GEOMETRICAL ASPECTS OF AIRBORNE LASER SCANNING AND TERRESTRIAL LASER SCANNING Norbert - PDF document

GEOMETRICALASPECTSOFAIRBORNELASERSCANNINGANDTERRESTRIALLASERSCANNINGNorbertPfeiferandChristianBrieseInstituteofPhotogrammetryandRemoteSensing,ViennaUniversityofTechnology,Gußhausstrasse27-29/E122,1040,Vienna,Austria,np@ipf.tuwien.ac.at,cb@ipf.tuwien.ac.atChristianDopplerLaboratory“SpatialDatafromLaserScanningandRemoteSensing”KEYWORDS:AirborneLaserScanning,TerrestrialLaserScanning,LiDAR,Geometry,Orientation,Registration,Segmentation,Calibration,DigitalTerrainModel,DTMABSTRACT: IAPRS Volume XXXVI, Part 3 / W52, 2007 311 2DATAACQUISITIONThissectionpresentsthestateoftheartinlaserscanningsystems,theircalibrationandtransformationofthepointcloudsacquiredintoasuperior,possiblyglobal,coordinatesystem.Duetothefactthatthesensortechnologyisdevelopingfastwewillnotdescribespecicscannersofdifferentvendors,asatthemomentofprintingthearticle,theinformationmaybeoutdatedalready.Therefore,ratherthemainperformanceparametersaregiven.Surveysofcurrentlyavailabledevicesareregularlypub-lished(GIM2007;POB2007).Airborneandterrestrial(alsocalledclose-range)deploymentwillbetreatedseparately,asthedifferentdeploymentofthescanners,atxedpositionsoverpe-riodsoftimevs.onamovingplatform,hasalargeimpactontherststepsofdataprocessing.However,withtheadventofscan-ningfrommovingplatformsontheground(alsotermedmobilelaserscanning),andthelongerhistoryofusingprolescannersontrains,itwouldbemoreappropriatetodistinguishbetweennamicandstaticscanning.Intherstcasescanningisperformedbyaunivariatebeamdeectionunitandareawisedataacquisi-tionisestablishedbythedynamics,i.e.themovement,ofthescanningplatform(aircraft,landvehicleoraboat).Inthesec-ondcasetheexteriororientationoftheplatformisconstantforonescanposition,andtwodimensionalcoverageintheangulardomainisperformedbyrotatingcomponentsofthedevice(e.g.,amirrorortheupperinstrumentpart).Proling(Lindenberger1989),ontheotherhand,iswhatisobtainedbyunivariatebeamdistribution,e.g.obtainedfromsatelliteplatforms(Zwallyetal.2002)orusedforcontinuousmonitoringorelongatedstructures(Hesse,Neuner,andKutterer2005).2.1AirborneLaserScanning2.1.1CurrentSystemsALSsystemsusealmostsolelythepulsetimeofightmeasurementprincipleforranging(Riegl2007;Optech2007;Leica2007;TopEye2007;TopoSys2007;Fli-Map2007).OneexceptionistheresearchsystemScaLARS,whichappliesthephasedifferencemeasurementprinciple(HugandWehr1997).Currently,therearetwodifferenttypesofcom-mercialALSsensorsystemsavailable:discreteechoandfull-waveformscanners.Whilediscreteechoscannersdetectarep-resentativetriggersignalformultipleechoesinrealtimeusinganaloguedetectors,full-waveformALSsystemsdigitizetheen-tireanalogueechowaveform,i.e.thetime-dependentvariationofreceivedsignalpower,foreachemittedlaserpulse.Digitizationisperformedtypicallywithanintervalof1ns(correspondingto15cmone-waydistance)andthedeterminationoftheindividualechoeshastobeperformedinpost-processing(Wagner,Ullrich,Melzer,Briese,andKraus2004).InALSmainlytwolaserwave-lengthsareinuse:1.06mand1.5m.Thepulserepetitionrate(PRR)ofcurrent“topend”devicesis100kHzto200kHz.Theoperatingaltitudeofthesystemsisdifferent,withsomesystemsrestrictedtoayingheightoflessthan1000maboveground,whereasotherscanbeused5kmabovegroundlevel.ManyALSsystemsarecurrentlyonlyabletorecordthereectionsofonelaserpulsebeforethenextisemitted.ThisrestrictshighPRRtoloweryingheights(notmorethan100kHzfor1500mmax-imumone-wayslantrange).Recentsensordevelopmentsleadtotheabilityofmultipulsesystemswhichallowtohavemultiplelasersignalsintheairsimultaneously(Optech2007;Leica2007).ThemaximumeldofviewinALSdataacquisition,measuredperpendiculartotheforwardmovementoftheaircraftisdepend-ingonthescannerusedandreachesfrom.ALS Nexttoincreasingthepulsegenerationrateofthelaser,anoptiontoincreasethemeasurementrateismountingtwolaserscannersononeplatform,asofferedcurrentlye.g.by(DiamondAirborneSensing2007).systemsareusedonxed-wingaircraftaswellasonhelicopters.Whilexed-wingedaircraftaretypicallyusedfortheacquisitionoflargeprojectareas,helicoptersarepreferredforfollowingalinearfeature(e.g.forcorridormapping)orfordifculttopogra-phy.Thescanningmechanismsappliedaremainlythosedeectingthelaserbeaminaplaneperpendiculartotheyingdirection,usinganoscillatingoramulti-facetedrotatingmirror(Latypov2005).ForrotatingmirrorscannersthePRRistypicallyonlyaburstmeasurementrate,andthenumberofpulsesusedformeasuringranges,i.e.theeffectivemeasurementrate,islower(Riegl2007).Therayswithlargernadiranglesarenotprovidedtotheusersorarereectedwithinthescannerhousing.Oscillatingmirrorshavetheadvantagethattheturningpointscanbesettoanglesappro-priateforaspecicproject.However,asthemirrorshavetobeacceleratedthepointdistributiononthegroundcanbelesshomo-geneousthanforrotatingmirrorscanners,especiallywhenusingaharmonicangleacceleration.Byusingmirrorswithdifferentanglesatthefacets,forward,nadir,andbackwardlookingcanbeperformedwithonescanner(Fli-Map2007).TheberscannerusedinoneoftheTopoSysscannersisspecialinthesensethatnoanglepositionofthemirrorhastobemeasuredastheemissiondi-rectionisxedandgovernedbythesinglebersdirectly.Palmerscanners(WehrandLohr1999)areusedbyTopEye,ScaLARSandNASA'sALTM(Finneganetal.2005).Whilegeneratingalessregulargroundpointpatterntheyofferanadvantageincali-bration,aseach“point”ismeasuredtwice.2.1.2CalibrationandStripAdjustmentForthetransfor-mationoftheALSdata(rangeandangleobservations)intoonecommoncoordinatesystemthepositionandangularattitude,i.e.theexteriororientation,oftheplatformhavetobeknowninor-dertoallowdirectgeo-referencing.Typically,thisisrealizedbyacombinationofaglobalnavigationsatellitesystem(GNSS)re-ceiverandaninertialmeasurementunit(IMU).Togetherwiththelaserscannertheyformamultisensorsystem.Duringdataac-quisitiondatastreamsarerecordedbyeachinstrumentatdiffer-entfrequencyandaresynchronizedviatheGNSStimemeasure-ments.Calibrationofthismultisensorsystemistheprocessofdeterminingtherelativeorientation,i.e.shiftsandrotations,be-tweenthecomponents(GNSSantenna,IMU,andlaserscanner)andtimelagsinthesynchronization.Tosomeextentthesepa-rameterscanbedeterminedbytotalstationmeasurementsontheground,butanumberofparameters,e.g.theIMU–laserscan-nerrelativeorientationortimelags,arebetterdetermineddur-ingight.Scannervendorsprovidespecialsoftwarethatallowderivationoftheseparametersifdedicatedightpatternsareper-formed.Typicallyonlyatsurfacesareusedforthisalignment,butasFilin(2003)hasshown,inclinedsurfaceswithdifferentaspectareaprerequisitefordeterminingallrelativeorientationparametersofthemultisensorsystem.Approachesforthecalibrationhavebeenpresentedin(Burman2002;Filin2003;Kager2004).InSkaloudandLichti(2006)amethodfordedicateddeterminationofthethreebore-sightan-glesandtherangenderoffsetisdescribed.Thesemodelsareallbasedontheobservationsrange,angle(ofbeamdeection),andobservationsofexteriororientation(i.e.positionandangularattitude).ThepointsmeasuredbytheALSsystemareeitherre-latedtogroundtruthand/ortopointsofanotherstrip(controlandtieinformationintheformofsurfacepatches,respectively).Thediscrepanciesencounteredinthoseareminimizedbydeterminingthecalibrationparameters.Incalibrationthetaskis,asdescribedabove,toreconstructthegeometriclayoutofthemultisensorsystem.Forapplicationsof 312 ALSdata,theeffectsofaninsufcientcalibrationandoferrorsintheexteriororientationdeterminationaremoreofinterest.Thetaskofstripadjustmentistocorrecttheseerrors.Thiscanbedoneeitherbycalibration,orbyapplyingcorrectionstothepoints.Thepointwithindexiscorrectedbyapplyingacorrectionfunctionforstrip.Therstpublicationsonstripadjustmentchosethatapproach.Inthesimplestcasethefunctionsareonlyshiftvectors,;y;zanddonotdependonthelocationwithinthestrip.In(Crombaghs,Brugelmann,anddeMin2000)and(KrausandPfeifer2001)thecorrectionfunctionappliestotheheightcom-ponentonly,usingalinearfunction(verticaloffsetandtiltsinandacrossightdirection),andpolynomials,respectively.TheapproachofKrausandPfeifer(2001)allowscorrectingshorterwavelengthdeformations,too.Amethodthatisnotrestrictedtoverticalcorrection,butalsoremovesdiscrepanciesinplanimetrywasdevelopedbyKilian,Haala,andEnglich(1996),wherethehasparametersforconstantoffsetandtimedepen-dentdriftsforshiftinandrotationaroundthethreecoordinateaxes,requiringthatthetimeofthemeasurementisknown.Vos-selmanandMaas(2001)describeasimilarmethod,mentioning,thatthismodeldoesnotallowtocorrectshorttimeeffectscausedbythelimitedGNSSaccuracy.Knowledgeonthemeasurementtimeisnotrequiredbutreplacedbyparameterizationalongthestripaxes.Contrarytotheseapproacheswhichmodeleffects,notcauses,calibrationprocedurescanbeextendedtoperformstripadjust-ment,too.ThishasbeendemonstratedbyBurman(2002)andKager(2004).InALSthereisstillaprocessofmodelidenticationgoingon.Calibrationmethodsandstripadjustmentshouldbegenericen-oughtoallowhandlingallairbornelaserscannersontheonehand,andcorrectthecausesoftheerrors,andnotonlyeffects,ontheotherhand.Specializeddevelopmentsareoftenbenecialfordataacquisition(e.g.therollanglecompensationofOptech),butlesspracticalforimplementationofon-the-jobsensorcali-brationtechniques.Especiallythelackintheavailabilityoftheoriginalobservations(i.e.,trajectory,angles,andranges)compli-catestheseeffortsfromascienticpointofview.Althoughmethodshavebeenpublished,andtheincreaseinpreci-sionisnotable,on-the-jobcalibrationisnotstandardyet.Thereisalackofsoftwareavailabletodataproviders.Nexttocalibration,alsoeffortstoimprovetheightpatharenecessary.TheglobalnavigationsatellitesystemcanbeseenascorrectionforthelowfrequencyerrorsinthehighfrequencyobservationsoftheightpathandsensorattitudebyIMUmeasurements.However,theGNSScomponentitselfissubjecttoerrorsthatoccurovertheen-tirestriporpartsofit(e.g.,wrongambiguityxes).Undersuchcircumstancesanoffsetanddriftcomponentasunknownsinstripadjustmentarenotenough(Ries,Kager,andStadler2002).Alsopolynomialmodelsdonotworksatisfyingly,andsplinemodelsformodelingthedynamicexteriororientationrequirecarefulbal-ancingofthenumberofknotsandtheirplacing.Summarizing,solutionsmodelingtheightpathhavenotbeenverysuccessfulyet,whichmaybeattributedpartlyalsotounexpectedbehaviorencounteredinightpathinformation,e.g.“jumps”intheightpath(Ries,Kager,andStadler2002).Theauthorsholdtheview,thatatighterintegrationofthede-terminationofthesensortrajectorywithKalmanlteringofthe Itshallbenotedthatsatellitepositioningisperformedinageomet-ricalcoordinatesystem(geocentriccartesianorellipsoidalcoordinates)whereasIMUmeasurementsdependalsoonthelocalgeoid.GNSS/IMUdatawiththedeterminationofsensorcalibrationandexploitationofhomologouspatchesonthegroundwillprovidethemostprecisesolution.Itallowstointroduceredundancyinthedeterminationoftheightpath,whichisabsentindirectgeo-referencing(Skaloud2006).Introducingredundancyincreases,atleastintheory,reliabilityandallowsestimatingtheprecision.Thereisalsoapotentialtoaccountforchangingsatelliteconstel-lations,GNSSoutages,orperiodswithlessthanfoursatellitesvisible.Inaero-triangulationthecombinedadjustmentofimagesandGNSSobservationsisinvestigatedin(Schmitz,Wuebbena,Bagge,andKruck2001)and(EllumandElSheimy2006).ForwelldenedsurfacestheprecisionofALS,applyingarigor-ousmodeloflaserstripcalibrationasdescribedabove,canreachafewcentimeter.ThedeterminationoftheightpathwithGNSSgivesaprecisionof5cmto10cmineachcoordinatedirec-tionandbecomesalimitingcomponentofALSprecision(CsanyiandToth2007).SolutionsmaycomefromimprovementsintheGNSS,alsobyusingmultiplereferencestations,orfrommoregroundcontrol.Thelattercanbeintheformofsurfacepatches,whichiseconomicallyandpracticallylessviable.Alternativenavigationsystemsmayemerge,althoughthecurrentalternativestoGNSSfornavigationincitiesorinsidebuildings,e.g.basedonmobilecommunicationandotherwirelessnetworks(KarimiandHammad2004)arefarfromtheaccuracyprovidedbyGNSSyet.ForfullexploitationofthemeasurementsofALSnotonlythegeometricaspectsshouldbeconsidered,butalsotheradiometry.Thebackscatteredenergy,intheformofphotons,istypicallyconvertedtoavoltageorcurrentandthenconvertedfurtherintoadigitalnumber,notnecessarilybyalinearfunction.Thisisdiscussedin(Wagneretal.2006;Ahokasetal.2006;HandPfeifer2007).Manyairbornesystemshavetworeceivers(so-calledlowchannelandhighchannelfordetectionofechoeswithsmallandlargeamplitude,respectively),whichhasnotbeenconsideredincalibrationeffortssofar.2.2TerrestrialLaserScanning2.2.1CurrentSystemsIncontrasttoALSsystemsmorevari-ationinthesensordesignofTLSsystemscanbeobserved.Thewavelengthsusedarebetween0.5mand1.5m.Longerwave-lengthsareaffectedlessbytheatmosphere,butshorterwave-lengthscanprovidesmallerfootprints.Terrestriallaserscannersusethepulsetimeofightmeasurementprinciple(Riegl2007;Leica2007;Trimble2007;Optech2007;Callidus2007;I-SiTE2007)aswellasphasebasedranging(Zoller+Frohlich2007;Faro2007;3rdTech2007).Thesecondsystemsusethephasediffer-encebetweentheemittedandreceivedbackscatteredsignalofanamplitudemodulatedcontinuouswave(AMCW)toinfertherange.Pulsetimeofightrangingscannersaresuitedbetterforoutdooroperationwherelongerrangeshavetobemeasuredandaretypicallypanoramicscanners,withaeldofviewof360e.g.80.ThePRRofthesesensorsisaround10kHzandless,andprecisionliesbetween5mmandSomesystemsofferthepossibilitytoeithermeasuretherstorlastecho,butsimultaneousrecordingisusuallynotavailable.Scannersapplyingthephasebasedrangingaretypicallyhemi-sphericalscannersthatallowtoscanintoalmostalldirections(e.g.360by135).However,duetotheirrangingprinciple(lim-itedrangeuniqueness,mostlybelow100m)theyarewellsuitedforindoorusageandoutdoorenvironmentswithalargernumberofobjects(e.g.pipinginstallations,innercityareas),restricting Leicarecentlyintroducedaterrestriallaserscannerwith50kHzPRR,butatthetimeofwriting(August2007)noindependentreportsofper-formancewereavailable. 313 theview.Themeasurementrateistypicallyabove100,000pointspersecond,andtheprecisionis2mmorbetter.Withthisrang-ingprincipleonlyonedistancecanbedetermined,becausethebackscatteredsignalsfromdifferentreectorsarealwaysover-layingeachother.Thisresultsinaphaseanglecorrespondingtoadistancebetweenthetwoormorereectors.Triangulatingscanners,similartostructuredlightsystems,arenotdiscussedhere.In(Blais2004)areviewisgiven.Addi-tionallyitshallbenotedthatcurrentlyeffortsareonthewaytogeneratestandardsforterrestrialscanners(Beraldinetal.2007;Breuckmannetal.2007).2.2.2CalibrationTerrestriallaserscannersare,fromthecon-structionprinciple,similartotheodolitesandtotalstations.Thisholdsespeciallyforstrictlymonostaticsystems,wheretheaxisoflaserbeamemissionandtheopticalaxisofthereceiverarethesame.Additionally,iftherotationaroundtheverticalaxisisperformedbyinstrumentrotationandthebeamdeectionintheverticalplaneisperformedbyarotatingmirrorinclined45againstthebeam,thenananalogybetweentheinstrumentaxesofaterrestriallaserscannerandatheodolitecanbeestablished.Lichti(2007)modelsdeviationsintheobservationsbycorrectionfunctions,someofwhicharebasedonthesensormodel(e.g.,trunnionaxiserror),whereasotherparametersarefoundempiri-cally(e.g.,sinusoidalerrorinhorizontalangleasafunctionofelevationangle).Thephysicalcorrectionsfortherangemea-surementerrorsareaconstantoffsetandharmonicfunctionsatthewavelengthsusedintheamplitudemodulationforthephasebasedranging(Rueger1990).ThisapproachisdrivenbythemodelofthesensorandforaFarolaserscanneritresultedinanimprovementofabout30%ineachcoordinatedirection(LichtiAbmayretal.(2005)usethesimilarityofaterrestriallaserscan-nertoatheodoliteanddetermineforaZ+FImager5003consec-utivelytrunnionaxiserror,collimationaxiserrorandverticalcir-cleindexerror.In(ParianandGruen2005)adifferentapproachforthesamescannerispresented.TheTLSobservationsinthesphericalcoordinatesystemaretransformedtoobservationsofacylindricalcoordinatesystem,whichispossibleifnottheentirehemisphericaleldofviewisused.Thenacalibrationapproachforpanoramiccamerasisapplied,reducingsystematicerrorsintheangleobservations.BythismethodresidualsattargetpointsidentiedinintensityimagesforaZ+FImager5003arereducedby90%.Theapproachespresentedsofarrelyontargetedpoints.Whilesuchawell-controlledexperimentallowstomakeobservationsintheentire(angularandrange)domain,itisnottypicalforprojectexecution.Asthestabilityoftheparameterscannotbeguaranteed(Lichti2007),thedevelopmentofon-the-jobcalibrationmethodsappearstobenecessary(Reshetyuk2006).Itshouldalsobenotedthatspecialdeviceconstructions,e.g.thedualwindowdesignoftheLeicaScanstations,havenotbeeninvestigated,yet.2.2.3Registration/OrientationInTLSrelativeorientation,alsotermedregistration,iscurrentlyperformedstandard-wisebyeitheroftwomethods:ICPtypealgorithmsontheonehandandexplicittiefeaturesontheotherhand.Withasufcientnumberofhomologoustiefeatures(points,lines,orsurfaces)thetransfor-mationparameterscanbecomputed.Forpointsthisispossiblewithoutapproximatevalues(Horn1987).ICPalgorithmsdonotrequirehomologouspoints,andtheexactcorrespondenceisre-placedbyiterativelydeterminedapproximatecorrespondenceofpointsorsmallsurfaceelements.Thetermsrelativeorientationandregistrationareusedalmostsynonymously.Brenner,Dold,andRipperda(2007)notethat`registration'isputtingmoreemphasisontheactiveroleofthepointcloudintheprocessitself(Brenner,Dold,andRipperda2007).Theterm`relativeorientation',ontheotherhand,alsoreferstotherelationbetweendevicecoordinatesystems.Nexttoregistrationandorientationalsotheterms(co)-alignment,con-solidation,andstitchingareregrettablyinuse.Ifonlytheobjectitselfisofinterest,itissufcienttodeterminetherelativeorientationbetweenscans.Iftheobjectalsohastobeplacedinasuperiorcoordinatesystem,absoluteorientationbecomesnecessary,too.Ifthesuperiorcoordinatesystemisearthxeditbecomesthetaskofgeo-referencing.Usinghomologousfeaturesforrelativeorientation,theyhavetobeextractedrst.Thisbecomessimple,ifarticialtargetsareplacedinthescene,e.g.withretroreectivematerial.Inthatcase,duetothehighintensityvalue,theycanbefoundautomati-cally.Naturaltieelementscanbeidentiedwithloweraccuracyintheintensityimagesbyvisualinspectionorautomaticproce-dures.Alternatively,objectsurfacescanbeusedastieelements(e.g.,cylindersandplanes).AmethodforautomaticextractionofthesepatchesandcomputationoftransformationparametershasbeenpresentedbyRabbani,Dijkman,vandenHeuvel,andVos-selman(2007),DoldandBrenner(2006),andBrenner,Dold,andRipperda(2007).Findingthecorrectcorrespondencesbetweenfeaturesoftwoscansautomaticallycanbeseenasagraphsearchproblemandmethodsforpruningthegraphbecomenecessarytoreducethesearchtime,e.g.bycomputingandcomparingparam-etersaspatchboundarylength.Anotherwaytoincreaseautoma-tionisrelyingonhighresolutionimages,wherethetaskhasbeenstudiedforalongertimeand(e.g.)thetechniqueofcodedtargetshasbeendeveloped.Al-ManasirandFraser(2006)presentedanapproachwherearticialtargetsareautomaticallyfoundinadig-italimage,takenwithacamerawithknownrelativeorientationtothelaserscanner.BohmandBecker(2007)suggestsusingtheSIFToperator(Lowe2004)tondhomologouspointsinthein-tensityimage.Fortwoscansfromnotablydifferentviewpointsofahouse,includingevenrepetitivetexture,therelativeorientationcouldbecomputedcorrectly,althoughwithlimitedprecision.TheICP(iterativeclosestpoint)methoddoesnotrequirehomol-ogouspointsandperformstheorientationoftwoscans,givenapproximatevaluesofsufcientquality,entirelyautomatically.Thisisadvantageous,becauseplacingtargetscanbeimpossi-ble,especiallyiftheobjectisnotaccessible,additionallyitcanbecometimeconsuming.ICPhasbeensuggestedbyBeslandMcKay(1992)andvariantsarestudiedin(RusinkiewiczandLe-voy2001).MuchresearcheffortiscurrentlyspentinordertoautomatendingapproximateparametersforICP.Thisleadstondingcorrespondingfeaturesasdescribedabove,possiblywithlowerqualityrequirements.Theauthorsbelievethattheregistrationtaskwillrunfullyauto-maticallyforcertainapplicationsinsomeyears.However,inthegeneralcase(includingterrestrialscanninginaforest,etc.),ornotrelyingondomainknowledge,thetaskwillremaindifcult.Analternativemaycomefromcheapexteriororientationdevices,allowingtoobtainapproximateexteriororientation,whichcanbeusedforreducingsearchspaces.Inordertotransformoneormultiplescans,generallyonepointcloud,intoasuperiorcoordinatesystemcontrolpointsand/orpatchesarerequired.Thiscontrolinformationcaneitherbedis-tributedinthesceneorthecoordinatesofalaserscannerstandpointcanbeobserved,e.g.bycenteringoveraknownpointorbymountingaGNSSantennaontopofthescanner.Deviationof 314 thestandaxisfromthevertical,denedbythelocalgravityeld,maybeobservedandcorrectedwithanelectronicspiritlevel(in-clinometer).Suchadeviceisbuiltintomanyterrestriallaserscanners.Concerningthetargetsusedforabsoluteorientationthesameasmentionedforrelativeorientationapplies.3GEOMETRICALASPECTSOFSIGNAL–OBJECTINTERACTIONInlaserscanningbackscatteredenergyisusedforrangemeasure-ment.Ifthebackscatteringsurfaceisat,reectingdiffusely,orientedorthogonaltothelaserbeam,reectionappearingonlyatthematerialtopsurface(i.e.,thereisnopenetrationofthein-cidentenergyintothematerial),andthesurfaceisnottooclosetothescanner(especiallyinthecaseofbistaticsystems),thesystemsinusemeasurethequantityofinterest.Furthermore,noothertargetsmaybeintheinstantaneouseldofview.InmanycasesinALSandTLStheserequirementsdonothold.Dependingontheechodetectionmethod(Fox,Accetta,andShumaker1993;Katzenbeisser2003;JutziandStilla2003b)used,theangleofincidenceorsurfaceroughnessmayhaveanimpactontherange.Forat,slantedtargets,discretereturnsystemsthatanalyzetheleadingedgeofthesignalmayreportrangesshorterthantherangetothebeamcenter(JutziandStilla2003a).Thiseffectisdiminishedforsmallerfootprintsandshorterpulses.InTLSthedynamicrangeofthebackscatteredenergyisquitehigh.Thisoriginatesinthelargerrangespectrumof1:100to1:1000(e.g.,minimumdistance2m,maximumdis-tance1km),andinthevarietyofbackscatteringsurfaces,too.Thesurfacesreachfromdarkmaterialstoretroreec-tivetargets.QuantizedintermsofLambertianscatterers,thedetectablereectedenergymayvaryfrom5%to1000%Thisresults,usingthelaserrangeequation(Jelalian1992),inreturnenergieswitharatioof1:10.Effectsdependingontheintensityhavebeenreportedbymanyauthorsforter-restrialscanning(Hanke,Grussenmeyer,Grimm-Pitzinger,andWeinold2006;ValanisandTsakiri2004).Itisalsore-portedthatrunwaymarkersfoundonair-stripshavecausedsimilareffectsinALS.Itshallbenotedthatmostairborneandterrestrialsystemsrequireameasureofthereturnenergyforapplyingarangecorrection.Somepreliminaryresultsontherelationbe-tweenobservedintensityandrange(andbetweenintensityandobjectreectivity)forapulsetimeofightterrestriallaserscannerhavebeenpresentedbyPfeifer,Dorninger,Haring,andFan(2007).InTLSpenetrationoftheenergyintodifferentmaterialsisgenerallynotverywellstudied.FormarblesurfacesandredlightGodinetal.(2001)havedemonstratedtheeffect.Thiseffectcanreachsignicantmagnitude,intheorderofmillimeter,forcloseranges,typicallyencounteredfortrian-gulatinglaserscannersandforphase-basedrangemeasuringlaserscanners.InALStheterrainand(vegetation)objectsuponitaretheobjectofinterest,butoftenitisimpossibletomeasureonlyone,eitherthegroundorthevegetation.Low(herbaceous) Forbistatic(two-eyed)systemstheemitterandreceivereldofviewoverlaponlyafteracertaindistance.Lambertianscatterershaveamaximumbackscatterof100%(noab-sorption),butretro-reectivematerialscattersbackmoreenergyintothedirectionofthesource.vegetationoffsetsthegroundmeasurements.Whilethecauseisbasicallywellunderstood,i.e.scatteringatallobjectswithinthefootprintandmultiplescattering,theamountandthein-uencingfactorsarenotverywellknown.Theycanbere-portedforaspecicexperiment,butpredictionisnotpos-sibleyet.Inanycase,theeffectsareintheorderofcmtoAhokas,Kaartinen,andHyyppa(2003)havereportedsys-tematicinuencesofgrassonthemeasuredrange.Boll-weganddeLange(2003)reportedsystematicupwardshiftsforlongdensegrass.In(OudeElberinkandCrombaghs2004)itisshownthatupwardshiftsoccurredupto15cmonlowvegetationareas(creepingredfescue,thrift).Arelationcouldbeseenbetweenthedensityofthevegetationcover-ageandthesystematicerror:0%coveragemeantnoupwardshift,100%coverageshoweda15cmshift.Thestudyof(HodgsonandBresnahan2004)tslesswellintothatpic-ture,asthesystematicshiftsreportedareallverysmall,i.e.below6cm.Pfeifer,Gorte,andOudeElberink(2004)re-portedshiftsof7cmforlongdensegrassand10cmforayoungforest.Hopkinsonetal.(2004)havecorrelatedheightoflowveg-etationwithstandarddeviationofheightsandothertexturalcharacteristics.Concentratingontheexperimentsoverlowvegetation(below0.5m)theerrorsarecomparativelylargewithrespecttothevegetationheightandafunctionalrela-tionshipisnotobvious.Contrarytotheresearcheffortsandsolutionspresentedinthesections2.1.2and2.2.2(ALSandTLScalibration)theseprob-lemscannotbeconnedtothemeasurementsystemitself,buttargetpropertieshavetobeconsidered,too.Evenwhenrecord-ingthefullwaveformofthebackscatteredechoes,notmuchin-formationbeyondspatialandabsorption/scatteringcharacteris-ticscanbeextracted.Theechowidthholdsinformationontherangedistributionwithinthefootprint,butthisisnotnecessarilyconnectedtothediscrepancybetweengroundelevationandsys-tematicallyshortenedrangemeasurement.Thus,materialprop-ertiescanonlybederivedifadditionalknowledgeisprovidedbyexternalsourcesasimagery,maps,orpossiblyrangeinformationatanotherwavelength.Phantompoints,alsocalledvirtualpoints,alsohinderautomaticexploitation,especiallyinTLSdatasets.Thesepointsareen-countered,whenthefootprintisdistributedoverdifferent,hardtargetsincloseproximity.Themeasuredrangeisthenbetweenthosetwoormoresurfaces.Firststepsforautomaticremovalofthesepointsweremade(e.g.Sotoodeh(2006))byanalyzingthespatialdistributionofpoints(withoutconsiderationofthescannerposition).Thereisnoreliablemethodavailableyet.Consideringthegeometryofthemeasurementsetupcancontributeinidenti-fyingthesepoints,astheyarealignedalongbundlesorrayswiththeorigininthelaserscanner.Furthermore,multi-pathreectionscanoccur(notonlyinTLS,butalsoinALSdata),resultingintoolongranges.Atypicalsitu-ationisthatasurfacealongthepropagationpathofthelaserbeamfeatures(some)specularreectionontoanother,diffuselyreect-ingsurface.Aportionofitsdiffusebackscattertravelsviathespecularlyreectingsurfacebacktothedetector.In(Lichti,Gor-don,andTipdecho2005)anoverviewoferrorsinTLS,includingtheinuenceofgeo-referencingandbeamwidth,isgiven. Forpulsetimeofightsystemsthisdependsonthepulseduration,whereasphase-basedsystemsarealwaysaffected. 315 Webelievethattheaboveitemswillcontinuetoplayarole.Fornaturalsurfaces,thusratherintheairbornecase,the“rangeer-rors”inducedchangespatially,asthevegetationisnotentirelyhomogeneous.Surfacesencounteredinterrestrialscanningareoftenmorehomogenous,especiallycomparedtothemeasure-mentdensity,andthe“error”ismoreofsystematicnature.Ap-plicationsrequiringhigherprecisionwillbecomepossiblewhentacklingandsolvingtheseproblems.4DATAPROCESSINGTheearlystepsofdataprocessing,whicharenotdirectlylinkedtoanapplication,aretypicallysegmentationorclusteringofthelaserscanningpointcloud,removaloferroneouspoints,andthin-Segmentationandclusteringaremeanstoorganizepoints,mea-suredbylaserscanning,intohomogeneousgroups.Pointsofagroupshallbeneighbors,andinthiswaylargerentitiesaregen-eratedandthedataisorganizedonahigherlevel.Aofsuchgroupsorsetsistheninthedomainofacertainapplica-tion,whichprovidesa“meaning”,aclassattribute(e.g.“houseroof”,“vegetation”,or“tree”)foreachsegment.Inmanypub-licationssegmentationispresentedasonestepforacertainap-plication,e.g.buildingreconstruction.Inthissectionwewanttospecicallyconcentrateongeneralpurposeapproaches.Segmentationandclusteringhavebeenstudiedforalongtimeinimageprocessingwheretheneighborhoodofelements,i.e.pixel,isgivenimplicitlybythematrixlayout.Forpointcloudsoflaserscanning,ontheotherhand,neighborhoodisoftendenedviaEuclideandistance,TINtopology,oranumberdeningthenearestpointsasneighbors(kNN).Anoverviewforneighbor-hoodinALSdataisgivenin(FilinandPfeifer2005).Ageneraloverviewonsegmentationalgorithmsisprovidedby(Hooveretal.1996),andoverviewsdedicatedtolaserscanningdataaregivenin(Vosselman,Gorte,Sithole,andRabbani2004)and(GeibelandStilla2000).Mostoftenregiongrowingfromaseedpointisapplied(Vosselman,Gorte,Sithole,andRabbani2004)wherethefeaturesusedassimilaritymeasureareheightdifferenceforairbornelaserscanningdata,normalvectorsimilarity,ordistancetoaplane.Differencesandsimilaritymayeitherbemeasuredfromtheseedpointtoacurrentlyinvestigatedpoint,orfromthepreviouslyac-ceptedsegmentpointtoitsnewneighbors.Thelatterstrategyallowstogrowoverbentsurfaces,therstonenot.Theseregiongrowingapproachesgenerallydeliversmooth(gentlycurved)orplanar(at)regions.Thewatershedtransformisusedtosegmentdigitalsurfacemod-els,notpointclouds,byanotablydifferentapproach.Inforestryitisonestandardmethodtoextractthesingletreesfromacanopymodel.Whilemostgeneralpurposeapproachesarepreformedonthepointcloud,VogtleandSteinle(2004)andRottensteiner,Trinder,Clode,andKubik(2005),forexample,apply2.5Dtech-niquesongriddedversionsoftheoriginaldata.Thisisonlyap-plicableforALSdataandreducestherangeofextractablestruc-tures.Operatingonthepointcloudenablesalsotheextractionofverticalplanesorplanesstackedontopofeachother,thusthefull3Dcontentofthedata.performsthegroupingofpointsetsnotinobjectspace,butinafeaturespace.Thefeaturesusedmaybeanestimatednormalvectorforeachpoint,alocalroughnessmeasure,thein-tensitymeasure,etc.Theconnectivityinobjectspaceisrealizedbyaddingthecoordinatesofthepointsaselementsofthefea-turevector.Suchapproacheshavebeenpresentedforlaserpointcloudsby,e.g.,FilinandPfeifer(2006)andMelzer(2007).Thenormalvector,afrequentlyusedfeature,isoftenestimatedbycomputinganorthogonalregressionplane.Inthiseigenvec-tor/eigenvalueapproachallthreeeigenvaluescanbeusedtoclas-sifypointsasbelongingtoasurface,avolumetricdistributionofpoints,asinglepointoralinearfeature(Medioni,Lee,andTang2000).In(BeltonandLichti2006)alsotherecognitionofsurfaceboundariesisdiscussed.VotingschemessuchastheHoughtransformarehardlyappliedonlargedatasets.Theirdisadvantageisthatconnectivityisnotconsidered.Suchapproachesaremoretypicallyused,ifsomeorganizationoftheentirepointcloudintosmallerentitieshasal-readybeenperformed.RabbaniandvandenHeuvel(2005),e.g.,rstapplysegmentationbasedregiongrowing,andthenuseaHoughtransformtodetectandreconstructcylindersinthein-dividualsegments.vonHansen,Michaelsen,andThonnessen(2006)applytheRANSACalgorithm(FischlerandBoller1982)fordetectingplanes.Inordertoovercometheproblemofcon-nectivity,theyrstdividethespaceintolarger3DcellsandapplyRANSACplanedetectiontothepointcloudwithinthecell.Thenagroupingstepconnectssimilarplanesofneighboringcells.Thereisanumberofstrategiestoreducethevolumeofthedata.Modelingitself,especiallymodelreconstructionwiththehelpofanalyticalsurfaces,canbeseenasameanstoreducethedatavolume,andbyttingsurfacesalsoameansofreducingnoise.Thesameholdsfortheinterpolationofadigitalsurfacemodel(DSM)oraDTMbyqualiedinterpolationmethodsthatcon-siderthestochasticpropertiesofthedata.Methodstodecimatedensepointcloudsandreducenoisearegivenin(Pauly,Gross,andKobbelt2002).Anoverviewondecimationofpolygonalmeshesisgivenin(HeckbertandGarland1997).Closerangescanningsystemsbasedonthephaseshiftmeasure-mentprinciplearecapableofproducingverydensepointclouds,e.g.5pointspercm.Thefootprintsofthelaserbeamontheobjectsurfacearethenoverlapping.Itisthereforejustiedtore-ducethevolumeofthedataandalsoreducethenoiseinthedatainonestep.Accordingtotheauthorsviewthereiscurrentlyalackofstudiesthatinvestigatethesepossibilitiesconsideringthepropertiesoflaserscanningdata(nexttonoisee.g.measurementposition,ormissingpoints)andnottreatingthemeasurementsasasetofdiscretepoints.5DTMDETERMINATIONFROMALSDATADuringtheALSdataacquisitionprocessnointerpretationorclas-sicationofthedeterminedechoes,whichwerereectedfromdifferentobjects,isperformed.However,forthegenerationofaDTMtheclassicationoftheALSdataintoterrainandoff-terrainpointsisessential.Thisseparation,whichisimportantforotherapplications(e.g.vegetationandpowerlinemapping),isoftenalsoentitledas“ltering”.Inthepast,manydifferentsolutionsforthelteringoftheALSdatawerepublished(cf.SitholeandVosselman(2004)).Ononehandthesemethodscanbeclassiedbytheinputdatatheyuse(onetypeofmethodsusesrasterizedALSdatawhileothersusetheoriginalALSpointcloud)whereasontheotherhandtheycan 316 begroupedbythedifferentconceptstheyuseinordertoclas-sifythedata.Onegroupofalgorithmsarethemorphologicallters(e.g.Vosselman(2000)),whichuseastructureelement,describingadmissibleheightdifferencesasafunctionofthehor-izontaldistance.Anothergrouparetheprogressivedensicationmethods(Axelsson2000;vonHansenandVogtle1999).TheystartwitharoughapproximationoftheDTMwithinitialterrainpoints(typicallythelowestpointwithinacertaingridcell)anditerativelydensifytheDTMbytheevaluationofasetofrules(e.g.maximaldistancetotheDTMapproximation,anglecrite-ria,etc.).Thethirdgroupofltermethodsworksurfacebased(KrausandPfeifer1998;Elmqvist,Jungert,Lantz,Persson,andoderman2001).Theyuseasurfacemodelthatiterativelyap-proachestheDTMcalculatedbasedontheentirepointsetbyadaptingtheinuenceoftheindividualinputpoints.Finally,re-centlyasetofsegmentationbasedmethodswerepublished(e.g.SitholeandVosselman(2005)andTovariandPfeifer(2005)).Intherststep,thesemethodssegmenttheALSdatawithalocalneighborhoodanalysisandsubsequentlyclassifythesegmentsbydifferentstrategies.Mostoftheexistingmethodsdonotconsiderfurtherinputdata(e.g.orthophotos)andonlyanalyzethegeo-metricrelationbetweenneighboredALSpoints.Acomparisonoftheperformanceofdifferentmethodscanbefoundin(SitholeandVosselman2004).DoneusandBriese(2006)studiedtheadvancedpossibilitiesforDTMgenerationusingfull-waveformALSdata.Theyusedtheechowidth,whichwasdeterminedwiththehelpofaGaussiandecompositionofthefull-waveform(FWF)signalforeachecho(Wagner,Ullrich,Ducic,Melzer,andStudnicka2006).Thepo-tentialofthisfurtherinformationfortheeliminationoflowveg-etationcouldbedemonstrated.Withthehelpofapre-lterstepthateliminatesechoeswithahigherechowidthasignicantim-provementoftheDTMcouldbeachieved.However,uptonowitisnotstudiedindetailhow(andif)theadditionalFWFinforma-tioncanbeusedforadvancedmodelingtasks.6SUMMARYInthispaperanoverviewondataacquisitionandtherstpro-cessingstepswasgivenforairborneandterrestriallaserscanning.Thereisasmallnumberofstandardproducts,e.g.theDTM,thatareproducedroutinelyandefciently.Inordertoincreaseau-tomationforotherapplications,furtherdevelopmentoftherstprocessingsteps,especiallyregistration,segmentation,ander-ror/outlierremoval,isnecessary.Forcalibration,geometricandphysicalaspectswillhavetobeconsideredsimultaneously.Alsotheapplication-specicapproachesarestillmatterofresearch,e.g.buildingreconstruction.Thehardwaredevelopmentintherecentyearshasbeenfast,consideringforexampletheincreaseinpulserepetitionrate.However,thesuccessrateinobjectrecon-structiondidnotgrowlinearlywithit.Therefore,mostresearcheffortwillhavetobespendintheseapplicationspecicelds.=InternationalArchivesofPhotogrammetryandRemoteSensingConf.onO3D=ConferenceonOptical3-DMeasurementTechniques3rdTech,2007.www.3rdtech.com.Homepageofthecompany3rdTech,accessed:June2007.Abmayr,T.,Dalton,G.,Hines,D.,Liu,R.,Hartl,F.,Hirzinger,G.,andohlich,C.,2005.Standardizationandvisualizationof2.5dscanningdataandcolorinformationbyinversemapping.In:GruenandKahmen7thConf.onO3D,Vienna,Austria,pp.164–173.Ahokas,E.,Kaartinen,H.,andHyyppa,J.,2003.Aqualityassessmentofairbornelaserscannerdata.In:IAPRS,XXXIV,3/W13,Dresden,Ger-many.Ahokas,E.,Kaasalainen,S.,Hyyppa,J.,andSuomalainen,J.,2006.Cal-ibrationoftheOptechALTM3100laserscannerintensitydatausingbrightnesstargets.In:IAPRS,XXXVI,1,Paris,France.Al-Manasir,K.andFraser,C.,2006.Registrationofterrestriallaserscan-nerdatausingimagery.ThePhotogrammetricRecord21(115),255–268.Axelsson,P.,2000.DEMgenerationfromlaserscannerdatausingadap-tiveTINmodels.In:IAPRS,XXXIII,B4,Amsterdam,Netherlands,pp.Behan,Maas,andVosselman,2000.Stepstowardsqualityimprovementofairbornelaserscannerdata.In:Proceedingsofthe26thannualconfer-enceoftheRemoteSensingSociety,Leicester,UK.Belton,D.andLichti,D.,2006.Classicationandsegmentationofter-restriallaserscannerpointcloudsusinglocalvarianceinformation.In:IAPRS,XXXVI,5,Dresden,Germany.Beraldin,J.-A.,El-Hakim,S.,andBlais,F.,2007.Traceable3dimag-ingmetrology:evaluationof3ddigitizingtechniquesandcommercial3dsystems/scannersinadedicatedmetrologylaboratory.In:8thConf.on,Zurich,Switzerland.Besl,P.andMcKay,N.,1992.Amethodforregistrationof3Dshapes.IEEETransactionsonPatternAnalysisandMachineIntelligence14239–256.Blais,F.,2004.Reviewof20yearsofrangesensordevelopment.JournalofElectronicImaging13(1),231–240.ohm,J.andBecker,S.,2007.Automaticmarker-freeregistrationofterrestriallaserscansusingreectancefeatures.In:8thConf.onO3DZurich,Switzerland.Bollweg,A.anddeLange,R.,2003.Watruisterdoorhetstruikge-was.TechnicalReportAGI-GAR-2003-22(inDutch),AdviesdienstGeo-InformatieenICT,Rijkswaterstaat,TheNetherlands.Brenner,C.,Dold,C.,andRipperda,N.,2007.Coarseorientationofter-restriallaserscansinurbanenvironments.ISPRSJournalofPhotogram-metryandRemoteSensing.Acceptedforpublication.Breuckmann,B.,Rein,H.,Pfeifer,J.,andMutsch,J.,2007.Acceptancetestoftopometric3d-scanners:experienceswiththeguidelinevdi/vde2634.In:8thConf.onO3D,Zurich,Switzerland.Burman,H.,2000.Adjustmentoflaserscannerdataforcorrectionofori-entationerrors.In:IAPRS,XXXIII,B3/1,Amsterdam,TheNetherlands,pp.125–132.Burman,H.,2002.Laserstripadjustmentfordatacalibrationandveri-cation.In:IAPRS,XXXIV,3,Graz,Austria,pp.67–72.Callidus,2007.www.trimble.com.HomepageofthecompanyCAL-LIDUSprecisionsystemsGmbH,accessed:June2007.Crombaghs,M.,Brugelmann,R.,anddeMin,E.,2000.Onthead-justmentofoverlappingstripsoflaseraltimeterheightdata.In:XXXIII,3A,Amsterdam,Netherlands,pp.230–237.Csanyi,N.andToth,C.,2007.Improvementoflidardataaccuracyusinglidar-specicgroundtargets.PhotogrammetricEngineering&RemoteSensing73(4),385–396.DiamondAirborneSensing,2007.www.diamond-air.at.HomepageofthecompanyDiamondAircraft,accessed:August2007,inGerman.Dold,C.andBrenner,C.,2006.Registrationofterrestriallaserscanningdatausingplanarpatchesandimagedata.In:IAPRS,XXXVI,5,Dresden,Germany,pp.78–83.Doneus,M.andBriese,C.,2006.Digitalterrainmodellingforarchae-ologicalinterpretationwithinforestedareasusingfull-waveformlaser-scanning.In:The7thInternationalSymposiumonVirtualReality,Ar-chaeologyandCulturalHeritageVAST,Zypern.Ellum,C.andElSheimy,N.,2006.Newstrategiesforintegratingpho-togrammetricandGNSSdata.In:IAPRS,XXXVI/5,Dresden,Germany,pp.103–108.Elmqvist,M.,Jungert,E.,Lantz,F.,Persson,A.,andSoderman,U.,2001.Terrainmodellingandanalysisusinglaserscannerdata.In:XXXIV,3/W4,Annapolis,MD,USA.Faro,2007.www.faro.com.HomepageofthecompanyFAROTechnolo-gies,Inc.,accessed:June2007.Filin,S.,2002.Surfaceclusteringfromairbornelaserscanningdata.In:IAPRS,XXXII,3A,Graz,Austria,pp.119–124. 317 Filin,S.,2003.Recoveryofsystematicbiasesinlaseraltimetrydatausingnaturalsurfaces.PhotogrammetricEngineering&RemoteSens-ing69(11),1235–1242.Filin,S.andPfeifer,N.,2005.NeighborhoodsystemsforairbornelaserPhotogrammetricEngineering&RemoteSensing71(6),743–755.Filin,S.andPfeifer,N.,2006.Segmentationofairbornelaserscanningdatausingaslopeadaptiveneighborhood.ISPRSJournalofPhotogram-metryandRemoteSensing60(2),71–80.Finnegan,D.C.,Krabill,W.,Lichvar,R.W.,Ericsson,M.P.,Freder-ick,E.,Manizade,S.,Yungel,J.,Sonntag,J.,andSwift,R.,2005.Us-ingNASA`sAirborneTopographicMapperIVtoQuantifyGeomorphicChangeinAridSouthwesternStreamSystems.In:AGUFallMeetingAbstractsFischler,M.A.andBoller,R.C.,1982.Randomsampleconsensus:Aparadigmformodelttingwithapplicationstoimageanalysisandauto-matedcartography.CommunicationsoftheACM24(6),381–395.Fli-Map,2007.www.imap.nl.HomepageofFLI-MAP,accessed:Au-gust2007.Flood,M.andGutelius,B.,1997.Commercialimplicationsoftopo-graphicterrainmappingusingscanningairbornelaserradar.Photogram-metricEngineering&RemoteSensing63,327–365.Fox,C.S.,Accetta,J.S.,andShumaker,D.L.,1993.ActiveElectro-OpticalSystems,Volume6ofTheinfraredandelectro-opticalsystems.SPIEOpticalEngineeringPress.Geibel,R.andStilla,U.,2000.Segmentationoflaseraltimetrydataforbuildingreconstruction:differentproceduresandcomparison.In:IAPRS,XXXIII,B3,Amsterdam,theNetherlands,pp.326–334.GIM,2007.www.gim-international.com/productsurvey.ProductsurveysofGIMinternational,accessed:August2007.Godin,Rioux,Beraldin,Levoy,Cournoyer,andBlais,2001.Anassess-mentoflaserrangemeasurementonmarblesurfaces.In:5thConf.on,Vienna,Austria.Hanke,K.,Grussenmeyer,P.,Grimm-Pitzinger,A.,andWeinold,T.,2006.Firstexperienceswiththetrimblegxscanner.In:IAPRS,XXXVI,,Dresden,Germany.Heckbert,P.S.andGarland,M.,1997.Surveyofpolygonalsurfacesim-plicationalgorithms.Technicalreport,SchoolofComputerScience,CarnegieMellonUniversity,Pittsburgh,PA.TechnicalReport.Hesse,C.,Neuner,H.,andKutterer,H.,2005.Statisticalanalysisofkinematiclasersscans.In:7thConf.onO3D,Vienna,Austria,pp.103–Hodgson,M.E.andBresnahan,P.,2004.Accuracyofairbornelidar-derivedelevation:Empiricalassessmentanderrorbudget.Photogram-metricEngineering&RemoteSensing70(3),331–339.oe,B.andPfeifer,N.,2007.Correctionoflaserscanningintensitydata:Dataandmodel-drivenapproaches.ISPRSJournalofPhotogram-metryandRemoteSensing.Acceptedforpublication.Hoover,Jean-Baptiste,Jiang,Flynn,Bunke,Goldgof,Bowyer,Eggert,Fitzgibbon,andFisher,1996.Anexperimentalcomparisonofrangeim-agesegmentationalgorithms.IEEETransactionsonPatternAnalysisandMachineIntelligence18(7),673–689.Hopkinson,D.,Chasmer,L.,Zsigovics,G.,Creed,I.,Sitar,M.,Treitz,P.,andMaher,R.,2004.Errorsinlidargroundelevationandwetlandvege-tationheightestimates.In:IAPRS,XXXVI,8/W2,Freiburg,Germany.Horn,B.,1987.Closed-formsolutionofabsoluteorientationusingunitJournaloftheOpicalSocientyofAmericaA4,629–642.Hug,C.andWehr,A.,1997.Detectingandidentifyingtopographicob-jectsinimaginglaseraltimetrydata.In:IAPRS,XXXII,4/W2,pp.16–29.a,J.,Hyyppa,H.,Litkey,P.,Yu,X.,Haggren,H.,Roennholm,P.,Pyysalo,U.,Pitkaenen,J.,andMaltamo,M.,2004.Algorithmsandmethodsofairbornelaser-scanningforforestmeasurements.In:XXXVI,8/W2,Freiburg,Germany.I-SiTE,2007.www.isite3d.com.HomepageofthecompanyI-SiTEPtyLtd,accessed:June2007.Jelalian,A.V.,1992.LaserRadarSystems.ArtechHouse,Boston.Jutzi,B.andStilla,U.,2003a.Analysisoflaserpulsesforgainingsurfacefeaturesofurbanobjects.In:2ndGRSS/ISPRSJointWorkshoponRemoteSensinganddatafusiononurbanareas.URBAN2003,IEEE,pp.13–17.Jutzi,B.andStilla,U.,2003b.Laserpulsesanalysisforreconstructionandclassicationofurbanobjects.In:IAPRS,XXXIV,3/W8,Munich,Germany,pp.151–156.Kager,H.,2004.Discrepanciesbetweenoverlappinglaserscanningstrips-simultaneousttingofaeriallaserscannerstrips.In:IAPRS,XXXV,B/1Istanbul,Turkey,pp.555–560.Karimi,H.A.andHammad,A.,2004.Telegeoinformatics:location-basedservicesandcomputing.CRCPress,BocaRaton.Katzenbeisser,R.,2003.Technicalnoteon:Echodetection.TopoSysGmbH,Germany.Kilian,J.,Haala,N.,andEnglich,M.,1996.Captureandevaluationofairbornelaserscannerdata.In:IAPRS,XXXI,B3,Vienna,Austria,pp.Kraus,K.andPfeifer,N.,1998.Determinationofterrainmodelsinwoodedareaswithairbornelaserscannerdata.ISPRSJournalofPho-togrammetryandRemoteSensing53,193–203.Kraus,K.andPfeifer,N.,2001.AdvancedDTMgenerationfromLIDARdata.In:IAPRS,XXXIV,3/W4,Annapolis,MD,USA,pp.23–30.Latypov,D.,2005.Effectsoflaserbeamalignmenttoleranceonlidaraccuracy.ISPRSJournalofPhotogrammetryandRemoteSensing59Leica,2007.www.leica-geosystems.com.HomepageofthecompanyLeicaGeosystems,accessed:August2007.Lichti,D.,2007.Errormodelling,calibrationandanalysisofanamcwterrestriallaserscannersystem.ISPRSJournalofPhotogrammetryandRemoteSensing61(5),307–324.Lichti,D.,Gordon,S.,andTipdecho,T.,2005.Errormodelsandpropaga-tionindirectlygeoreferencedterrestriallaserscannernetworks.JournalofSurveyingEngineering131(4),135–142.Lin,1997.Waveformsamplinglidarapplicationsincomplexterrain.ternationalJournalofRemoteSensing18Lindenberger,1989.Testresultsoflaserprolingfortopographicterrainsurvey.TechnicalReport13,InstituteofPhotogrammetry,UniversityofStuttgart.Lindenberger,J.,1993.Laser-ProlmessungenzurtopographischenGel-.Ph.D.thesis,UniversityStuttgart,InstituteofPhotogram-metry.Lohr,U.andEibert,1995.Thetoposyslaserscanner-system.In:FritschandHobbie(Eds.),PhotogrammetricWeekLowe,D.G.,2004.Distinctiveimagefeaturesfromscale-invariantkey-InternationalJournalofComputerVision60(2),91–110.Medioni,G.,Lee,M.-S.,andTang,C.-K.,2000.Acomputationalframe-workforsegmentationandgrouping.Elsevier,Amsterdam.Melzer,T.,2007.Themeanshiftalgorithmforclusteringpointclouds.JournalofAppliedGeodesy.Acceptedforpublication.Naesset,E.,1997.Determinationofmeantreeheightofforeststandsusingairbornelaserscannerdata.ISPRSJournalofPhotogrammetryandRemoteSensing52(2),49–56.Optech,2007.www.optech.ca.HomepageofthecompanyOptech,ac-cessed:June2007.OudeElberink,S.andCrombaghs,M.,2004.LaseraltimetrievoordehoogtemetingenvandekweldersWaddenzee.TechnicalReportAGI-GAP-2003-50(inDutch),AdviesdienstGeo-InformatieenICT,Rijkswa-terstaat,TheNetherlands.Parian,J.A.andGruen,A.,2005.Integratedlaserscannerandintensityimagecalibrationandaccuracyassessment.In:IAPRS,XXXVI,3/W19Enschede,theNetherlands,pp.18–23.Pauly,M.,Gross,M.,andKobbelt,L.,2002.Efcientsimplicationofpoint-sampledsurfaces.In:ProceedingsoftheconferenceonVisualiza-tion'02,Boston,Massachusetts,pp.163–170.IEEEComputerSociety.Pfeifer,N.,Dorninger,P.,Haring,A.,andFan,H.,2007.Investigatingterrestriallaserscanningintensitydata:Qualityandfunctionalrelations.8thConf.onO3D,Zurich,Switzerland.Pfeifer,N.,Gorte,B.,andOudeElberink,S.,2004.Inuencesofvegeta-tiononlaseraltimetry–analysisandcorrectionapproaches.In:XXXVI,8/W2,Freiburg,Germany,pp.283–287. 318 POB,2007.www.pobonline.com.ProductsurveysofPointofBeginning,accessed:August2007.Rabbani,T.,Dijkman,S.,vandenHeuvel,F.,andVosselman,G.,2007.AnintegratedapproachformodellingandglobalregistrationofpointISPRSJournalofPhotogrammetryandRemoteSensing61,355–Rabbani,T.andvandenHeuvel,F.,2005.Efcienthoughtransformforautomaticdetectionofcylindersinpointclouds.In:IAPRS,XXXVI,,Enschede,theNetherlands,pp.60–65.ReedandLynch,1996.Neareldairborneremotesensingusingalasermappingsystemonelectrictransmissionlinecorridorsurveysandcapac-ityanalyses.In:InternationalAirborneRemoteSensingConferenceandExhibition–Technology,Measurement&Analysis,SanFrancisco,USA.Reshetyuk,Y.,2006.Calibrationofterrestriallaserscannersforthepur-posesofgeodeticengineering.In:3rdIAGSymposiumonGeodesyforGeotechnicalandStructuralEngineeringand12thFIGSymposiumonDeformationMeasurements,Baden,Austria.Riegl,2007.www.riegl.com.HomepageofthecompanyRIEGLLaserMeasurementSystemsGmbH,accessed:August2007.Ries,C.,Kager,H.,andStadler,P.,2002.High-qualitygeo-referencingofGPS/IMU-supportedmulti-spectralairbornescannerdata-experiencesandresults.In:IAPRS,XXXIV/3,Graz,Austria,pp.222–226.RitchieandPachepsky,1998.Seasonalchangesinfractallandscapesur-faceroughnessestimtedfromairbornelaseraltimetrydata.JournalofRemoteSensing19Rottensteiner,F.,Trinder,J.,Clode,S.,andKubik,K.,2005.Automateddelineationofroofplanesfromlidardata.In:IAPRS,XXXVI,3/W19Enschede,theNetherlands.Rueger,J.,1990.ElectronicDistanceMeasurement:anintroductioned.).Springer-Verlag,Heidelberg,Germany.Rusinkiewicz,S.andLevoy,M.,2001.Efcientvariantsoftheicpalgo-rithm.In:ThirdInternationalConferenceon3-DDigitalImagingandModeling(3DIM'01),LosAlamitos,CA,USA,pp.145.Schmitz,Wuebbena,Bagge,andKruck,2001.Benetofrigorousmodellingofgpsincombinedat/gps/imu-bundleblockadjustment.In:OEEPEWorkshop,Hannover,Germany.Sithole,G.andVosselman,G.,2004.Experimentalcomparisonofl-teringalgorithmsforbare-earthextractionfromairbornelaserscanningpointclouds.ISPRSJournalofPhotogrammetryandRemoteSens-ing59(1-2),85–101.Sithole,G.andVosselman,G.,2005.Filteringofairbornelaserscan-nerdatabasedonsegmentedpointclouds.In:IAPRS,XXXVI,3/W19Enschede,TheNetherlands,pp.66–71.Skaloud,J.,2006.Reliabilityofdirectgeoreferencingphase0–anoverviewofcurrentapproachesandpossiblities.In:EuroSDROfcialPublication51,pp.143–168.Skaloud,J.andLichti,D.,2006.Rigorousapproachtoboresightself-calibrationinairbornelaserscanning.ISPRSJournalofPhotogrammetryandRemoteSensing61(1),47–59.Sotoodeh,S.,2006.Outlierdetectioninlaserscannerpointclouds.In:IAPRS,XXXVI,5,Dresden,Germany.TopEye,2007.www.blomasa.com.HomepageoftheBLOMGROUP,accessed:Augsut2007.TopoSys,2007.www.toposys.com.HomepageofthecompanyTopoSysGmbH,accessed:June2007.ovari,D.andPfeifer,N.,2005.Segmentationbasedrobustinterpola-tion-anewapproachtolaserdataltering.In:IAPRS,XXXVI,3/W19Enschede,TheNetherlands,pp.79–84.Trimble,2007.www.trimble.com/.HomepageofthecompanyTrimble,accessed:June2007.Valanis,A.andTsakiri,M.,2004.Automatictargetidenticationforlaserscanners.In:ProceedingsofXXthISPRSCongress,Istanbul,Turkey.ogtle,T.andSteinle,2004.Detectionandrecognitionofchangesinbuildinggeometryderivedfrommultitemporallaserscanningdata.In:IAPRS,XXXV,Istanbul,Turkey.vonHansen,W.,Michaelsen,E.,andThonnessen,U.,2006.Clusteranal-ysisandprioritysortinginhugepointcloudsforbuildingreconstruction.18thInternationalConferenceonPatternRecognition(ICPR'06),pp.vonHansen,W.andVogtle,T.,1999.ExtraktionderGelausugzeuggetragenenLaserscanner–Aufnahmen.PhotogrammetrieFernerkundungGeoinformation4,229–236.Vosselman,G.,2000.Slopebasedlteringoflaseraltimetrydata.In:IAPRS,XXXIII,B3,Amsterdam,Netherlands,pp.935–942.Vosselman,G.,Gorte,B.G.H.,Sithole,G.,andRabbani,T.,2004.Recognisingstructureinlaserscannerpointclouds.In:IAPRS,XXXVI,,Freiburg,Germany.Vosselman,G.andMaas,H.-G.,2001.Adjustmentandlteringofrawlaseraltimetrydata.In:ProceedingsofOEEPEWorkshoponAir-borneLaserscanningandInterferometricSARforDetailedDigitalTer-rainModels,Stockholm,Sweden.Wagner,W.,Ullrich,A.,Ducic,V.,Melzer,T.,andStudnicka,N.,2006.Gaussiandecompositionandcalibrationofanovelsmall-footprintfull-waveformdigitisingairbornelaserscanner.ISPRSJournalofPhotogram-metryandRemoteSensing60(2),100–112.Wagner,W.,Ullrich,A.,Melzer,T.,Briese,C.,andKraus,K.,2004.Fromsingle-pulsetofull-waveformairbornelaserscanners:potentialandpracticalchallenges.In:IAPRS,XXXV,Istanbul,Turkey.Wehr,A.,Hemmleb,M.,Thomas,M.,andMaierhofer,C.,2007.Mois-turedetectiononbuildingsurfacesbymulti-spectrallaserscanning.In:8thConf.onO3D,Zurich,Switzerland.Wehr,A.,Kleusberg,A.,andSchiele,O.,2006.CW-laser-scanningstate-of-the-artatINSandfuturemultispectralsensors.In:L.GruendigandO.Altan(Eds.),Turkish-GermanJointGeodeticDays”GeodesyandGeoinformationintheServiceofourDailyLife”,Berlin,Germany.Wehr,A.andLohr,U.,1999.Airbornelaserscanninganintroductionandoverview.ISPRSJournalofPhotogrammetryandRemoteSensing543),68–82.ohlich,2007.www.zofre.de.HomepageofthecompanyohlichGmbH,accessed:June2007.Zwally,H.,Schutz,B.,Abdalati,Abshire,Bentley,Brenner,Bufton,Dezio,Hancock,Harding,Herring,Minster,Quinn,Palm,Spinhirne,andThomas,2002.Icesat'slasermeasurementsofpolarice,atmosphere,ocean,andland.JournalofGeodynamics34(3),405–445. 319 ISPRS Workshop on Laser Scanning 2007 and SilviLaser 2007, Espoo, September 12-14, 2007, Finland