基于虹膜識別的身份鑒別方法研究.pdf

1、中國科學院自動化研究所博士學位論文基于虹膜識別的身份鑒別方法研究姓名：馬力申請學位級別：博士專業(yè)：模式識別與智能系統(tǒng)指導教師：譚鐵牛王蘊紅20030501PersonalIdentificationBasedonIrisRecognitionAbstractWiththeincreasingrequirementsforsecuritybiometricsbasedpersonalidentificationmethodshaverec

2、eivedextensiveattentionRecentlyirisrecognitionisbecomingallactivetopicinbiometricsduetoitshighreliabilityforpersonalidentificationAgreatdealofprogressinirisrecognitionhasbeenachievedinthepastdecade，butmanyeffortsremainto

3、betakentofurtherimprovetheperformanceofirisrecognitionsystemsAimingtodevelopallirisrecognitionprototypesystem，weinvestigatesomeofthekeyissuesrelatedtoirisrecognitioninthisthesis，includingirisimaging，irislivenessdetection

4、andirisfeatureextractionandmatchingThemaincontributionsofourworkreportedinthisthesisareasfollows：1Anirissensori8designedandCASIAirisimagedatabaseiSconstructedWehavereleasedthisdatabasetothepublicdomaininordertopromoteres

5、earchonirisrecognition2AnexperimentplatformisbuiltwhereonecandevelopandtestnewalgorithmsforirisrecognitionThisplatformincludesthedesignedirissensorandsomebasicirisimageprocessingmethodsWealsoproposeaneffectiveimagequalit

6、yassessmentalgorithmbasedonFourierspectrumanalysisaswellasanewimagepreprocessingalgorithmExperimentalresultshaveshowntheeffectivenessandaccuracyofthesealgorithms，3Arealirisrecognitionsystemshouldbe曲1etodeterminewhetherth

7、einputimagesarefromalivesubjectinsteadofanirisphotograph，avideoplayback，aglasseyeorotherartifactsInthisthesis，aneffectivemethodisimplementedtodetectliveirisesbyanalyzingembodimentsofopticalandphysiologicalcharacteristics

8、oftheliveeyeinanimagesequence4SeveralhovelalgorithmsareproposedforirisfeatureextractionandmatchingByourexperiments，twoimportantfindingscanbesummarized：1)theirischaracteristicsalongtheangulardirectionhavehigherinformation

9、densityanddiscriminability2)featuresreflectinglocalsharpvariationsoftheirisplayanimportantroleinrecognitionTheresultingconclusionswillbehelpfulforfurtherresearch5Extensiveperformancecomparisonofexistingschemesforirisreco

10、gnitionisconductedonCASIAirisimagedatabaseTothebestofourknowledge，thisisthefirstcomparativestudyonirisrecognition，TheproposedirisrecognitionmethodbasedonlocalvariationanalysishasahighperformanceBasedonthisalgorithmandwor