智能交通ⅹ圖像閾值分割方法研究.pdf

1、南京航空航天大學碩士學位論文智能交通ⅹ圖像閾值分割方法研究姓名：潘喆申請學位級別：碩士專業(yè)：信號與信息處理指導教師：吳一全201012智能交通圖像閾值分割方法研究IIABSTRACTWiththeincreaseofviheclestheintelligenttrafficbecomestoanewresearchdirection.Applyingimageprocessingtechnologytointelligenttraffi

2、csystemisachallengingfieldwhichhasgreattheeticalsignificancepracticalvalue.Thispaperconcernsonthethresholdingsegmentationontheimagesofliceseplatevihecleroad.Firstlytheintelligenttraffictechinologyisintroduced.Onthebasiso

3、freviewingthemethodofimagethresholdingsegmemntationwepointouttheshtagesoftheexistingalgithmsproposeseveralnewalgithms.ThetwodimensionalShannon’sinfmationentropyisaclassicalcommonlyusedimagesegmentationmethodbuttherearest

4、illsomedisadvantagesinvolved.ThenatwodimensionalexponentinfmationentropyfthresholdionisproposedherewhichcouldovercomethedisadvantagesintheShannon’sentropy.Meaanwhileafastalgithmoftwodimensionalexponententropythresholding

5、methodisalsogivenwhichchangedthetwodimensionalthresholdintoonedimensional.Theresultsoftheexperimentindicatethattheproposedalgithmhashighspeedofcalculationgoodsegmentationquality.TheMinimumWithinClusterVariancealgithm(Ots

6、u)hasagoodsegmentationqualitywidesuitablescopewhichisactuallytheLeastSquaresalgithm(L2Nm).TwoalgithmfthresholdingareproposedinthispaperwhicharebasedonMinimumWithinClusterAbsoluteDifference(L1Nm)MinimumWithinClusterMaximu

7、mDifference(L∞Nm).thecrespondingtwodimensionalalgithmsofthosetwonewmethodsarealsopresented.Theresultsshowthatthosetwonewmethodshavemuchbetterperfoemancefsomekindsofimageseachofthetwodimensionalalgithmsisbetterthanitsowno

8、nedimensionalalgithm.Thresholdingalgithmbasedonentropyisoneofthemostfamousmethods.InthispapertwofastrecurringtwodimensionalRenyientropicthresholdingalgithmswhosecomputationalcomplexitiesarebothonlyO(L2)areproposedwhileth

9、ecomputationalcomplexityoftheiginalalgithmisO(L4).Experimentalresultsshowthatthesetworecurringalgithmscanbothgreatlyreducetheprocessingtimeofimageswhichislessthan0.1%oftheiginalalgithm.Baedontheobviouswrongsegmentationin

10、theexistingtwodimensionalhistogramverticalsegmentationmethodatwodimensionalhistogramobliquesegmentationmethodisproposed.ThenthefmulaitsfastrecursivealgithmofthemaximumShannonentropythresholdingarededucedbasedonthetwodime

11、nsionalhistogramobliquesegmentation.Experimentalresultsshowthattheproposedmethodmakestheinnerpartunifmtheedgeaccurateinthethresholdimageithasabetterantinoisepropertywiththeincreaseofitsspeed.Inthispaperthefishswarmalgith

12、mofswarmintelligenceisalsointroducedinimagethresholdingthetwodimensionalOtsuthresholdingalgithmbaesonthefishswarmalgithmisproposed.Comparitionwithsinglegeicalgithmtheelitiststrategygeicalgithmshowsthatthisalgithmcouldthe

13、bestthresholdaccuratelywithafasterconvergentspeed.KeyWds:intelligenttrafficimagethresholdingsegmentationexponententropyMinimumWithinClusteralgithmfastrecurringalgithmtwodimensionalhistogramobliquesegmentationfishswarmalg