搜索
Vol.16,No.6©2005JournalofSoftware1000-9825/2005/16(06)1046∗1,2,1,2+1(,100080)2(,100049)AFingerprintMatchingAlgorithmwithRegistrationPatternInspectionCHENHong1,2,TIANJie1,2+1(CenterforBiometricResearchandTesting(CBRT),KeyLaboratoryofComplexSystemsandIntelligenceScience,InstituteofAutomation,TheChineseAcademyofSciences,Beijing100080,China)2(GraduateSchool,TheChineseAcademyofSciences,Beijing100080,China)+Correspondingauthor:Phn:+86-10-62532105,Fax:+86-10-62527995,E-mail:tian@doctor.com,(6):1046−1053.DOI:10.1360/jos161046Abstract:The“registrationpattern”betweentwofingerprintsistheoptimalregistrationofeachpartofonefingerprintwithrespecttotheotherfingerprint.Registrationpatternsgeneratedfromimposter’smatchingattemptsaredifferentfromthosepatternsfromgenuinematchingattempts,althoughtheymaysharesomesimilaritiesintheaspectofminutiae.Thispaperpresentsanalgorithmthatutilizesminutiae,associateridgesandorientationfieldstodeterminetheregistrationpatternbetweentwofingerprintsandtheirsimilarity.Theproposedmatchingschemehastwostages.Anofflinetrainingstagederivesagenuineregistrationpatternbasefromasetofgenuinematchingattempts.Then,anonlinematchingstageregistersthetwofingerprintsanddeterminestheregistrationpattern.Onlyifthepatternmakesagenuineone,afurtherfinematchingisconducted.ThegenuineregistrationpatternbaseisderivedusingasetoffingerprintsextractedfromtheNISTSpecialDatabase24.ExperimentalresultsonthesecondFVC2002databasedemonstratetheperformanceoftheproposedalgorithm.Keywords:registrationpattern;fingerprintfeaturerepresentation;automaticfingerprintmatching:“”.,..:;,∗SupportedbytheNationalNaturalScienceFoundationofChinaunderGrantNo.60332010();theProjectforYoungScientists’FundofNationalNaturalScienceFoundationofChinaunderGrantNo.60303022();theNationalScienceFundforDistinguishedYoungScholarsofChinaunderGrantNo.60225008():(1976),,,,;(1960),,,,,,.:1047,.NIST24.FVC2002DB2,.:;;:TP18:A,,..,,,.,.;.1.,Ratha[1],Dorai[2].,,.,(),.Cappelli[3].,,..Senior[4]“”.,...,,,.,.Bazen[5][6](thin-platespline,TPS).,TPS“”.,.,TPS.“”,.Kovács-Vajna[7].,,.,,(),.,,“”.,“”“”,(),().,.“”,“”“”.2,,.;,,.137pixelbiasOverlappingFig.1Overlappingtwodeformedfingerprints11048JournalofSoftware2005,16(6)GenuineregistrationpatternsetTrainingregistrationpatternTestingInputregistrationpatternEvaluatingmatchingscoreOffline:RegistrationpatternlearningOnline:RegistrationpatterntestFig.2Flowchartofregistrationpatternmatching21.,.,:},...,,{21inivvvV=,Tjjjyxv),(=(inj,...,2,1=,bai,=),bann=.)(vtopt((1)):}{}{:iim′a,2|)(|∑′−=jbjoptajvtvd.⎟⎟⎟⎠⎞⎜⎜⎜⎝⎛⎟⎟⎟⎠⎞⎜⎜⎜⎝⎛−=1100cossinsincos)(yxyxvtopt∆αα∆αα(1),,,,.,,.,:baVV,(bann=),topt)}({}{:mjjNjiim∈′a,},...,2,1{nNm′=,),min(bannn≤′,boxboundbjajthrvvsim_'),(′(mNj∈∀)(2),)(bjoptbjvtv′′=′|Nm|.7,,3.7:TtttRP),...,,(610=(3)Fig.3Overlappingareaoftwofingerprints3,:1049NIST24[8],.FVC2002DB2[9],.22.1,.,,.Jain[10,11]Gabor,Fan[12]20.[13,14]..,.4.Fig.4Fingerprintfeatures42.2(),.NP-,.Jiang[15].,..,,.Jiang,,,.,,5.Fig.5Minutiaeandsampledpointsontheirassociatedridge5,.Jiang,,tk,Σsl.,,,.,.,,..RidgeendingSampletype1Sampletype2…SampletypenBifurcation1050JournalofSoftware2005,16(6)ww×,)(,jioO=:⎩⎨⎧′′−=′other,0),(),,(),(,,,,,,booodoodboooslqpjiqpjiqpjiboφφ(4))(TckTOtO=′,∑′=′jijijiboOooslOOsl,,,),(),((5),,.6.,FinputFtemplate,:),(templateinputoptFFalignt=(6)RejectedMatchingparametercandidatesetNextmatchingparametercandidateEmploytransformationintoorientationestimationComparingitwithinputorientationMatchedGetfinalmatchingparametersYesNoFeedbackNoFig.6Matchingmethodwithfeedback62.3FinputFtemplate,),(templateinputglobalFFalignt=,)(templateglobalFt,()Rgni,Rgnt,iinputF,′(i=0,1,…,6).F′input,0Ftemplate),(0,0templateinputFFalignt′=,t0)(0templatetemplateFtF=′,6),(,templateiinputiFFalignt′′=′(i=1,2,…,6).,00=′t.),...,,,0(621tttRP′′′=.2.4(NIST)NIST24“”.NIST24100.,MPEG-2..100..,,RP=(t0,t1,…,t6)T.,}{kRPGRPB=.7.:1051(a)Templatefingerprint(b)Inputfingerprint(c)Matchingpattern(a)(b)(c)Fig.7Genuinematchingpattern72.5,.,.,,,.,.(7).,(7)RPiRPj.3333|)|(||||),(∑−×+−+−=kjkikjkikjkikjiRyyxxRPRPdαα∆∆∆∆(7)(7):}),(,|{gspaceiiGRPthrRPRPdGRPBRPRPS∈∃=(8),,,,.8,.(a)Templatefingerprint(b)Inputfingerprint(c)Matchingpattern(a)(b)(c)Fig.8Impostermatchingpattern82.6,:∑×=slnnmMtemplateinput),max((9)m,ninput,ntemplate,sl.,0.3FVC20022.FVC20021052JournalofSoftware2005,16(6).,,2.:LinHong[16]Gabor,,[17].9.,,,20.,.16×16,.FVC2002DB2880,110,8.FVC1,.(110×8×7)/2=3080(880×872)/2=383680.Fig.9Enhancementperformance9,.1.[15],.ROC10.,,,EER(1.1%).,,,ZFMR(17%).2.,.569dpi,NIST24(500dpi).:⎟⎠⎞⎜⎝⎛××=′α∆∆,500569,500569yxti(10)2EER=0.51%,ZeroFMR=2.3%.10ROC.,,(8).FVC2002[9]31.,.Fig.10ROCcurveofourmatchingovertheDB2fingerprintdatabaseofFVC200210FVC2002DB2ROC4,:1053.,,,..,:,,,,,.,.,,,.,.,,,..References:[1]RathaNK,BolleRM.Effectofcontrolledacquisitiononfingerprintmatching.In:Proc.ofthe14thInt’lConf.onPatternRecognition.Brisbane:IEEEComputerSociety,1998.1659−1661.[2]DoraiC,R