引用格式：张文元,陈江媛,谈国新.基于3D基元拟合的复杂屋顶点云三维重建[J].地球信息科学学报,2023,25(8):1531-1545.[ZhangWY,ChenJY,TanGX.Complexroofstructurereconstructionby3Dprimitivefittingfrompointclouds[J].JournalofGeo-informationScience,2023,25(8):1531-1545.]DOI:10.12082/dqxxkx.2023.220927基于3D基元拟合的复杂屋顶点云三维重建张文元，陈江媛*，谈国新华中师范大学国家文化产业研究中心，武汉430079ComplexRoofStructureReconstructionby3DPrimitiveFittingfromPointCloudsZHANGWenyuan,CHENJiangyuan*,TANGuoxinNationalResearchCenterofCulturalIndustries,CentralChinaNormalUniversity,Wuhan430079,ChinaAbstract:Geometricandsemanticintegrationof3Dbuildingmodelsareimportantinfrastructuredataforsmartcity,theyareconduciveforpromotingtherefinedmanagementandintelligentapplicationofbuildingfacilities.However,mostoftheexistingpointcloud-basedmodelingmethodsfocusonthereconstructionofgeometricmodelswithsimpleroofstructure,andsemanticandtopologicalrelationsareignored.Moreover,thesemethodsaresensitivetonoise,whicharedifficulttoassuretopologicalconsistencyandgeometricaccuracy.Tosolvetheseproblems,thispaperproposesa3Dprimitivefittingalgorithmforautomaticallyreconstructingbuildingmodelswithcomplexroofstructurefrompointclouds.Firstly,a3Dbuildingprimitivelibraryisdesigned,includ-ingvarious3Dbuildingprimitiveswithsimpleandcomplexrooftypes.Secondly,anindividualbuildingpointcloudinputissegmentedintomultipleplanesusingRANSACalgorithm.TheRoofTopologyGraph(RTG)isthengeneratedaccordingtotherelationshipofroofplanes,andtherooftypeofpointcloudissubsequentlyrec-ognizedbycomparisonofRTGbetweenpointcloudandbuildingprimitives.Thirdly,thereconstructionisformu-latedasanoptimizationproblemthatminimizesthePoint-to-MeshDistance(PMD)betweenthepointcloudandthecandidatemeshedbuildingprimitive.Thesequentialquadraticprogrammingoptimizationalgorithmwithnec-essaryconstraintsisadoptedtoperformholisticallyprimitivefitting,soastoestimatetheshapeandpositionpa-rametersofa3Dprimitive.Finally,theparameterizedmodelisautomaticallyconver...
