基于图像处理的地铁隧道裂缝识别算法研究-王耀东

35720147ChineseJournalofScientificInstrumentVol.35No.7Jul．20142014-01ＲeceivedDate2014-01*M14JB00020、NCET-11-0572*王耀东1，余祖俊2，白彪1，许西宁1，朱力强21．1000442．100044。、、。。。。。TP242．8A520．20ＲesearchonimageprocessingbasedsubwaytunnelcrackidentificationalgorithmWangYaodong1，YuZujun2，BaiBiao1，XuXining1，ZhuLiqiang21SchoolofMechanicalElectronicandControlEngineeringBeijingJiaotongUniversityBeijing100044China2．StateKeyLaboratoryofＲailTrafficControlandSafetyBeijingJiaotongUniversityBeijing100044ChinaAbstractAsakindoftunneldefectscracksseriouslyaffectthesafetyofthetunnel．Itisessentialtodetecttunnelcrackseffectively．Howeverlowcontrastunevenilluminationandseverenoisepollutionexistcommonlyintunnelcrackimages．Astraditionalimagepro-cessingalgorithmscannotdetectthetunnelcracksanewalgorithmisproposedinthispapertodetectsubwaytunnelsurfacecracks．Thealgorithmutilizesthepreprocessingalgorithmcombiningglobalandlocalmethodssimultaneouslyandthemultiplestagedfilterpro-cessingbasedontheconnectedregions．Thepreprocessingalgorithmcanrestraintheinfluenceoflowcontrastandunevenillumination．Themultiple-stagedfilteringprocessingmethodbasedontheconnectedregionscaneliminatedifferenttypesofnoisesinthesubwaytun-nelimageswell．Thispaperalsoproposesanalgorithmbasedonthelocalmeanandstandarddeviationofthecracktocalculatethecrackwidth．Theexperimentresultsshowthatthisalgorithmcaneffectivelydetectandmarksubwaytunnelsurfacecrackfeatureareas．Keywordscrackdetectionimageprocessingconnectedregioncrackwidth1、、。。、、、1。、、。。2PCNN3149035245Chen“”67。。89。AＲT-2510Landstrm11Zhu12Oliveira213C14SIFT15。。。2。1。1Fig．1Flowchartofcrackextraction2．1、。。2。Ixy1Gxy=minIx+x'y+y'－Sx'y'|x'y'∈DS1Sx'y'DSSx'y'Gxy。。。。WH。W、H。。UixyVixy2Vixy=Uixy－minUixymaxUixy－minUixyi=12…K2K。01。Mi3Mi=1W·H∑x∑yVixy3。Lixy4λ=4。Lixy=11+MiVixyλVixy≠00Vixy={04。Qixy5。Qixy=Lixy×2555。71491。2OtsutiPxy。6。Pixy=0Qixy≤ti1Qixy＞t{i62．2。。。1CkxyPxypCkxyB3。7。Zj=Zj－1B∩Pxyj=012…N7ZZ0p。Zj=Zj+1Zj。。2。。nkNnFxy。Tn8Ckxy=1nk＜Tn0{k=12…Nn8nk=∑x∑yCkxy93。。FxyDkxy10Ｒk=∑x∑yDkxySM10SM。NＲYxy。TＲ11Dkxy=1Ｒk＞TＲ0{k=12…NＲ11422。YxyEkxywmaxhmax2。23。2Fig．2SpecialshapednoisesNw3Tw、ThTr12Ekxy=1wmax＞Twhmax＞ThＲk＞Tr0{12k=12…Nw。3。3。3．1。Zhang16。。01。3。3Fig．3Schematicdiagramofthinningalgorithm1492353×3P1ijP2P3…P94。12≤ZP1≤62TP1=131－P2×1－P4×1－P8=0TP1≠141－P2×1－P4×1－P6=0TP4≠1ZP1P2P3…P90TPP801。3．2。。。。3。1。。。3×33。8。。4○0×1。3×3。4Fig．4Templatesofendpointdetection2Nxy。1317。L=ANe+BNo+CNc13NeNoNc。A=1B槡=2C=0。3Tl。14Nkxy=1Lk＜Tl0{k=12…NT14NT。1155。5Fig．5Blurremovingresults3．3。。2。1。2。6。2－16。6Fig．6Calculationofthenormallinesatcrackpoints71493。。。5×5。2Ixy。2。DA15A=Ix0y0Ix1y1…IxDyD…Ix2D－1y2D－1Ix2Dy2D15IxDyDxiyi。abrabprmrmμσ16、17μ=∑bm=armprm16σ=∑bm=arm－u2prm槡177ab。7Fig．7Crackwidthmeasurementpk118pk1=IxD－k1yD－k1k1=01…D18k10Dk1。pk1≤μ－σpk1+1＞μ－σk1=01…D19pk220pk2=IxD+k2yD+k2k2=01…D20k20Dk2。pk2≤μ－σpk2+1＞μ－σk2=01…D2121width=k1+k22248。8bMask18NDHM198c。8Fig．8CrackextractionalgorithmcomparisonMask。。。9149435。。2112．78362．4168。9Fig．9Crackwidthcalculation1。10。Mask。10cMask-NDHM。10Fig．10Crackextractionalgorithmcomparison。。。11。11。11Fig．11Crackwidthcalculation50×50Tn=300TＲ=0．3Tw=20Th=30Tr=0．18D15。1150。。1216。。1235。121、2、4、7。71495。12Fig．12Crackextractionresults5。。。。。1。。。1．J．20051103-106．LIX．AnalysisofthecausesofcracksintunnelliningJ．ＲailwayConstructionTechnology2005Suppl．1103-106．2．J．20124819163-166219．LIWZHUPZH．ImagedetectionalgorithmresearchforasphaltpavementcrackJ．ComputerEngineeringandApplication20124819163-166219．3．J．2012333637-642．ZHAOHJGEWQLIXD．Detectionofcrackdefectbasedonminimumerrorandpulsecoupledneuralnet-worksJ．ChineseJournalofScientificInstrument2012333637-642．4．J．201339176-82．WANGXMFENGXDANGJW．Pavementcrackdetec-tionmethodbasedonmulti-imageandmulti-resolutionJ．JournalofLanzhouUniversityofTechnology201339176-82．5．J．201330103121-31233132．DONGANGLIANGMM．CrackdetectionalgorithmbasedongraycorrelationJ．ApplicationresearchofComputers201330103121-31233132．6CHENCWANGJZOULetal．Anovelcrackdetec-tionalgorithmofunderwaterdamimageC．SystemsandInformaticsICSAI20121825-1828．7．J．201318169-77．XUWTANGZHMLVJY．PavementcrackdetectionbasedonimagesaliencyJ．JournalofImageandGraphics201318169-77．8．J．20123381850-1855．YANGSSHAOLTGUOXXetal．SkeletonandfractallawbasedimagerecognitionalgorithmforconcretecrackJ．ChineseJournalofScientificInstrument20123381850-1855．9．149635J．201031102260-2266．XUZHGZHAOXMSONGHSH．AsphaltpavementcrackrecognitionalgorithmbasedonhistogramestimationandshapeanalysisJ．ChineseJournalofScientificIn-strument201031102260-2266．10．AＲT-2J．20093071420-1425．LIUGHJIANGZHJ．ＲecognitionofporcelainbottlecrackbasedonmodifiedAＲT-2networkandinvariantmomentJ．ChineseJournalofScientificInstrument20093071420-1425．11LANDSTＲMATHUＲLEYM．Morphology-basedcrackdetectionforsteelslabsJ．IEEESignalProcess-ingSociety201211866-875．12ZHUZGEＲMANSBＲILAKISI．Visualretrievalofconcretecrackpropertiesforautomatedpost-earthquakestructuralsafetyevaluationJ．AutomationinConstruc-tion2011207874-883．13OLIVEIＲAHCOＲＲEIAPL．AutomaticＲoadCrackDetectionandCharacterizationJ．IEEEIntelligentTransportationSystemsSociety20133155-168．14．FCMJ．201349431-34．SONGBBWEIN．FCM