英文文献.docx

英文文献.docx

《英文文献.docx》由会员分享，可在线阅读，更多相关《英文文献.docx（12页珍藏版）》请在冰豆网上搜索。

英文文献

TheTechnologyofVehicleIdentification

BasedonVideo

IntelligenttransportationtechnologyoverviewoftheIntelligentTransportationSystem（IntelligenceTransportSystem,ITS,）istheworldforefrontofthefieldoftransportationresearchprojects,combiningelectronicandinformationtechnology,communicationstechnology,automaticcontroltheory,computertechnologyandtraditionaltrafficengineeringtheoryandmanyotherthetheoryofthesubject,andusedinmoderntransportmanagementsysteminordertoachievetransportservicesandintelligentmanagement.Thetrafficmonitoringsystemisanimportantpartofintelligenttransportationsystems,thissubsystemismainlyresponsibleforthecollectionofroadtrafficflowparameters,suchastrafficvolume,speed,models,queuingtimeandlength.Currently,thedetectionmethodoftheroadparameters,ultrasonicdetection,infrareddetection,inductionloopdetectionandvideo-baseddetection.Ultrasonicdetectionaccuracyisnotsusceptibletotheblockofvehiclesandpedestrians,thedetectiondistanceisshort;infrareddetectionbytheheatsourceofthevehicleitself,andanti-noisecapabilityisnotstrong,andthusthedetectionaccuracyisnothigh;althoughtoasensecoildetectionaccuracyrelativehigher,buttherequirementssetinthepavementstructure,andontheroaddamage,constructionandinstallationisrelativelyinconvenient,lifeisrelativelyshort,easytodamageandothershortcomings.Inrecentyears,computertechnology,imageprocessing,artificialintelligenceandpatternrecognitiontechnologycontinuestoevolvebasedonvideodetectionmethodinthedetectionoftrafficflowhasbeenmorewidelyused,relativetoothertrafficflowdetectiontechnology,Ithasthefollowingadvantages:

1.Videotestcandetectthetrafficscenearea;

（2）relativetootherdetectionmethods,investment,lowcost;videosensorsandotherdevices,suchascamera,easeofinstallationandcommissioning,andontheroadthefacilitydoesnotproducedamage;usingvideodetectiontechnologycanbecollected.moretrafficflowparameters.

Theintroductionofrelatedtechnologiesandimagepreprocessing,thebackgroundextractionusingbackgroundsubtractionmethodforvehicledetection,ingeneral,notcollectedthewholelotofvideoimageprocessing.Here,thedestinationaccordingtothelane,inthewholeimageinafewareasofinterest（alsoknownasvirtualcoils）,andregionalreal-timeprocessing,detectionofthevehicleandtrafficflowparameters.Continuousvideoimagetoextractanumberofframes,andthepixelgrayvalues​​inthevirtualcoilframeserialnumberstoredinorderinthearray,eachpixelinthevirtuallooppointbypointtofindthegrayvaluehistogram,selectthehighestnumberofgrayvalues​​asthebackgroundimagepixelgrayvalue.Ingeneral,selectthemostfrequentlyoccurringgrayvalueofeachpixelasabackgroundimagecorrespondingtothepixelgrayvalueisveryreasonable,Ifyouhitavehicleintensive,canbeappropriatelyincreasedthenumberofframescollected,inordertogetamoregoodresults.Thebackgroundextractionrequiresaninitializationprocessintheextractionprocessinthebackground,nottoreadintothevideoimagevehicledetection.Thebackgroundupdatingfromtimetotimebackgroundsubtractionandbackgroundupdatingmethod:

setatimerintheprogram,fromtimetotime,theprogrambegananewroundofbackgroundextractionintheextractionprocessinthebackground,theprogramvehiclestesting.However,thistimewiththebackgroundforthelasttimetoextractthebackground,Extractafterwhenthecurrentcontext,coveringthebackgroundwiththecurrentbackgroundimagenextframeofthevehicledetectionusingtheupdatedbackground,thecurrentbackgroundfortesting.Thismethodcaneffectivelysuppresstheslowchangesoflightandnaturalconditions,andtoimprovethebackgroundsubtractionalgorithmtodetecttheeffectofthevehicle.Whenthevehiclepassesthroughthevirtualcoil,virtualcoilwithintheabsolutedifferencebetweenthecurrentgrayscalevalueofallpixelsandthebackgroundofthecorrespondingpixelgrayvaluesumofthechangeprocess,andtheimageintensityofallpixelsinthecoilthesumofabsolutedifferencebetweenthevalueofthebackgroundimagegrayvaluevariables.Virtualcoilnovehiclepassesthroughthecase,thevirtualcoilcurrentimageinformationremainedrelativelyconstant,smallchangesofthegrayvalueofpixelswithinthecoil.Then,whenthevehiclebegantoenterthevirtualcoilregion,duetothehugedifferencebetweenthepixelgrayvalueofbackgroundandvehicleswillleadtograduallyincreasing;vehiclestoleavethetimevalueofthevirtualloopareawillalsograduallydecrease.Whenthevehicleleftthetimevaluebecomessmall.Thisapproachmayhaveshortcomingsandweaknesses:

framedifferenceisused,threeinarow22differential,althoughthismethodhasastrongself-adaptive,butthedifferenceofsuccessiveframestochoosethetimingrequirementsofhigh,butalsodependsonthevelocityofmovingobjects,ifthemovementspeedisfaster,andtheselectedtimeintervalistoolong,itwillresultincoverageareabetweentwomovingobjects,whichcannotbedivided;Ifthemotionistooslow,andthetimeselectedistoosmallwillcauseexcessiveoverlap,theworstcase,theobjectisalmostcompletelyoverlap,thereisanobjectnotdetected.

Computerimagefilteringtwocategories:

aclassmethodinthespatialdomainprocessingavarietyofprocessing:

theimageintheimagespace;othermethodsofspaceimagesafterthechanges,suchasFourier-transform,avarietyofprocessinginthefrequencydomain,andthenchangebacktotheimagespacedomain,theformationoftheprocessedimage.FrequencydomainprocessingmethodsFouriertransformandinversetransform,andavarietyofwavelettransformandinversewavelettransform.Thesemethodsusethecomputermemoryandcomputingtimeisexpensive,notsuitableforreal-timesystemoftheintelligentvehicle.Therefore,thespatialdomainmedianfilteringapproach,thewaysalocalaveragesmoothingtechnique,pulseinterferenceandimpulsenoisesuppressioneffect.Undercertainconditions,toovercomethelinearfilter,suchasminimummeansquarefilterandmeanfilterimagedetailisblurred,theeffectiveprotectionoftheedgesoftheimage.Thestatisticalcharacteristicsdonotneedanimageinactualoperation,sothisisagreatdealofconveniencefortheimagepreprocessing.Imagepreprocessing-edgeenhancementbasedonvisualtheoryshows:

identificationofanobjectisfromtheedgeofthebeginningofanimagedifferentpartsoftheedgeisoftenthemostimportantfeaturesofthepatternrecognition.Theedgeofthesurroundingpixelgrayscalestepchangeorroofchangesthesetofpixels,whichwidelyexistsbetweenobjects,betweenobjectsandbackground,betweentheprimitiveandprimitiveintheimagecollectedbythemachinevisionsystemtheedgeofthelaneinformationislostinthebackgroundbetween.Theedgeenhancementisaimedathighlightingtheedgeoftheroad,inordertofacilitateroadboundaryidentification.

Inaddition,theedgeenhancementalgorithmalsohelpovercometheeffectsoftheroadunevenillumination.CommonlyusedinedgeenhancementoperatorRobertoperator,Sobeloperator,Krischoperator,Prewittoperator,theLaplaceoperator.Here,edgedetectionusingSobeloperator,itisactuallyafirst-orderdifferentialoperator,whichcaneffectivelyeliminatemostoftheuselessinformationintheroadimage.Discrete-BCDFalgorithmisdefinedasthefollowingformula$ofSobeloperatorwithastrongabilitytosuppressthenoise,infacttheessenceoftheSobeloperatortoreflectadjacentorsomedistanceawayfromthepixelgray-scaledifferencesincharacteristics.Roadintermsofthephysicalpropertiesclosetonormalontheroad,andthelightisgenerallyuniform.Inthiscase,thegrayvalues​​ofneighboringpixelsorlessafteraSobeloperator,putthisclosetoextentintoacertainvalue,thevalueisusuallyclosetothevalue.ThethesisoftheroadboundaryandotherpartsoftheroadwithacertaingrayleveldifferencesofSobeloperatortohighlighttheborderpixelvaluedifferences,relativetootherroadpartofthevalue.Whiletheedgeofthegradientdirectioninformation,andthealgorithmissimple,easytoimplement.AftertheimageaftertheSobeloperatoroperation,borderstandoutfromthewholeimage.Imagepreprocessing-Binarizationimagebinarizationprocessingkeyisreasonableselectionofthethreshold,thethresholdissettooeasytoproducenoisethresholdissetlowerresolutionoftheGeneralAssembly,non-noisesignalisconsiderednoisefilteredout,takingintoaccountthegeneraltestinamoreuniformlightorstrongchangesinlightconditions,soselecttheoveralloptimalthresholdmethodforimagebinarizationprocessingprincipleoftheoveralloptimalthresholdisastatisticaleachpieceofimagegraydistributioncharacteristics,theuseofcategoryvarianceasacriteriontoselectthemaximumvalueofbetween-classvarianceastheselectedthresholdbinaryimageusingtheoptimalthresholdalgorithmtofurtherreducenoiseforsubsequentHoughtransformrelativelycleanimagedata,imagepreprocessing-cameracalibrationandsetupvideocamerascanbeinstalledontheoverpassonthehigh-risebuilding,orasufficientlyhighpole.Asforthecamera'sperspectiveisdeterminedinthecalibrationtime.cameratheconfigurationfromthetwo-dimensionalspacetothree-dimensionalimagemappingtheactualacquisitionofthree-dimensionalimagethree-dimensionalcoordinatesinthetwo-dimensional