车牌识别外文翻译文献综述.docx

《车牌识别外文翻译文献综述.docx》由会员分享，可在线阅读，更多相关《车牌识别外文翻译文献综述.docx（5页珍藏版）》请在第一文库网上搜索。

1、车牌识别外文翻译文献综述pfe1ement1icenseP1ateRecognitionBasedOnPriorKnow1edgeAbstract-Inthispaper、anewa1gDri1hmbasedonimprovedBP(backpnopagation)neura1networkforChinesevehic1e1icensepEaterecognition(1PRisdescribed.JhcproposedapproachprovidesaSo1u1icnifcrthevehic1e1icensep1a1e(V1Pwhichweretiegradedseverity.WhaEi

2、1nMnarkab1ydiffersfromthetiadi1iona1methodsisEhcapphcadonofpriorknow1edgeof1icenseph1ctotheprocedureofIoca1ioti,SiigmentaiiotandneeoignitioTi.CohrCu11acd1ionisuedto1ocate1he1icensep1ate?in1hcimage.DiinensionsofeachC1iarat1crarecorih1ant.whichisuacd.tosegmenttheCharae1erofV1Ps.The1ayuuioftheChineseV1

3、Pisanimportantfeature,whichisusedIOconstructadEassiGerfbrnxDgnizing.TheCxpcrimcn1d1resu1bshtnvthatTheimprovedaigcrithmisef1ectiveundcr1hecondi1ioEthat(he1icensep1aEtsweredegradedsevere1y.IndexTcrmi-1icensep1aterecognition,priorknowkiige.vehic1e1icensep1ates,neura1network.1INTRODUCTIONVehic1e1iccf1se

4、-P1ate(V1P)reccgni1iouisaxcryinterestingbutdi1cuhprob1cin.Itisimportantinanumberofapp1icationssuchaswcigh1-and-speed-1imitredraff1einfrirtgementruadsurveysandparksceuri1y(U.V1Precognitionsystemconsistsoftheph1cIucattcn,thecharacterss(Tncn1a1ionandthecharactersrecogni1icjn.Thesetasksbecomemarcsephiti

5、eacedwhendea1ingwithp1a1cimagescakcninvaricusinc1inedang1esOTundervarious1ighting,weatherCondi1iotiandc1ean1inessofIhep1aitsBecausethisprob1emusttaHyusedinrea1-timesys1cms1Unequircano1on1yaeeuFseybuta1sofhs1processing.MostexistingV1Preccgni1ioHmethadi2.3,4.5reducethecomp1exityandincreaseIherecogniti

6、onratebyusingsomespecificica1urcsofIotaiV1Psandestab1ishingsomeconstrainsontheposi1iondistancefromthecameraIovehic1es,andtheinc1inedangks.In3ddi1iohneura1ne1workwasusedtoincreasetheFctogni1icinra1e6J.7butthetraditiona1recogni1ionmethodsse1domconsiderthepriorknow1edgeoftheteea1V1Ps.Iethispaper,wcprop

7、osedartewimproved1earningmethodof13Pa1gorithmbasedonspecificfeaturesofChineseV1PS1heproposeda1goriihmovercomesthe1ewspeedconvergenceofHPtieuraJnetwork8andremarkab1eificneascsthereccgnitionra1etspecta11yundertheconditionIha1the1icensep1ateimagesweredegradesevere1y.I1SPECIFICFEATURESOfCH15J1-SEV1PSA.D

8、imensionsaxesoftwoadjoiningetiaractersisfixedandEheratiobetweenwidthandheightis;near1yeunstant1TheseIcaturescanbeusedto1ocatethep1atecindsegmenttheindividua1character.Accordingtotheguide1ineforvehic1einspection9,a11Iitctisjepia1csmustbereciangii1arjudhaveIhediHicMisionbiandhaveajj7Ctidracterswrit1ef

9、tinasing1e1ine.Underp血Eica1Cnviranrncn1s*thedistanceirom1IhecameraIovehic1esandtheinc1inedang1esarcconstant,soa11charactersofthe1icensep1atehaveafixedwidth,andthedistancebetweenthemediumB.Co1orCoRoca1iotiofthep1ateTherearefourkindsofco1oreo1iocationfor1heChinesevehic1e1icensep1ate-Theseco1oreciiocat

10、ionsareshownintab1e1.TAb1E1CategoryufIiccnstp1ateCo1orCo1Iocaitioiisma11horsepowerp1ateb1uebackgnoimt1sint1whi1echaractersmci1cjrtruckp1ateyd1owbackgroundandbEaekChaiBetertimdiEaryvthideandpo1icewionp1ateb1ackbackgroundandIhewhitecharactersembassyvehic1ep1atewhitebackgroundandb1ackcharactcriMoreover

11、,mjIharyvehic1eandpo1icewagonp1atesccn1ainaredcharac.与toaspecificCharad1erset.Thisfeaturecanbeusedtoimprovetherecognitionratbook1IBComC.1ayoutoftheChineseV1PSTtJisJS?X3TheeriIerionofIhevehic1e1icensep1a1edefines;theehamc1ers1ayoutofChineseHccrtscp1ate.A11standard1icencecontaitiChintsccharacters,numb

12、ersandk1tcrswhichareshcftvninFig.1ThefirstoneaChinesecharacterwhichisanabhreviaticnofChineseprovinces.ThesecondoneisaJe1ttrranging(roinAEGZexceptthe1etterI-ThethirdandfourthonesarcIc11ersornumbers.Thefifthtoiseventhonesarenumbersrangingfrem0o9on1y.HowevertheHtxtortheseventhernesmayberedchanicteninsp

13、ecia1p1a(asshowninfig.IAfiergmenta1ionprotesttheindiviJua1ehatae1erisextracted.akicigadvantageofIheIayoki1andco1orco11ocationpriorknow1edge,theindhidua1charetdwi11enteroneofthedashes：abbreviationsofChineseprovincessek1etters1exersornumbersset,numbersetspecia1charactersse1.oftheChinese1icensep1ateII1

14、THEPROPOSEDA1GORITHMThistitgorithmconsistsoffourmcdu1eti：V1PNation,charactersegmentation,charac1erc1ashHeadonandcharacterrceognitiGn.Themainstepsafthef1owchartof1PRsys1emareshowninFigj2.Firi1Irthe1icensep1ateisioca1edinaninputimageandcharactersaresegmented.Theneveryindividua1characterimageentersthec

15、1assifiertodecidewhichctassitbe1ongs(,andfina11yIhcBPnetworkdecidetiwhichehaiacterthecharacterimagerepresents.A.Preprocessingthe1icensep1ateI)V1P1oca1ionanddistnbu1iauinthep1ateregion,whichThisprocesssufficient1yuti1izesIheco1cnrIeaturesuchasco1orCoU(itaiun_co1orcen1ersadescribedinsectionII.Teseco1o

16、rfeaturescanbeusedtoetimina1ethediurbaneeofthefakep1atesregions.Thef1owchartofthep1atehcaticnisshowninFig.3.ITieregionswhichStnictiincandtcjcturesimi1artcthevehic1ep1atearcextracted.Theprincessisdescribedasfo11owed:P1=/“方jiP,th原创力文档Herc。theGaussiativarianceIrissetEobeEcssthanW/3(Wisthecharacteri下栽高清无水印io八getsitstnaximumva1ueMa1thecert1crofthestroke.Aft