汽车牌照定位与字符分割的研究及实现

中南民族大学硕士学位论文汽车牌照定位与字符分割的研究及实现姓名：李国立申请学位级别：硕士专业：计算机应用技术指导教师：段汕20090523II,OtsuSobelIIIAbstractVehicleLicensePlateRecognitionisakeytechnologyofintelligenttransportationSystemChargesonthehighway,vehicledetection,monitoringandmanagementofcarparks,Roadvehiclemonitoringinareashavebroadapplicationprospects.Licenseplatelocationandlicenseplatecharactersegmentationofthelicenseplaterecognitionsystemismainlystudiedinthispaper.Firstly,studyofimagepreprocessing,gray-scaleimages,respectively,andgray-scalestretch,binarization,edgedetection,etc.werestudied.grayandgray-scalehistogramequalizationprocessingofthecollectedimages,toovercomenon-idealinthelightofthecontrastcausedbytoolargeortoosmalltoimagetheimpactofeffectiveinformation.andthenusetheOtsualgorithmforimagebinarization,itgetabetterbinarizationresults.Furtheranalysisofcommonlyusedinimageprocessingedgedetectionoperator,thenselecttheSobeloperatorfortheedgeoperatorofthisarticle.Secondly,licenseplatelocationtostudy,firstintroducedthemethodcommonlyusedinlicenseplatelocation,Andthenthelicenseplatelocationmethodbasedonmathematicalmorphologyisgiven.Themethodistoimagesofpreprocessing,gradientandbinarization,thenuseofclosedandopenofmathematicalmorphologytoformanindependentcharacterregionconnecteddomain,Thebodythroughthecharacteristicsofalicense:Aspectratioandsizeofthescopetoachievelicenseplatelocation.Theexperimentalresultsshowthatthehigheraccuracyofpositioningmethods,runfaster.Finally,theresearchdivisionofthelicensecharacters.Licenseplatecharactersegmentationincludinglicensespretreatmentandcharactersegmentation.Licensespretreatmentincludinganglecorrectionlicenseandfilteringofnon-characterregionandremovalofupperandlowerframeandrivets.Verticalprojectionmethodusingtheinitialpartitionwheninthecharactersegmentation,thenbasedonasinglelicenseandtheoverallscaleofthegeometricrelationshiptoremovenon-charactersoftheadhesionbetweenthecharactersandtheeliminationofover-segmentationofChinesecharactersandsoon.Theexperimentalresultsshowthatthismethodofcharactersegmentationaccuracy,especiallyforthecharactersbetweentheplateadhesionornoisecanplayabettersegmentationKeywordsBinarizationEdgeDetectionMathematicalMorphologyLicensePlateLocationCharacterSegmentation1______2111.1(LicensePlateRecognitionLPR)LPRLPR,LPRLPR[1]LPR1.11.11.2LPR,,LPR1.2.1LPR2LPR[2-8](1)/(2)(3)/(4)(5),[9]1.2.21)[2][3]Robert[4]Otsu2)[5-8]HoughHough3)[10-13]3[14-16]1.3121OtsuSobel2.3.1.241.21.4:LPR,52,[16,17]1.2.3.2.1RGB,,RGBRGBRGB6[18]0.229*0.587*0.114*IRGB=++(2.1)I2.1.12.12.22.32.42.1-2.22.3-2.4,2.32.4,,72.2[1]N[19]1.,0,1,2,1iriL=−L2.(),0,1,2,1inriL=−3.()(),0,1,2,1iinrPriLN==−4.0()(),0,1,2,1ijijnrTriLN===−∑5.ig[]maxminmin()()0.5iigINTggTrg=−++0,1,2,1iP=−PINT6.(),0,1,2,1ingiP=−7.()(),0,1,2,1iingQgiPN==−8.irig82.52.62.72.52.4,,,2.62.72.60200−,0100−,2.70255−2.1.2[19-21](,)fxy{}T•(,)gxy{}(,)(,)gxyTfxy=(2.2)92.1.2.1(,)fxy[,]ab[,]cd2.8(,)gxy0(,)(,)[(,)](,)(,)cfxyadcgxyfxyacafxybbadfxyb≤−=−+≤≤−(2.3)0ac==2.92.82.9(0ac==)(,)(,)gxykfxy=dkb=(2.4)1k=−[0,1]L−1gLf=−−(2.5)1k=1k1k102.1.2.22.102.10(,)fxy[0,]fM2.10(,)gxy[0,]gMgMfM(,)0(,)(,)[(,)](,)[(,)](,)gffcfxyfxyaadcgxyfxyacafxybbaMdfxybdbfxyMMb≤−=−+≤−−−+≤≤−(2.6)2.102.1.2.3111.ln[(,)1](,)*lnfxygxyabc+=+(2.7)abc2.112.[(,)](,)1cfxyagxyb−=−(2.8)abc2.122.112.122.132.42.142.6(0,200)(128,255)122.132.142.225601''0''''1''(,)fxy(,)xy(,)gxy0(,)(,)1(,)fxyrgxyfxyr=≥(2.9)r:[18-22],Otsu[22]Otsu,Otsurrr2Aσ2Bσ2Tσ13212ABQσσ=222ATQσσ=232TBQσσ=(2.10),1Q,1Qr()iPrir0,1,2,1iL=−LOtsu1.0r=2.maxr3.k2Aσ2Bσ2221122AWWσσσ=+(2.11)221212()BWWMMσ=−(2.12)(2.11)(2.12)10()kiWPi==∑max21()kikWPi=+=∑(2.13)101()kiiPiMW==∑max212()kikiPiMW=+=∑(2.14)21101()()kiiMPiWσ=−=∑max22212()()kikiMPiWσ=+−=∑(2.15)4.1rr=+rmaxr5.rrOtsu2.132.15,142.152.3[1,19]2.3.1[1,23](,)fxy(,)xy15[(,)]fxGfxyfy∂∂=∂∂(2.20)fx∂∂fy∂∂(,)fxyxy[(,)]Gfxy(,)fxy(,)xy22[(,)]()()ffGfxyxy∂∂=+∂∂(2.21)22[(,)][(,)(1,)][(,)(,1)]Gfijfijfijfijfij=−++−+(2.22)(2.22)(2.22)[(,)](,)(1,)(,)(,1)Gfijfijfijfijfij=−++−+(2.23)(2.23)(,)xyGijGG=+(2.24)xGyG1100xG−=1010yG=−2.3.2RobertsSobelPrewittLaplacian[22,23]1.RobertsRoberts(,)(1,1)xffijfij∆=−++(2.25)16(,1)(1,)yffijfij∆=+−+(2.26)yf∆xf∆Roberts(,)Rij(,)xxRijff=∆+∆(2.27)10(,):01xfxy∆01(,):10yfxy∆−Roberts2.PrewittPrewitt22×33×Prewitt22(,)xyPxyGG=+(2.28)xGyG(1,1)(,1)(1,1)(1,1)(,1)(1,1)xGfxyfxyfxyfxyfxyfxy=−++++++−−−−−−+−(1,1)(1,)(1,1)(1,1)(1,)(1,1)yGfxyfxyfxyfxyfxyfxy=+−+++++−−−−−−−+Prewitt101(,):101101xfxy−∆−−111(,):000111yfxy−−−∆3.SobelSobelPrewitt33×Sobel22(,)xySxyGG=+(2.29)xGyG(1,1)2(,1)(1,1)(1,1)2(,1)(1,1)xGfxyfxyfxyfxyfxyfxy=−++++++−−−−−−+−(1,1)2(1,)(1,1)(1,1)2(1,)(1,1)yGfxyfxyfxyfxyfxyfxy=+−+++++−−−−−−−+Sobel17101(,):202101xfxy−∆−−121(,):000121yfxy−−−∆4.LaplacianLaplacian,(,)fxy(,)xyLaplacian22222(,)fffxyxy∂∂∇=+∂∂(2.30)010141010−−−−111181111−−−−−−−−LaplacianLaplacianLaplacianLaplacian2.3.3LPRRoberts2.152.16,RobertsRobertsPrewitt2.152.17Prewitt2.182.15LaplacianRoberts18LaplacianRoberts2.192.15SobelSobelSobelSobel2.16Roberts2.