1997,19:775-779.
[28]Liu C.,Wechsler H.Evolutionary pursuit and its application to face recognition.IEEE Trans.Patt.Anal.Math.Intell.,2000,22:570-582.
[29]Liu C.J.Gabor-based kernel PCA with fractional power polynomial models for face recognition.IEEE Trans.On Patt.Anal.and Mach.Intelt.,2004,26(5):572-581.
[30]田欣.基于不同色彩空间的肤色模型[J].西安科技学院学报,2001,21(4):369-371.
[31]Stem H.Efros B.Adaptive color space switching for face tracking in multi-colored lighting environments.Fifth IEEE International Conference,Automatic Face and Gesture Recognition.2002:236-241.
[32]Wang Yanjiang,Yuan Baozong.Human face tracking using Gaussian mixture models and fuzzy shape analysi.ICSP'04 7th International Conference,Signal Processing.2004,2:1322-1325.
[33]梁路宏,艾海舟,何克忠,张钹.基于仿射模板匹配的多角度单人脸定位[J].计算机学报,2000,23(6):641-646.
[34]梁路宏,艾海舟,何克忠,张钹.基于多关联模板匹配的人脸检测[J].软件学报,2001,12(1):94-102.
[35]王洪群,彭嘉雄,强赞霞.基于边缘和纹理特征相结合的快速人脸精确定位方法[J].计算机工程与应用,2004,40(7):27-32.
[36]Rowley H.A.Neural Network-Based Face Detection.IEEE Transactions on Pattern Analysis andMachine Intelligence,1998,20(1):23-38.
[37]E.Osuna R.Freund F.Girosi.Training support vector machines:An application to face detection.IEEE Conf,Computer Vision and Pattern Recognition.Hilton Head Island,South Carolina,2000:130-136.
[38]陈茂林,戚飞虎.自组织隐马尔可夫模型的人脸检测研究[J].计算机学报,2002,25(11):1165-1169.
[39]张九龙,张毅坤.人脸检测的贝叶斯特征判别方法[J].计算机工程与应用,2004,40(29):164-165.
[40]Schapire R E.The strength of weak learn ability.Machine Learning.1990.5(2):197-227
[41]Kearns M.Valiant L G.Learning Boolean Formulae or Factoring:[Technical Report TR-1488].Cambridge.MA:Havard University Aiken Computation Laboratory.1988
[42]Keams M,Valiant L G.Crytographic Limitation on Learning Boolean Formulae and Finite Automata.In:Proc of the 21 annual ACM Sytoposlum on Theory of Computing.New York NY:ACM press.1989:433-444
[43]Valiant L G.A theory of the learnable.Communications of the ACM.1984,27(11):1134- 1142
[44]Freund Y.Boosting a weak learning algorithm by majority.Information 1994 and computation,1995,141(2):256-285
[45]Keams M.I.Vazirani L G.Learning Boolean formulae or finite automata is as hard as storing:[Technical Report TR -14- 88]Harvard University Aiken Computation Laboratory,Aug.1988
[46]Kearns M,Valiant L G.Cryptographic limitations on learning Boolean formulae and finite automata.Journal of the Association for Computing Machinery,41(1):67-95
[47]Freund Y.Schapire R E.A decision theoretic generalization of online learning and an application to boosting.Journal of Computer and System Science,1997,55(1):119- 139
[48]Viola P.,Jones M.J.Robust Real-Time Face Detection.International Journal of Computer Vision 2004,57(2):137-154.
[49]Peer P,Kovac J,Solina F.2003 Human Skin Colour Clustering for Face Detection.In:Submined to Eurocon 2003 International Conference on Computer as a Tool,2003.
[50]涂承胜,刁力力,鲁明羽,陆玉昌.Boosting家族AdaBoost系列代表算法[J].计算机科学,2003 Vol.30 No.3
[51]涂承胜,陆玉昌.BoostingN论基础[J].计算机科学,2004,Vol.31 No.10
[52]Bo Wu,HaizhouAi,Chang Huang,Shihong Lao.Fast rotation invariant multi-view face detection based on real Adaboost.Sixth IEEE International Conference,Automatic Face and Gesture Recognition.2004:79-84.
[53]邓亚峰,苏光大,傅博.计算机工程[J].2006(11):222-224.
[54]陈启泉,邱文字,陈维斌.标准正面人脸图像的特征提取[J].华侨大学学报,2000,2l(4):413-418.