GIMlnternational

Feature

the

GIS and Computer Aided Architectural Design (CAAD) follow distinct

modelling paradigms tor the generation and management of virtual 3D city

models. New applications and the increasing demand tor quality make it

necessary to integrale models !rom both disciplines. Although in recent limes

the functionalities of 3D GIS and CAAD systems have been slowly

converging, fundamental differences between the modelling approaches make

it difficult to reach full interoperability. The authors treat in greater detail the

differences between the 'worlds' of geometry representation, spatial

referencing, multiple representations, international standards and semantic

modelling. In addition, possibilities ofbridging the gap are discussed.

struction, and Facility Management(AEC/FM). However, althoughGIS and Computer Aided Architec-tural Design (CAAD) are separatedomains, each would profit flomthe integration of models flomthe other discipline:. The GIS domain would get ac-

cess to semantically-rich objects,gain access to the inside of build-ings and generallybecome bettelplatforms für planning

. CAAP would profit flom thespatial context provided by thehuge amount of geo~objectsmaintained in GIS

In the past, neit~r GIS norCAAD systems provided fullsupport für the representation,management and visualisation oflarge 3D city models. The currenttrend is für CAAD systems to in-corporate GIS functionality, suchas the support of geodetic coordi-natereference systems and data-

By Thomas H. Kolbe andLutz Plümel; University of Bonn, Germany

Virtual 3D city models have be-come increasingly important forurban planning, architectural de-sign, disaster management and

3D cadastre. They have been es-tablished and used both in theGIS domain and in the domain ofA!chitecture, Engineering, Con-

Figure 1,Integration of 3D

CAAD and GISmodels in an

urban planningscenario. Whereas

architecturemodels provide

detailedinformation about

the plannedobjects, GIS

models set theenvironmental

context. Image:City of Solingen,

Stadtdienst621/623. Planning

version fOT therebuilding of the

Graf- Wilhelm-Platz. Design by

Quick / Beckmann/ Quick Architects,

Berlin

12 July 20{)4

GIM International.

Feature

base functionalities. On the GISside, the main development con-cerns the integration of full sup-port for 3D coordinates and 3Dgeometric primitives.

typically restricted to Cartesiancoordinate systems, During con-struction, 3D objects are oftenmod~lled usihg local model coor-dinates without referring to aCRS. Geo-referencing is düne(often manually) by the explicitapplication of an affine transfor-mation which translates, scalesand rotates all objects in a projectto their world coordinates.Whereas the explicit representa-tion of absolute coordinates inGIS models enables efficient in-cJexing and immediate process-ing, it inhibits the use of proto-type objects. Prototypes aredefined using local coordinatesand can be instantiated at differ-ent locations in a 3D-city modelby a simple transformation.Therefore, classical GIS datamodels make it necessary to ex-plicitly represent every street-light, traffic light, tree and othervegetation object.

Two Worlds

Integration of 3D CAAD and GISmodels is increasingly carried outfür purposes such as urban plan-ning (Figure 1). Although GISand CAAD systems are thus get-ting closer, fundamental differ-ences between the disciplineshinder the coherent integration ofcity models from these dornains.The major difference originales intheir distinct modelling para-digm. This is a result of the wayin which 3D models are acquired.In the GIS domain 3D objects areacquired from aerial images,laser-scanner data or terrestrialrneasurements: that is, observa-tions. Therefore an accumulativemodelling principle is appliedwhich allows easy description ofthe observed features. In contrast,the CAAD domain follows a gen-erative modelling approach.CAAD models are the result of aconstruction process and focus onmanmade artefacts rather than onrepresentation of topography.

tries, since it can describe evencomplex polyhedra by just onegeometric object.In CSG models, the object geome-try is composed of volumetricand parametric primitives thatare combined using the set theo-retical operations union, intersec-tion and difference. Foilowingthe generative modeiling para-digm, CSG is weil suited tomodel construction as applied inarchitectural design. A look onthe system side reveals that l1lostCAAD systems realise a hybridmodeiler supporting both CSGand B-Rep. Problems arise sincethese systems:. Are not designed to handle the

large number of objects in a citymodel

. Offer as yet only limited GISfunctionality

However, 3D GIS is currently lim-ited by the fact that it realisesonly B-Rep. This makes it difficultto incorporate AEC/FM modelsin a war that would ailow theirsubsequent editing using genera-tive modeiling tools.

Multiple Representations3D-city models are often classi-fied with respect to their resolu-tion and generalisation level (Fig-ure 3). In many existing modelsthree discrete Levels 01 Detail

Geometry RepresentationOf all the different modellingprinciples tor 3D geometry, theBoundary Representation (B-Rep)and the Constructive Solid Geome-try (CSG) are most important tor3D city models (Figure 2). The B-Rep model is defined as the accu-mutation of all surfaces enclosingthe volume of an object. Becausesurfaces are explicit, textures canbe draped directly onto them. 3Dpolygons can efficiently be ren-dered using hardware accelera-tion from 3D graphic cards. TheB-Rep modelling approach is es-pecially suited t<;> representationof measured 3D object geome-

Spatial Reference

When dealing with geo-informa-tion, the exact spatial reference isof the utmost importance. Geo-information systems traditional-ly offer strang support für thevarious geodetic Coordinate Refer-enGe Systems (CRS). Besides geo-graphic and projected coordi-nates, same GIS models alsoallow different CRS für planime-try and height. Embedded proce-dures for coordinate transforma-tions in GIS make it possible tocombine geo-objects with differ-ent CRS.Geo-referencing has played onlya minor role in CAAD systemsso far. Architectural models are

Figure 2, 3D GISand theAEC/FMdomain preferdifferent geometryrepresentations.Whereasarchitecturemodels aretypicallyconstructed fromvolumetricprimitives, GISmodels arecomposed of theobservable objectsurfaces

2004 13

GIM Interhationol

Feature

Whereas such ontology hag beendeveloped für the AEC/FM do-main by the IAl, no common on-tology hag Jet been defined für3D-city models in the context ofGIS.In an initiative designated theGeodata Infrastructure North-Rhine Westphalia (GDI NRW) weare currently working on the def-inition of a unified 3D-city modelfür GIS. About fifty parties inGerman industry, academia, ad-ministration, municipalities andcadastre are participating in thedevelopment. The model com-prises differentiated concepts fürthe different types of real worldobjects in cities, like buildings,streets, wafers bodies, vegetation,terrain etc. The model is based onISO standards flom the ISO 191xxfamily. In an forthcoming test-bed a GML3 application schemawill be derived.

{LoD) are distinguished:. LoD 1 comprises simple build-

ings with flat roofs. LoD 2 comprises building mod-

els with roof structures and tex-tured facades

. LoD 3 comprises highly de-tailed models also containingsmaller features like chimneys,stairs and balconies

Sihce often the dat" für each LoDis obtained by specific acquisitionmethods, single object~ may havemultiple representations that aresimultaneously valid. The contin-uation process of city models im-plies a second representation di-mension. Geo-objects changeover time and therefore have dif-ferent representations at different

Bridging the Gap

3D-city models from the GIS andABC/FM domains differ substan-tially in many aspects. Thus itcannot be expected that eitherGIS or CAAD systems will solveall above problems in the mid-term. Instead, semantic model-ling is the key to the integrationof (geo-)objects from both do-mains. Such objects will be pre-sented both in terms of the GISand the CAAD model. The map-ping of the two different ontolo-gies then allows interpretation ofobjects from one domain in thecontext of the other. It is a ques-tion of future research how to en-sure consistency between coexist-ing models. The long-term goalshould be definition of a commonontology für 3D-city models withspecific profiles für GIS applica-tions and the ABC/FM domain.A first step would be the broaderdiscussion between GIS andCAAD communities on the onehand and the users of 3D-citymodels on the other.

international standards up to ISOlevel. The bad news is that thesestandards have been developedwithin separate contexts by dis-tinct organisations that have thensubmitted their work to differentISO technical committees (Figure4). In the AEC/FM domain the In-ternational Alliance foT Interoper-ability (lAI) hag developed the Ih-du$try Foundation Classes (IFC,ISO/PAS 16739). IFC defines aproduct model in the sense that itcovers both the geometry and thesemantic properties of buildingsand their components. For exam-pIe, IFC contains definitions fürbuildings, storeys, walls, roofs,rooms, stairs and so on. With theGeography Markup Language 3

(GML3), the Open GISConsortium (OGC) hagreleased an XML-basedstandard für the repre-sentation of 0-40 geo-objects. GML3 is basedon different 191xx ISOstandards. It is a 'metaformat', which me ansthat it can be tailored todifferent application do-mains.In February 2003 theOGC and the lAI an-nounced their official co-operation. An ongoingproject is investigating

- the conversion of mod-els between the formats.

However, such conversions willbe problematic because IFC sup-ports both CSG and B-Rep geom-etry, whereas GML3 is restrictedto B-Rep. Furthermore, IFC cur-rently offers oi:11y little CRS sup-port.

Figure 3, In 3Dcity models eachobject can have

multiplerepresentations

according to therespective point in

time, differentplanning versionsand specific level

of detail

points in time (in any LoD!). Thisaspect is especially important fürthe development of 3D cadastres,where the complete history ofspatial objects has to be main-tained. In contrast to the LoD caseabove, at any time exactly onerepresentation of an object isvalid. Furthermore, in a planningscenario some objects may existin different concurrent versions.Since these alternative represen-tations may exist in differentLoDs and at any point in time,versions add another dimensionto the problem of multiple repre-sentations. Currently, neither GISnor CAAD systems providemechanis~s für the realisation ofall three different repfesenta~ondimensions.

Semantic Modelling

The modelling aspect, which hasprogressed best in recent years,concems the explicit representa-tion of object semantics. With IFCand GML3, both the AEC/FMand GIS domain nowadays pro-vide data models and exchangeformats which are not only ahleto represent object geometries hutalso their type, structure, functionanc;i relations to othet öbj~<;:~s. ~order to create a semantic model,ontology has first to be devel-oped. Ontology defines an terms,the meaning of different obj~ctsand their interrelationships für aspecific application domain.

Further Reading

.. Cox, 5., Daisy, P., Lake, R.,Potteie, G., Whiteside, A., 2003,OperiGIS Geography MarkupLanguage (GML 3), Implemen-tation Specification Versiori3.00, OGC Doc. No. 02-023r4

. Adachi, Y., Forester, J., Hyvari-

Internatianal Standards

The good news conceming the in-teroperability of 3D-city modelsis that there exist comprehensive

14 Julv 2004

GIM International

Feature

nen, J., Karstila, K., Liebich, T.,Wrx, J., 2003, Industry Founda-tion Classes IFC2x Edition 2, In-ternational Alliance for Interoper-ability, www.iai-international.~rg

. Kolbe, T. H., Gröger, G., 2003, 10-wards unified 3D city models. In:

Schiewe, J., Hahn, M., Madden,M., Sester, M., Eds.: Proceedingsof the ISPRS Comm. IV JointWorkshop on Challenges inGeospatial Analysis, Integrationanp VlSualization 11 in Stuttgartwww.gin-online.de/isprs/

Websites

www.ikg. uni-bonn.de/ sig3d/(homepage of the SIG 3D of theinitiative Geodata InfrastructureNorth-Rhine Westphalia GD!NRW)+

Figure 4, GIS and the AEC/FM domain have their own industry standardisationorganisations that submit proposals to different technical committees 01 the ISO. Since20a3 OGC and lAI have co-opera ted with the lang-term goal 01 seamless integration 01

architecture and GIS modelsDr Thomas H. Kolbe, Institute for Cartography and Geoin-formation, University of Bonn, Meckenheimer Allee 172,D - 53115 Bonn, Germany, e-mail: kolbe@ik~.uni-bonn.de