欢迎来到天天文库
浏览记录
ID:36455763
大小:2.39 MB
页数:71页
时间:2019-05-10
《基于神经网络的指纹识别》由会员上传分享,免费在线阅读,更多相关内容在学术论文-天天文库。
1、北京工业大学硕士学位论文基于神经网络的指纹识别姓名:杨静申请学位级别:硕士专业:电路与系统指导教师:孙光民2003.5.1北京工业大学工学硕士学位论文出了将指纹自动识别过程分为两步进行,即采用不同的神经网络模型分别对指纹进行分类和识别,首先采用BP网络对指纹进行分类,并在此基础上,根据BP分类器求出的指纹类别,分别将指纹的细节点特征送入相应指纹类别对应的LVQ网络进行识别,最后输出识别结果。实验结果表明,该方法取得了较好的识别结果。关键词:指纹预处理;指纹识别;神经网络IIABSTRACTWitllnledevelopmemoftheco埘lputerandinf0
2、衄ationtechnologies,identityidentmcationtechnologybycomputermakesquietgreatprogress.BecauSeofmeiⅡVariabilityanduIliquenessof血ef.mgerprints,metechⅡologyoffingerp血tVerincationandidentmcationhasbecomeoneofmeWidestusedidenti付identificationtecllllologies.nrela_testoalotoffields,includingpatt
3、emrecogIlition,digitalimageprocessiIlg,digitalsi印alprocessing,aniflcialiIltell追enceandsoon.nhasgreatWonhincrirninalidcntification,ne咖rks疵吼im肌ationfacilitys疵tyetc,andhasnot0111ymeoreticalsi卿ficancebutaIsopracticalvalue.NoW,AutomatedFingerpmIde而flcationSystems(AFIS)iss伽iIlmestageofexp甜me
4、nt.Whatmal(es“最lr矗Dmmatureisonen船cribedtoitsd锄and0n1a唱enumbersoffinger埘nt-matchingprocesses,andtlleconsiderableinnuenceofimages’noiseandskins’el枷citydllringtheprocedureofautomated丘ngerprimident讯cation.TheselimitationsresUdtinalowrecog芏1itionrateaIldaslowfingerprint-matcllingSpeed.IIl吐l
5、ispaper,wenytocombine也eneuralne如,orkwitIlpattemrecogIlitionandfindseveralnewme如odsinf.mgerprintautomaticidentmcationas也epmcticables01utiontothe仃。讧ble.Ingeneral,wemainlyresearchonfir唱erprintprepmcess洫g,featureex协lctionandneuml一n出vork.basedfingerprimidernificationin也ispap%Int11eprocedure
6、ofpreprocessing,、vcresearchonme叩e洲。璐ofenhancement,segment,bir血zation,thimlillg觞weUasno肌alization,aIld也enachjeveaprogramforits印plication.Wbdesignanormalizationoperation船theadditionalsteptomefingeIprimpr印mcessingope嘶on.Thestepoff.mgerprintnonnalization110tomynonnalizesthesizeoffingeIprin
7、timage,faci王itatesthetestonAFIS,butalsoreducestheAFISnlnningtiIIle趴d触ensAFISreco鲥tionspeed.Additionally,tIlispaperdoesapilot曲udyonfImgerprintthiminga_lgoritllIIlWllichisbasedonneuralnet、vork.IIl廿1epmcedllreoffiI培erprintfeaturee斌raction,cl够si母ingfeatureandidenti匆ingf色aturehavebeenextr
此文档下载收益归作者所有