毕业论文外文翻译指纹识别和验证的匹配系.docx
《毕业论文外文翻译指纹识别和验证的匹配系.docx》由会员分享,可在线阅读,更多相关《毕业论文外文翻译指纹识别和验证的匹配系.docx(15页珍藏版)》请在冰豆网上搜索。
毕业论文外文翻译指纹识别和验证的匹配系
外文翻译
毕业设计题目:
指纹识别与研究
原文:
FingerprintIdentificationand
VerificationSystemusing
MinutiaeMatching
译文:
指纹识别和验证的匹配系统
原文:
FingerprintIdentificationandVerificationSystemusingMinutiaeMatching
Abstract:
Fingerprintsarethemostwidelyusedbiometricfeatureforpersonidentificationandverificationinthefieldofbiometricidentification.Fingerprintspossesstwomaintypesoffeaturesthatareusedforautomaticfingerprintidentificationandverification:
(i)globalridgeandfurrowstructurethatformsaspecialpatterninthecentralregionofthefingerprintand(ii)Minutiaedetailsassociatedwiththelocalridgeandfurrowstructure.ThispaperpresentstheimplementationofaminutiaebasedapproachtofingerprintidentificationandverificationandservesasareviewofthedifferenttechniquesusedinvariousstepsinthedevelopmentofminutiaebasedAutomaticFingerprintIdentificationSystems(AFIS).Thetechniqueconferredinthispaperisbasedontheextractionofminutiaefromthethinned,binarizedandsegmentedversionofafingerprintimage.Thesystemusesfingerprintclassificationforindexingduringfingerprintmatchingwhichgreatlyenhancestheperformanceofthematchingalgorithm.Goodresults(~92%accuracy)wereobtainedusingtheFVC2000fingerprintdatabases.
1.INTRODUCTION
Fingerprintshavebeeninuseforbiometricrecognitionsincelongbecauseoftheirhighacceptability,immutabilityandindividuality.Immutabilityreferstothepersistenceofthefingerprintsovertimewhereasindividualityisrelatedtotheuniquenessofridgedetailsacrossindividuals.Theprobabilitythattwofingerprintsarealikeis1in1.9x1015[1].Thesefeaturesmaketheuseoffingerprintsextremelyeffectiveinareaswheretheprovisionofahighdegreeofsecurityisanissue.Themajorstepsinvolvedinautomatedfingerprintrecognitionincludea)FingerprintAcquisition,b)FingerprintSegmentation,c)FingerprintImageEnhancement,d)FeatureExtractione)MinutiaeMatching,f)FingerprintClassification.
Fingerprintacquisitioncaneitherbeoffline(inked)orOnline(Livescan).Intheinkedmethodanimprintofaninkedfingerisfirstobtainedonapaper,whichisthenscanned.Thismethodusuallyproducesimagesofverypoorqualitybecauseofthenon-uniformspreadofinkandisthereforenotexercisedinonlineAFIS.Foronlinefingerprintimageacquisition,capacitativeoropticalfingerprintscannerssuchasURU4000,etc.areutilizedwhichmakeuseoftechniquessuchasfrustratedtotalinternalreflection(FTIR)[2],ultrasoundtotalinternalreflection[3],sensingofdifferentialcapacitance[4]andnoncontact3Dscanning[5]forimagedevelopment.Livescanscannersoffermuchgreaterimagequality,usuallyaresolutionof512dpi,whichresultsinsuperiorreliabilityduringmatchingincomparisontoinkedfingerprints.
Segmentationreferstotheseparationoffingerprintarea(foreground)fromtheimagebackground[6].Segmentationisusefultoavoidextractionoffeaturesinthenoisyareasoffingerprintsorthebackground.ASimplethresholdingtechnique[7]provestobeineffectivebecauseofthestreakednatureofthefingerprintarea.Thepresenceofnoiseinafingerprintimagerequiresmorevigoroustechniquesforeffectivefingerprintsegmentation.Agoodsegmentationmethodshouldexhibitthefollowingcharacteristics[8]:
·Itshouldbeinsensitivetoimagecontrast
·Itshoulddetectsmudgedornoisyregions
·Segmentationresultsshouldbeindependentofwhethertheinputimageisanenhancedimage
orarawimage
·Thesegmentationresultsshouldbeindependentofimagequality
Renetal.[8]proposedanalgorithmforsegmentationthatemploysfeaturedots,whicharethenusedtoobtainaclosesegmentationcurve.Theauthorsclaimthattheirmethodsurpassesdirectionalfieldandorientationbasedmethods[9,10,11]forfingerprintimagesegmentation.Shenetal.[12]proposedaGaborfilterbasedmethodinwhicheightGaborfiltersareconvolvedwitheachimageblockandthevarianceofthefilterresponseisusedbothforfingerprintsegmentationandqualityspecification.Xianetal.[13]proposedasegmentationalgorithmthatexploitsablock’sclusterdegree,meanandvariance.Anoptimallinearclassifierisusedforclassificationwithmorphologicalpost-processingtoremoveclassificationerrors.Bazenetal.[14]proposedapixelwisetechniqueforsegmentationinvolvingalinearcombinationofthreefeaturevectors(i.e.gradientcoherence,intensitymeanandvariance).Afinalmorphologicalpost-processingstepisperformedtoeliminateholesinboththeforegroundandbackground.Inspiteofitshighaccuracythisalgorithmhasaveryhighcomputationalcomplexity,whichmakesitimpracticalforrealtimeprocessing.Kleinetal.[15]proposedanalgorithmthatemploysHMMstoremovetheproblemoffragmentedsegmentationencounteredduringtheuseofdifferentsegmentationalgorithms.
Foragoodqualityfingerprintfeatureextractionismucheasier,efficientandreliableincomparisontoarelativelylowerqualityfingerprint.Thequalityoffingerprintsisdegradedbyskinconditions(e.g.wetordry,cutsandbruises),sensornoise,non-uniformcontactwithsensorsurface,andinherentlylowqualityfingerprintimages(e.g.thoseofelderlypeople,laborers).Asignificantpercentageoffingerprintsareofpoorquality,whichmustbeenhancedfortherecognitionprocesstobeeffective.Therearetwomajorobjectivesoffingerprintenhancementi.e.i)toincreasethecontrastbetweenridgesandvalleysandii)toconnectbrokenridges.Theseobjectivescanbefulfilledbyusingacontextualfilterwhosecharacteristicsvaryaccordingtothelocalcontexttobeusedforfingerprintenhancementinsteadofconventionalimagefilters.Thefiltershouldpossesthefollowingcharacteristics:
·Itshouldprovidealowpass(averaging)effectalong
theridgedirectionwiththeaimoflinkingsmallgapsandfillingimpuritiesduetoporesornoise.
·Itshouldhaveabandpass(differentiating)effectinthedirectionorthogonaltotheridgesinordertoincreasethediscriminationbetweenridgesandvalleysandtopseparateparallellinkedridges.
Sherlocketal.[16]proposedanalgorithmforfingerprintimageenhancementthatemploysposition-dependentFourier-domain-filtering-basedorientationsmoothingandthresholdingtechnique.Greenbergetal.[17]proposedtwoschemesforfingerprintenhancement.Onemethoduseslocalhistogramequalization,Wienerfiltering,andimagebinarizationwhereastheothermethodusesauniqueanisotropicfilterfordirectgrayscaleenhancement.O’Gormanetal.[18,19]proposedacontextualfilterbasedapproachthatutilizesfourmainparametersoffingerprintimagesatagivenresolutioni.e.maximaandminimaoftheridgeandvalleywidthstoformamotherfilterwhoserotatedversionsarethenconvolvedwiththeimagetoyieldtheenhancedoutput.Hongetal.[20]proposedaneffectivemethodbasedonGaborfiltersforimageenhancement.Gaborfiltersfulfilltherequirementsofagoodfingerprintenhancementfiltermentionedearlierastheypossessboththedifferentiatingandaveragingeffects.SlightmodificationstothistechniqueweremadebyGreenbergetal.[21].Jiangweietal.[22]modifiedthetechniquegivenby[20].Theirapproachmodelsthesinusoidalshapeoftheridge-valleystructureinaclosermannerandgivesbetterresults.Ticoetal.[23]madeuseofaridgedetectingtechniquebasedontheseconddirectionalderivativeoftheimagetocarryoutfingerprintenhancement.Watsonetal.[24]multipliedtheFouriertransformofeach32x32blockbyitspowerspectrumraisedtoapowerktoproduceanefficienttechniquethatcanfinditsuseinonlinefingerprintrecognitionsystems.Fingerprintspossesstwomajortypesoffeatures:
specialtypeofpatternformedbytheridgeandfurrowstructuresinthecentralregionofthefingerprintsandminutiaedetailsassociatedwithlocalridgesandfurrows.Minutiaeareminutedetailssuchasridgeendings(apointwherearidgeendsabruptly)oraridgebifurcation(wherearidgebreaksupintotworidges).Minutiaearecharacterizedbytheposition,directionandtype(Ridgeendingorbifurcation).Theglobalfeaturesareusedforfingerprintclassificationintosixmajorclasseswhereastheminutiaedetailsareusedforfingerprintbasedpersonidentification.FingerprintanalystsutilizetheminutiaeinformationforfingerprintidentificationanditisthemostestablishedtechniqueinthefieldofAFIS.Theaccuracyofthetechniqueisdependentupontheprecisionintheextractionofminutiae.Therearetwomajortypeofmethodsthatareusedforminutiaeextractioni)binarizationbasedmethodsandii)directgrayscalemethods.Inbinarization-basedmethodssomeinformationislostduringbinarizationthatcandegradetheperformanceoftheminutiaeextractor.Directgrayscalemethodsovercometheseproblemsbutmaybedifficulttoimplementandtimeconsumingtooperate.Thetypicalapproachforabinarization-basedmethodinvolvesapriorienhancement,binarizationandthenthinning.Variousbinarizationandthinningapproachesarediscussedin[25,26,27,28,29].Onceabinaryskeletonhasbeenobtained,asimpleimagescanallowsthepixelscorrespondingtotheminutiaetobedetectedbyfindingthecrossingnumber.Theminutiaeobtainedasaresultofminutiaeextractionneedtobefilteredinordertoremovethefalseminutiaeintroduced.Variousminutiaefilteringtechniquesareproposedin[30,31].
Duringenrollmenttheminutiaesetobtainedfromanindividual’sfingerprintisstoredasatemplateforthatsubject.Intheauthenticationmodule,thefinger