ResearchonImpactofGroundControlPointDistributiononImageGeometricRectificationBasedonVoronoiDiagramAbstractWiththedevelopmentofremotesensingsatellites,thedataquantityofremotesensingimageisincreasingtremendously,whichbringsahugeworkloadtotheimagegeometriccorrectionthroughmanualgroundcontrolpoint(GCP)selection.GCPsautomaticallyselectedbasedonsoftwareisoneoftheeffectivemethodstocutdownmanualoperation.TheGCPsobtainedfromthatwayisgenerallyredundant.Inthispaper,aftersomeexistingmethodsaboutautomaticoptimizationofGCParesummarized,anewmethodofautomaticoptimizationofGCPbasedonVoronoiDiagram(Thiessenpolygon)isproposed.ThismethodusestheratioofThiessenPolygonmax-minareageneratedbyGCPasthebasisandfilterGCPfromtheoverfulloneswithoutmanualsubjectivityforbetteraccuracy.ExperimentsinthispaperalsodemonstratedtherelationshipbetweentheaccuracyofgeometriccorrectionandthedistributionofGCP.?2011PublishedbyElsevierLtd.Selectionand/orpeer-reviewunderresponsibilityoftheIntelligentInformationTechnologyApplicationResearchAssociation.?2011PublishedbyElsevierLtd.Selectionand/orpeer-reviewunderresponsibilityof[nameorganizer]Keywords:GeometricCorrection;GCPDistribution;DelaunayTriangulation;ThiessenPolygon1.IntroductionRemotesensingisbeingusedinthedevelopmentofnationaleconomywidely.Duetotheremotesensingimagingmechanism,theimagescannotbeusedwithoutaseriesofpre-processes,suchasnoisefiltering,radiometriccorrection,geometriccorrection.Groundcontrolpoint(GCP)selectionisanimportantstepinthegeometriccorrectionofremotesensingimage.Theprecisionofgeometriccorrectiondependsonthequantity,accuracyanddistributionofthegroundcontrolpoint(GCP)[1].GCPisanimportantsourceoffundamentaldataintheaerialphotographandremotesensingimage'sGeometricCorrection[2].ThenumberofGCPsthatisneededintheprocessofGeometricCorrectionisdeterminedbydifferentkindsofreasons,including:methodoftheoriginalimagecorrection,thetypeoftheoriginalimageandtherequiredmappingaccuracyofthesoftwarethatisused.TherearemanywaystocollectGCP,mainlydistinguishedbythesourceofthedatasource,including:DigitalRasterGraphic(DRG),DigitalOrthophotoMap(DOM),DigitalLineGraphic(DLG)orfieldsurveying.ThedistributionofGCPhasagreatinfluenceontheGeometricCorrection'saccuracyoftheremotesensingimage.ItisgenerallyrequiredthatthedistributionofGCPshouldbeuniform,sotheycanfullycontroltheaccuracyofmappingregions[3].ItisalsorequiredthateachcorneroftheimageshouldhaveaGCP,soalltheGCPscancontroltheaccuracyofmappingareas.Althoughmountainousterrainiscomplex,wealsoneedtoselectsomeGCPsaccordingtotheactualsituation.GCPmustbearrangedintheclearimagetextureandtheplacecanbeeasilylocated.Wecannottakeitforgrantedandweshouldputqualitybeforequantity.ExperimentsdesignedinthispaperaretoverifytheinfluenceofdistributionofGCPontheresultsofGeometricCorrection.IfGCPisredundant,thispaperalsopresentedamethodtoautomaticallyoptimizetheseGCPandmakethoseGCPwelldistributed.2.MethodilogyInthispaper,originaldatawascollectedonMay6,2005inBeijing.ItwasaTMimagewithresolutionof30meters.Thereferenceimageisanortho-correctedimageproducedbytheU.S.Earthsatcompanywithresolutionof30meters,anditwascollectedonSeptember7,1992.Theexperimentwasdividedintotwoparts:Experiment1:distributionofGCP;Experiment2:optimizationexperimentofGCPbasedonThiessenPolygon.2.1.ExperimentI:ErdasandPCIareusedinthispapertoanalyzetheimpactofGCPdistributiononimagerectification.·Experimentprocedure·SelectionGCPsandcheckpoints(CPs)bymanualorcomputer.Inthispaper,pointsareselectedbyErdasautomatically.SevendifferentkindsofdistributionsofGCPsandcheckpointsaredesigned,asshowninFigure1,whereGCPsareplacedintheshadowarea,checkpointsinblankarea·DesignappropriatenumberofGCPandcheckpointsaccordingtothesizeofblankandshadowarea.ThecorrespondingnumberofpointsandcheckpointsofthesevendistributionsareshowninTableⅠ.TableINumberofGCPandcheckpoints·GeneratetextdocumentdataoftheGCPsandcheckpoints,andprocessingthisdatawithPCItogetaresidualerrorreportforeachdistribution.Becauseofusingorthorectificationmodelasthegeometriccorrectionmodelintheexperiment,theimpactofelevationfactorsonthecalibrationresultsmustbeconsidered.TheelevationofcontrolpointsandcheckpointscanbecalculatedbytheDEMofTMimage.Figure2isGCPandcheckpointsdistributionintheimageofdistributiona.·Reportontheresidualerrorgeneratedbysortingthedata.Residualerrorof7casesofcontrolpointsandcheckpointsareshowninTableⅡ:TableIIResidualerrorof7casesofGCPandcheckpoints·GeneratetheresidualerrorvectordiagramofGCPandcheckpoints.Inthispaper,MATLABisusedtogeneratetheseresidualerrorvectordiagrams,wherecontrolpointsaredrawninblueandcheckpointsinred,asshowninFigure3:·ResultofexperimentFromabovedata,wecandrawconclusionsasfollows:·Unevendistributionofcontrolpointswillleadtoalargerresidualerrorinareawherethereisnocontrolpoint,whichmeansthatcalibrationresultswiththesepointsarefarfromsatisfactory.·Placingaproperdistributionofcontrolpointsontheedgeofremotesensingimagescaneffectivelyimprovetheaccuracyandreliabilityofimagecorrectionresults.·Well-distributedcontrolpointintheentireimagecanimprovetheaccuracyandreliabilityofimagecorrectionresults.·Theresidualerrorofcheckpointsisproportionaltothedistancebetweencheckpointsandtheareaofcontrolpoints.2.2.ExperimentII:AutomaticallyselectedGCPforthemomenthadthreeoptimizationmethods:regulargridmethod,longestdistancemethod,Voroniodiagrammethod(ThiessenPolygonMethod)[4].RegulargridmethodusesregulargridtodivideROI(AreaofInterest).Inordertoachievetheevendistribution,eachgridretainsthesamenumberofGCPswhichhavecomparativelyhigheraccuracy.Themethodissimpleandcanbeactualizedeasily.ButROIisnotnecessarilyaregularregion.Inaddition,thereisnotalwaysanobviousfeaturepointineachgrid.Thereforetheeffectofusingregulargridtocontroldistributionisnotsatisfactory.Thelongestdistancemethod'sresultdependsontheorderofreadingpoints.Thecomputer'sdifferentreadingorderproducesdifferentresult.Inordertoovercomethesedefects,thispaperusesthemethodofautomaticoptimizationofGCPbasedonVoronoidiagram,andmeasuresthedistributionsituationthroughtheindexof“importantvalue”.TheplatformofthisexperimentissoftwareErdasandMapinfo.·Procedureofexperiment·ImportoriginalimageandreferenceimageinErdaswhichcanautomaticallyselectedGCP.·FormatconversionofGCPdata.ChangeGCPdataintoEXCELsheetformat(Note:ReserveIDofGCP,GCP'spixelcoordinatesandgeographiccoordinates.Findoutthefourcorners'pixelcoordinatesandGeographicalcoordinateswithPCIsoftware).·UseGCPtogenerateThiessenpolygons.SoftwareMapinfoautomaticallygeneratedVoronoidiagrambasedonGCP,asshowninFigure4.·GenerateimportantvalueforeachGCP.Usetheconceptof"importantvalue"[5]asanobjectiveevaluationindex.Basedontheareameasurementfunctionofmapinfo,andrecordtheareaofeachpolygon.EachGCPhasaThiessenpolygon,andthisareacanbeseenastheaffectedareaoftheGCP.Thelargeraffectedareais,thehigherimportanceofGCPinGeometricCorrectionis.WeimvcallthepercentageofeachGCP'sareaintotalareaasimportantvalue,andwerepresentofimvimportantvalue.Therefore,weusetheratioofminimumandmaximumofGCP’sasanobjectivestandardtoevaluatewhetherGCPisuniform.TheidealstateofuniformdistributionisthatalltheimportantvaluesofGCPareequal.Itmeanstheratiois1.·OptimizeGCP.Theexperimentretained20,25,30GCPseparately,andthenumberofconvexhullpointsis8.GenerateThiessenpolygonwiththeoptimizedGCPasshowninFigure5:·ResultofexperimentAfterGCPwasoptimized,wecanseethatGCPismoreuniformthanbefore,butit'sonlysubjectivefeeling.Wecoulddescribeitquantitativelythroughtheratioofmaximumandminimum.Afteroptimization,theratioiscloseto1.AsshowninTableⅢ:TableⅢRatiobeforeandafteroptimizationNowweusetheresultofexperimentIItoanalyzeexperimentTableⅣandratioofdifferentsituationsTableⅤandresidualerror·TherelationshipbetweendistributionofGCPandtheaccuracyofGeometricCorrection.ThedataofdistributionofGCPandtheaccuracyofGeometricCorrectionareinTableⅤ.Thereisapositiverelationshipbetweentheratioofmaximumandminimumandresidualerrorofcheckpoints.3.DiscussionAndConclusionInthispaper,thefirstexperimentisaboutthedistributionofGCP.ThereareseventypesofGCPdistributionintheexperiments.Foreachsituation,theorthorectificationmodelisusedforthegeometriccorrection.TheresidualsofGCPsandcheckpointsareusedtoevaluatetheresultsofGeometricCorrection.Someconclusionscanbedrawnbyanalyzingthedata.IftheaccuracyoftheGCPsisunderthesamecircumstances,themoreevendistributionofGCPis,thehigheraccuracyis.ThecorrectionaccuracyisrelatedtothedistanceofpixelfromGCP.TheareaneartotheGCPismoreaccuratethantheotherarea.GCPshouldbeevenlydistributedwithintheimage.Fortheunderstandingofthesubjectivityofuniformdistribution,thereisnoobjectiveconclusion.Inthispaper,weuseimportantvalueasanobjectiveevaluationindex.EachGCPhasaThiessenpolygon,andit'spercentageofpolygon'sareacanmeasuretheimportancevalue.TheresultofmanyexperimentsshowsthattheratiobetweentheThiessenPolygonmax-minareasgeneratedbyGCPisabout4.TherearesomanyfactorsthataffectGeometricCorrection,anddistributionisonlyonefactor.TostudytheimpactofGeometricCorrection,otherfactorsmustbeconsidered,forexample,GCP'samount,GCP’saccuracyandtheirinteraction.AcknowledgmentThisworkwassupportedbytheNationalHigh-TechResearchandDevelopmentPlanofChinaunderGrantNo.2006AA12Z118,theNationalNaturalResourcesandGeo-spatialBasicInformationDatabaseunderGrantNo.JCXXK-HT2008-015andtheNationalNaturalScienceFoundationofChinaunderGrantNo.60972142.ThanksareduetomyfriendsSONGXiaoluandLONGTengfeifortheirassistance.References[1]ZhouJiaxiang,ZhuXiaodong,ZhouXiaoguang.SelectionofGroundControlPointinGeometricCorrectionofRemoteSensingImageBasedonVoronoiDiagram[J].GeographicInformationWorld,2007,(02)[2]HuangJian,WangJi.SelectionofGroundControlPointinGeometricCorrectionofRemoteSensingImage[C].EnvironmentalRemoteSensingAnnualConferenceProceedings.2004[3]ZhaoYingshi.ThePrincipleandMethodofAnalysisofRemoteSensingApplication[M].SciencePress.2003.P176-179[4]ThiessenAH.PrecipitationAveragesforLargeAreas,MonthlyWeatherReview,1911(39):1082～1084[5]YingLi,BoCheng.ResearchonAutomaticOptimizationofGroundControlPointsinImageGeometricRectificationBasedonVoronoiDiagram.ProceedingsofSPIE–TheInternationalSocietyforOpticalEngineering,v7146,2008基于C8051F單片機(jī)直流電動(dòng)機(jī)反饋控制系統(tǒng)的設(shè)計(jì)與研究基于單片機(jī)的嵌入式Web服務(wù)器的研究MOTOROLA單片機(jī)MC68HC（8）05PV8/A內(nèi)嵌EEPROM的工藝和制程方法及對(duì)良率的影響研究基于模糊控制的電阻釬焊單片機(jī)溫度控制系統(tǒng)的研制基于MCS-51系列單片機(jī)的通用控制模塊的研究基于單片機(jī)實(shí)現(xiàn)的供暖系統(tǒng)最佳啟停自校正（STR）調(diào)節(jié)器單片機(jī)控制的二級(jí)倒立擺系統(tǒng)的研究基于增強(qiáng)型51系列單片機(jī)的TCP/IP協(xié)議棧的實(shí)現(xiàn)基于單片機(jī)的蓄電池自動(dòng)監(jiān)測(cè)系統(tǒng)基于32位嵌入式單片機(jī)系統(tǒng)的圖像采集與處理技術(shù)的研究基于單片機(jī)的作物營養(yǎng)診斷專家系統(tǒng)的研究基于單片機(jī)的交流伺服電機(jī)運(yùn)動(dòng)控制系統(tǒng)研究與開發(fā)基于單片機(jī)的泵管內(nèi)壁硬度測(cè)試儀的研制基于單片機(jī)的自動(dòng)找平控制系統(tǒng)研究基于C8051F040單片機(jī)的嵌入式系統(tǒng)開發(fā)基于單片機(jī)的液壓動(dòng)力系統(tǒng)狀態(tài)監(jiān)測(cè)儀開發(fā)模糊Smith智能控制方法的研究及其單片機(jī)實(shí)現(xiàn)一種基于單片機(jī)的軸快流CO〈,2〉激光器的手持控制面板的研制基于雙單片機(jī)沖床數(shù)控系統(tǒng)的研究基于CYGNAL單片機(jī)的在線間歇式濁度儀的研制基于單片機(jī)的噴油泵試驗(yàn)臺(tái)控制器的研制基于單片機(jī)的軟起動(dòng)器的研究和設(shè)計(jì)基于單片機(jī)控制的高速快走絲電火花線切割機(jī)床短循環(huán)走絲方式研究基于單片機(jī)的機(jī)電產(chǎn)品控制系統(tǒng)開發(fā)基于PIC單片機(jī)的智能手機(jī)充電器基于單片機(jī)的實(shí)時(shí)內(nèi)核設(shè)計(jì)及其應(yīng)用研究基于單片機(jī)的遠(yuǎn)程抄表系統(tǒng)的設(shè)計(jì)與研究基于單片機(jī)的煙氣二氧化硫濃度檢測(cè)儀的研制基于微型光譜儀的單片機(jī)系統(tǒng)單片機(jī)系統(tǒng)軟件構(gòu)件開發(fā)的技術(shù)研究基于單片機(jī)的液體點(diǎn)滴速度自動(dòng)檢測(cè)儀的研制基于單片機(jī)系統(tǒng)的多功能溫度測(cè)量儀的研制基于PIC單片機(jī)的電能采集終端的設(shè)計(jì)和應(yīng)用基于單片機(jī)的光纖光柵解調(diào)儀的研制氣壓式線性摩擦焊機(jī)單片機(jī)控制系統(tǒng)的研制基于單片機(jī)的數(shù)字磁通門傳感器基于單片機(jī)的旋轉(zhuǎn)變壓器-數(shù)字轉(zhuǎn)換器的研究基于單片機(jī)的光纖Bragg光柵解調(diào)系統(tǒng)的研究單片機(jī)控制的便攜式多功能乳腺治療儀的研制基于C8051F020單片機(jī)的多生理信號(hào)檢測(cè)儀基于單片機(jī)的電機(jī)運(yùn)動(dòng)控制系統(tǒng)設(shè)計(jì)Pico專用單片機(jī)核的可測(cè)性設(shè)計(jì)研究基于MCS-51單片機(jī)的熱量計(jì)基于雙單片機(jī)的智能遙測(cè)微型氣象站MCS-51單片機(jī)構(gòu)建機(jī)器人的實(shí)踐研究基于單片機(jī)的輪軌力檢測(cè)基于單片機(jī)的GPS定位儀的研究與實(shí)現(xiàn)基于單片機(jī)的電液伺服控制系統(tǒng)用于單片機(jī)系統(tǒng)的MMC卡文件系統(tǒng)研制基于單片機(jī)的時(shí)控和計(jì)數(shù)系統(tǒng)性能優(yōu)化的研究基于單片機(jī)和CPLD的粗光柵位移測(cè)量系統(tǒng)研究單片機(jī)控制的后備式方波UPS提升高職學(xué)生單片機(jī)應(yīng)用能力的探究基于單片機(jī)控制的自動(dòng)低頻減載裝置研究基于單片機(jī)控制的水下焊接電源的研究基于單片機(jī)的多通道數(shù)據(jù)采集系統(tǒng)基于uPSD3234單片機(jī)的氚表面污染測(cè)量儀的研制基于單片機(jī)的紅外測(cè)油儀的研究96系列單片機(jī)仿真器研究與設(shè)計(jì)基于單片機(jī)的單晶金剛石刀具刃磨設(shè)備的數(shù)控改造基于單片機(jī)的溫度智能控制系統(tǒng)的設(shè)計(jì)與實(shí)現(xiàn)基于MSP430單片機(jī)的電梯門機(jī)控制器的研制基于單片機(jī)的氣體測(cè)漏儀的研究基于三菱M16C/6N系列單片機(jī)的CAN/USB協(xié)議轉(zhuǎn)換器基于單片機(jī)和DSP的變壓器油色譜在線監(jiān)測(cè)技術(shù)研究基于單片機(jī)的膛