基于人脸识别的防盗视频监控系统研究
|
摘要
基于人脸识别的数字视频监控系统由于具有隐蔽性、非接触性等优势,近年来得到了广泛关注。本文运用人脸识别技术设计并实现了一套基于固定背景的防盗视频监控系统。系统的中心内容分为三部分,分述如下:
1.系统平台的设计:借助微软公司的多媒体处理软件开发包DirectShow搭建了整个系统的框架,同时应用其功能完成视频采集及格式转换工作。
2.人脸检测和定位:提出了一种复合的光照补偿策略,其中对对数变换算法进行了合理的改进;结合视频监控的特点,提出了一种基于固定背景的快速的人脸检测新方法,实验结果表明,该方法检测速度快且具有较强的鲁棒性,能够很好地应用于视频监控系统中。
3.人脸的快速识别:首先介绍了用于人脸识别的主成分分析方法和线性判别分析方法,然后针对线性判别分析的缺陷提出了一种改进的方案,并应用主成分分析和改进后的线性判别分析相结合来提取人脸的特征,最后分别用最小距离分类器和三层BP神经网络分类器实现了人脸识别。实验表明,识别的性能基本满足了系统要求。
本文实验图像均由摄像头采集而来,包括了单人脸、多人脸及人脸不同表情姿态等情况,实验环境为Pentium IV,内存256M的PC机。实验结果显示,对80幅大小为320×240像素,包含92个人脸的小型图像库,检测率达到93.5%。平均的检测时间为0.5s。在人脸识别阶段,当人脸的类别不多的情况下,识别率达到了95%以上。平均的识别时间为0.3s。两个功能模块的准确性和处理时耗基本满足了系统的要求。
The video surveillance system based on face recognition has the advantage of concealment and non-contact, so it has received extensive attention in recent years.In this paper,a security video surveillance system which is used face recognition technology has been designed, and it will be used in fixed-background environment.System includes three main parts. Each is described as follows:
1.System platform,s design: In this paper, we adopt Microsoft's multimedia software development kits --DirectShow structures in the framework of the whole system,at the meanwhile, the kits can be also used to accomplish video capture and format conversion.
2 . Face detection and locating: A composite light compensation strategy is proposed,and a new logarithmic transform algorithm is proposed. According to the character of video surveillance system, a rapid face detection approach based on fixed-background is proposed. The testing results show that the algorithm is fast and the approach is high robust. This approach can be used in video surveillance system well.
3.Quick face recognition: The principal component analysis and linear discriminant analysis are introduced. And we have put an improved method against linear discriminant analysis of defect, then use the method of principal component analysis and improved linear discriminant analysis combining to finish feature extraction of face image. Finally we have achieved face recognition using a minimum distance classifier and a three layer BP neural network classifier. Experiments show that properties meet the basic system requirements.
The experimental images are all collected by Camera, including single face, muti-faces and different face expression. Experimental environment is the Pentium IV, the PC memory 256M PC. Experimental results demonstrates that as for a small image library ,including 92 individuals face, each size 320×240 pixels, face detection correctness rate has reached to 93.5%.The average detection time is 0.5s. In face recognition stage, when the category is not too much, face recognition correctness rate has reached to more than 95%.And average recognition time is approximately 0.3s in laboratory. The accuracy and time-consuming performance of these two functional modules can meet general requirements of the important things anti-theft surveillance system.
1.系统平台的设计:借助微软公司的多媒体处理软件开发包DirectShow搭建了整个系统的框架,同时应用其功能完成视频采集及格式转换工作。
2.人脸检测和定位:提出了一种复合的光照补偿策略,其中对对数变换算法进行了合理的改进;结合视频监控的特点,提出了一种基于固定背景的快速的人脸检测新方法,实验结果表明,该方法检测速度快且具有较强的鲁棒性,能够很好地应用于视频监控系统中。
3.人脸的快速识别:首先介绍了用于人脸识别的主成分分析方法和线性判别分析方法,然后针对线性判别分析的缺陷提出了一种改进的方案,并应用主成分分析和改进后的线性判别分析相结合来提取人脸的特征,最后分别用最小距离分类器和三层BP神经网络分类器实现了人脸识别。实验表明,识别的性能基本满足了系统要求。
本文实验图像均由摄像头采集而来,包括了单人脸、多人脸及人脸不同表情姿态等情况,实验环境为Pentium IV,内存256M的PC机。实验结果显示,对80幅大小为320×240像素,包含92个人脸的小型图像库,检测率达到93.5%。平均的检测时间为0.5s。在人脸识别阶段,当人脸的类别不多的情况下,识别率达到了95%以上。平均的识别时间为0.3s。两个功能模块的准确性和处理时耗基本满足了系统的要求。
The video surveillance system based on face recognition has the advantage of concealment and non-contact, so it has received extensive attention in recent years.In this paper,a security video surveillance system which is used face recognition technology has been designed, and it will be used in fixed-background environment.System includes three main parts. Each is described as follows:
1.System platform,s design: In this paper, we adopt Microsoft's multimedia software development kits --DirectShow structures in the framework of the whole system,at the meanwhile, the kits can be also used to accomplish video capture and format conversion.
2 . Face detection and locating: A composite light compensation strategy is proposed,and a new logarithmic transform algorithm is proposed. According to the character of video surveillance system, a rapid face detection approach based on fixed-background is proposed. The testing results show that the algorithm is fast and the approach is high robust. This approach can be used in video surveillance system well.
3.Quick face recognition: The principal component analysis and linear discriminant analysis are introduced. And we have put an improved method against linear discriminant analysis of defect, then use the method of principal component analysis and improved linear discriminant analysis combining to finish feature extraction of face image. Finally we have achieved face recognition using a minimum distance classifier and a three layer BP neural network classifier. Experiments show that properties meet the basic system requirements.
The experimental images are all collected by Camera, including single face, muti-faces and different face expression. Experimental environment is the Pentium IV, the PC memory 256M PC. Experimental results demonstrates that as for a small image library ,including 92 individuals face, each size 320×240 pixels, face detection correctness rate has reached to 93.5%.The average detection time is 0.5s. In face recognition stage, when the category is not too much, face recognition correctness rate has reached to more than 95%.And average recognition time is approximately 0.3s in laboratory. The accuracy and time-consuming performance of these two functional modules can meet general requirements of the important things anti-theft surveillance system.
引文
[1]毕文斌.基于Web视频服务器的监控系统设计[J].现代有线传输,2002,02:20-31.
[2]余萍.基于Internet的实时视频传输技术浅析[J].电力系统通信,2002,04:23-28.
[3]杨宏林,吴陈.指纹识别方法的综述[J].华东船舶工业学院学报(自然版),2003,17(3):37-42.
[4]邬向前,王宽全,张大鹏.一种用于掌纹识别的线特征表示和匹配方法[J].软件学报.2004,15(6):869-880.
[5]浦昭邦,扬帆等.虹膜识别技术的发展与应用[J].光学精密工2004,12(3):316-322.
[6]杨子华,吴敏,凌江荣.基于纹理分析的笔迹识别系统[J].2004,14(2):67-69.
[7] M.H. Yang,D. Kriegman and N.Ahuja.Detecting Faces in Images:ASurvey[J].IEEE Trans.Pattern Analysis and Machine Intelligence,2002,24(1):34-58.
[8]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,2004,09:1-4.
[9] G. Yang and T. S. Huang . Human Face Detection in Complex Background environment[J].Pattern Recognition,1994,27(1):53-63.
[10] C.Kotropoulos and I. Pitas.Rule-Based Face Detection in Frontal Views.Proc.Int’l Conf.Acoustics,Speech and Signal Processing,1997,4:2537-2540.
[11] Craw H Ellis.J Lishman. Automatic Extraction of Face Features [J].Pattern Recognition Letters,1987,5:183一187.
[12] J Miao,B Yin,K Wang.A Hierarchical Multiscale and Multiangle System for Human Face Detection in a Complex Background Using Gravity-Center Template[J] . Pattern Recognition.1999,32(7):1237-1248.
[13]梁路宏,艾海舟,何克忠.基于多模板匹配的单人脸检测[J].中国图像图形学报,1999,823-830.
[14]周激流,张晔等.基于先验模板的人脸面部特征提取的研究[[J].计算机辅助设计与图形学学报,2000,12(5):337-339.
[15] Liu Chengjun.A Bayesian Discriminating Features Method for face Detection[J].IEEE Trans. Pattern Analysis and Machine Intelligence,2003,25(6):725-740.
[16] Turk M,Pentland A.Eigenfaces for Recognition[J].Journal of Cognitive Neuroscience,1991,3:71-86.
[17] LüX G,Wang Y H,Jain A K.Combining Classifiers for Face Recognition[C].In:Proc ofthe IEEE International Conference on Multimedia and Expo.Baltimore,USA,2003,3: 13-16.
[18]陆其明.DirectShow开发指南[M].清华大学出版社,2003.
[19]卢春雨,张长水,闻芳等.基于区域特征的快速人脸检测法[J].清华大学学报(自然科学版),1999,39(1):101-105.
[20]许华彬.复杂背景图像中的多人脸检测与定位.硕士学位论文[D],南京,南京航空航天大学,2004:50-59.
[21] K.K.Sung and T.Poggio . Example-based learning for view-based human face detection[J].IEEE Trans.on PAMI,1998,20:39-51.
[22]刘红,李锦涛,苗军.多方法融合来解决人脸检测中的光照补偿[J].系统仿真学报,2001:486-490.
[23] M.J.Donahue and S.I.Rokhlin,On the Use of Level Curves in Image Analysis[J].Image Understanding,March,1993:185-203.
[24]黄凤贤,彩色序列图像中的人脸检测与跟踪.硕士学位论文[D].西安,西安电子科技大学,2004.
[25] LiangTao , H K Kwan . Automatic localization of human eyes in complex background[C].IEEE International Symposiumon Circuitsand Systems,2002:669-672.
[26]陈健,周利莉,史红刚.一种基于多分量信息融合的人脸检测方法[J].计算机工程与应用,2005,36:57-59,65.
[27]朱夏君,王勋,李必威.人脸识别中的眼睛定位[J].电路与系统学报,2007,12(2):98-100
[28] P.N.Belhumeur,J.P.Hespanha and D.J.Kriegman.Eigenfaces vs. fisherfaces: recognition using class specific linear projection [J].IEEE Trans.on PAMI,1997,19(7):711-720.
[29]贾永红.数字图像处理[M].武汉大学出版社,2003:63-72,132-141.
[30]周志明,王以治,黄文芝等.基于小波和支持向量基的人脸识别技术[J].计算机工程与应用,2004,12:52-54.
[31] Richard O. Duda,Peter E. Hart,David G. Stork.模式分类(李宏东,姚天翔等译)[M].北京:机械工业出版社,中信出版社.2003:96-101.
[32] D.Sewets,J.Weng.Using discriminant eigenfeatures for image retrieval [J].IEEE Trans.on PAMI 1996,18(8):100-110.
[33]王洪元,史国栋.人工神经网络技术及其应用[M].北京:中国石化出版社,2002.
[34]张立明.人工神经网络的模型及其应用[M].复旦大学出版社,1993:43-46.
[35]韩力群.人工神经网络理论、设计及应用[M].化学工业出版社,2002:55-56.
[2]余萍.基于Internet的实时视频传输技术浅析[J].电力系统通信,2002,04:23-28.
[3]杨宏林,吴陈.指纹识别方法的综述[J].华东船舶工业学院学报(自然版),2003,17(3):37-42.
[4]邬向前,王宽全,张大鹏.一种用于掌纹识别的线特征表示和匹配方法[J].软件学报.2004,15(6):869-880.
[5]浦昭邦,扬帆等.虹膜识别技术的发展与应用[J].光学精密工2004,12(3):316-322.
[6]杨子华,吴敏,凌江荣.基于纹理分析的笔迹识别系统[J].2004,14(2):67-69.
[7] M.H. Yang,D. Kriegman and N.Ahuja.Detecting Faces in Images:ASurvey[J].IEEE Trans.Pattern Analysis and Machine Intelligence,2002,24(1):34-58.
[8]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,2004,09:1-4.
[9] G. Yang and T. S. Huang . Human Face Detection in Complex Background environment[J].Pattern Recognition,1994,27(1):53-63.
[10] C.Kotropoulos and I. Pitas.Rule-Based Face Detection in Frontal Views.Proc.Int’l Conf.Acoustics,Speech and Signal Processing,1997,4:2537-2540.
[11] Craw H Ellis.J Lishman. Automatic Extraction of Face Features [J].Pattern Recognition Letters,1987,5:183一187.
[12] J Miao,B Yin,K Wang.A Hierarchical Multiscale and Multiangle System for Human Face Detection in a Complex Background Using Gravity-Center Template[J] . Pattern Recognition.1999,32(7):1237-1248.
[13]梁路宏,艾海舟,何克忠.基于多模板匹配的单人脸检测[J].中国图像图形学报,1999,823-830.
[14]周激流,张晔等.基于先验模板的人脸面部特征提取的研究[[J].计算机辅助设计与图形学学报,2000,12(5):337-339.
[15] Liu Chengjun.A Bayesian Discriminating Features Method for face Detection[J].IEEE Trans. Pattern Analysis and Machine Intelligence,2003,25(6):725-740.
[16] Turk M,Pentland A.Eigenfaces for Recognition[J].Journal of Cognitive Neuroscience,1991,3:71-86.
[17] LüX G,Wang Y H,Jain A K.Combining Classifiers for Face Recognition[C].In:Proc ofthe IEEE International Conference on Multimedia and Expo.Baltimore,USA,2003,3: 13-16.
[18]陆其明.DirectShow开发指南[M].清华大学出版社,2003.
[19]卢春雨,张长水,闻芳等.基于区域特征的快速人脸检测法[J].清华大学学报(自然科学版),1999,39(1):101-105.
[20]许华彬.复杂背景图像中的多人脸检测与定位.硕士学位论文[D],南京,南京航空航天大学,2004:50-59.
[21] K.K.Sung and T.Poggio . Example-based learning for view-based human face detection[J].IEEE Trans.on PAMI,1998,20:39-51.
[22]刘红,李锦涛,苗军.多方法融合来解决人脸检测中的光照补偿[J].系统仿真学报,2001:486-490.
[23] M.J.Donahue and S.I.Rokhlin,On the Use of Level Curves in Image Analysis[J].Image Understanding,March,1993:185-203.
[24]黄凤贤,彩色序列图像中的人脸检测与跟踪.硕士学位论文[D].西安,西安电子科技大学,2004.
[25] LiangTao , H K Kwan . Automatic localization of human eyes in complex background[C].IEEE International Symposiumon Circuitsand Systems,2002:669-672.
[26]陈健,周利莉,史红刚.一种基于多分量信息融合的人脸检测方法[J].计算机工程与应用,2005,36:57-59,65.
[27]朱夏君,王勋,李必威.人脸识别中的眼睛定位[J].电路与系统学报,2007,12(2):98-100
[28] P.N.Belhumeur,J.P.Hespanha and D.J.Kriegman.Eigenfaces vs. fisherfaces: recognition using class specific linear projection [J].IEEE Trans.on PAMI,1997,19(7):711-720.
[29]贾永红.数字图像处理[M].武汉大学出版社,2003:63-72,132-141.
[30]周志明,王以治,黄文芝等.基于小波和支持向量基的人脸识别技术[J].计算机工程与应用,2004,12:52-54.
[31] Richard O. Duda,Peter E. Hart,David G. Stork.模式分类(李宏东,姚天翔等译)[M].北京:机械工业出版社,中信出版社.2003:96-101.
[32] D.Sewets,J.Weng.Using discriminant eigenfeatures for image retrieval [J].IEEE Trans.on PAMI 1996,18(8):100-110.
[33]王洪元,史国栋.人工神经网络技术及其应用[M].北京:中国石化出版社,2002.
[34]张立明.人工神经网络的模型及其应用[M].复旦大学出版社,1993:43-46.
[35]韩力群.人工神经网络理论、设计及应用[M].化学工业出版社,2002:55-56.