結(jié)合顏色、紋理和先驗(yàn)形狀的車輛檢測(cè)技術(shù)的研究.pdf

1、浙江大學(xué)信息科學(xué)與工程學(xué)院碩士學(xué)位論文結(jié)合顏色、紋理和先驗(yàn)形狀的車輛檢測(cè)技術(shù)的研究姓名：趙璐申請(qǐng)學(xué)位級(jí)別：碩士專業(yè)：信息與通信工程指導(dǎo)教師：于慧敏20080501浙江大學(xué)碩士學(xué)位論文AbstractWiththedevelopmentofcityandeconomyandtheimprovementofthepeople’Slivingstandard，thenumberofvehiclesontheroadincreasedquick

2、lyandtheburdenoftheroadtransportationsystembecomeshigherandhigherTherefore，moreattentionhasbeenpaidtothevision—basedintelligenttransportationsystemInthispaper，wedosomeresearchesonvehicledetectionwhichisthekeytechnologyof

3、theintelligenttransportationsystem，andproposealgorithmsonbackgroundextractionandupdate，andshadowsuppressionResearchworksinthispaperaresummarizedasfollows：Firstlyweproposeabackgroundextractionandupdatemethodbasedonblockfe

4、aturesbyimprovinganexistingalgorithmThemethoddividesimagesintoblocks，andcomputesstatisticallikelihoodforeachblockinthetimedomainBlockswitllslightchangeareclassifiedasthestaticblocksandputintoabufferTheseblocksareusedtoex

5、tractandupdatethebackgroundimageComparedwiththebackgroundmodelbasedonpixel，thismethodCanreducethecomputingcomplexsignificantlyandCanalsogetgoodresultSecondlywepresentanalgorithmtodovehiclesegmentationandcastshadowremoval

6、usingthecolorandtextureinformationInthefirststep，themethodusescolorinformationtodetectshadowsinobjectsInthesecondstep，texturefeaturesareusedtofindedgesofvehiclesandshadowsThenwecombinecharactersoftheshadowedgesandtheresu

7、ltinthefirststagetemovetheshadowedgesExperimentalresultsshowthatthemethodCancopewithshadowsinanydirectionsButforafewofvehicles，thismethodstillCan’tobtainthewholecontourFinallyavehicledetectionmethodbasedonthepriorshapekn

8、owledgeisproposedThismethodusestheresultoftheshadowremovalalgorithmastheinitialcontourandrestoresthevehiclecontour、析tllpriorshapeknowledgeAnimplicitshapemodelisbuiltandanactivecontourenergyfunction、析t11therestrictionofth

9、eexistingshapepriorsisconstructedThenweapplytheshapealignmentandlevelsetmethodtoevolvingtheinitialcontouruntilconvergenceResultsintheexperimentsshowthatwecanovercomethedefectoftheshadowremovalalgorithmusingthismethodandg