期刊名称: |
Pattern Recognition |
全部作者: |
Feng Liu, Qijun Zhao, David Zhang |
出版年份: |
2011 |
卷 号: |
44 |
期 号: |
|
页 码: |
|
查看全本: |
|
Fingerprintmatchingisanimportantandessentialstepinautomatedfingerprintrecognitionsystems (AFRSs). Thenoiseanddistortionofcapturedfingerprintsandtheinaccurateofextractedfeaturesmake fingerprintmatchingaverydifficultproblem.Withtheadventofhigh-resolutionfingerprintimaging techniquesandtheincreasingdemandforhighsecurity,sweatporeshavebeenrecentlyattracting increasingattentioninautomaticfingerprintrecognition.Therefore,thispapertakesfingerprintpore matchingasanexampletoshowtherobustnessofourproposedmatchingmethodtotheerrorscaused by thefingerprintrepresentation.Thismethoddirectlymatchesporesinfingerprintsbyadoptinga coarse-to-finestrategy.Inthecoarsematchingstep,atangentdistanceandsparserepresentation- based matching method(denoted as TD-Sparse)is proposed to compare pores in the template and test fingerprint image sand establish one-to-many pore correspondences between them.Theproposed TD-Sparsemethodisrobusttonoiseanddistortionsinfingerprintimages.Inthefinematchingstep,false porecorrespondencesarefurtherexcludedbyaweightedRANdomSAmpleConsensus(WRANSAC) algorithminwhichtheweightsofporecorrespondencesaredeterminedbasedonthedis-similarity betweentheporesinthecorrespondences.Theexperimentalresultsontwodatabasesofhigh-resolution fingerprintsdemonstratethattheproposedmethodcanachievemuchhigherrecognitionaccuracy compared with other state-of-the-art pore matching methods.