车型与牌照识别系统的实现
|
摘要
在大力加快高速公路建设的同时，实现不停车自动收费势在必行。基于非接触IC卡的不停车收费系统(No Stop Electronic Toll Collection System，简称NSETCS或ETC)作为自动收费方式的代表，可以有效提高高速公路的通行能力和服务水平。
车型识别及牌照识别系统是高速公路不停车收费系统的重要组成部分，是近几年发展起来的基于图像和字符识别技术的智能化交通管理系统，是目前国内外模式识别应用研究领域的一个热点。
本文对系统中图像预处理、特征提取和识别方法等环节涉及的新算法、新技术以及系统整体设计作了一个比较全面的论述，同时针对目前的研究现状，对一些关键的技术问题进行了深入探讨。
本文的研究，为整个不停车收费系统快速步入实用化阶段奠定了基础，也推动了基于非接触IC卡的收费方式在国内的逐步推行，不停车收费系统具有巨大的经济效益和社会效益。
本文主要完成了以下几方面的工作：
一、提出了利用差图像法识别车型，主要是利用两幅图像的差别来提取汽车外形轮廓，进而求取车长、车宽、车高，并以其作为系统软件划分车型的标准。
二、利用模糊模板匹配法对牌照区域进行定位，该方法依据牌照区域边缘丰富的特点进行区域搜索，定位准确。
三、探讨了字符识别的预处理方法，采用了轮廓投影的方法来切分字符。
四、采用了线性插值法来克服由于对字符进行归一化时造成的马赛克现象。并且针对汉字，数字和字母的笔划疏密不同，采用了两种不同的字符尺寸。
五、讨论了两种统计特征的提取方法，一种是网格特征，另外一种是交叉点特征，实验中通过最小距离法进行字符的分类识别。
六、对将近100个不同字符进行识别时，考虑到车牌字符的汉字，字母及数字的排列顺序，采用了优先级的办法进行匹配识别。
七、提出了牌照识别结果和车型识别结果的融合模型，提高了车辆识别的准确率，并完成了相关的融合算法设计。
The project is supported by the Innovation Foundation of Excellent Intellectual in HeNan Province.
It will be imperative under the situation that we build more and more speedways to realize the vehicle identification and intelligent management. The ETC ( No Stop Electronic Toll Collection System )based on noncontact IC card as the representative of the automatic toll collection improves the traffic capacity and service level efficiently.
The vehicle style and the license plate recognition system is the important part of the system of the automatic charging in speedway toll stations. Based on the technology of recognizing images and characters, the vehicle style and license plate recognition system is one of the focused researches on the application of pattern recognition home and abroad.
Some new algorithms and methods are introduced comprehensively in such stages as images preprocessing, feature extraction , character recognition and vehicle style recognition. Meanwhile, some key technologies are discussed in detail according to the current research status quo.
It lays foundation for the whole No Stop Electronic Toll Collection System to march into the practical period, it also propels the popularity of this special toll collection in our country. The ETC has great economical and social profit.
The thesis mainly fulfills such assignments as follows:
Firstly, the method of two images' subtraction is used to recognize the vehicle style, it can so extract the car's silhouette that the length, the width and the height of
the vehicle thus can be calculated, which specifies the vehicle style.
Secondly, based on the characteristic that vehicle license plates have plentiful edge information, the method of fuzzy template matching is carried out to locate the vehicle license plate, which has high location ratio.
Thirdly, the method of characters preprocessing is discussed in detail. The method of silhouette projection is used to segment characters.
Fourthly, the method of linear interpolation is applied in order that the mosaic phenomenon brought out by size normalization should be overcome, moreover, two
types of character size are used because of Chinese characters'and letters'different stroke density.
Fifthly, it introduces two kinds of statistical feature extraction methods, one is the gridding feature, the other is the crossing feature, the minimum-distance-matching algorithm is used in the recognition system of characters.
Sixthly, when recognizing the approximate one hundred different characters, a high premium on their priority is put considering the regular character arrangement in the vehicle license plate.
Finally, the fusion model is put forward, which combines the recognition results of the vehicle style with those of the license plate systematically and enhances the precision of vehicle recognition, it also accomplishes the pertinent fusion algorithm design.
车型识别及牌照识别系统是高速公路不停车收费系统的重要组成部分，是近几年发展起来的基于图像和字符识别技术的智能化交通管理系统，是目前国内外模式识别应用研究领域的一个热点。
本文对系统中图像预处理、特征提取和识别方法等环节涉及的新算法、新技术以及系统整体设计作了一个比较全面的论述，同时针对目前的研究现状，对一些关键的技术问题进行了深入探讨。
本文的研究，为整个不停车收费系统快速步入实用化阶段奠定了基础，也推动了基于非接触IC卡的收费方式在国内的逐步推行，不停车收费系统具有巨大的经济效益和社会效益。
本文主要完成了以下几方面的工作：
一、提出了利用差图像法识别车型，主要是利用两幅图像的差别来提取汽车外形轮廓，进而求取车长、车宽、车高，并以其作为系统软件划分车型的标准。
二、利用模糊模板匹配法对牌照区域进行定位，该方法依据牌照区域边缘丰富的特点进行区域搜索，定位准确。
三、探讨了字符识别的预处理方法，采用了轮廓投影的方法来切分字符。
四、采用了线性插值法来克服由于对字符进行归一化时造成的马赛克现象。并且针对汉字，数字和字母的笔划疏密不同，采用了两种不同的字符尺寸。
五、讨论了两种统计特征的提取方法，一种是网格特征，另外一种是交叉点特征，实验中通过最小距离法进行字符的分类识别。
六、对将近100个不同字符进行识别时，考虑到车牌字符的汉字，字母及数字的排列顺序，采用了优先级的办法进行匹配识别。
七、提出了牌照识别结果和车型识别结果的融合模型，提高了车辆识别的准确率，并完成了相关的融合算法设计。
The project is supported by the Innovation Foundation of Excellent Intellectual in HeNan Province.
It will be imperative under the situation that we build more and more speedways to realize the vehicle identification and intelligent management. The ETC ( No Stop Electronic Toll Collection System )based on noncontact IC card as the representative of the automatic toll collection improves the traffic capacity and service level efficiently.
The vehicle style and the license plate recognition system is the important part of the system of the automatic charging in speedway toll stations. Based on the technology of recognizing images and characters, the vehicle style and license plate recognition system is one of the focused researches on the application of pattern recognition home and abroad.
Some new algorithms and methods are introduced comprehensively in such stages as images preprocessing, feature extraction , character recognition and vehicle style recognition. Meanwhile, some key technologies are discussed in detail according to the current research status quo.
It lays foundation for the whole No Stop Electronic Toll Collection System to march into the practical period, it also propels the popularity of this special toll collection in our country. The ETC has great economical and social profit.
The thesis mainly fulfills such assignments as follows:
Firstly, the method of two images' subtraction is used to recognize the vehicle style, it can so extract the car's silhouette that the length, the width and the height of
the vehicle thus can be calculated, which specifies the vehicle style.
Secondly, based on the characteristic that vehicle license plates have plentiful edge information, the method of fuzzy template matching is carried out to locate the vehicle license plate, which has high location ratio.
Thirdly, the method of characters preprocessing is discussed in detail. The method of silhouette projection is used to segment characters.
Fourthly, the method of linear interpolation is applied in order that the mosaic phenomenon brought out by size normalization should be overcome, moreover, two
types of character size are used because of Chinese characters'and letters'different stroke density.
Fifthly, it introduces two kinds of statistical feature extraction methods, one is the gridding feature, the other is the crossing feature, the minimum-distance-matching algorithm is used in the recognition system of characters.
Sixthly, when recognizing the approximate one hundred different characters, a high premium on their priority is put considering the regular character arrangement in the vehicle license plate.
Finally, the fusion model is put forward, which combines the recognition results of the vehicle style with those of the license plate systematically and enhances the precision of vehicle recognition, it also accomplishes the pertinent fusion algorithm design.
引文
[1] 李峰．智能交通系统在国外的发展趋势．国外公路，1999，19(1)：1
[2] 蔡华．ETC系统DSRC技术标准化动态．公路交通科技，1998，15(3)：49～51
[3] D hensher.Electronic toll collection (ETC). Transportation Research, 1991, 25(1): 9～16
[4] Davidson Levinson, Elva Chang. A model for optimizing electronic toll collection systems. Transportation Research Part A: Policy and Practice, 2003, 37(4): 293～301
[5] Zarrillo M.L, Radwan A.E, Al-Deek H.M. Modeling Traffic Operations at Electronic Toll Collection and Traffic Management Systems. Computers & Industrial Engineering, 1997, 33(3): 857～860
[6] 张海峰，段颖妮．道桥收费站车型识别技术．现代电子技术，2001年，第9期：42～43
[7] 王振峰．图像处理技术在汽车类型自动识别中的应用研究：[硕士学位论文] ，2000年．
[8] 白洁，王洪庆，陶祖钰．GMS卫星红外云图强对流云团的识别与追踪．热带气象学报，1997，13(2)：158～167
[9] 陶青萍，陶白云．基于模糊神经网络的汽车类型自动识别分类系统．计算机工程与应用，1998年，第11期：78～81
[10] 任彬，汪炳权，黄勇．医学B超图像的模糊增强处理方法．中国图象图形学报，1997，1(2，3)：133～136
[11] 詹劲峰，皮德富．多尺度下低对比度红外热图像增强技术．红外与激光工程，1997，26(5)：28～32
[12] 李宏贵，李兴国等．非线性滤波器在红外图像增强中的应用．数据采集与处理，1999，14(3)：302～306
[13] 张引，潘云鹤．面向车辆牌照字符识别的预处理算法．计算机应用研究，1999年，第7期：85～87
[14] 廖金周，宣国荣．车辆牌照的自动分割．微型电脑应用，1999年，第7期：32～34
[15] Haralick R.M, Shapiro L.G. Survey:image segmentation techniques.Computer Vision, Graphics&Image Processing, 1985, 29(2): 100～132
[16] 刘效静，成瑜．汽车牌照自动识别技术研究．南京航天航空大学学报，1998，30(5)：573～576
[17] Lee J Nelson. Recent Advances in License Plate Recognition.Advanced Imaging ,2002,17(1):18～21,53
[18] N.Suqanuma, I.S.Kweon, Y.Bao, N.Fujiwara.Dynamic detection of a license plate using neural network.JSME International Journal,Series C:Mechanical Systems,Machine Elements and Manufacturing, 2001, 44(2): 431～436
[19] Cui Yuntao, Huang Qian. Automatic License Extraction from Moving Vehicles. Siemens Corporate Research, Proceedings of the 1997 International Conference on Image Processing. Part 3 (of 3) Santa Barbara, USA, Oct 26-29 1997, IEEE Comp Soc, Los Alamitos, USA, 1997,10 (3): 126～129
[20] Jun-Wei Hsieh,Shih-Hao Yu, Yung-Sheng Chen. Morphology-based License Plate Detection in Image of Differently Illuminated and Oriented Cars.Journal of Electronic Imaging, 2002, 11 (4): 507～516
[21] Park, Sung -Wook, Ku Yong -Hea. A License plate region extraction method using edge properties of wavelet subband. Proceeding of the International Conference on Imaging Science,Systems and Technology, CISST'03. Las Vegas,NV, United States,2003. CSREA Press, 2003. Volume 1: 8～12
[22] 求是科技，张宏林．Visual C＋＋数字图像模式识别技术及工程实践．北京：人民邮电出版社，2003．422～442
[23] Cao Guangzhi,Chen Jiaqian,Jang Jingping. An Adaptive Approach to Vehicle License Plate Location. The 29~(th) Annual Conference of the IEEE Industrial Electronics Society.Roanoke,VA, United States.2003. Institute of Electrical and Electronics Engineers Computer Society ,2003.Volume 2: 1786～1791
[24] 叶晨洲，廖金周．一种基于纹理的牌照图像二值化方法．微型电脑应用，1999年，第6期：28～30
[25] 周宏文，李见为，许盛．基于统计模式识别的彩色图像分割方法．光电工程，1999，26(5)：53～57
[26] 丁震，胡钟山等．一种基于模糊聚类的快速二值化方法．计算机学报，1998年，第21卷，增刊：331～334
[27] Richard G Casey, Eric Lecolinet. Survey of methods and strateies in character segmentation. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1996, 18(7): 690～706
[28] Reed T.R, du Buf J M.H. A review of recent texture segmentation and feature extraction techniques. CVGIP: Image Understanding,1993, Issue 57:359～372
[29] Seong-Whan Lee, Dong-June Lee, Hee-Seon Park. New methodlogy for gray-scale character segmentation and recognition. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1996, 18(10): 1045～1050
[30] 叶晨洲，廖金周，梅帆．车辆牌照字符识别系统．计算机系统应用，1999年，第5期：10～13
[31] 曹迪铭，宣国荣．牌照字符分割中的区域分裂与合并．计算机工程，2000，26(10)：18～21
[32] 王建平，盛军，朱程辉．基于小波分析的视频图像字符特征提取方法研究．微电子学与计算机，2002年，第5期：51～54
[33] 郭军，马跃，盛立东，钟义信．发展中的文字识别理论与技术．电子学报，1995，23(10)：184～187
[34] 王年，任彬，黄勇， 汪炳权．人工神经网络在公路车辆管理中的应用．电子技术应用， 1998年，第9期：11～12
[35] 董林，陈锡先，唐远炎，吴善培．一种新的手写体数字识别方法．北京邮电大学学报，1997，20(1)：79～83
[36] YeChenzhou, JangJie, XuanGuorong. Number plate character recognition. Shanghai Jiaotong Daxue Xuebao, 2000, 34(5): 672～675
[37] 宋加涛，刘济林．车辆牌照上英文和数字字符的结构特征分析及提取．中国图象图形学报，2002，7(9)：945～949
[38] T.Naito, T.Tsukada. Robust recognition methods for inclined license plates under various illumination conditions outdoors. 1999 IEEE/IEEJ/JSAI International Conference on Intelligent Transportation Systems. Tokyo, Jpn .Oct 5-Oct 8 1999, IEEE, Piscataway, NJ, USA, 1999. 697～702
[39] J.A.G Nijhuis, M.H ter Brugge, K.A Helmholt.Car license plate recognition with neural networks and fuzzy logic. Proceedings of the 1995 IEEE International Conference on Neural Networks. Part 5 (of 6). Perth, Aust. Nov 27-Dec 1 1995. IEEE, Piscataway, NJ, USA,1995. 2232～2236
[40] 刘俊，付敬奇，董新平．数据融合在目标识别中的应用．传感器技术，2001， 20(6)：9
[41] 钟子发，罗明．多传感器数据融合中基于目标识别的DS算法研究．电讯技术，2000年，第4期：57～58
[42] 李圣怡，吴学忠等．多传感器融合理论及在智能制造系统中的应用．长沙：国防科技大学出版社，1998．159～165
[43] 王联明，刘文刚．高速公路不停车收费系统设计．山西电子技术,2000年，第5期：25～26
[44] 孙涛，张宏建．目标识别中的信息融合技术．自动化仪表，2001，22(2)：1
[45] S.S Iyengar, B.Jones. Information fusion techniques for pattern analysis in large sensor data networks:An overview. Journal of The Franklin Institute, 2001,338 (5): 571～582
[46] 严学强，刘济林，郭小军，顾伟康．基于空间频率域的纹理分割算法．浙江大学学报，1998，32(11)：726～731
[47] 姜珊，孙玉方．手写汉字的集群识别．中文信息学报，1998，12(4)：1～8
[2] 蔡华．ETC系统DSRC技术标准化动态．公路交通科技，1998，15(3)：49～51
[3] D hensher.Electronic toll collection (ETC). Transportation Research, 1991, 25(1): 9～16
[4] Davidson Levinson, Elva Chang. A model for optimizing electronic toll collection systems. Transportation Research Part A: Policy and Practice, 2003, 37(4): 293～301
[5] Zarrillo M.L, Radwan A.E, Al-Deek H.M. Modeling Traffic Operations at Electronic Toll Collection and Traffic Management Systems. Computers & Industrial Engineering, 1997, 33(3): 857～860
[6] 张海峰，段颖妮．道桥收费站车型识别技术．现代电子技术，2001年，第9期：42～43
[7] 王振峰．图像处理技术在汽车类型自动识别中的应用研究：[硕士学位论文] ，2000年．
[8] 白洁，王洪庆，陶祖钰．GMS卫星红外云图强对流云团的识别与追踪．热带气象学报，1997，13(2)：158～167
[9] 陶青萍，陶白云．基于模糊神经网络的汽车类型自动识别分类系统．计算机工程与应用，1998年，第11期：78～81
[10] 任彬，汪炳权，黄勇．医学B超图像的模糊增强处理方法．中国图象图形学报，1997，1(2，3)：133～136
[11] 詹劲峰，皮德富．多尺度下低对比度红外热图像增强技术．红外与激光工程，1997，26(5)：28～32
[12] 李宏贵，李兴国等．非线性滤波器在红外图像增强中的应用．数据采集与处理，1999，14(3)：302～306
[13] 张引，潘云鹤．面向车辆牌照字符识别的预处理算法．计算机应用研究，1999年，第7期：85～87
[14] 廖金周，宣国荣．车辆牌照的自动分割．微型电脑应用，1999年，第7期：32～34
[15] Haralick R.M, Shapiro L.G. Survey:image segmentation techniques.Computer Vision, Graphics&Image Processing, 1985, 29(2): 100～132
[16] 刘效静，成瑜．汽车牌照自动识别技术研究．南京航天航空大学学报，1998，30(5)：573～576
[17] Lee J Nelson. Recent Advances in License Plate Recognition.Advanced Imaging ,2002,17(1):18～21,53
[18] N.Suqanuma, I.S.Kweon, Y.Bao, N.Fujiwara.Dynamic detection of a license plate using neural network.JSME International Journal,Series C:Mechanical Systems,Machine Elements and Manufacturing, 2001, 44(2): 431～436
[19] Cui Yuntao, Huang Qian. Automatic License Extraction from Moving Vehicles. Siemens Corporate Research, Proceedings of the 1997 International Conference on Image Processing. Part 3 (of 3) Santa Barbara, USA, Oct 26-29 1997, IEEE Comp Soc, Los Alamitos, USA, 1997,10 (3): 126～129
[20] Jun-Wei Hsieh,Shih-Hao Yu, Yung-Sheng Chen. Morphology-based License Plate Detection in Image of Differently Illuminated and Oriented Cars.Journal of Electronic Imaging, 2002, 11 (4): 507～516
[21] Park, Sung -Wook, Ku Yong -Hea. A License plate region extraction method using edge properties of wavelet subband. Proceeding of the International Conference on Imaging Science,Systems and Technology, CISST'03. Las Vegas,NV, United States,2003. CSREA Press, 2003. Volume 1: 8～12
[22] 求是科技，张宏林．Visual C＋＋数字图像模式识别技术及工程实践．北京：人民邮电出版社，2003．422～442
[23] Cao Guangzhi,Chen Jiaqian,Jang Jingping. An Adaptive Approach to Vehicle License Plate Location. The 29~(th) Annual Conference of the IEEE Industrial Electronics Society.Roanoke,VA, United States.2003. Institute of Electrical and Electronics Engineers Computer Society ,2003.Volume 2: 1786～1791
[24] 叶晨洲，廖金周．一种基于纹理的牌照图像二值化方法．微型电脑应用，1999年，第6期：28～30
[25] 周宏文，李见为，许盛．基于统计模式识别的彩色图像分割方法．光电工程，1999，26(5)：53～57
[26] 丁震，胡钟山等．一种基于模糊聚类的快速二值化方法．计算机学报，1998年，第21卷，增刊：331～334
[27] Richard G Casey, Eric Lecolinet. Survey of methods and strateies in character segmentation. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1996, 18(7): 690～706
[28] Reed T.R, du Buf J M.H. A review of recent texture segmentation and feature extraction techniques. CVGIP: Image Understanding,1993, Issue 57:359～372
[29] Seong-Whan Lee, Dong-June Lee, Hee-Seon Park. New methodlogy for gray-scale character segmentation and recognition. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1996, 18(10): 1045～1050
[30] 叶晨洲，廖金周，梅帆．车辆牌照字符识别系统．计算机系统应用，1999年，第5期：10～13
[31] 曹迪铭，宣国荣．牌照字符分割中的区域分裂与合并．计算机工程，2000，26(10)：18～21
[32] 王建平，盛军，朱程辉．基于小波分析的视频图像字符特征提取方法研究．微电子学与计算机，2002年，第5期：51～54
[33] 郭军，马跃，盛立东，钟义信．发展中的文字识别理论与技术．电子学报，1995，23(10)：184～187
[34] 王年，任彬，黄勇， 汪炳权．人工神经网络在公路车辆管理中的应用．电子技术应用， 1998年，第9期：11～12
[35] 董林，陈锡先，唐远炎，吴善培．一种新的手写体数字识别方法．北京邮电大学学报，1997，20(1)：79～83
[36] YeChenzhou, JangJie, XuanGuorong. Number plate character recognition. Shanghai Jiaotong Daxue Xuebao, 2000, 34(5): 672～675
[37] 宋加涛，刘济林．车辆牌照上英文和数字字符的结构特征分析及提取．中国图象图形学报，2002，7(9)：945～949
[38] T.Naito, T.Tsukada. Robust recognition methods for inclined license plates under various illumination conditions outdoors. 1999 IEEE/IEEJ/JSAI International Conference on Intelligent Transportation Systems. Tokyo, Jpn .Oct 5-Oct 8 1999, IEEE, Piscataway, NJ, USA, 1999. 697～702
[39] J.A.G Nijhuis, M.H ter Brugge, K.A Helmholt.Car license plate recognition with neural networks and fuzzy logic. Proceedings of the 1995 IEEE International Conference on Neural Networks. Part 5 (of 6). Perth, Aust. Nov 27-Dec 1 1995. IEEE, Piscataway, NJ, USA,1995. 2232～2236
[40] 刘俊，付敬奇，董新平．数据融合在目标识别中的应用．传感器技术，2001， 20(6)：9
[41] 钟子发，罗明．多传感器数据融合中基于目标识别的DS算法研究．电讯技术，2000年，第4期：57～58
[42] 李圣怡，吴学忠等．多传感器融合理论及在智能制造系统中的应用．长沙：国防科技大学出版社，1998．159～165
[43] 王联明，刘文刚．高速公路不停车收费系统设计．山西电子技术,2000年，第5期：25～26
[44] 孙涛，张宏建．目标识别中的信息融合技术．自动化仪表，2001，22(2)：1
[45] S.S Iyengar, B.Jones. Information fusion techniques for pattern analysis in large sensor data networks:An overview. Journal of The Franklin Institute, 2001,338 (5): 571～582
[46] 严学强，刘济林，郭小军，顾伟康．基于空间频率域的纹理分割算法．浙江大学学报，1998，32(11)：726～731
[47] 姜珊，孙玉方．手写汉字的集群识别．中文信息学报，1998，12(4)：1～8