高分辨率遙感衛(wèi)星影像中城市人工地物提取方法研究.pdf

1、中圖分類號：TP751UDC：學(xué)校代碼：10004密級：公開北京交通大學(xué)碩士學(xué)位論文高分辨率遙感衛(wèi)星影像中城市人工地物提取方法研究StudyonUrbanMan—madeObjectsExtractionMethodsinHighResolutionRemoteSensingSatelliteImages作者姓名：付卓導(dǎo)師姓名：胡吉平學(xué)位類別：工學(xué)學(xué)號：04121200職稱：副教授學(xué)位級別：碩士學(xué)科專業(yè)：攝影測量與遙感研究方向：遙感技術(shù)

2、與應(yīng)用北京交通大學(xué)2006年12月韭塞鑾顯叁堂亟堂焦淦塞ABSTRACrABSTRACrRecentyears。withtheimprovementofspatialresolutionofthesatelliteSensors，highresolutionremotesensingimageshavebecomeoneofthemostimportantdata姍“：esforurbaninformationextractionFor

3、supplyingtechnicalsupporttocollectingandupdatingurbandataurbanmallmadeobjects(roedsandbuildings)extractionmethodsinllighresolutionremotesensingsatelliteimagesarestudiedinthispaperAndtheworkinthispaperfocusesOnfollowingas

4、pects：1Consideringthecharacteristicsofhi曲resolutionimages，thetechnicaldifficultiesofextractionfromhighresolutionremotesensingimageareanalyzedAndtheproblemsofthetraditionalsegmentationmethodsbasedonpixelspectralattributea

5、regivenTheadvantagesofthesegmentationmethodsbasedOnobjects’severalkindsofattributesaleillustrated2TheQuickbirdremotesensingimagesofYanchengareselectedasthemainresearchimagesBasedOiltheseimagesthefeaturesofbuildingsandroa

6、dsarcanalyzedindifferentspectralimagesAndaccordingtothosefeatures，single—bandremotesensingimageOrmultibandfusionimagewhichisbestforobjectsextractionisselected鷂thefIlrtherprocessingimage3ObjectOrientedMultisolesegmentatio

7、nmethodisfurtherstudiedAndtheoptimalscaleparametersofroadsandbuildingsaleobtainedrespectivelyt如oughtheexperimentsThenconsideringthefeaturesofroadsandbuildings，theoptimalsegmentationparametersofroadsandbuildingsa肥obtained

8、incorrespondingimagelevelthroughtheexperiments4MeanShiftsegmentationbasedonconfidenceisstudiedAndtheoptimalsegmentationparametersofroadsandbuildingsinthismethodaleobtainedbyexperiments5Theinitialcontoursofmanmadeobjectsa

9、reobtainedbytwomethodsmentionedabovethroughexperiments硼1eresultshowsthatcomparedtoObjectOrientedMulti—scalesegmentationmethod，MeanShiftsegmentationbasedonconfidencehasmoreadvantagesinhi【gllresolutionremotesensingimageseg

10、mentationTherefore，MeanShiftsegmentationbasedonconfidenceisavailableandeffectiveinextractingtheurbanmanmadeobjectsfromhighresolutionremotesensingsatelliteimamsKEYWORDS：HiglIResolution；RemoteSensingImage；Manmadeobjects；Ex