首页 Vision-Based Path Generation Method.pdf

Vision-Based Path Generation Method.pdf

Vision-Based Path Generation Me…

上传者: 流浪客老爷 2013-12-25 评分1 评论0 下载0 收藏0 阅读量783 暂无简介 简介 举报

简介:本文档为《Vision-Based Path Generation Methodpdf》,可适用于高等教育领域,主题内容包含VisionBasedPathGenerationMethodforaRobotBasedArcWeldingSystemTheodorePPach符等。

VisionBasedPathGenerationMethodforaRobotBasedArcWeldingSystemTheodorePPachidisJohnNLygourasReceived:JuneAccepted:AugustPublishedonline:October#SpringerScienceBusinessMediaBVAbstractInthispaperavisionbasedintegratedmethodintendedforpathgenerationforarobotbasedarcweldingsystem,ispresentedThedescribedsystemiscomposedoftherecentlydevelopedpseudostereovisionsystem(PSVS)oranordinarystereovisionsystemandtherelatedsoftwareAdesiredpathcanbegenerated,usingapartortheentireedgeofanimagecapturedfromasceneoftheroboticenvironment,alinemanuallydesignedintheimage,acombinationoflinesofthepreviouscasesorlinesbelonginginsuccessiveimagescapturedfromdifferentscenesAusercaninitiallyprocessimagesselectingbymeansofpulldownmenusavarietyoffilters,edgedetectionmethodsandoperationsThenthedesiredpathasacombinationoflinesisselectedfromimagesApplyingourcorrespondencealgorithm,correspondingedgescanbefoundFinally,anumberofsuccessivepathpointsarecalculatedbymeansoftheproposedpathpointcalculationalgorithmInonlineoperation,thevisionsystemmountedontheendeffectorcancaptureimageswiththedesiredbestview(weldingview)ofascenebymovingorrotating(usingpushbuttons)theendeffectoroftheroboticmanipulator–PUMAOtherfacilitiesofthedescribedsystemaretheselectionofavarietyofcolorsandshapes,histogramview,desiredmagnification,systeminformationandautomaticexecutionofuserselectedoperationsThegraphicaluserinterfaceisdevelopedinVisualC,itrunsinapersonalcomputerandcommunicateswiththeroboticmanipulator(PUMA)throughALTERcommunicationportKeywordsarcweldinggraphicaluserinterfaceimageprocessingpathgenerationpseudostereovisionroboticmanipulatorJIntellRobotSyst():–DOIsyTPPachidis(*):JNLygourasDepartmentofElectricalandComputerEngineering,DemocritusUniversityofThrace,VasSofias,Xanthi,Greeceemail:patedotenetgrJNLygourasemail:ilygoureeduthgrIntroductionAroboticmanipulatorcouldbeusedinanarcweldingsystemInasystemlikethismanyproblemsmaybeencounteredSomeofthemare()theaccuracyofmovementofatorchmountedontheendeffectoroftheroboticmanipulator,()thegenerationofaproperpaththetorchmustfollow,()theflexibilityofthesystemtoadaptthespecifiedpathtoanychangesinshapeofmetalpiecestobewelded,()thepossibilityofthesystemtochangethemovementrate,orthepatternofmovementrelativelytometaltypesorweldingsurfacesshapeTheaboveproblemscouldbetterbemanipulatedwiththeaidofastereovisionsystemandatorchtogethermountedontheendeffectoroftheroboticmanipulatorAstereovisionsystemcouldmeasurewithbetteraccuracythedesiredpathpointsandadaptaninitiallyspecifiedpathtoanychangesofthemetalpiecestobeweldedInsuchasystemtheinitialpathgenerationmaybequiteeasyandfast,withgoodaccuracyThepreviousfeatureswillpermitarcweldingofdifferentinshapeorinmaterialmetalpiecesandtorepeattheprocedurewhenitisnecessarySomeresearchersinrelevantpapersproposemethodsfortheinitialpathgenerationIn–,AlesUdeandRudigerDillmanemployedthe“teachingbyshowing”programmingparadigmAuserspecifiesthedesiredtrajectorybymovingtheobjecttobemanipulatedwithhishandTheperformanceismeasuredwithastereovisionsystemIn,BBrunneretalproposedthesocalledTeleSensorProgrammingconceptasaneasywaytoprogramarobotofflinevialearningbyshowinginavirtualenvironmentRMVoylesandPKKhoslainproposedagesturebasedprogrammingmethodofarobotIn,JohnsonandMarshpresentedthebasicmathematicalandcomputationalframeworkforanewmethodofmodelingrobotmanipulatorworkspacesInhisthesis,Sabesstudiedtheplanningofvisuallyguidedarmmovementsintwocases:feedbackperturbationandobstacleavoidanceIn,LiFangGustudiedthevisualguidanceofarobotbymeansofastereovisionsystemandusingasfeaturescornerpointsPaschkeandPauliinimplementedtheprogrammingbydemonstrationmethodtoreconstructasmoothDtrajectoryofagripper,usingastereovisionsystemIn,RufandHoraudproposedamethodologicalframeworkfortrajectorygenerationinprojectivespaceIntheirpaper,ZhaandDupresentedanewapproachforthegenerationandoptimizationofrobotictrajectoriesinCartesiantaskspaceSimulationismadeinavirtualCADbasedofflineprogrammingenvironmentInthispaper,anintegratedsystemfortheinitialpathgenerationofaroboticmanipulatorisproposedUsingthissystemaninitialpathcouldbecalculatedquiteeasyandfastandwithgoodaccuracyTheinputofthesystemisapairofstereoimagescapturedbymeansofastereovisionsystemoracompleximagecapturedbymeansofourpseudostereovisionsystem(PSVS),Adesiredpathisgeneratedbymeansofthecurrentsceneasastraightorcurvededge,combinationofedgesormoreoveruserdesignedlinesconnectingforexamplethetorchwithanobjectofthesceneForbetteraccuracyincalculations,theselectionofthebestviewofthedesiredsceneispossiblebytranslatingandrotatingtheendeffectorusingpushbuttonsanduserselectedrobotvelocitiesThegraphicaluserinterface(GUI)canbeusedinanyotherapplicationwheregenerallyimageprocessingorparticularlytheextractionofadesiredpathisrequiredThispaperisorganizedasfollowsInSection,thebasicconceptsusedinthesoftwareapplication,aswellastheoreticalnotesofthem,arebrieflyexplainedInSection,abriefdescriptionoftheGUI,whichispartofHumanPTroboticapplication,ispresentedInSection,experimentalresultsareprovidedInSection,ashortdiscussionisgivenFinallytheconclusionsofthisworkarepresentedinSectionJIntellRobotSyst():–BasicConceptsPseudoStereoVisionSystem(PSVS)IntheGUIofthepresentedsystem,asacapturingsystemcanbeusedanordinarystereovisionsystemwithtwocameras,orourpseudostereovisionsystem(PSVS)DetailsforPSVSarepresentedin,Itisamonocularstereovisionsystemcomposedofacameraandfourmirrors(Figure)ThefirstmirrorisathinbeamsplitterTheothermirrorsarefirstsurfacemirrorsUsingPSVSacheapersystemiscreatedcombiningmanygoodfeaturesTheseare:()creationoftwovirtualcameraswiththesamegeometricproperties,fieldofviewandactivefocallength()Theangularfieldofview(FOV)oftheapparatusisthesamewiththeangularFOVoftherealcamera()Ithasslightlybetteraccuracythananordinarystereovisionsystem()ItcapturesacompleximageinasingleslotThiscompleximageiscomposedoftheleftandrightimagesofanordinaryvisionsystemandconsequentlythiscompleximagecanbefasterprocessedThecameracoordinatesystemhasasorigintheopticalcenterOoftherealcameraanditsZCaxiscoincideswiththeopticalaxisfromOTocalculatepathpointcoordinateswithrespecttothecameracoordinatesystem,theintrinsicparametersaswellasthedistortioncoefficientsofthecameramustbefoundHere,thecamerahasbeencalibratedbymeansofZZhangcalibrationmethodwhileforcornerdetectionthealgorithmproposedinhasbeenpreferredThetransformationmatrixwiththeintrinsiccameraparametersis:AaucuavvðÞWhereuo,vo,aretheestimatedcoordinatesoftheimagecenterinpixels,au,avaretheestimatedscalefactorsinimageuandvaxes,inpixelsagainandcistheparameterdescribingtheskewofthetwoimageaxesTheequationsgivingthecoordinatesofapointP(xp,yp,zp)inspace,ingeneralform,forparallelcameras,are:xPuRuRðÞauRzPzRðÞþxRorxPuLuLðÞauLzPzLðÞþxLðÞyPvRvRðÞavRzPzRðÞþyRoryPvLvLðÞavLzPzLðÞþyLðÞzPauLuRuRðÞzRauRuLuLðÞzLþauLxLxRðÞauLuRuRðÞauRuLuLðÞðÞTheaboveequationsbymeansofthesimplificationsbelowareappliedtoPSVS:auLauRauavLavRavxLbmyLzLbþlxRbyRzRbðÞJIntellRobotSyst():–Themodifiedequationsarethefollowing:xPauzPþbuRuRðÞþborxPlauzPþbluLuLðÞbmðÞyPavzPþbvRvRðÞoryPlavzPþblvLvLðÞðÞzPaubþmðÞþluLuLðÞuLuLðÞuRuRðÞbðÞFigure(a)PSVS,mirrorsandvirtualcameras(b)MirrorarrangementinPSVSandrefractionphenomenaduetomirrorJIntellRobotSyst():–InEq–,auistheestimatedscalefactorofvirtualcameras,bisthebaselinelengthanduL,vL,uR,vR,arecoordinatesofapointintheimageplaneTheconstantmistheparalleldisplacementofthevirtualopticalaxisfromOandlisthedisplacementoftheopticalcenterO(Figureb)duetorefractionphenomenainmirror()Theconstantmisequalto:mdcoswþqrðÞcosqrðÞTheconstantlisequalto:ldcosqrsinwþqrðÞþcoswþqrðÞtanwðÞðÞFormirror()thicknessequaltomm(asintheprototypemodelofPSVS),m=mmandl=mm(refractionindices,nglass=andnair=)bisthelengthofbaselinebetweenthetwovirtualcamerasFigurebcontinuedJIntellRobotSyst():–JIntellRobotSyst():–CorrespondenceAlgorithmTheproposedcorrespondencealgorithmbelongsinhighlevelfeaturebasedalgorithmsandparticularlyinalgorithmsthatcanfindcorrespondencesincurvesThealgorithmusedisdescribedinItisbasedontheconceptofseedsToimplementtheproposedcorrespondencealgorithm,acompleximageorastereopairofimagesareinitiallyprocessedIntheapplicationdevelopedinVisualC,avarietyoffiltersandedgedetectionmethodsmaybeusedInthefinaledgeimages,thedesirededgesareselectedasleftviewedges,inasemiautomaticprocedureie,bycoloringapixelmanuallyandthenbypropagatingthepixeltothewholeedgeWhenallthedesirededgesarecolored,withdifferentcolorvalues,thecorrespondingedgesaredetectedInanautomaticprocedure,eachleftviewedgeisautomaticallyselectedfirst,thecorrespondingedgeisdetectedandthewholeprocedureisrepeateduntilallthepairsofcorrespondingedgesaredetectedThreecriteriaareusedtoselectthecorrespondingedge:ThehorizontalupperandlowerlimitsoftheinitialedgeplusasmallpermissibledeviationmeasuredinpixelsThenumberofpixelsineachinitialedgeextendedbyapredefinedpercentageoftheinitialnumberofpixelsThecriterionofthemostprobableedgeUsingthisalgorithmtheselectionandprocessingofindependentedgesorlineswithdifferentcolors,ofanimageofthesamescene,ispossibleTheresults,namely,thecolorofpoints,theirimageplanecoordinatesandthedisparitiesarestoredinamatrixandatthesametimeinafileforfutureuseBelowthecorrespondencealgorithmispresentedintheformofaparadigmTworealweldingworkpiecesandaproperlyformedjointbetweenthemareused(Figurea)Theseworkpiecesarecarefullypreparedsothattoobtainatypicaljointshapewiththeproperangle,rootgapandrootfaceAccordingtotheproposedcorrespondencealgorithmacompleximageorastereopairofimagesareinitiallyacquiredInourparadigmacompleximageiscapturedForcompatibilitywithotherstereosystemsthisimageisdoubledandthentheresultingimagesareproperlysimultaneouslyprocessedprovidingfinallytwoedgeimageswherethedifferentviewsoftheweldingpiecesandthejointbetweenthemareappeared(Figureb)Fromtheleftview(firstimage)theoperatorselectsthedesirededgeoredgesLetconsiderherethatthedesirededgeisonlytheedgeconcerningjointlineSelectionismadebyplantingintheedgemanuallyaproperlycoloredpixelandthepropagationofthispixelalongtheedgeoftheleftvieworbymeansoftheareacoloringtool(itdependsonthedensityofedgesintheareasurroundingthedesirededge)TheresultisacolorededgeAfterthispreparationthethreecriteriaofcorrespondencealgorithmareimplementedAccordingtothefirstcriterion(searchinginthesecondimage)differentinupperandlowerlimitedgesareexcludedAccordingtothesecondcriterionedgeswithadifferentnumberofpixelsarealsoexcludedInthesetwopreviouscriteriaexclusionispossiblebycoloringtheundesirableedgeswithaneutralcolorThenthethirdcriterionisimplementedAbasicconceptofthethirdcriterionistofindonlyasmallnumberofpixels(calledseeds)oftheFigure(a)Compleximageoftwoweldingworkpiecesandofthejointbetweenthem(b)AfterimageprocessingtheresultingedgeimageisillustratedInthisimagethedesiredandthecorrespondingedge,whichisdetectedaftertheimplementationofthethreecriteriaofcorrespondencealgorithmarealsopresentedasleftandrightviews,respectivelyRJIntellRobotSyst():–correspondingedge(subsampling)andthentopropagatetheseseedsalongthewholeedgeThepropagationisapassiveprocedurewithsmallercomputationalcostThusitisfasterandusingpredefinedconditions(ie,processingofasmallnumberoflinesineachimage)couldbeusedinrealtimeapplicationsTheconstraintusedforthecorrespondenceeitherwithanordinarystereosystemorwithPSVSisonlyscanninginascanlineeachtimeCandidateedgepixelsaredetectedandstoredinamatrixwithcountersaccordingtotheirdistancefromtheinitialedgepixelWhenthescanningoftheimageisfinished,distancecorrespondingtothemaximumpopulationofpixelsisconsideredasthemostprobableforthecandidateedgeandatthisdistanceatleastonepixelofthecandidateedgeisdetectedbelonginginthecorrespondingedgeEachoneofthesepixelsisaseedTheseseedshavethesamecolorwiththecolorofthedesirededgeTheyarepropagatedalongthecorrespondingedgecoloringthiswaythewholeedgewiththesamecolorConsequentlytheresultisthecorrespondingedgefoundinthesecondimage(rightview)Thealgorithmwiththeaboveprocedurecouldfindlessseedsthanthepredefinedseednumber(thismaximumnumberispreselectedbytheuser)butduringpropagationalledgepixelsaredetectedScanningthenimages,correspondingalgorithmfindspairsofcorrespondingpixels,storestheircolor,coordinatesanddisparityinamatrixandatthesametimeinafileforofflineprocessingPathPointCalculationAlgorithmBymeansofpixelpairsfoundfromthecorrespondencealgorithm,DpointcoordinateswiththeaidofthestereovisionequationswithrespecttothecameracoordinatesystemcanbecalculatedHoweverthesepointshaven’tthedesiredorder,thuscannotbeusedasroboticpathpointsForthisreasonforeachpairofedgesorlinesthesepointshavetobeclassifiedsothatfromeachpairofedgesorlines,aroboticpathtobegeneratedThispathhasastartandanendpointPointdensity(relatedwithmovementvelocityandacceleration)ineachpathhastobedeterminedandforeachpointanorthonormalcoordinatesystemwithoriginthispointhastobealsoestablishedThus,informationnecessaryforproperlydrivingtheendeffectorofaroboticmanipulatorisensuredBymeansofpathpointcalculationalgorithm,thegenerationofaroboticpathforadesiredtaskispossibleTheendeffectororientationwillchangeaccordingtothesepoint’sdataInthiswaytheendeffectorcanfollowaroboticpaththatitisnotnecessarilyastraightlineConsequentlyarcweldingoncurvedsurfacesispossibleTheproposedalgorithmhasasresulttogeneratefromlinesegmentsdesiredroboticpathsandcombiningthemproperlytogeneratethepathforaspecifictaskCalculationofcoordinatesofpathpointsinvolvestwostages:()rearrangementofdesiredpixelssothattobesuccessivepathpoints,and()necessarycalculationsforeachpairofcorrespondingpixelsprovidingtheDpointcoordinatesThealgorithmisdescribedintheflowchartofFigureAlgorithmDescriptionTheobjectiveoftheproposedalgorithm,asithasbeenreferred,istorearrangepixelsfoundthroughcorrespondencealgorithm,toselectthespecifiednumberofpixelsandtocalculatethefinalpathpointsIndeterminingapath,itisdesirable,tohaveastartandanendpointThus,firstthealgorithmfindsendpointsofeachcolorededgeaswellascrosspointsThealgorithmcanJIntellRobotSyst():–findpointsofinterestinalmosteverytypeofedges(ie,aline,aclosedline,alinewithcrosspoints,combinationofpreviouscases)Thentherearrangementofpixelsbetweentwoendpointsofalineeachtime(endorcrosspoints)followsTherearrangementismadeforallpathpixelsThenextstepistheselectionofthenumberNTofthedesiredpointsineachdesiredpath(thatistoselectthepointsdensity)ThisnumberisuserselectedanddeterminesanequidistantdistributionofpathpointsTheselectedpixelshaveequaldistancesThenforeachpairofpixels,thecalculationofthecoordinatesoftherealpathpointismadeCalculationsarepresentedinthenextsubsectionThedesiredpathpointsaredesignedtemporaryonthescreenassmallcircleswithcenterthespecifiedpointResultsareappearedinthelistboxoftheGUIandaresavedautomaticallyinafileasasetofsixvaluesperline:x,y,z,yaw,pitch,roll(yaw,pitch,rollaretheEulerangles)TheoreticalAnalysisAsithasbeenpreviouslymentionedtheendeffectororientationisnotnecessarilyfixedInordertocalculatethevectorX(x,y,z,y,θ,φ)withrespecttotheworldcoordinatesystemoftheroboticmanipulator,Dcoordinatesofapointwithrespecttothecameracoordinatesystem,aswellas,therotationmatrixofacoordinatesystemattachedwithoriginthispointmightbefoundWeintroduceamethod,betteradaptedtotherequirementsoftheproposedmethodforarobotbasedarcweldingsystemThemethodissupportedbyPSVSbutcanalsobeusedwithanordinarystereovisionsystemwithparallelcamerasAssumingthatthedesirededgeisalinesegmentdefinedbytwopoints,P(x,y,z)asthestartpointandP(x,y,z)astheendpoint,thethreeanglesroll(φc),pitch(θc),yaw(yc)withrespecttocameracoordinatesystemmightbecalculatedTheobjectiveistofindFigureBlockdiagramofthepathpointcalculationalgorithmJIntellRobotSyst():–rotationanglesaroundXc,Yc,ZcaxesofthecameracoordinatesystemsothatYcaxistobeparalleltothelinesegmentBydefinition,insequence,angleyawrepresentsrotationaroundXcaxis,anglepitchrotationaroundYcaxisoftheoncerotatedsystemandfinally,anglerollrotationaroundZcaxisofthetwicerotatedsystemThentheresultingrotationmatrixwillbeR(yc,θc,φc)=R(φc)IR(θc)IR(yc

职业精品

(汽车)产品营销策划书范文.doc

HH牙膏营销方案策划书.doc

加班管理人力资源考勤管理系统方案.doc

物品采购管理制度-正式.doc

用户评论

0/200
    暂无评论
上传我的资料

精彩专题

相关资料换一换

资料评价:

/ 25
所需积分:5 立即下载

意见
反馈

返回
顶部