指纹识别的算法研究

华中科技大学硕士学位论文指纹识别的算法研究姓名：胡佳曦申请学位级别：硕士专业：通信与信息系统指导教师：王芙蓉20050401IGaborPoincareHough3()On2()OnIIAbstractFirstly,thedevelopmentandapplicationofBiometricAuthenticationisfullyreviewedinthispaper,aswellasFingerprintAuthentication,oneofthemostmaturetechniquesofBiometricAuthentication.Basedontheabroadresearches,asystemalgorithmdesignisdescribedhere,whichconsistsofFingerprintPreprocess,CharacteristicExtraction,andFingerMatching.FingerprintPreprocessincludesFingerprintForegroundExtraction,ImageEnhancementandThinning.FingerprintForegroundExtractioninthispaperisimportantforthefollowingalgorithms,whichisbasedonEdgeExtractionOperatorandMathematicalMorphology.Asaresult,foregroundandbackgroundoffingerprintimagescanbedistinguished,whichreducemuchnoisefromthebackgroundforthefollowingfingerprintalgorithms.Thustheaccuracyisenhanced,thesearchingareaisreduced,andthecomputingtimeissaved.AlsotheforegroundisvaluableinformationforFingerprintMatching.BasedonaGaborwaveletimageenhancement,anadaptiveenhancementalgorithmisproposed,whichalsoconsiderstheinformationofSingularPointsandForeground,andgetsbetterperformance.Becauseofthehexagonalgrid,theconnectivityparadoxintherectangulargridcanbeavoided.Thetemplatesherearesimpler,andeasiertoimplementonsoftwareandhardware.Thealgorithmismorerobusttothenoiseontheedges,whicharebetterthanitscounterpartontherectangulargrid.CharacteristicExtractionheremeansSingularPointExtraction.Thepositions,types,andorientationsofSingularPointsareextractedfromgrayfingerprintimages,inthestepofSingularPointExtraction,bycomputingthePoincareIndexandsquareorientationfield.Algorithminthispaperownsmoreaccuracy,computingspeedandrobustness,becauseofhierarchicallydesigned.TherearetwodifferentlevelsforFingerMatching--CoarseMatchingandPreciseMatching.Atthecoarselevel,fingerprintsarefilteredbythestructureinformationsuchasSingularPoints,whichsavesmuchcomputingtime.Andthen,atthepreciselevel,fingerprintsarematchedbytheminutiatemplatesbasedonHoughTransform.Becauseofthemoreminutiainformationisconsidered,thecomplexityisreducedto2()Onfrom3()On.IIIKeyword:Fingerprint;SingularPoint;ImageEnhancement;HexagonalGrid;BiometricAuthentication.££1;1.1;20.001%[1]2070[1,2];;;;;()()31.1.11-11-11-112EnrollmentVerification[3]4On-Line[3]1.1.21234567[4]1.1.31Verification2IdentificationVerificationICIdentification1.1.151.1.4(1)(genuine);(2)(genuine);(3)(imposter);(4)(imposter)(1)(3)(2)(4)FAR(falseacceptancerate)FRR(falserejectrate)FARFRR1-2FARFRRFRRFAREnrollmentTimeMatchingTime61-2FARFRR1.2[4]1.1.21122372045handgeometry6PalmPrint7fingerprint828DNA1.31-11-1[4]91.4BioAPI1.5237345Gabor678102.17000600018239187718892070(AutomatedFingerprintIdentificationSystemAFIS)1.50192.3.4.5.611ICATM()2.2200IBMIntelNECIdentiixDigitalpersona(CaliforniaInstituteofTechnology)(MichiganStateUniversityPatternRecognitionandImageProcessingLab)12Bologna(NISTVisualImageProcessingGroup)IBM(IBMT.JWatsonResearchCenter)802.21-12-12-12-113IDIDIC(1)(2)ATMATMATMATM2-125[4]142.31.(1)[5][6][7][8](2)01[9](3)[10]Hildth;82.153.Kohonen[11][12][13](hiddenMarkovmodelclassification)4.(NIST()((FBI)();RathaChenJain[14]HoughHoughJ.PStafnkE.BakeRanadRosenfeld[15]S.Sobaji[16]2.4162-22-22-238172.51.1.4FARFRR1.1.4FARFRRFARFRR2-32-3FARFRR[4]FARFRR2-3FRRFRR2.618[4][17]canny[17]1/53.11985ErosionDelationOpeningClosingBaBaBaXaaXBE(X)={a|BacX}=XB3-1a19(a)b3-1BaBaBaXaaXBD(X)={a|BaX}=XB3-1b3-23-33-23-2XBaBaXXB203-33-3XBaBaXXB3-43-53-43-4XB213-53-5XB3.23.2.11.2.3.4.3.2.21222N3M23.3400300300256FVC2004Celeron500MHZ128MSDRMatlab6.5[17]3-63-6abc[17]3-6[17][17]CannySobel3-1233-10.51[17]2.5162N3MNMN23-73-7CannySobelLaplace3-1CannySobelSobel3.4[17]1/524[5,18][19]4-14-1Poincare[18,20,21,22]PauilRamoMariusTicoVesaOnniaandJukkaSaarinen[23][23][18,20,21,22][23][23]27%4.1254.1.1[18,20]4-28Poincare[21]0Poincare0[21]4.1.2[23]4-34-44-3a4-3b826844a464babcd4-3abcd4.2[23]0[23][23][23]127),(),(),(),(),(),(jijijiPYjijijiPXyxyx∂∂=∂-∂=(4.1)(i,j)33Poincare332PXPY333PXPY9PXPY4354.3[20][23][23][20][23]34[20]4-1[20]40.18[20]40.17[20]4[20]4Matlab6.5[20]3.801.270.522.01[23]5.19[23]28[23][5][5]4-1.[20][23](Matlab6.5)[20]30.80.20.02N/A[20]40.50.130.05N/A[23]2.20.700.495.19s.0.430.150.023.80s4.4Poincare4-5Matlab6.5Matlab6.5JavaC/C++0.98294.530D.M.Weber[24]D.C.DouglasHuang[25]B.G..Sherlock[26]Mehtre[27][28]O’GormanJ.V.Nickerson[29][30]LinHong[31]Gabor5.1GaborLinHong[31]Gabor31Gabor[32,33,34]XudongJiang[32][31]GaborX-SignatureGaborGabor[33]Gabor8[34]LingHong[31]LingHong[31]GaborGabor)2cos(21exp),:,(2222fffpddffxyxfyxhyx⎪⎭⎪⎬⎫⎪⎩⎪⎨⎧⎥⎥⎦⎤⎢⎢⎣⎡+-=(5.1)ffffffcossinsincosyxyyxx+-=+=fGaborfyxdd,GaborDeltayx==dd5-1[31]5-25-35-4Delta=345DeltaDelta32Delta5-15-2Delta=35-3Delta=45-4Delta=55-1[31]5-55-65-75-811131517335-15-5BlockSize=115-6BlockSize=135-7BlockSize=155-8BlockSize=17GaborGabor345.2LingHong[31]GaborMatlabGaborGaborDelta=1/(2f)fGaborGabor4⎪⎪⎪⎩⎪⎪⎪⎨⎧---+=otherwiseVARMjiIVARMMjiIifVARMjiIVARMjiG,)),((),(,)),((),(200200(5.2)),(jiG0VAR0MVARM),(jiI1717xyyx∂∂,),(jiq),(jif(5.3)[35]35∑∑+-=+-=∂∂=2/2/2/2/),(),(2),(wiwiuwjwjvyxxvuvujin)),(),((),(2/2/2/2/22∑∑+-=+