路面破损图像的边缘检测研究-控制科学与工程专业论文.docx

THEALGORITHMRESEARCHOFROADSURFACEDISTRESSIMAGEEDGEDETECTIONABSTRACTAsanimportantcontentofdigitalimageprocessing,thetechnologyofedgedetectionbas beenwidelyusedintheautomaticdetectionandidentificationofroadsurfacedistressimage. Theprocessofroadimageedgedetectionincludesthestepsofimagepreprocessing,edge detection,edgeconnectivityandrefinement.Thispaperfirstlyintroducestheimage pre-processingmethod,andthenintroducesthetraditionaledgedetectionmethods.These methods,whichisgotthelongestresearch,getstheedgeaccordingtothevarietyofthepixel gray.ThemaintechniquesareRobert,Sobel,Laplace,andLogalgorithm.Thesetechniquesare simpleandeasytooperate,butthey aresensitivetonoise.Especiallyforroadimagescontaining alargenumberofbackgroundtextures,stainsandcracks,thesetechniquescannotproducegood effect.Basedontheanalysisoftheclassicaledgedetectionalgorithm，Cannypresentedtheverypopularthreecriteriaofdetectors:gooddetection,goodlocalizationandonlyoneresponsetoa singleedge.Andalsobasingonthethreecriteria,hedevelopedtheoptimaledgedetection operatorsimilartothefirstderivativeofGaussianfunction.AlthoughtheCannyalgorithmhasa goodtheoreticalbasis,theclassicalCannyalgorithmalsohasmanydrawbacks.Inthe applicationofthisalgorithmforedgedetection,afterNon-MaximalSuppression(NMS),we needtosetdoublethresholdstodetectandconnectedges.Atpresent,thedoublethresholdsare selectedmainlybyartificialmethods,whichisalackofobjectivityandefficiency.Inorderto avoidmanualinterventionandtoimprovetheefficiencyofthealgorithm,manyameliorative methods areput forward.Forexample, wecandeterminethedoublethresholds through thegrads histogram. But the ratio factors in this algorithm need to bedetermined byartificial methods.Inthispaper,weintroduceandproposeimprovementmeasuresforthe drawbacks of Cannyalgorithm.Thek-meansalgorithmisoneofthemostwidelyusedclusteringalgorithm andthisalgorithmiscorrespondinglyscalableandefficient;Fuzzyc-means(FCM)clustering algorithmisthebestknownandpowerfulmethodinfuzzysegmentationsmethods.However,ii