Automated Texture Extraction from Multiple Images to Support Site Model Refinement and Visualization

Please download to get full document.

View again

of 10
24 views
PDF
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Document Description
Automated Texture Extraction from Multiple Images to Support Site Model Refinement and Visualization
Document Share
Document Tags
Document Transcript
  AutomatedTextureExtractionfromMultipleImagestoSupport SiteModelRenementandVisualization   XiaoguangWang,JonathanLim,RobertT.Collins,andAllenR.Hanson DepartmentofComputerScience Box34610,UniversityofMassachusettsAmherst,MA.01003-4610,USA Email:xwang@cs.umass.edu  Abstract  Texturemappinghaswideandimportantapplicationsinvisualizationandvirtualreality.Sur-facetextureextractionfromasingleimagesuersfromperspectivedistortion,datadeciency,andcorruptioncausedbyshadowsandocclusions.Inthispaper,asystemisdevelopedforautomatedacquisitionofcompleteandconsistenttexturemapsfrommultipleimagesinor-dertosupportsubsequentdetailedsurfaceanalysisandscenerendering.Givencameraand lightsourceparametersforeachimage,andageometricmodelofthescene,thetexturesofobjectsurfacesaresystematicallycollectedintoanorganizedorthographiclibrary.Occlusionsandshadowscausedbyobjectsinthescenearecomputedandassociatedwitheachretrieved surface.A\BestPieceRepresentation"algorithmisdesignedtocombineintensitiesfrommul-tipleviews,resultinginauniquesurfaceintensityrepresentation.Detailedsurfacestructures,suchaswindowsanddoors,areextractedfromtheuniquelyrepresentedsurfaceimagestore-nethegeometricmodel.Experimentsshowsuccessfulapplicationsofthisapproachtomodelrenementandscenevisualization. Keywords: texturemapping,CADandGISSystems,virtualreality,computervision  1Introduction  Texturemappinghasbecomeanincreasinglyimportantaspectofcomputergraphicswith thewideavailabilityofspecializedhardwareandsoftware.Visualizationandvirtualreality applicationsarenowcapableofgeneratinghigh-qualityrenderingsandanimationsoftextured geometricobjects.However,incertainapplicationssuchaslandscapearchitecture,somelevelofrealismissacricedinusingarticialtexturesasopposedtoonessynthesizedfromimagesofanactualsceneorobject(forexample,usingagenericbricktextureonabuildingfacein placeofanimageofthesurfaceitself).Synthesizingsurfacetexturesfromimagesofascene isbynomeansaneasytask;whathasnotbeenwidelyaddressedisanautomaticandecientmeansoftextureacquisitionandmanagement.Thispaperfocusesontheautomatedacquisitionofcompleteandconsistenttexturemapsfrommultipleimagesinordertosupportsubsequentdetailedsurfaceanalysis,scenerendering,andothergoals.Givenapolygonalmodel,acameraviewpoint,andatexturemap,rendering thetextureontothemodelsurfacesisawell-understoodtechniqueincomputergraphics1,2,3].  ThisworkwasfundedbytheRADIUSprojectunderARPA/ArmyTECcontractnumberDACA76-92-C-0041.  However,extractionofacompleteandconsistenttexturemapforallsurfacesofa3Dobjectisachallengingtask.Figure1isanimageportionfromasetofeightimagesofasinglesite.Considertheproblemsinvolvedingeneratingtexturemapsforthelargestbuildinginthesite,whichappearsatthetopoftheimage.Thetexturesarenotcomplete:somewallsareoccluded bythebuilding,somewallsarebeyondtheviewoftheimage.Somultipleimagesmustbeused.Thereareonlyafewpixelsintheverticaldirectionofthebuildingwall;other,moreoblique views,mayhavehigherresolution.Buildingsareoftenbuiltfairlyclosetooneanotheranditmaynotbepossibletogetagoodview-occlusionandshadowingmaybreakupatexturemap anditmaybenecessaryto\piecetogether"themapfromseveralimagesinordertoremove theoccludedorshadowedportions.Sincetheimagesmaybetakenatvarioustimesofday,fromdierentpositions,andundervariedweatherconditions,thebrightnessofsurfaceintensity mapsmayvaryconsiderablyfromimagetoimage.Allofthesefactorsmustbeconsideredwhen generatingaconsistenttexturemapfromtheimagesequence.Traditionalgraphicsalgorithmsprovidenodirectwaytocollectandcombineintensitiesundervariousconditionsfortexture consistencyandcompleteness.Figure1:PartofsiteimageJ1fromModelBoard1 Inthefollowingsections,wedescribeamulti-imagetexturemappingapproach.Theem-phasisisonobtainingconsistentandcompleteintensityinformationwithbesttexturequality frommultipleimages.TheapplicationforthisarchitectureistheORD/ARPARADIUSproject,whosegoalistodevelopsoft-copymodel-basedimageexploitationtoolsandinfrastructureforimageanalysts.Animportantfunctionoftheevolvingsystemisthedevelopmentofa  site model ofanareafrommultipleimages( siteimages  ).Usesofthesitemodelinclude3Dsite visualizationandfamiliarization,missionplanningandassessment,andchangedetection4].Consequently,animportantcomponentofthesitemodelisanaccurategeometricrepresenta-tionofsignicantobjectsinthesite,suchasbuildings,aspolygonalmodelswithassociated surfacetexturemapsderivedfromthesiteimages.Experimentalresultsindicatethatthetech-niquespresentedbythepaperprovideaconvenientwayinvirtualrealityrenderingandinthe meantimefacilitate  modelrenement ,wheretheacquiredtexturemapsdriveananalysisofthe buildingsurfacestodetectdetailedstructuressuchaswindows,doors,androofvents,anden-richtheoriginalgeometricmodel.Potentialusesofthepresentedapproacharenotconnedto   RADIUS.Itcanbeappliedtoanyscenecontainingobjectsapproximatedbypolygonalfacetsofarbitraryshapeandorientation.Section2describesanorthographicfacetimagelibraryasanarchitectureofmulti-image texturemapping.Section3discussesmodelingofocclusionsandshadowsandhowthisinfor-mationisusedtodetermineusableportionsofindividualtexturemaps.Section4describeshow asingle,consistentrepresentationofthetexturemapforafacetisobtainedfromthetagged libraryoffacets.Section5discussesrelatedissuesinmodelrenementandscenerendering,andSection6describesfuturework. 2OrthographicFacetImageLibrary  Thetexturemapextractionstartsfromaninitial,coarseunderstandingofthe3Denvi-ronment,ofwhichthetexturesareinterested.Collins,etal.5]describerecentprogressin RADIUSsitemodelacquisition.Asetofimageunderstandingalgorithmshavebeendeveloped toextractthegeometricsitemodelfromtheRADIUSModelBoard1siteimagesequenceJ1-J8.Asaresult,25buildingsthatrepresentmostofthestructuresinthesitehavebeenextracted intheformof3Dpolygons.Figure2isaCADdisplayofthesitemodel.Thesitemodelacquisitionmoduleonlygivesacoarsegeometricdescriptionofthestructuresinthesite.Due toperspectivedistortionsandcorruptioncausedbyocclusionsandshadows,detailstructuresareverydiculttobedetermineddirectlyfromtheoriginalsiteimagesequence.Figure2:ACADdisplayoftheacquiredsitemodelTheacquisitionofthesitemodelprovidesanaturalandconvenientwayoftextureex-tractionandmanagementforthegoalsofgeometricmodelrenementandtexturedmodelvisualization.Thearchitectureofourautomatedtextureextractionandmanagementsystem isan  orthographicfacetimagelibrary  (OFIL).A   facet inthesitemodelisamodeledpolygonalsurfaceofabuilding,suchasawalloraroof.A   facetimage  isanintensityimageofthefacetasseenunderorthographicprojection.AnOFILforasitemodelstoresindexedorthographic imagesofallthepolygonalbuildingfacetsthathavebeenmodeledinthesite.Theintensity valuesofeachfacetimagearesampledfromasiteimageusingthetraditionalinversetexture mappingalgorithm.Ifthefacetappearsinmorethanonesiteimage,thelibrarywillholdallthefacetimages( versions  )forthefacet.Thesemultipleversionsarewell-indexedtofacilitate libraryaccess.Forexample,ahorizontalrooffacetusuallyappearsinalltheaerialsiteimagesandthushasacompletesetoforthographicversionsinthelibrary,whereasotherfacetslike verticalwallsonlyappearinsomeofthesiteimages.Thustheavailabilityofafacetversion   isanimportantpieceofinformationtobeindexedinthelibrary.Otherinformationlikelocal-ization(howthefacetisalignedintheorthographicimage)andvisibility(theobliquenessand lightingconditionsofthefacetinthesiteimage)needtoberecordedaswell.Insummary,an OFILisanimagedatabasewhoserecordsareorthographicallyprojectedfacetimagestogetherwithrelevantinformationtoaidretrievalandanalysisoftheseimages.ConstructionofanOFILasanintermediate-levelrepresentationhasadvantagesoverthe mechanismofstoringrawsiteimageintensities.First,individualfacetsarestoredseparately sothatspecicsurfacestructureextractiontechniquescanbeappliedonlytorelevantsurfaces:windowextractiononwallimages,roofventcomputationsonroofimages,etc.Second,many man-madestructuresrelatedtobuildingshaverectilinear,repetitivepatterns,likethelatticesofwindowsonbuildingwalls.Theorthographicfacetimageprovidesaviewthatisfreefrom perspectivedistortion,whichiscriticaltothedevelopmentofecienttechniquesforextracting thesepatterns.Third,thecollectionandalignmentofallthevisibleversionsofabuildingfacetprovidesamechanismforcomparingandcombiningintensitiesfrommultipleviewstoproduce abetter,orclearer,viewofeachfacet.Finally,asetofseparatelystoredfacetimagesisa naturalandconvenientcomponentofasystemforrenderingtexture-mapped3Dperspective views. 3OcclusionandShadowModeling  OneimportantfeatureoftheOFILarchitectureisitsabilitytohandle  occlusions  and  shadows  thatariseontheobjectsurfacesinascene.Occlusionsinasiteimageoccurwhen objectsstandbetweenthecameraandtheobjectsurfaceofinterest.Duringtexturemapping,intensityvaluesfromtheoccludingobjectsgetmappedontotheorthographicfacetimageaswell.Shadowareasoccuronthesurfaceduetoobjectsstandingbetweenthelightsource andthesurfaceofinterest.Generallyspeaking,occlusionsdonotprovideusefulintensity informationforsurfacetextureanalysis,whileshadowareasmaystillbeuseful,providedthatenoughdynamicrangeexistsintheshadowareatoreconstructthetextureofthesurface.Typically,unocclude,sunlitpartsofthesurfacearethebestsourcesofintensityinformation.Toavoidthenegativeeectsofocclusionsandshadowsonsubsequentfacetimageanalysis,anextrarecordisassociatedwitheachfacetimage,explicitlyindicatingwhichpixelsinthe imageareoccludedandwhichareinshadow.IntheOFIL,eachorthographicfacetimage versionisassociatedwithanorthographic  labelingimage  ofthesamesize,inwhicheachpixeliscomposedofanumberof\attribute"bitsthatrecordwhetherthecorrespondingpixelonthe facetimageisoccludedorinshadow.Figure3showsanexampleofthiskindoflabeling. shadowocclusionsunlit region Figure3:Occlusionandshadowlabeling(left:siteimage,middle:facetimage,right:facetlabelingimage)  ThecomputationofocclusionsandshadowsinthecurrentOFILsystemisperformedina model-drivenway,usingthegeometricdatacontainedinthesitemodel,thecameraparametersofthesiteimage,andlightsourceparameters.Thisisaclassicproblemofhiddensurfaceand shadowcomputationincomputergraphics1].Labelingimagesplayanimportantroleinindexingthepixelattributesoforthographic facetimages.Inthecurrentsystem,labelingimagesalsoprovideotherinformationbesidesocclusionandshadow.Thecompletesetofattributebitsinalabelingimageis:   FacetBit .Thesystemisabletohandlearbitrarypolygonalfacets.Thisbittellswhetherapixelintherectangularfacetimageiscontainedinthefacetpolygon.   PresenceBit .Abuildingsurfacemaypartlylieoutsideoftheboundariesofasiteimage.Thisbitlabelswhetherapixel'sintensityispresentinthesiteimage.   OcclusionBit .Tellswhetherthepixelisoccluded.   ShadowBit .Tellswhetherthepixelisinshadow.ToprovideaglimpseoftheOFIL,Figure4(a)showsasetoforthographicfacetimages,withlabeling,foraparticularbuildingfacetinModelBoard1.Thisrectangularfacetisthe rightwallofthelargestbuildingshownontopofsiteimageJ1inFigure1.ThiswallappearsonlyinsiteimagesJ1,J2,J6,andJ8,andthusonlythesefourversionsareavailable.Insite imageJ1,partofthewalliscutbytheimageborder,asismarkedinthelabelingimagefortheversionfromJ1.FacetversionsfromJ6andJ8lookdarkerbecausetheyareself-shadowed,i.e.orientedawayfromthelightsource.InsiteimageJ2andJ6,thiswallisviewedfromsuch anobliqueanglethatthetexturesmappedfromthesetwoimagesprovideverylittleadditionalinformationovermuchofthewallsurface.However,nearthelowerleftofthewallthereisanothersmallbuildingthatoccludesthewallinversionsJ1andJ8,butnotinJ2andJ6due totheextremeobliquenessoftheviewingangle.Fromthisexamplewecanseethatmultiple imagesarenecessarytoseealltheportionsofthisparticularbuildingface,andthattheOFIL hascollectedandorganizedtheavailableinformationaboutthiswallfacet. 4UniqueIntensityRepresentation  TheOFILcollectsalltheintensityinformationabouteachbuildingfacetintheformofseparate,butaligned,orthographicimageversions.Formanytasksitisdesirabletoproducea single,uniqueintensityrepresentationofthefacet.Asimpleapproachistoselectone\good"versionfromthefacetimagesastheuniquerepresentation.Thedrawbackofthisapproachisthatanyocclusionsorshadowsinthatfacetversionwillbeincludedasartifactsintheresulting representativetexturemap.Inaddition,theversionthatisleastcorruptedbyocclusionsand shadowsisnotnecessarilytheclearestone.InFigure4,theversionthatcontainstheleastocclusionsandshadowsistheonefromsiteimageJ2,butthatversionistooblurredtobea goodrepresentationduetotheobliquenessoftheviewingangle.Inthissectionwepresenta  bestpiecerepresentation  (BPR)methodforcombininginten-sitiesintoauniquerepresentationofafacet.Thismethodisbasedontheobservationthatdierentregionsonafacetmayonlyhavegoodvisibilityindierentimageversionsduetothe existenceofocclusionsandshadows.ThenalrepresentationisacombinedimagewhosepixelintensitiesareselectedfromamongmultipleimageversionsintheOFIL.Arepresentativefacetimagesynthesizedinthiswayiscalleda  BPRimage  .
Similar documents
View more...
Search Related
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks