Radar Remote Sensing of Urban Areas Remote Sensing and Digital Image Processing VOLUME15 SeriesEditor: EARSelSeriesEditor: FreekD.vanderMeer AndréMarçal DepartmentofEarthSystemsAnalysis DepartmentofAppliedMathematics InternationalInstituitefor FacultyofSciences Geo-InformationScienceand UniversityofPorto EarthObservation(ITC) Porto,Portugal Enchede,TheNetherlands & DepartmentofPhysicalGeography FacultyofGeosciences UtrechtUniversity TheNetherlands EditorialAdvisoryBoard: EARSelEditorialAdvisoryBoard: MichaelAbrams MarioA.Gomarasca NASAJetPropulsionLaboratory CNR-IREAMilan,Italy Pasadena,CA,U.S.A. MarttiHallikainen PaulCurran HelsinkiUniversityofTechnology UniversityofBournemouth,U.K. Finland ArnoldDekker HåkanOlsson CSIRO,LandandWaterDivision SwedishUniversity Canberra,Australia ofAgriculturalSciences Sweden StevenM.deJong DepartmentofPhysicalGeography EberhardParlow FacultyofGeosciences UniversityofBasel UtrechtUniversity,TheNetherlands Switzerland MichaelSchaepman RainerReuter DepartmentofGeography UniversityofOldenburg UniversityofZurich,Switzerland Germany Forothertitlespublishedinthisseries,goto http://www.springer.com/series/6477 Radar Remote Sensing of Urban Areas Uwe Soergel Editor LeibnizUniversitätHannover InstituteofPhotogrammetryandGeoInformation,Germany 123 Editor UweSoergel LeibnizUniversitätHannover InstituteofPhotogrammetryandGeoInformation NienburgerStr.1 30167Hannover Germany [email protected] Coverillustration:Fig.7inChapter11inthisbook ResponsibleSeriesEditor:AndréMarçal ISSN1567-3200 ISBN978-90-481-3750-3 e-ISBN978-90-481-3751-0 DOI10.1007/978-90-481-3751-0 SpringerDordrechtHeidelbergLondonNewYork LibraryofCongressControlNumber:2010922878 (cid:2)c SpringerScience+BusinessMediaB.V.2010 Nopartofthisworkmaybereproduced,storedinaretrievalsystem,ortransmittedinanyformorby anymeans,electronic,mechanical,photocopying,microfilming,recordingorotherwise,withoutwritten permissionfromthePublisher,withtheexceptionofanymaterialsuppliedspecificallyforthepurpose ofbeingenteredandexecutedonacomputersystem,forexclusiveusebythepurchaserofthework. Coverdesign:deblik,Berlin Printedonacid-freepaper SpringerispartofSpringerScience+BusinessMedia(www.springer.com) Preface One of the key milestones of radar remote sensing for civil applications was the launchoftheEuropeanRemoteSensingSatellite1(ERS1)in1991.Theplatform carriedavarietyofsensors;theSyntheticApertureRadar(SAR)iswidelyconsid- ered to be the most important.This active sensing technique providesall-day and all-weather mappingcapability of considerablyfine spatial resolution. ERS 1 and its sister system ERS 2 (launch 1995) were primarily designed for ocean appli- cations, but soon the focus of attention turned to onshore mapping. Examples for typical applications are land cover classification also in tropical zones and moni- toringofglaciersorurbangrowth.Inparallel,internationalSpaceShuttleMissions dedicated to radar remote sensing were conducted starting already in the 1980s. ThemostprominentweretheSIR-C/X-SARmissionfocussingontheinvestigation ofmulti-frequencyandmulti-polarizationSARdataandthefamousShuttleRadar TopographyMission (SRTM). Data acquired during the latter enabled to derive a DEM of almostglobalcoverageby meansof SAR Interferometry.Itis indispens- ableeventodayandformanyregionsthebestelevationmodelavailable.Differential SARInterferometrybasedontimeseriesofimageryoftheERSsatellitesandtheir successorEnvisatbecameanimportantanduniquetechniqueforsurfacedeforma- tionmonitoring. The spatial resolution of those devices is in the order of some tens of meters. Imageinterpretationfromsuch datais usuallyrestrictedto radiometricproperties, which limits the characterization of urban scenes to rather general categories, for example,thediscriminationofsuburbanareasfromcitycores.Theadventofanew sensorgenerationchangedthissituationfundamentally.Systemslike TerraSAR-X (Germany)andCOSMO-SkyMed(Italy)achievegeometricresolutionofabout1m. In addition,these sophisticatedsystems are moreagileand provideseveralmodes tailored for specific tasks. This offers the opportunity to extend the analysis to individual urban objects and their geometrical set-up, for instance, infrastructure elements like roads and bridges, as well as buildings. In this book, potentials and limits of SAR for urban mapping are described, including SAR Polarimetry and SARInterferometry.Applicationsaddressedcompriserapidmappingincaseoftime critical events, road detection, traffic monitoring, fusion, building reconstruction, SARimagesimulation,anddeformationmonitoring. v vi Preface Audience This bookis intendedto providea comprehensiveoverviewof the state-of-theart of urban mapping and monitoring by modern satellite and airborne SAR sensors. The reader is assumed to have a background in geosciences or engineering and to be familiar with remote sensing concepts. Basics of SAR and an overview of differenttechniquesandapplicationsaregiveninChapter1.Allchaptersfollowing thereafterfocusoncertainapplications,whicharepresentedingreatdetailbywell knownexpertsinthesefields. In case of natural disaster or political crisis rapid mapping is a key issue (Chapter 2). An approach for automated extraction of roads and entire road net- worksispresentedinChapter3.Atopiccloselyrelatedtoroadextractionistraffic monitoring.Incase ofSAR, Along-TrackInterferometryis apromisingtechnique for this task, which is discussed in Chapter 4. Reflections at surface boundaries mayalterthepolarizationplaneofthesignal.InChapter5,thiseffectisexploited for objectrecognitionfrom a set of SAR imagesof differentpolarizationstates at transmitandreceive.Often,up-to-dateSARdatahastobecomparedwitharchived imageryofcomplementingspectraldomains.AmethodforfusionofSARandop- tical images aiming at classification of settlements is described in Chapter 6. The opportunitytodeterminetheobjectheightabovegroundfromSARInterferometry is of course attractive for building recognition. Approaches designed for mono- aspectandmulti-aspectSARdataareproposedinChapters7 and 8, respectively. Such methods may benefit from image simulation techniques that are also useful foreducation.InChapter9,amethodologyoptimizedforreal-timerequirementsis presented.Monitoringofsurfacedeformationsuffersfromtemporalsignaldecorre- lation especiallyin vegetatedareas. However,in cities manytemporallypersistent scatteringobjectsarepresent,whichallowtrackingofdeformationprocesseseven for periodsof severalyears. Thistechniqueis discussed in Chapter10. Finally,in Chapter11,designconstraintsofamodernairborneSARsensorarediscussedfor thecaseofanexistingdevicetogetherwithexamplesofhigh-qualityimagerythat state-of-the-artsystemscanprovide. UweSoergel Contents 1 ReviewofRadarRemoteSensingonUrbanAreas....................... 1 UweSoergel 1.1 Introduction............................................................. 1 1.2 Basics................................................................... 2 1.2.1 ImagingRadar................................................ 3 1.2.2 Mappingof3dObjects....................................... 8 1.3 2dApproaches.......................................................... 11 1.3.1 Pre-processingandSegmentationofPrimitiveObjects..... 11 1.3.2 ClassificationofSingleImages.............................. 13 1.3.2.1 DetectionofSettlements.......................... 14 1.3.2.2 CharacterizationofSettlements.................. 15 1.3.3 ClassificationofTime-SeriesofImages..................... 16 1.3.4 RoadExtraction............................................... 17 1.3.4.1 RecognitionofRoadsandofRoadNetworks... 17 1.3.4.2 Benefit of Multi-aspectSAR ImagesforRoadNetworkExtraction............ 19 1.3.5 DetectionofIndividualBuildings ........................... 20 1.3.6 SARPolarimetry ............................................. 20 1.3.6.1 Basics.............................................. 21 1.3.6.2 SARPolarimetryforUrbanAnalysis............ 23 1.3.7 FusionofSARImageswithComplementingData ......... 24 1.3.7.1 ImageRegistration................................ 24 1.3.7.2 FusionforLandCoverClassification............ 25 1.3.7.3 Feature-Based Fusion of High-ResolutionData............................. 26 1.4 3dApproaches.......................................................... 26 1.4.1 Radargrammetry.............................................. 27 1.4.1.1 SingleImage...................................... 27 1.4.1.2 Stereo.............................................. 28 1.4.1.3 ImageFusion...................................... 29 vii viii Contents 1.4.2 SARInterferometry .......................................... 29 1.4.2.1 InSARPrinciple .................................. 29 1.4.2.2 AnalysisofaSingleSARInterferogram ........ 32 1.4.2.3 Multi-imageSARInterferometry................ 34 1.4.2.4 Multi-aspectInSAR............................... 34 1.4.3 FusionofInSARDataandOtherRemote SensingImagery.............................................. 36 1.4.4 SARPolarimetryandInterferometry........................ 37 1.5 SurfaceMotion......................................................... 38 1.5.1 DifferentialSARInterferometry............................. 38 1.5.2 PersistentScattererInterferometry........................... 39 1.6 MovingObjectDetection.............................................. 40 References...................................................................... 41 2 RapidMappingUsingAirborneandSatelliteSARImages ............. 49 FabioDell’AcquaandPaoloGamba 2.1 Introduction............................................................. 49 2.2 AnExampleProcedure................................................. 51 2.2.1 Pre-processingoftheSARImages .......................... 51 2.2.2 ExtractionofWaterBodies .................................. 52 2.2.3 ExtractionofHumanSettlements............................ 53 2.2.4 ExtractionoftheRoadNetwork............................. 54 2.2.5 ExtractionofVegetatedAreas ............................... 56 2.2.6 OtherSceneElements........................................ 57 2.3 ExamplesonRealData ................................................ 57 2.3.1 TheChengduCase............................................ 58 2.3.2 TheLuojiangCase............................................ 61 2.4 Conclusions............................................................. 64 References...................................................................... 66 3 FeatureFusion Based on Bayesian NetworkTheory forAutomaticRoadExtraction............................................. 69 UweStillaandKarinHedman 3.1 Introduction............................................................. 69 3.2 BayesianNetworkTheory............................................. 70 3.3 StructureofaBayesianNetwork...................................... 72 3.3.1 Estimating Continuous Conditional ProbabilityDensityFunctions ............................... 76 3.3.2 DiscreteConditionalProbabilities........................... 79 3.3.3 EstimatingtheA-PrioriTerm................................ 80 3.4 Experiments ............................................................ 81 3.5 DiscussionandConclusion............................................ 82 References...................................................................... 85 Contents ix 4 TrafficDataCollectionwithTerraSAR-X andPerformanceEvaluation................................................ 87 Stefan Hinz, Steffen Suchandt, Diana Weihing, andFranzKurz 4.1 Motivation.............................................................. 87 4.2 SARImagingofStationaryandMovingObjects..................... 88 4.3 DetectionofMovingVehicles......................................... 93 4.3.1 DetectionScheme ............................................ 94 4.3.2 IntegrationofMulti-temporalData.......................... 96 4.4 MatchingMovingVehiclesinSARandOpticalData................ 98 4.4.1 MatchingStaticScenes....................................... 98 4.4.2 TemporalMatching...........................................100 4.5 Assessment .............................................................101 4.5.1 AccuracyofReferenceData .................................101 4.5.2 AccuracyofVehicleMeasurementsinSARImages........103 4.5.3 ResultsofTrafficDataCollection withTerraSAR-X.............................................103 4.6 SummaryandConclusion..............................................107 References......................................................................107 5 ObjectRecognitionfromPolarimetricSARImages......................109 RonnyHa¨nschandOlafHellwich 5.1 Introduction.............................................................109 5.2 SARPolarimetry.......................................................111 5.3 FeaturesandOperators.................................................117 5.4 ObjectRecognitioninPolSARData ..................................124 5.5 ConcludingRemarks...................................................129 References......................................................................130 6 FusionofOpticalandSARImages.........................................133 FlorenceTupin 6.1 Introduction.............................................................133 6.2 ComparisonofOpticalandSARSensors.............................135 6.2.1 Statistics.......................................................136 6.2.2 GeometricalDistortions......................................137 6.3 SARandOpticalDataRegistration ...................................138 6.3.1 KnowledgeoftheSensorParameters........................138 6.3.2 AutomaticRegistration.......................................140 6.3.3 A FrameworkforSARandOpticalData RegistrationinCaseofHRUrbanImages ..................141 6.3.3.1 RigidDeformationComputation andFourier–MellinInvariant.....................141 6.3.3.2 PolynomialDeformation .........................143 6.3.3.3 Results.............................................144