搜索
上海交通大学硕士学位论文摘要I基于MATLAB的车牌识别系统研究摘要近几年，车牌识别系统作为智能交通的一个重要方向越来越受到重视。车牌识别系统可以应用于停车场管理系统、高速公路超速管理系统、城市十字路口的“电子警察”、小区车辆管理系统等各个领域，对国家的安全发展有很大的作用。虽然目前已有一些车牌识别系统相关产品出现，但是对其算法的研究发展从没有停止，仍有许多学者在做着进一步的研究改进。本文首先对车牌识别系统的现状和已有的技术进行了深入的研究，在研究的基础上开发出一个基于MATLAB的车牌识别系统。确定了整体设计方案，其中软件部分包括车牌定位、车牌字符切分及车牌字符识别三个模块。车牌定位模块中提出了基于小波变换的车牌边缘提取的算法，以及车牌二次定位的算法，提高了系统在光照条件较差的情况下的定位准确率，该算法对于各种底色的车牌具有良好的适应性；车牌的二值化采用了改进的Otus算法，重新划分了其两维直方图的区域，改进后的算法大大减少了运行时间，对于各种类型的车牌都能达到较好的二值化效果；针对BP神经网络字符识别算法，采用有动量的梯度下降法训练网络，减小了神经网络学习过程的振荡趋势，使得BP网络能够较快的达到收敛，完成车牌字符的识别。对模板匹配算法和BP网络算法进行对比，证明了BP网络算法要优于模板匹配算法。根据上述算法搭建了一个测试平台。整个测试平台的软件部分采用MATLAB的M语言编写。通过测试平台，对353幅卡口汽车照片进行车牌识别，测试系统的性能。测试结果表明，本课题设计的车牌识别系统可有效地实现车牌识别，为今后的产品化奠定了很好的技术基础。关键词：车牌识别，小波变换，Otsu算法，模板匹配，BP网络，MATLAB上海交通大学硕士学位论文ABSTRACTIIRESEARCHONPLATELICENSERECOGNITIONSYSTEMBASEDONMATLABABSTRACTInrecentyears,thedevelopmentofintelligenttransportationhasbecomemoreandmoreimportant.Asanimportantaspectinintelligenttransportation,platelicenserecognitionsystemhastakenmoreandmoreattention.Theplatelicenserecognitionsystemcanbeappliedtopublicparking,highwayspeedingmanagementsystem,crossingroad,districtvehiclemanagementsystem,andsoon.Althoughnowtherearealreadysomeexsitingplatelecenserecognitionsystems,theresearchanddevelopmentofarithmetichaveneverstopped,andtherearestillmanyscholarswhoaredoingfurtherresearchandimprovement.Firstly,thepapergivesadeepresearchonthestatusandtechniqueoftheplatelicenserecognitionsystem.Onthebasisofresearch,asolutionofplatelicenserecognitionsystemisproposed,andthepaperfocusedonthesoftwarepart.Thewholesystemconcludesthreemodules.Theyareplatelocation,platecharactersegmentation,andplatecharacterrecognition.Intheplatelocationmodule,thepaperputsforwardanarithmeticofplateedgerecognitionbywaveletdecomposition,andanarithmeticoflocatingtwice,whichimprovetheaccuracyinbadlightcondition,andarefitforplateswithdifferentgrounding.AnimprovedOtsuarithmeticisusedintheprocessofbinaryzating,whichreducestherunningtime,andcanachievegoodeffectfordifferentkindsofplate.Incharacterrecognitionpart,withthemomentumofthegradientdescentmethod,theBPneuralnetworkcanfastconvergence.ComparedtheBPneuralnetworkwithtemplatematchingarithmetic,whichimprovesthattheBPneuralnetworkarebetterthanthetemplatematchingarithmetic.上海交通大学硕士学位论文ABSTRACTIIIThen,atestplatformhasbeenbuiltwithMATLAB,forthetestofthesystem.Throughthetestof353monitoringcarphotographs,theresultsshowsthatthesystemcaneffectivelymeetstherequirement,andlayagoodfoundationoftechnologyforproductization.KEYWORDS:platelicenserecognition,wavelettransform,Otsu,templatematching,BPneuralnetwork,MATLAB上海交通大学硕士学位论文图目录VII图目录图1车牌识别系统····································································································1图2自选号牌车牌示例·····························································································3图3车辆牌照识别系统结构图···············································································10图4系统流程图······································································································13图5车牌定位的过程······························································································15图6（a）原始汽车图像（b）灰度图···································································16图7灰度变换的对比曲线·······················································································17图8（a）灰度图（b）灰度变换后的图像···························································17图9（a）灰度图（b）中值滤波后的图像···························································18图10小波分解树[10]································································································21图11小波变换的Mallat算法················································································23图12二维小波变换的Mallat算法········································································24图13车辆灰度图····································································································25图14X=214数据线的灰度图··················································································25图15用HAAR小波进行五层分解········································································26图16车牌图像的小波分解·····················································································27图17小波分解提取边缘··························································································27图18开闭运算后的图像·························································································28图19车牌区域标记································································································29图20初步提取的车牌····························································································29图22平滑后的水平差分累加投影图·····································································31图23水平定位后的图像·························································································31图24平滑后的垂直差分累加投影图·····································································32图25精确定位后的车牌·························································································32图26车牌定位算法··········································