基于神经网络的车牌自动识别系统(IJEM-V7-N4-3)

I.J.EngineeringandManufacturing,2017,4,26-35PublishedOnlineJuly2017inMECS()DOI:10.5815/ijem.2017.04.03Availableonlineat:AutomaticSystem,NeuralNetworks,Recognition,oflicenseplates.©2017PublishedbyMECSPublisher.Selectionand/orpeerreviewunderresponsibilityoftheResearchAssociationofModernEducationandComputerScience.1.IntroductionTheproblemofthetrafficcontroloftheintersectionsrequirestheunderstandingof,butalsoandsomewaytocheckthedecisionsandtheassessmentofitseffectiveness.Testingofthereal,intersectionswillbesodiffi-cultespeciallywhenthedesignermustconstantlymakechangesinitsdevelopmentanduseofoutputdataforanalysisandotherpurposes.Anotherwayofmodelingisnotrequired,thedesignerflexibilitytotestitsdesignandtheresultsofthemucheasierandfaster.Oneofthesewaysusingstolenvehiclessearch[1].Constantlygrowinguseofvideosurveillancetechnology,troopsimultaneouslyandmultipleremotefromtheoperatorobjects.SpecialplaceamongsuchtechnologiesareintelligentvideosurveillanceIntelligentSurveil-lanceSystems,capableinautomaticmodetodetectthesituationandtheachievementoftheaimsofmonitor-ing.Suchsystemsdemand,primarilybecausetheorderstoreducetheuseofhumanresourcesastheoperatordo*Correspondingauthor.Tel.:E-mailaddress:[e-mail],[e-mail]itionofLicensePlatesusingNeuralNetworks27notneedtobeacontinuousviewdatafrommultiplecamerastoconfirmthatthesystemspecialsituations.“Subcategoryintelligentvideosurveillancesystemsarethesystemofautomaticlockingviolationsoftherulesoftheroad,capabletoidentifyviolationssuchasexcessivecollisionsdisplaysthespeedoftheplannedforthesolidorstoptheline,traveltoredlight.Inthisarticleisthejournalistsstudywithaviewtotheestablishmentofthedetectionunitvehiclestocorrectthefailureofthebenefitsofplannedunregulatedtransition[2].Identifyingsuchviolationsrequiressimultaneoustrajectorytrackingofpedestriansandvehiclesthatit'sharderthanlockingsystemcitedabove,itwasdecidedtostartbyexaminingtheapplicabilityofalgorithmsbasedonbackgroundsubtractionandfollow-uptofiltertheareasofprocessingandobjectrecognitioninimag-esisoneofthemostdifficulttasksinthefieldofinformationtechnology.Duetotheimportanceofthisissue,researchfacilityofrecognition,imageandspeechanalysisareincludedinthelistofprioritydirectionsofsci-ence,technologyanddevelopmentofFederalcriticaltechnologies.Theproposedsystemacquiresacarimagewithadigitalcamera.Snapshotimprovedbyremovingnoiseprocessingstagebyapplyingthemedianfilter.Contrastenhancementiscarriedoutonthefilteredimagetoreducevariouslightingeffectday.Alsoimprovedthetextandbackgroundcontrastregistrationnumber.Edgedetectorappliedonthetestimage,maximumamountandrejectingedgesfoundbyusingtheproposedmethod,whichgivesthelocationoftheplateatthelaststageoftheproposedgrowingwindowalgorithmisusedtoremovethefalsenumberplateareas.Aftermarkingplatesareaisdonefortheconvenienceoftheuse[3,4,5].2.RelatedWorksTheWorkwillbeheldModernmethodsofcharacterrecognitioninimagesareusedforawiderangeofscopeofworksuchastextrecognitionInfrastructureofImplementationThisworkhasbeensurfacesofvariouswork;CorrespondingblockdiagramoftotalStepsofthealgorithminthispartareshowninFig.1.thefollowingsteps:-Startand-Localization-CharacterandNumbersSegmentation-FeatureExtractionofSegmentedImage-RecognizetheExtractedFeatures-ShowtheLicensePlate-EndFig.1.CharacterSegmentationforExtractionofNumberPlates28AutomaticSystemRecognitionofLicensePlatesusingNeuralNetworksCurrently,thesetechnologiesareimplementedinthreetraditionalmethods,theyare:structural,featureandtemplatemethods.Eachofthesemethodsisfocusedonitsconditionsofusewhichitiseffectivefor.However,allthesemethodshavedisadvantages.Whenimagerecordingthegreatesttransformations,affectingtherecog-nitionresult,aremadebytheaffineandProjectivetransformationsoccurringduetochangeintheimagingan-glechangeofscaleandweatherconditions.Also,thepresenceofforeignobjectsinimageswithcomplexbackgroundsignificantlyreducethereliabilityoftherecognitionmethodsusedinmodernsystemsofca