




版权说明:本文档由用户提供并上传,收益归属内容提供方,若内容存在侵权,请进行举报或认领
文档简介
ConstructingRulesandSchedulingTechnologyfor3DBuildingModels
ZhengweiSUI,LunWU,JingnongWENG,XingLIN,XiaoluJI
Abstract
3Dmodelshavebecomeimportantformofgeographicdatabeyondconventional2Dgeospatialdata.Buildingsareimportantmarksforhumantoidentifytheirenvironments,becausetheyareclosewithhumanlife,particularlyintheurbanareas.Geographicinformationcanbeexpressedinamoreintuitiveandeffectivemannerwitharchitecturalmodelsbeingmodeledandvisualizedinavirtual3Denvironment.Architecturalmodeldatafeatureswithhugedatavolume,highcomplexity,non-uniformrulesandsoon.Hence,thecostofconstructinglarge-scalescenesishigh.Meanwhile,computersarelackofprocessingcapacityuponalargenumberofmodeldata.Therefore,resolvingtheconflictsbetweenlimitedprocessingcapacityofcomputerandmassivedataofmodelisvaluable.Byinvestigatingthecharacteristicsofbuildingsandtheregularchangesofviewpointinvirtual3Denvironment,thisarticleintroducesseveralconstructingrulesandschedulingtechniquesfor3Dconstructingofbuildings,aimingatthereductionofdatavolumeandcomplexityofmodelandthusimprovingcomputers’efficiencyatschedulinglargeamountofarchitecturalmodels.Inordertoevaluatetheefficiencyofproposedconstructingrulesandschedulingtechnologylistedintheabovetext,theauthorscarryoutacasestudyby3DconstructingthecampusofPekingUniversityusingtheproposedmethodandthetraditionalmethod.Thetworesultsarethenexaminedandcomparedfromaspectsofmodeldatavolume,modelfactuality,speedofmodelloading,averagerespondingtimeduringvisualization,compatibilityandreusabilityin3Dgeo-visualizationplatforms:ChinaStar,oneChina’sownplatformfor3DglobalGISmanufacturedbytheauthorsofthispaper.Theresultofcomparisonrevealsthatmodelsbuiltbytheproposedmethodsaremuchbetterthanthosebuiltusingtraditionalmethods.Fortheconstructingofbuildingobjectsinlarge-scalescenes,theproposedmethodscannotonlyreducethecomplexityandamountofmodeldataremarkably,butcanalsoimprovingcomputers’efficiency.
Keywords:Constructingrules,Modelscheduling,3Dbuildings
I.INTRODUCTION
Inrecentyears,withthedevelopmentof3DGIS(GeographicalInformationSystem)softwarelikeGoogleEarth,Skyline,NASAWorldWind,large-scale3Dbuildingmodelswithregionalcharacteristicshavebecomeimportantformofgeographicdatabeyondconventional2Dgeospatialdata,likemulti-resolutionremotesensingimagesandvectordata[1].Comparedtotraditional2Drepresentation,geographicinformationcanbeexpressedinamoreintuitiveandeffectivemannerwitharchitecturalmodelsbeingmodeledandvisualizedinavirtual3Denvironment.3Drepresentationandvisualizationprovidesbettervisualeffectandvividurbangeographicinformation,andthusplaysanimportantroleinpeople'sperceptionsoftheirenvironment.Meanwhile,the3Dbuildingdataisalsoofgreatsignificancefortheconstructionofdigitalcities.
Buthowtoefficientlyvisualizethousandsof3Dbuildingmodelsinavirtual3Denvironmentisnotatrivialquestion.Themostdifficultpartofthequestionistheconflictsbetweenlimitedprocessingcapacityofcomputerandmassivevolumeofmodeldata,particularlyintheprocedureofmodelrendering.Takingthe3Dmodelingofacityfortheexampleusingtraditional3Dmodelingmethod,supposethereare100000buildingstomodelintheurbanareaandtheaveragesizeofmodeldataforeachbuildingisroughly10M.SothetotaldatavolumeofbuildingmodelsinthecitycouldreachaTBlevel.However,thecapacityofordinarycomputermemoryisonlyintheGBscale.Basedonthisconcern,theauthorsproposedtheschedulingtechnologyforlarge-scale3Dbuildingsmodelsinaspectsofmodelloadingandrendering.Duetothelackofbuildingconstructingrulesandstandard,modelsofbuildingsvaryinaspectsofconstructingmethods,texturescollectionandmodeldatavolume,especiallyinaspectsofmodelreusabilityandfactuality.Suchalargeamountofdatawithoutuniformconstructingrulesbecomesahugechallengefordatastorage,processingandvisualizationincomputers.Italsobringstheproblemofincompatibilityamongdifferent3DGISsystems.
AfteryearsofresearchinGIS(GeographicInformationSystem),peoplehaveaccumulatedanumberofwaystosolvetheaboveproblems[3].Howeverinvirtual3Denvironment,becauseofthedifferenceindataorganizationandmannersofhumancomputerinteraction(HCI),weneedtoapplyanewstandardizedmethodofmodelingandschedulingfor3Dmodels.Atpresent,thereisnosuchauniformmethodastheconstructingspecificationorstandardforthemodelingof3Dbuildings.Existingapproachesareinsufficientandinefficientintheschedulingoflarge-scalebuildingmodels,resultinginpoorperformanceorlargememoryoccupancy.Inresponsetosuchquestions,theauthorsproposedanewmethodfortheconstructionof3Dbuildingmodels.Modelsbuiltusingtheproposedmethodscouldbemuchbetterthanthosebuiltusingtraditionalmethods.Forthe3Dmodelingofbuildingobjectsinscenesoflargescale,theproposedmethodscannotonlyremarkablyreducethecomplexityandamountofmodeldata,butcanalsoimprovingthereusabilityandfactualityofmodels.Concerningtheschedulingoflarge-scalebuildingmodels,theModelLoadingJudgmentAlgorithm(MLJA)proposedinthispapercouldsolvetheoptimaljudgmentproblemofmodelloadingin3Dvisioncone,particularlyincircumstancewithuncertainuserinteractions.
Thispaperfirstexaminesandanalyzesexistingproblemsinconstructingandschedulingstepsof3Dbuildingmodels.Thentheauthorsproposeasetofconstructingrulesfor3Dbuildingmodelstogetherwithmethodsofmodeloptimization.Besides,specialschedulingtechnologyandoptimizationmethodformodelrenderingisalsoappliedinthispaperforlarge-scale3Dbuildingmodels.Inordertoevaluatetheefficiencyofproposedrulesandmethods,acasestudyisundertakenbyconstructinga3DmodelforthemaincampusofPekingUniversityandShenzhenusingboththeproposedmethodandthetraditionalmethodrespectively.Thetworesulting3DmodelsofPekingUniversitycampusandShenzhenarethenexaminedandcomparedwithoneotherinaspectsofmodeldatavolume,modelfactuality,speedofmodelloading,averagerespondingtimeduringvisualization,compatibilityandreusabilityinvarious3Dgeo-visualizationplatformslikeChinaStar(oneChina’sownplatformfor3DglobalGISmanufacturedbytheauthors),Skyline,etc.Resultofcomparisontellsthatprovidedsimilarfactualityofmodels,usingtheproposedmethodofus,thedatavolumeofmodelswasreducedby86%;thespeedofmodelloadingwasincreasedby70%;theaveragerespondingtimeofmodelduringvisualizationandinteractionspeedwasreducedby83%.Meanwhile,thecompatibilityandreusabilityof3Dmodeldataarealsoimprovediftheyareconstructedusingourapproach.
II.MODELINGRULESOF3DBUILDINGS
3Dsceneisthebestformofvisualizationfordigitalcitysystems.Whileconstructing3Dmodelsforbuildingsobjects,propermethodsandrulesshouldbeused,whicharemadewithfullconcernsofthecharacteristicsof3Dbuildingmodels[2].Theresultingmodelsshouldberobust,reusableandsuitableenoughfortransmissionovercomputernetwork,andshouldatthesametimebeautomaticallyadaptedtosystemcapability.
Generallyspeaking,methodsofconstructing3Dbuildingmodelscanbeclassifiedintothreetypes:wireframemodeling,surfacemodelingandsolidmodeling.Innormalcircumstances,tomodelbuildingsin3Dformat,theframeworkofbuildingshouldbeconstructedfirstaccordingtothecontourfeatures,numberoffloors,floorheight,aerialphotographandliveactionphotosofbuildings.Then,gatherthecharacteristicsofscenethatthebuildingstomodelarerepresenting.Importantcharacteristicsincludebuildingsaerialphotographorliveactionshootingphotos.Finally,mapthegatheredtexturetomodelframework,optimizethemodelandcreatedatabaseofthe3Dbuildingmodels.
Althoughtherehavealreadybeenmanyapproachesfortheconstructionof3Dbuildingmodels,aunifiedmodelingmethodandrulesarestillneededtoimprovetheefficiency,quality,facilitatechecking,reusabilityandarchivingofconstructedmodels.Byinvestigatingthecharacteristicsofbuildings,wefoundthatbuildingshaveregulargeometricsolidformodeling,similartextureonthesurfacesofdifferentdirections,highsimilarityinsmall-scalemodelsofbuildings,etc.Accordingtothese,thisarticlegivesadiscussiononthemodelingrulesfromthreeaspects,includingconstructingrulesofthe3Dbuildingmodels,texturemappingrulesof3Dbuildingmodelsandoptimizationmethodforconstructedmodelsbasedonmentionedconstructingrules.
A.Constructingrulesofthe3Dbuildingmodels
The3Dbuildingmodelingreferstotheprocedureofrepresentingtruebuildingsfromtherealworldintocomputerintheformof3Dobjects[4].Humanbeings,asthecreatorandatthesametimepotentialusersofmodels,playakeyroleinthisprocedure.Peoplearedifferentfromeachotherintheunderstandingofthebuildingobjects,methodsofmodelingandthesoftwaretoolstheyuseformodeling.Suchdifferencesamongpeoplewhocarryoutmodelingworkatthesametimeleadtothe3Dmodelsofdiversequalityandlowefficiency.Sothe3Dbuildingconstructingrulesproposedinthisarticlebecomenecessaryandhelpfultosolvetheaboveproblems.
1)Combinesimilarfloorsasawholeandkeeptheroofindependent
2)Sharesimilarmodelsandprocessthedetailsespecially
3)Constructingintheunitofmeters
4)Definecentralpointofthemodel
5)Unifiedmodelcodes
6)Reducenumberofsurfacesinasinglemodel
7)Reducecombinationofthemodels
8)Rationalsplitofmodels
B.Texturemappingrulesof3Dbuildings
Basedontheframeworkof3Dmodels,weneedtoattachthesemodelswithpropertexturestocreateabettervisualizationeffectfor3Dbuildings.Thequalityoftexturemappinghasadirectimpactonthevisualeffectofthescenewhilingbeingrendered[5].Sincethegraphicscardofcomputerwillloadallthetexturestogetherwhenrenderingamodel,texturemappingrulesandthequalityofthetexturemappingcandirectlyinfluencetheefficiencyofrenderingaswell.
C.Optimizationofmodelsbasedonconstructingrules
Basedonconstructingrulesandthecharacteristicsof3Dbuildingmodels,theauthorsdevelopasoftwaretooltooptimizethe3Dbuildingmodelsautomatically.Theoptimizationsimplementedinthesoftwaretoolcontainthedeletionofmodels’internaltextures,mergingadjacentvertices/lines/surfaces,removingun-mappedframeworkandsoon.Besides,thesoftwarecanenhancetheshapeofthewholemodel,texturepositionandmodelfacticityintheprocedureofmodeloptimization.
III.SCHEDULINGTECHNOLOGYOFLARGE-SCALE3DBUILDINGMODELS
Forthe3Dvisualizationoflarge-scalearchitecturalmodels,aseriesofmeasurescouldbeappliedtoensuretheefficientrenderingofmodels.Importantmeasuresincludesthesceneorganization,visionconecutting,eliminationoftexturesonthebacksideofmodels,Shaderoptimization,LODAlgorithm,mathlibraryoptimization,memoryallocationoptimization,etc..
Howtodisplaythousandsof3Dcitybuildings’modelsinavirtual3Denvironmentisnottrivial.Themainproblemistheschedulingofmodels[7].Itdetermineswhenandwhichmodelstobeloaded.Thisproblemcanbedividedintotwosmallerproblems:Findvisiblespatialregionofmodelsin3Denvironment,andoptimizationmethodofmodelrenderingefficiency.
A.Findvisiblespatialregionofmodelsin3Denvironment
Accordingtooperatingmechanismofcomputersduring3Dvisualizationandthecharacteristicsoflarge-scale3Dscene,weneedtodeterminethepositionofcurrentviewpointfirstbeforeloadingsignalmodelsorurban-unitmodels.Theninresponsetotheregularchangesofviewpointinvirtual3Denvironment,thesystemwillpreloadthe3Dmodeldataintomemoryautomatically.Inthisway,frequentIOoperationscanbereducedandthusoverallefficiencyofsystemgetsimproved.AnewalgorithmnamedMLJA(ModelLoadingJudgmentAlgorithm)isproposedinthispaperinordertofindoutvisibleregionofmodelsinthe3Denvironment.Thealgorithmintegratesthegraticulesandelevationinformationtodeterminethecurrentviewpointofusersinthe3Dspace.Andwiththemovementofviewpoint,thealgorithmschedulestheloadingofmodelcorrespondinglyandefficiently.
B.Optimizationmethodofmodelrenderingefficiency
Theschedulingmethodoflarge-scale3Dbuildingmodelsproposedaboveisaneffectivewaytosolvetheproblemcausedthecontradictionbetweenlargemodeldatavolumeandlimitedcapacityofcomputers.Accordingtothealgorithm,wecanavoidloadingthewholelarge-scale3Dbuildingmodelsatonetimeforthesakeoflimitedcomputermemory,andthenimprovesystemefficiencyintheprocedureofmodelloadingandabandoning.DuetothelimitedcapacityofGPUandlocalvideomemory,weneedafurtherresearchonhowtodisplaytheloadedmodeldatainmoreefficientmanner.Intheremainingpartofthispaper,theauthorswillcontinuetointroduceseveralmethodsontheoptimizationofmodelrenderinginthevisioncone.
1)Eliminationoftexturesonthebacksideofmodels
Thebacksideofthe3Dmodelisinvisibletotheusers.Ifweomitthetexturemappingforthe3Dmodelonthebackside,theprocessingloadofgraphiccardwillbereducedasmuchasatleast50%.Besides,accordingtoaninvestigationonprocedureofactualmodelrendering,theauthorsfoundthatonthebacksideofthe3Dmodel,theinvisibletextureisrenderedinacounter-clockwisemanneragainstthedirectionofeyesight,whilethevisibletexturemappingisrenderedinclockwisemanner.Sowecanomittherenderingofmodelswhichisintendedtoberenderedincounterclockwisemanner.Therefore,thetextureswon’texistonthebackof3Dmodels.Thegraphiccardcouldthenworkmorerapidlyandefficiently.
2)Eliminatetheshieldedmodel
Bycalculatingthegeometricrelationshipbetween3Dmodelsinthescene,theshieldedmodelscanbeomittedwhiledisplayingthescenewithappropriateshieldingpatches.Throughthisway,wecaneffectivelyreducetheusageofgraphicscardmemory,andthusachievehigherrenderingefficiencyandfaster3Dvirtualsystem.
Inthevirtual3Dgeographicinformationsystem,weoftenobserve3Dmodelsfromahighaltitude.Itisespeciallytrueforlarge-scaleoutdoor3Dmodels.Theusualarrangementof3Dbuildingmodelsarealwayssparse,howevertherealblockisverysmall.Therefore,establishinganindexforvisualcontrol,whichissimilartotheBSPtree,doesn’tamounttomuch.ThroughcarefullystudyingDirectX,wefoundthatwecantakeadvantageofthelatestZ-bufferingtechnologyofDirectXtoimplementtheshieldingcontrolofmodels.
3)OptimizationmethodoftheShaderinstructions
Inshader3.0technology,SM(ShaderModel)isamodelwhichcanoptimizetherenderingengine.A3Dsceneusuallycontainsseveralshaders.Amongtheseshaders,somedealwiththesurfacesandskeletonsofbuildings,andothersdealwiththetextureof3Dbuildingmodels.
Geometrycanbehandledquicklybyshaderbatchprocess.Theshadercancombinesimilarculminationin3Dbuildingmodels,dealwiththecorrelationoperationofasinglevertex,determinethephysicalshapeofthemodel,linkthepoint,line,triangleandotherpolygonsforarapidprocessingwhilecreatenewpolygons,etc.WecanassignthecomputingtasktoshaderandlocalvideomemorydirectlyinaveryshorttimewithoutbotheringtheCPU.Inthiscase,visualeffectsofsmoke,explosionsandotherspecialeffectsandcomplexgraphicsarenolongernecessarytobeprocessedbytheCPUofcomputer.SuchfeaturesofshadercanspeedupboththeCPUandgraphiccardinprocessinghugeamountof3Dmodels.
4)LODalgorithmoflarge-scale3Dscene
LOD(LevelofDetail)isacommonandeffectivesolutiontoresolvetheconflictsbetweenrealtimevisualizationandtheauthenticityofmodels[8].ByinvestigatingthemainfeaturesandtypicalalgorithmsofLODtechnology,theauthorsproposedanewstructurefordynamicmulti-leveldisplay.Thisstructuren
温馨提示
- 1. 本站所有资源如无特殊说明,都需要本地电脑安装OFFICE2007和PDF阅读器。图纸软件为CAD,CAXA,PROE,UG,SolidWorks等.压缩文件请下载最新的WinRAR软件解压。
- 2. 本站的文档不包含任何第三方提供的附件图纸等,如果需要附件,请联系上传者。文件的所有权益归上传用户所有。
- 3. 本站RAR压缩包中若带图纸,网页内容里面会有图纸预览,若没有图纸预览就没有图纸。
- 4. 未经权益所有人同意不得将文件中的内容挪作商业或盈利用途。
- 5. 人人文库网仅提供信息存储空间,仅对用户上传内容的表现方式做保护处理,对用户上传分享的文档内容本身不做任何修改或编辑,并不能对任何下载内容负责。
- 6. 下载文件中如有侵权或不适当内容,请与我们联系,我们立即纠正。
- 7. 本站不保证下载资源的准确性、安全性和完整性, 同时也不承担用户因使用这些下载资源对自己和他人造成任何形式的伤害或损失。
最新文档
- 重要法规:国际物流师试题及答案
- 2024年物流师考试失败原因试题及答案
- 肉牛场床一体化建设养殖技术
- 生物体的水分代谢:试题及答案
- CPSM考试中的案例分析技巧试题及答案
- 2024年供应链管理师的项目管理挑战及试题及答案
- CPSM考试内容综述试题及答案
- 保安基本知识培训课件
- 2024年CPMM试题及答案全景透视
- 生态恢复的理论与实践探讨试题
- 年产10万吨聚氯乙烯生产工艺设计毕业设计
- 高中18岁成人仪式主题活动设计
- 《婚姻家庭纠纷调解》课件
- 高中数学培优讲义练习(必修二):专题8.1 基本立体图形(重难点题型精讲)(教师版)
- 兵团红色经典文化在新疆高校思想政治教育中的运用研究
- 《珠穆琅玛峰》课件
- 注塑机定期保养记录表2016
- 3.28百万农奴解放纪念日演讲稿
- 全科医学科疾病诊疗指南全集诊疗规范
- 安全教育教程大学生安全教育PPT完整全套教学课件
- 2023年东方航空技术应用研发中心有限公司招聘笔试题库含答案解析
评论
0/150
提交评论