基于BP神经网络的车型识别外文翻译.docx
- 文档编号:24828873
- 上传时间:2023-06-01
- 格式:DOCX
- 页数:26
- 大小:335.29KB
基于BP神经网络的车型识别外文翻译.docx
《基于BP神经网络的车型识别外文翻译.docx》由会员分享,可在线阅读,更多相关《基于BP神经网络的车型识别外文翻译.docx(26页珍藏版)》请在冰豆网上搜索。
基于BP神经网络的车型识别外文翻译
、外文资料
LicensePlateRecognitionBasedOnPriorKnowledge
Abstract-Inthispaper,anewalgorithmbasedonimprovedBP(backpropagation)neural
networkforChinesevehiclelicenseplaterecognition(LPR)isdescribed.Theproposedapproachprovidesasolutionforthevehiclelicenseplates(VLP)whichweredegradedseverely.
Whatitremarkablydiffersfromthetraditionalmethodsistheapplicationofpriorknowledgeoflicenseplatetotheprocedureoflocation,segmentationandrecognition.Colorcollocationisusedtolocatethelicenseplateintheimage.Dimensionsofeachcharacterareconstant,whichisusedtosegmentthecharacterofVLPs.TheLayoutoftheChineseVLPisanimportantfeature,whichisusedtoconstructaclassifierforrecognizing.Theexperimentalresultsshowthattheimprovedalgorithmiseffectiveundertheconditionthatthelicenseplatesweredegradedseverely.
IndexTerms-Licenseplaterecognition,priorknowledge,vehiclelicenseplates,neuralnetwork.
I.INTRODUCTION
VehicleLicense-Plate(VLP)recognitionisaveryinterestingbutdifficultproblem.Itisimportantinanumberofapplicationssuchasweight-and-speed-limit,redtrafficinfringement,roadsurveysandparksecurity[1].VLPrecognitionsystemconsistsoftheplatelocation,thecharacterssegmentation,andthecharactersrecognition.Thesetasksbecomemoresophisticatedwhendealingwithplateimagestakeninvariousinclinedanglesorundervariouslighting,weatherconditionandcleanlinessoftheplate.Becausethisproblemisusuallyusedinreal-timesystems,itrequiresnotonlyaccuracybutalsofastprocessing.MostexistingVLPrecognition
methods[2],[3],[4],[5]reducethecomplexityandincreasetherecognitionratebyusingsome
specificfeaturesoflocalVLPsandestablishingsomeconstrainsontheposition,distancefrom
thecameratovehicles,andtheinclinedangles.Inaddition,neuralnetworkwasusedtoincreasetherecognitionrate[6],[7]butthetraditionalrecognitionmethodsseldomconsiderthe
priorknowledgeofthelocalVLPs.Inthispaper,weproposedanewimprovedlearningmethodofBPalgorithmbasedonspecificfeaturesofChineseVLPs.Theproposedalgorithm
overcomesthelowspeedconvergenceofBPneuralnetwork[8]andremarkableincreasesthe
recognitionrateespeciallyundertheconditionthatthelicenseplateimagesweredegradeseverely.
II.SPECIFICFEATURESOFCHINESEVLPS
A.Dimensions
Accordingtotheguidelineforvehicleinspection[9],alllicenseplatesmustberectangularandhavethedimensionsandhaveall7characterswritteninasingleline.Underpracticalenvironments,thedistaneefromthecameratovehiclesandtheinclinedanglesareconstant,soallcharactersofthelicenseplatehaveafixedwidth,andthedistaneebetweenthemediumaxesoftwoadjoiningcharactersisfixedandtheratiobetweenwidthandheightisnearlyconstant.Thosefeaturescanbeusedtolocatetheplateandsegmenttheindividualcharacter.
B.Colorcollocationoftheplate
TherearefourkindsofcolorcollocationfortheChinesevehiclelicenseplate.Thesecolor
collocationsareshownintableI.
TABLEI
Categoryoflicenseplate
Colorcollocation
smallhorsepowerplate
bluebackgroundandwhitecharacters
motortruckplate
yellowbackgroundandblackcharacters
militaryvehicleandpolicewagonplate
blackbackgroundandthewhitecharacters
embassyvehicleplate
whitebackgroundandblackcharacters
Moreover,militaryvehicleandpolicewagonplatescontainaredcharacterwhichbelongstoaspecificcharacterset.Thisfeaturecanbeusedtoimprovetherecognitionrate.
C.LayoutoftheChineseVLPS
ThecriterionofthevehiclelicenseplatedefinesthecharacterslayoutofChineselicenseplate.AllstandardlicenseplatescontainChinesecharacters,numbersandletterswhichareshowninFig.1.ThefirstoneisaChinesecharacterwhichisanabbreviationofChineseprovinces.ThesecondoneisaletterrangingfromAtoZexcepttheletterI.Thethirdandfourthonesarelettersornumbers.Thefifthtoseventhonesarenumbersrangingfrom0to9
only.Howeverthefirstortheseventhonesmayberedcharactersinspecialplates(asshowninFig.1).Aftersegmentationprocesstheindividualcharacterisextracted.Takingadvantage
ofthelayoutandcolorcollocationpriorknowledge,theindividualcharacterwillenteroneof
theclasses:
abbreviationsofChineseprovincesset,lettersset,lettersornumbersset,numberset,specialcharactersset.
(a)Typicallayout
(b)Specialcharacter
Fig.1ThelayoutoftheChineselicenseplate
III.THEPROPOSEDALGORITHM
Thisalgorithmconsistsoffourmodules:
VLPlocation,charactersegmentation,characterclassificationandcharacterrecognition.ThemainstepsoftheflowchartofLPRsystemareshowninFig.2.
Firstlythelicenseplateislocatedinaninputimageandcharactersaresegmented.Theneveryindividualcharacterimageenterstheclassifiertodecidewhichclassitbelongsto,andfinallytheBPnetworkdecideswhichcharacterthecharacterimagerepresents.
Chinesecharacter
Fig.2TheflowchartofLPRsystem
A.Preprocessingthelicenseplate
1)VLPLocation
Thisprocesssufficientlyutilizesthecolorfeaturesuchascolorcollocation,colorcenters
anddistributionintheplateregion,whicharedescribedinsectionII.Thesecolorfeaturescanbeusedtoeliminatethedisturbanceofthefakeplate'sregions.TheflowchartoftheplatelocationisshowninFig.3.
Fig.3Theflowchartoftheplatelocationalgorithm
Theregionswhichstructureandtexturesimilartothevehicleplateareextracted.The
processisdescribedasfollowed:
Here,theGaussianvarianceissettobelessthanW/3(Wisthecharacterstrokewidth),soRgetsitsmaximumvalueMatthecenterofthestroke.Afterconvolution,binarizationisperformedaccordingtoathresholdwhichequalsT*M(T<0.5).Medianfilterisusedtopreservetheedgegradientandeliminateisolatednoiseofthebinaryimage.AnN*Nrectanglemedianfilterisset,andNrepresentstheoddintegermostlyclosetoW.
Morphologyclosingoperationcanbeusedtoextractthecandidateregion.Theconfidencedegreeofcandidateregionforbeingalicenseplateisverifiedaccordingtotheaspectratioandareas.Here,theaspectratioissetbetween1.5and4forthereasonofinclination.Thepriorknowledgeofcolorcollocationisusedtolocateplateregionexactly.ThelocatingprocessofthelicenseplateisshowninFig.4.
[vT【八"⑴小]
(f)Plateexttacting
(e>SrinctiireveHtlcation
(c)Medinnifillerinf
Fig.4Thewholeprocessoflocatinglicenseplate
2)Charactersegmentation
Thispartpresentsanalgorithmforcharactersegmentationbasedonpriorknowledge,usingcharacterwidth,fixednumberofcharacters,theratioofheighttowidthofacharacter,andsoon.TheflowchartofthecharactersegmentationisshowninFig.5.
Fig.5Theflowchartofthecharactersegmentation
Firstly,preprocessthelicensetheplateimage,suchasunevenilluminationcorrection,contrastenhancement,inclinecorrectionandedgeenhancementoperations;secondly,
eliminatingspacemarkwhichappearsbetweenthesecondcharacterandthethirdcharacter;thirdly,mergingthesegmentedfragmentsofthecharacters.InChina,allstandardlicenseplatescontainonly7characters(seeFig.1).Ifthenumberofsegmentedcharactersislargerthanseven,themergingprocessmustbeperformed.TableIIshowsthemergingprocess.Finally,extractingtheindividualcharacter'imagebasedonthenumberandthewidthofthecharacter.Fig.6showsthesegmentationresults.(a)Theinclineandbrokenplateimage,(b)theinclineanddistortplateimage,(c)theseriousfadeplateimage,(d)thesmutlicenseplateimage.
TABLEII
GetNf
IfNF>MaxF
Foreachcharactersegments
CalculatethemediumpointMi
Foreachtwoconsecutivemediumpoints
CalculatethedistaneeDk
Calculatetheminimumdistanee
Mergethecharactersegmentkandthecharactersegmentk+1
NF=NF-1
Endofalgorithm
whereNfisthenumberofcharactersegments,MaxFisthenumberofthelicenseplate,andiistheindexofeachcharactersegment.
Themediumpointofeachsegmentedcharacterisdeterminedby:
(3)
whereistheinitialcoordinatesforthecharactersegment,andSi2isthefinalcoordinate
forthecharactersegment.Thedistaneebetweentwoconsecutivemediumpointsiscalculatedby:
(4)
B.Usingspecificpriorknowledgeforrecognition
Fig.6The
segmentationresults
ThelayoutoftheChineseVLPisanimportantfeature(asdescribedinthesectionII),
whichcanbeusedtoconstructaclassifierforrecognizing.Therecognizingprocedureadopted
conjugategradientdescentfastlearningmethod,whichisanimprovedlearningmethodofBP
neuralnetwork[10].Conjugategradientdescent,whichemploysaseriesoflinesearchesinweightorparameterspace.Onepicksthefirstdescentdirectionandmovesalongthatdirection
untiltheminimuminerrorisreached.Theseconddescentdirectionisthencomputed:
thisdirectionthe“conjugatedirection”istheonealongwhichthegradientdoesnotchangeitsdirectionwillnot“spoil”thecontributionfromthepreviousdescentiterations.Thisalgorithmadoptedtopology625-35-NasshowninFig.7.Thesizeofinputvalueis625(25*25)andinitialweightsarewithrandomvalues,desiredoutputvalueshavethesamefeaturewiththeinputvalues.
InputXXIX2…Xi…K625
Fig.7Thenetworktopolog
- 配套讲稿:
如PPT文件的首页显示word图标,表示该PPT已包含配套word讲稿。双击word图标可打开word文档。
- 特殊限制:
部分文档作品中含有的国旗、国徽等图片,仅作为作品整体效果示例展示,禁止商用。设计者仅对作品中独创性部分享有著作权。
- 关 键 词:
- 基于 BP 神经网络 车型 识别 外文 翻译