欢迎来到天天文库
浏览记录
ID:33668106
大小:17.67 MB
页数:60页
时间:2019-02-28
《复杂环境下交通标志的检测》由会员上传分享,免费在线阅读,更多相关内容在学术论文-天天文库。
1、万方数据山东大学硕士学位论文ABSTRACTInrecentyears,intelligenttransportationsyStem(IntelligentTransportationSystem,ITS)hasarousedmoreandmoreinterestandattention.Thesystemcanreducepeople'sdrivingpressure,makepeopletravelmorefreedom,security,andreliability.Asanecessa
2、rypartofITS,TrafficSignsDetectionandRecognitionSystem(TrafficSignsDetectionandRecognitionSystem)isareliableguaranteeforintelligentvehicleordrivergettrafficinformationoutside.Trafficsignsdetectionisthekeytoidentify,real—timedetectionofregionofinterest
3、accuratelywilllayagoodfoundationforidentification.Fortheimagevehiclecapturedundertherunningstateint,hecomplexexternalenvironment,itrequirespreprocessingstagetoensuretheaccuracyofdetection.Thispaperfirstaimsatthemotiondeblurringandfogdehazingandthenth
4、edetectionofsignsTheproblemofimagedeblurringdividedintotwoclasses,uniformlinearmotionblurrestorationandnonuniformlinearmotionblurrestoration.Fortheformer,introducesthecommonmethodtoestimatethedirectionandscaleoftheblurkernelandthebasicalgorithmofimag
5、erestoration.Forthelatterweimprovetheblurkernelestimationusingstrongedgealgorithm.DenoisingwithguidedfiltersandoperationinR,G,Bthreechannel,ThealgorithmCandealdirectly埘thcolorimagesAtthecostofincreasedlesscomputation,Therestorationresultkeepstheimage
6、colorinformation.Fortheproblemofimagedehazingthispaperadoptsthelatestresearchresultsofimagedehazingarea,dehazingusingDarkChannelPriorPriori.IntheaspectsofoptimizingtransmissionfunctionputforwardthemethodofcombiningguidedfilteringandLinearinterpolatio
7、nAfternaturalimagespreprocessing,linksintothetrafficsigndetection,thispaperisdividedintotwosteps:(1)Forthered,blue,yellowthreecolorsestablishtwodimensionalnormaldistributionmodelfortheIcomponentandQcomponentinYIQcolorspace,III万方数据山东大学硕士学位论文calculatet
8、heconformityofapixelforthethreemodelstocompletethesegmentationofcolor,Thenthemorphologyprocessing,getthetwovalueimageastheinitialinterestregions(2)SelectHumomentinvariantsfeatureswhicharerobustnesstoshapesize,rotationandangle.GettheHumomentinvariants
此文档下载收益归作者所有