简介:De-duplicationusingbiometricshasgainedmuchattentionfromresearchcommunitiesasitprovidesauniqueidentityforeachandeveryindividualamongthelargepopulation.De-duplicationistheprocessofremovingtheinstancesofmultipleenrollmentsbythesamepersonusingtheperson’sbiometricdata.Animportantissueinthelarge-scalede-duplicationapplicationsisthespeedofmatchingandtheaccuracyofthematchingbecausethenumberofpersonstobeenrolledrunsintomillions.Thispaperpresentsanefficientmethodtoimprovetheaccuracyoffingerprintde-duplicationinde-centralizedmanner.De-duplicationaccuracydecreasesbecauseofthenoisepresentinthedata,whichwouldcauseimproperslapfingerprintsegmentation.Inthispaper,anattemptismadetoremovethenoisepresentinthedatabyusingbinarizationofslapfingerprintimagesandregionlabelingofdesiredregionswith8-adjacencyneighborhood.Thedistinctfeatureofthistechniqueistoremovethenoisepresentinthedataforanaccurateslapfingerprintsegmentationandimprovethede-duplicationaccuracy.Experimentalresultsdemonstratethatthefingerprintsegmentationrateandde-duplicationaccuracyareimprovedsignificantly.
简介:本文分析了视频预处理的De-interlace算法.在分析了线性滤波De-interlace算法缺点的基础上,提出了像素相关性判断改进方法,改善了处理的效果.针对主要是水平或者垂直方向运动的视频,运动补偿法的块匹配算法存在速度慢的缺点,提出了行搜索运动补偿De-interlace算法,改善了预处理效果,提高了预处理速度.最后,对这些算法进行了评估与比较.
简介:AnewfilteringmethodforSARdatade-noisingusingwaveletsupportvectorregression(WSVR)isdeveloped.OnthebasisofthegreyscaledistributioncharacterofSARimagery,thelogarithmicSARimageasanoisepollutedsignalistakenandthenoisemodelassumptioninlogarithmicdomainwithGaussiannoiseandimpactnoiseisproposed.Basedonthebetterperformanceofsupportvectorregression(SVR)forcomplexsignalapproximationandthewaveletforsignaldetailexpression,thewaveletkernelfunctionischosenassupportvectorkernelfunction.ThenthelogarithmicSARimageisregressedwithWSVR.Furthermoretheregressiondistanceisusedasajudgmentindexofthenoisetype.Accordingtothejudgmentofnoisetypeeverypixelcanbeadaptivelyde-noisedwithdifferentfilters.Throughanapproximationexperimentforaone-dimensionalcomplexsignal,thefeasibilityofSARdataregressionbasedonWSVRisconfirmed.AfterwardtheSARimageistreatedasatwo-dimensionalcontinuoussignalandfilteredbyanSVRwithwaveletkernelfunction.Theresultsshowthatthemethodproposedherereducestheradarspecklenoiseeffectivelywhilemaintainingedgefeaturesanddetailswell.
简介:Isospectralandnon-isospectralhierarchiesrelatedtoavariablecoefficientPainlev′eintegrableKorteweg-deVries(KdVforshort)equationarederived.Thehierarchiesshareaformalrecursionoperatorwhichisnotarigorousrecursionoperatorandcontainstexplicitly.Bythehereditarystrongsymmetrypropertyoftheformalrecursionoperator,theauthorsconstructtwosetsofsymmetriesandtheirLiealgebrafortheisospectralvariablecoefficientKorteweg-deVries(vcKdVforshort)hierarchy.
简介:Copperprocessorsusingsolventextraction/electrowinning(SX/EW)haveknownforyearsthateliminatingtheelectrolytebleedcouldbenefitprocesseconomicsdramaticallyUntilnow.therewasnopracticalsolutiontoeliminatethebleed.Availableprocesseseithercouldnorreduceironcontaminantssufficiently,orpulledouttoomuchvaluablecopperandcobaltalongwiththeiron.Basedonpilottestsofanewbreedofionexchangeseparatioinsystematthreedifferentminesitesonrepresentativeelectrolytes,analternmativetobleedingelectrolytesnowexists.Onthisbasis,processorscanshelvethewastefulelectrolytebleedingpractice.Already,onemajorcopperproducerisinstallingafullscalesystem.
简介:Facede-identificationhasbecomeincreasinglyimportantastheimagesourcesareexplosivelygrowingandeasilyaccessible.Theadvanceofnewfacerecognitiontechniquesalsoarisespeople'sconcernregardingtheprivacyleakage.Themainstreampipelinesoffacede-identificationaremostlybasedonthek-sameframework,whichbearscritiquesofloweffectivenessandpoorvisualquality.Inthispaper,weproposeanewframeworkcalledPrivacy-Protective-GAN(PP-GAN)thatadaptsGAN(generativeadversarialnetwork)withnovelverificatorandregulatormodulesspeciallydesignedforthefacede-identificationproblemtoensuregeneratingde-identifiedoutputwithretainedstructuresimilarityaccordingtoasingleinput.Weevaluatetheproposedapproachintermsofprivacyprotection,utilitypreservation,andstructuresimilarity.Ourapproachnotonlyoutperformsexistingfacede-identificationtechniquesbutalsoprovidesapracticalframeworkofadaptingGANwithpriorsofdomainknowledge.