网络不良图片过滤技术研究
|
摘要
随着互联网技术的不断发展,网络已经成为人们获取信息的主要途径之一。图像作为最直观的信息传递方式,也越来越受到网民的欢迎和重视,但也同时成为不法分子获利的一个主要手段。因此,网络图像的过滤也逐渐成为科学研究的一个热点问题。
本文介绍了一种基于实际网络环境的不良图像快速过滤算法。全文包括3大部分：肤色检验颜色模型的选取；图像中人脸的检测；在检测到人脸情况下的性别判定。共提出了3个创新点,肤色模型的选取,人脸检测中的超分辨算法和基于人脸的性别判定。
首先,在对图像进行肤色检测的模块中,本文提出了一种HSV, YUV和YIQ空间上的混合肤色模型的方法,并给出了实验分析,取得了较为理想的实验结果。然后,在人脸检测模块中,提出了PCA(主成份分析)与MAP(最大后验概率)相结合的人脸超分辨算法,是本文的核心与创新内容。最后,在性别判定模块中,本文引入了PCA-SIFT算法,通过模糊矢量量化网络进行分类。在40：1的实际非对称网络图像数据环境下进行试验,正常图像的识别率达到96.3%,误识率为3.5%,不良图像的识别率为69.2%,平均识别率达到95.8%。
实验结果表明了基于混合模型的过滤算法具有较高的平均识别率,但存在过滤速度慢,不良图像识别率不高等情况。如何改进检测率与提高算法速度是今后进一步研究的课题。
The network has become one of the main accesses to get information with the fast development of internet technology. Image is more and more popular for the intuitive features, but it also becomes a major means of profit for criminals. Therefore, the erotic image filtering is becoming a hot topic of scientific research.
The paper introduces a new method of erotic image filtering based on real internet. It includes three parts:skin color detection model; human face detection; and sex determination in case of the face is detected. We present three innovations, color model selection, and face detection with the super-resolution algorithm and the sex determination.
First, the HSV, YUV and YIQ color space model were present for the skin color detection and analysis the experimental results. Then a hybrid method mixed with PCA (principal component analysis) and MAP (maximum a posteriori probability) is proposed for face super-resolution in the human face detection modle, it is a key innovative content in the paper. Finally, in sex determination modle, we introduce the PCA-SIFT algorithm and LVQ (Learning Vector Quantization) for sex detection. With the experiment in the real conditions, the ratio of normal image to erotic image is 40:1, the recognition of erotic image can achieve 69.2%, the recognition of normal image can achieve 96.3%, and the average recognition rate can achieve 95.8%.
Experimental results show that the method we present get a higher average recognition rate, but recognition rate of erotic image need to be improve, and improve the speed of the algorithm is also the subject of further study.
本文介绍了一种基于实际网络环境的不良图像快速过滤算法。全文包括3大部分：肤色检验颜色模型的选取；图像中人脸的检测；在检测到人脸情况下的性别判定。共提出了3个创新点,肤色模型的选取,人脸检测中的超分辨算法和基于人脸的性别判定。
首先,在对图像进行肤色检测的模块中,本文提出了一种HSV, YUV和YIQ空间上的混合肤色模型的方法,并给出了实验分析,取得了较为理想的实验结果。然后,在人脸检测模块中,提出了PCA(主成份分析)与MAP(最大后验概率)相结合的人脸超分辨算法,是本文的核心与创新内容。最后,在性别判定模块中,本文引入了PCA-SIFT算法,通过模糊矢量量化网络进行分类。在40：1的实际非对称网络图像数据环境下进行试验,正常图像的识别率达到96.3%,误识率为3.5%,不良图像的识别率为69.2%,平均识别率达到95.8%。
实验结果表明了基于混合模型的过滤算法具有较高的平均识别率,但存在过滤速度慢,不良图像识别率不高等情况。如何改进检测率与提高算法速度是今后进一步研究的课题。
The network has become one of the main accesses to get information with the fast development of internet technology. Image is more and more popular for the intuitive features, but it also becomes a major means of profit for criminals. Therefore, the erotic image filtering is becoming a hot topic of scientific research.
The paper introduces a new method of erotic image filtering based on real internet. It includes three parts:skin color detection model; human face detection; and sex determination in case of the face is detected. We present three innovations, color model selection, and face detection with the super-resolution algorithm and the sex determination.
First, the HSV, YUV and YIQ color space model were present for the skin color detection and analysis the experimental results. Then a hybrid method mixed with PCA (principal component analysis) and MAP (maximum a posteriori probability) is proposed for face super-resolution in the human face detection modle, it is a key innovative content in the paper. Finally, in sex determination modle, we introduce the PCA-SIFT algorithm and LVQ (Learning Vector Quantization) for sex detection. With the experiment in the real conditions, the ratio of normal image to erotic image is 40:1, the recognition of erotic image can achieve 69.2%, the recognition of normal image can achieve 96.3%, and the average recognition rate can achieve 95.8%.
Experimental results show that the method we present get a higher average recognition rate, but recognition rate of erotic image need to be improve, and improve the speed of the algorithm is also the subject of further study.
引文
[1]我国互联网发展已形成一定规模仍有较大发展空间,http://news.sohu.com/20091102/n267899225.shtml
[2]《第 27 次中国互联网络发展状况统计报告》 ,http://tech.163.com/special/cnnic27/
[3]整治网络传播淫秽色情信息经验交流会召开 刘云山讲话,http://news.xinhuanet.com/politics/2010-12/13/c_12875807.htm
[4]九部门联合依法整治网络环境取得明显成效,http://www.scio.gov.cn/ztk/hlwxx/01/03/3/201002/t545852.htm
[5]美萍电脑软件工作室.美萍反黄专家：中国,ZL01117726.8[P/OL].(2002). http://www.msoft.net/shield.htm.
[6]中国科学院下属科大讯飞公司.火眼金睛反黄软件：中国,ZL01117731.4[P/OL]. (2002). http://www.iflytek.com.
[7]北京展翅鸟科技公司. 网络爸爸软件：中国,ZL01117730.6[P/OL]. (2002). http://www.tueagles.com/baba.
[8]FLECKM M, FORSYTH D A, BREGLER C. Finding naked people[C]//Proc of the 4th European Conference on Computer Vision. London:Springer-Verlag,1996: 593-602.
[9]WANG J Z,LI Jia,WEDERHOLD G,et al. System for screening objectionable images [j]. Computer Communications Journa 1,1998,21(15):1355-1360.
[10]JONES M J, REHG J M. Statistical color models with application to skin detection [C]//Proc of Computer Vision and Pattern Recognition.1999:274-280.
[11]DUAN Li-juan, CUI Guo-qin, GAO Wen, et al. Adult image detection method based on skin color model and support vector machine [C]//Proc of the 5th Asian Conference on Computer Vision.2002:797-800.
[12]徐欣欣,袁华.一种基于自适应小波不变矩的色情图像过滤方法[J].清华大学学报：自然科学版,2005,45(1)：1878-1881.
[13]杨金锋,傅周宇,谭铁牛等一种新型的基于内容的图像识别与过滤方法[J].通信学报,2004,25(7)：93-106.
[14]许强,江早,赵宏.基于图像内容过滤的智能防火墙系统研究与实现[J]计算机研究与发展,2000,37(4)458-464.
[15]胡冠宇.基于肤色之裸体影像侦测之研究[D]台南国立成功大学,2004.
[16]王宇石,李远宁,高文.运用局部形态SIFT描述法过滤网络淫秽图像方法的研究[J].刑事技术,2007(2)：9-11.
[17]HNC网络信息智能过滤器,http://www.hncit.com/product02.htm
[18]金惠反黄专家系统http://china.npicp.com/jishu/detail/17-109583.html
[19]蓝眼睛智能信息过滤系统软件http://www.soft6.com/business/5/50605.html
[20]绿坝·花季护航http://211.103.158.33/
[21]M.Storring, HJ Andersen, and E. Granum. Skin color detection under changing lighting conditions[C]. In H. Araujo and J.Dias, editors,7th International Symposium on Intelligent Robotic Systems, pages 187-195, Coimbra, Portugal, July 1999.
[22]Hunke M & Waibel. A Face locating and tracking for human-computer interaction[C]. Proc. Conference Record of the Twenty-Eighth Asilomar Conference on Signals, Systems and Computers, Pacific Grove, California, USA, 1994,2, pp.1277-1281.
[23]Yang J & Waibel A. A Real-time Face Tracker[C]. Proc.3 IEEE Workshop on Applications of Computer Vision, Florida, USA,1996, pp.142-147
[24]Alexandru F.Drimbarean,Image processing techniques to detect and filter objectionoable images based on skin tone and shape recognition,IEEE,2001.
[25]BOSSONA, CAWLEY G C, CHAN Y. Non-retrieval:blocking pornographic images[C]//Proc of International Conference on Image and Video Retrieval. Berlin: Springer,2002:50-60.
[26]刘益和,刘嘉勇,袁新峰,基于内容分析的特定图像过滤技术研究[J],计算机工程与应用,2006,24,56—58
[27]T.Ahonen, A.Hadid, M.Pietikainen, Face Recognition with Local Binary Patterns, Computer Vision Proceedings, ECCV 2004, Lecture Notes in Computer Science 3021, Springer,469-481.
[28]T.Ojala, M. Pietikainen, T.Maeopaa, Multiresolution Gray-Scale and Rotation Invariant Texture Classification with Local Binary Patterns, IEEE trans. Pattern Analysis and Machine Intelligence, vol.24, pp.971-987,2002
[29]T.Ojala, M.Pietikainen, D.Harwood, A Comparative Study of Texture Measures with Classification Based on Feature Distributions, Pattern Recognition, vol.29, pp.51-59,1996.
[30]C. Liu and H. Wechsler, "Gabor Feature Based Classification Using the Enhanced Fisher Linear Discriminant Model for Face Recognition," IEEE Trans.on Image Processing, Vol.11, No.4, pp.467-476, April.2002.
[31]M. Lades, J. C. Vorbruggen, J. Buhmann, J. Lange, C. v.d Malsburg, R. P. Wurtz, and W. Konen, "Distortion invariant object recognition in the dynamic link architecture," IEEE Trans.on Computer, Vol.42, No.3, pp.300-311,1993.
[32]K. Fukunnaga, "Introduction to Statistical Pattern Recognition," 2nd ed. New York, NY:Academic,1991
[33]X. Wang and X. Tang, "Hallucinating face by eigentransformation," IEEE Trans, on Systems, Man, and Cybernetics, Vol.35, No.3, Aug.2005, pp.425-434
[34]D.G.Lowe, Distinctive image features from scale-invariant keypoint. International Journal of Computer Vision,60(2):91-110,2004.
[35]罗俊.基于最小最大模块化分类器和人脸图像的性别分类研究,硕士论文,上海交通大学,2007.
[36]D.G.Lowe, Distinctive image features from scale-invariant keypoint. International Journal of Computer Vision,60(2):91-110,2004.
[37]Y Ke, R Sukthankar. PCA-SIFT:A more distinctive representation for local image descriptors. Computer Vision and Pattern Recognition,2004.
[38]Vapnik V N,The Nature of Statistical Learning Theory[M],New York:Springer-Verlag,1995
[39]John C. platt. Using Analytic QP and Sparseness to Speed Training of Sport Vector Machines.
[40]S.S.Keerthi etal. ImProvements to Platt's SMO Algorithm for SVM Classifler Design
[41]ChrisJ.C.Burges. Simplified Sport Decision Rules.
[42]ChrisJ.C.Burges. Beruhard Scholkopf. Improving the Accuracy and speed of Support Vector Machines. Neural Information Processing Systems, Vol.9, M.Mozer, M.Jordan,& T.Petsehe, eds. MIT Press, Cambridge, MA,1997.
[43]Corinna Cortes, V.VaPnik. Support Vector Network.Machine Leaning,20.273-297(1995).
[44]Nando de Freitas etal.Sequenttial Support Vector Machines Neural Networks for Signal Processing IX.31-40(1999).
[45]Olvi L.Mangasarian, David R.Musicant. Successive Over relaxation for Support Vector Machines.IEEE Trans.on Neural Networks.Vol.10, No.5, Sep 1999.
[46]P.S.Bradley O.L.Mangasarian. Feature Selection via Concave Minimization and Support Vector Machines.
[47]J.Platt.Sequential minimal optimization:A fast algorithm for training support vector machines.Advances in Kernel Methods-Support Vector learning. Cambridge, MA:MIT Press,1999, PP.185-208.
[48]S.S.Keerthi et al. A Fast iterative Nearest Point Algorithm for Support Vector Machine Classifier Design.
[49]边肇祺等,模式识别,清华大学出版社,1988.
[50]Vapnik V N. Statistical learning theory.NewYork:John Wiley and Sons,1998.
[51]T.Kohonen.The self-organizing map, Proc. IEEE, Vol.78, No.9, pp.1464-1480, Sep.1990.
[52]程剑锋.LVQ神经网络的研究及其在说话人识别中的应用,2004.
[53]董长虹.Matlab神经网络与应用[M].北京：国防工业出版社,2005.
[54]F.-L. Chung and T. Lee, "A Fuzzy Learning Model for Membership Function Estimation and Pattern Classification." Proceedings of the third IEEE Conference on Fuzzy Systems,1994, pp.426-431.
[55]F.-L. Chung and T. Lee, "Fuzzy Learning Vector Quantization." Proceedings of 1993 International Joint Conference on Neural Networks,1993, pp.2739-2743.
[56]N. B. Karayiannis, "Weighted Fuzzy Learning Vector Quantization and Weighted Fuzzy C-Means Algorithms." IEEE International Conference on Neural Networks, 1996, pp.1044-1049.
[57]Anil K J, Hsu R L, Mohamed A M, Face Detection in Color Image[A],ICIP'2001
[58]段丽娟,崔国勤,高文,张洪明.计算机辅助设计与图形学学报[J],2002,14(5),404-409
[59]B.K. Gunturk, A. U. Batur, Y. Altunbasak, et al, "Eigenface-domin super-resolution for face recognition," IEEE Trans, Image Processing,2003,12(5):597-605
[2]《第 27 次中国互联网络发展状况统计报告》 ,http://tech.163.com/special/cnnic27/
[3]整治网络传播淫秽色情信息经验交流会召开 刘云山讲话,http://news.xinhuanet.com/politics/2010-12/13/c_12875807.htm
[4]九部门联合依法整治网络环境取得明显成效,http://www.scio.gov.cn/ztk/hlwxx/01/03/3/201002/t545852.htm
[5]美萍电脑软件工作室.美萍反黄专家：中国,ZL01117726.8[P/OL].(2002). http://www.msoft.net/shield.htm.
[6]中国科学院下属科大讯飞公司.火眼金睛反黄软件：中国,ZL01117731.4[P/OL]. (2002). http://www.iflytek.com.
[7]北京展翅鸟科技公司. 网络爸爸软件：中国,ZL01117730.6[P/OL]. (2002). http://www.tueagles.com/baba.
[8]FLECKM M, FORSYTH D A, BREGLER C. Finding naked people[C]//Proc of the 4th European Conference on Computer Vision. London:Springer-Verlag,1996: 593-602.
[9]WANG J Z,LI Jia,WEDERHOLD G,et al. System for screening objectionable images [j]. Computer Communications Journa 1,1998,21(15):1355-1360.
[10]JONES M J, REHG J M. Statistical color models with application to skin detection [C]//Proc of Computer Vision and Pattern Recognition.1999:274-280.
[11]DUAN Li-juan, CUI Guo-qin, GAO Wen, et al. Adult image detection method based on skin color model and support vector machine [C]//Proc of the 5th Asian Conference on Computer Vision.2002:797-800.
[12]徐欣欣,袁华.一种基于自适应小波不变矩的色情图像过滤方法[J].清华大学学报：自然科学版,2005,45(1)：1878-1881.
[13]杨金锋,傅周宇,谭铁牛等一种新型的基于内容的图像识别与过滤方法[J].通信学报,2004,25(7)：93-106.
[14]许强,江早,赵宏.基于图像内容过滤的智能防火墙系统研究与实现[J]计算机研究与发展,2000,37(4)458-464.
[15]胡冠宇.基于肤色之裸体影像侦测之研究[D]台南国立成功大学,2004.
[16]王宇石,李远宁,高文.运用局部形态SIFT描述法过滤网络淫秽图像方法的研究[J].刑事技术,2007(2)：9-11.
[17]HNC网络信息智能过滤器,http://www.hncit.com/product02.htm
[18]金惠反黄专家系统http://china.npicp.com/jishu/detail/17-109583.html
[19]蓝眼睛智能信息过滤系统软件http://www.soft6.com/business/5/50605.html
[20]绿坝·花季护航http://211.103.158.33/
[21]M.Storring, HJ Andersen, and E. Granum. Skin color detection under changing lighting conditions[C]. In H. Araujo and J.Dias, editors,7th International Symposium on Intelligent Robotic Systems, pages 187-195, Coimbra, Portugal, July 1999.
[22]Hunke M & Waibel. A Face locating and tracking for human-computer interaction[C]. Proc. Conference Record of the Twenty-Eighth Asilomar Conference on Signals, Systems and Computers, Pacific Grove, California, USA, 1994,2, pp.1277-1281.
[23]Yang J & Waibel A. A Real-time Face Tracker[C]. Proc.3 IEEE Workshop on Applications of Computer Vision, Florida, USA,1996, pp.142-147
[24]Alexandru F.Drimbarean,Image processing techniques to detect and filter objectionoable images based on skin tone and shape recognition,IEEE,2001.
[25]BOSSONA, CAWLEY G C, CHAN Y. Non-retrieval:blocking pornographic images[C]//Proc of International Conference on Image and Video Retrieval. Berlin: Springer,2002:50-60.
[26]刘益和,刘嘉勇,袁新峰,基于内容分析的特定图像过滤技术研究[J],计算机工程与应用,2006,24,56—58
[27]T.Ahonen, A.Hadid, M.Pietikainen, Face Recognition with Local Binary Patterns, Computer Vision Proceedings, ECCV 2004, Lecture Notes in Computer Science 3021, Springer,469-481.
[28]T.Ojala, M. Pietikainen, T.Maeopaa, Multiresolution Gray-Scale and Rotation Invariant Texture Classification with Local Binary Patterns, IEEE trans. Pattern Analysis and Machine Intelligence, vol.24, pp.971-987,2002
[29]T.Ojala, M.Pietikainen, D.Harwood, A Comparative Study of Texture Measures with Classification Based on Feature Distributions, Pattern Recognition, vol.29, pp.51-59,1996.
[30]C. Liu and H. Wechsler, "Gabor Feature Based Classification Using the Enhanced Fisher Linear Discriminant Model for Face Recognition," IEEE Trans.on Image Processing, Vol.11, No.4, pp.467-476, April.2002.
[31]M. Lades, J. C. Vorbruggen, J. Buhmann, J. Lange, C. v.d Malsburg, R. P. Wurtz, and W. Konen, "Distortion invariant object recognition in the dynamic link architecture," IEEE Trans.on Computer, Vol.42, No.3, pp.300-311,1993.
[32]K. Fukunnaga, "Introduction to Statistical Pattern Recognition," 2nd ed. New York, NY:Academic,1991
[33]X. Wang and X. Tang, "Hallucinating face by eigentransformation," IEEE Trans, on Systems, Man, and Cybernetics, Vol.35, No.3, Aug.2005, pp.425-434
[34]D.G.Lowe, Distinctive image features from scale-invariant keypoint. International Journal of Computer Vision,60(2):91-110,2004.
[35]罗俊.基于最小最大模块化分类器和人脸图像的性别分类研究,硕士论文,上海交通大学,2007.
[36]D.G.Lowe, Distinctive image features from scale-invariant keypoint. International Journal of Computer Vision,60(2):91-110,2004.
[37]Y Ke, R Sukthankar. PCA-SIFT:A more distinctive representation for local image descriptors. Computer Vision and Pattern Recognition,2004.
[38]Vapnik V N,The Nature of Statistical Learning Theory[M],New York:Springer-Verlag,1995
[39]John C. platt. Using Analytic QP and Sparseness to Speed Training of Sport Vector Machines.
[40]S.S.Keerthi etal. ImProvements to Platt's SMO Algorithm for SVM Classifler Design
[41]ChrisJ.C.Burges. Simplified Sport Decision Rules.
[42]ChrisJ.C.Burges. Beruhard Scholkopf. Improving the Accuracy and speed of Support Vector Machines. Neural Information Processing Systems, Vol.9, M.Mozer, M.Jordan,& T.Petsehe, eds. MIT Press, Cambridge, MA,1997.
[43]Corinna Cortes, V.VaPnik. Support Vector Network.Machine Leaning,20.273-297(1995).
[44]Nando de Freitas etal.Sequenttial Support Vector Machines Neural Networks for Signal Processing IX.31-40(1999).
[45]Olvi L.Mangasarian, David R.Musicant. Successive Over relaxation for Support Vector Machines.IEEE Trans.on Neural Networks.Vol.10, No.5, Sep 1999.
[46]P.S.Bradley O.L.Mangasarian. Feature Selection via Concave Minimization and Support Vector Machines.
[47]J.Platt.Sequential minimal optimization:A fast algorithm for training support vector machines.Advances in Kernel Methods-Support Vector learning. Cambridge, MA:MIT Press,1999, PP.185-208.
[48]S.S.Keerthi et al. A Fast iterative Nearest Point Algorithm for Support Vector Machine Classifier Design.
[49]边肇祺等,模式识别,清华大学出版社,1988.
[50]Vapnik V N. Statistical learning theory.NewYork:John Wiley and Sons,1998.
[51]T.Kohonen.The self-organizing map, Proc. IEEE, Vol.78, No.9, pp.1464-1480, Sep.1990.
[52]程剑锋.LVQ神经网络的研究及其在说话人识别中的应用,2004.
[53]董长虹.Matlab神经网络与应用[M].北京：国防工业出版社,2005.
[54]F.-L. Chung and T. Lee, "A Fuzzy Learning Model for Membership Function Estimation and Pattern Classification." Proceedings of the third IEEE Conference on Fuzzy Systems,1994, pp.426-431.
[55]F.-L. Chung and T. Lee, "Fuzzy Learning Vector Quantization." Proceedings of 1993 International Joint Conference on Neural Networks,1993, pp.2739-2743.
[56]N. B. Karayiannis, "Weighted Fuzzy Learning Vector Quantization and Weighted Fuzzy C-Means Algorithms." IEEE International Conference on Neural Networks, 1996, pp.1044-1049.
[57]Anil K J, Hsu R L, Mohamed A M, Face Detection in Color Image[A],ICIP'2001
[58]段丽娟,崔国勤,高文,张洪明.计算机辅助设计与图形学学报[J],2002,14(5),404-409
[59]B.K. Gunturk, A. U. Batur, Y. Altunbasak, et al, "Eigenface-domin super-resolution for face recognition," IEEE Trans, Image Processing,2003,12(5):597-605