SEMANTIC AND TOPOLOGICAL REPRESENTATION

OF BUILDING INTERIORS

An Overview

B. Dom´ınguez,

A. L. Garc´ıa and F. R. Feito

Departamento de Inform´atica, Universidad de Ja´en, Campus Las Lagunillas s/n, Ja´en, Spain

Keywords:

Building Information Models, Geographic Information Systems, Indoor Representation.

Abstract:

Software related to urban environments has experimented a considerable growth over the last years. This trend

makes necessary the creation and management of urban models, both for cities and building indoors. There ex-

ists a huge variety of proposals dealing with the geometry, topology and semantics of building indoors. In this

paper we give an overview of a number of recent works from several research areas such as Geographic Infor-

mation Systems (GIS), Building Information Models (BIM), Building Product Models (BPM) and Computer

Aided Design (CAD) tools and propose a reference architecture which gathers the analyzed features.

1 INTRODUCTION

Building indoor information is needed for multiple

applications covering diverse areas such as GIS, BIM,

spatial databases or CAD. Each area uses different ap-

proaches to represent the information; for instance,

the most extended standard used to manage BIM

information in construction is the Industry Founda-

tion Classes (IFC) model. Even within the same re-

search area, a wide range of partial views of the same

model can be found, depending on the speciﬁc ap-

plication. A classiﬁcation of models and application

areas would be useful to have a starting point before

considering the development of new approaches.

2 BIM AND GIS MODELS

In 1992, Bj¨ork (Bjork, 1992) introduces an object-

oriented model to represent semantic data of build-

ings. His work focuses on the deﬁnition of a schema

including information about spaces and their enclos-

ing entities (walls, columns, doors and windows), be-

ing a precedent of the Industry Foundation Classes

(IFC) standard.

On the other hand, in the last years GIS systems

require more and more information about indoor of

buildings. Some works have researched on this topic.

Isikdag et al. (Isikdag et al., 2008) propose use case

scenarios to implement Building Information Models

(BIM) in a geospatial environment. Van Berlo

and Laat (Van Berlo and de Laat, 2010) introduce

an implementation of the conversion from IFC to

CityGML. Cerovsek makes an exhaustive research

study about BIM technology (Cerovsek, 2010). This

work offers a number of recommendations about how

BIM models should evolve in order to make easier the

development and standardization of BIM tools.

3 2D MODELS

Franz et al. (Franz et al., 2005) analyze building mod-

els under two different points of view: cognitive sci-

ences and architecture. They summarize various ex-

isting graph-based models used in both areas, and dis-

cuss the transfer of models between them.

Lamarche and Donikian (Lamarche and Donikian,

2004) propose a method to represent the topology of

an indoor space for the simulation of crowds of hu-

mans. They compute a set of graph-represented con-

vex cells using the constrained Delaunay triangula-

tion of the ﬂoor plan. This set is then represented as

a graph with nodes for the convex cells and edges for

the neighbor convex cells (see Figure 1). This topo-

logical representation of the space allow them to iden-

tify passages, crossroads and dead ends in order to

determine bottlenecks for pedestrians.

Pl¨umer and Gr¨oger (Pluemer and Groeger, 1996)

utilize another formal representation for the aggrega-

169

Domínguez B., L. García Á. and R. Feito F..

SEMANTIC AND TOPOLOGICAL REPRESENTATION OF BUILDING INTERIORS - An Overview.

DOI: 10.5220/0003930101690174

In Proceedings of the International Conference on Computer Graphics Theory and Applications (GRAPP-2012), pages 169-174

ISBN: 978-989-8565-02-0

 2012 SCITEPRESS (Science and Technology Publications, Lda.)

tion of 2D spatial objects: the nested maps, deﬁned

as planar graphs whose cycles are structured hierar-

chically, useful to model the hierarchical structure of

closed spaces.

Hierarchical region graphs are used by Stoffel et

al. (Stoffel et al., 2007) to model the structure of spa-

tial regions and the inclusion relations among them.

The data structure includes type parameters to spec-

ify the semantics of nodes (doors, windows, etc.) and

graphs (rooms, walls, ﬂoors, etc.).

Li et al. (Li et al., 2010) represent rooms, lobbies,

inner or outer walls, doors and windows using a set

of labeled cellular units (free, occupied). A regular

decomposition of the space is then made using a grid-

graph. Algorithms from graph theory allow to solve

some problems of space analysis and agent naviga-

tion.

Zhi et al. (Zhi et al., 2003) convert architectural

ﬂoor plans into an object graph representing the struc-

ture of walls and openings, such that loops represent

closed spaces. Since rooms are obtained as minimal

loops, this work uses spatial vectors of loops to com-

pute minimal area fundamental loops.

Hahn et al. (Hahn et al., 2006) deal with the real-

time generation of building interiors. The main char-

acteristics of the generator are: (1) the generation of

building interiors is lazy, and (2) a set of rules is fol-

lowed to ensure the correctness and realism of the re-

sults.

Merrell et al. (Merrell et al., 2010) use an ap-

proach based on bayesian networks to solve the gen-

eration of building interiors starting from high-level

requirements. An architectural program is created af-

ter training a bayesian network with real data; then the

architectural program is turned into a real ﬂoor plan

by applying optimization over the space of possible

building layouts.

4 2D MODELS WITH HEIGHT

Slingsby and Raper (Slingsby and Raper, 2007) deal

with pedestrian navigation in 3D city models. After

introducing a state of the art on 3D city modeling,

pedestrian navigation and pedestrian access within

buildings, this work proposes a model to represent

navigable spaces in cities consisting of a 2.5D rep-

resentation of building ﬂoors. In order to deal with

irregular morphology of ﬂoors, they propose the use

of four constraint elements: ramps, stairs, breaklines

and offsets.

Tutenel et al. (Tutenel et al., 2009) propose a rule-

based solver to generate indoor building scenarios au-

tomatically which works by using classes to represent

3D shapes (e.g. Sofa, Table, TV) with tags. They pro-

pose a rule layout planner, which consists of a back-

tracking mechanism executed to solve the placement

of each object. The solver computes a set of possible

locations for the current object according to the previ-

ously placed ones, assigning weights to them. Then,

it selects the most feasible location.

The arrangement of furniture is also solved by

Germer and Schwarz (Germer and Schwarz, 2009).

However, they use a different approach in which

agents are used to represent pieces of furniture. Each

agent is responsible for placing and orienting itself

properly, and ﬁnding a parent object. In order to

achieve this, each agent has three possible states: (1)

search, when it has not been processed, its parent has

been lost or the search for its position has failed; (2)

arrange, when the agent has found a possible parent;

and (3) rest, when the arrange is ﬁnally successful.

Regarding building rendering, Van Dongen (Don-

gen, 2008) proposes a technique to simulate building

interiors viewed from the street without any storage of

geometry. While buildings are modeled using single

cubes, the rendering process simulate the existence of

rooms, objects and people using a ray-tracing algo-

rithm with diffuse textures and billboard planes.

The structured ﬂoor-plan (Choi et al., 2007) con-

sists of a high-level semantic structure which accom-

plishes with nine principles about object orientation

of the model, existence of relationships among enti-

ties, managing of spatial information, levels of detail,

and automatic creation of 3D models.

5 3D MODELS

Choi and Lee (Choi and Lee, 2009; Lee, 2001) pro-

pose a graph structure called 3D Geometric Network,

constructed using the Straight Medial Axis Trans-

form. It represents the connectivity between rooms

of a building. Geometric networks also model the 3D

structure by means of adding edges between rooms

from different ﬂoors.

Van Treeck and Rank (Van Treeck and Rank,

2007) represent the topology using a radial-edge

structure (Weiler, 1988), which represents relations

among vertices, edges, co-edges, loops, faces and

bodies, and derives four different graphs: structural

components, the room faces, the relational objects

and rooms.

Borrmann and Rank (Borrmann and Rank, 2009)

propose a set of spatial operators to determine the rel-

ative position between the bounding boxes of two 3D

spatial objects among above, below, eastOf, westOf,

northOf and southOf.

GRAPP 2012 - International Conference on Computer Graphics Theory and Applications

170

Billen and Zlatanova (Billen and Zlatanova, 2003)

propose the dimensional model, a topological abstrac-

tion which analyzes complex relations between 3D

objects using four dimensional elements (0-D, 1-D,

2-D and 3-D) for each spatial object.

Boguslawski and Gold (Boguslawski and Gold,

2010) deal with the problem of representing non-

manifold CAD models using a data structure called

Dual Half-Edge (DHE), consisting basically of two

dual structures: a net of half-edges from solids and a

dual structure of connected solids.

Clemen and Gielsdorf (Clemen and Frank, 2008)

propose a systematic way to reduce the redundancy in

geometric models using a generalized representation

for models consisting of solids made up by faces con-

tained in planes, half edges and nodes, and ensuring

geometric constraints by referential integrity.

Van Berlo and Laat (Van Berlo and de Laat, 2010)

collaborate with the introduction of an implementa-

tion of the conversion from IFC to CityGML. In or-

der to achieve this, they introduce an extension for

CityGML called GeoBIM. Therefore, the underlying

geometric, semantical and topological models are the

same in IFC and in CityGML.

A schema with four levels of detail is proposed

by Hagedorn et al. (Hagedorn et al., 2009) to repre-

sent indoor building models. This schema has some

similarities to CityGML; however the authors include

features for indoor routing, not included in CityGML.

An example of the application of BIM for com-

puter games with indoor scenarios can be found in

(Yan et al., 2010). Yan et al. propose an architec-

ture consisting of three modules: BIM, crossover and

game. Information about the buildings is managed

by the BIM module, while the crossover module is

used to exchange information between the BIM mod-

ule and the game module.

Topological houses proposed by Paul and Bradley

(Paul and Bradley, 2003) constitute a purely mathe-

matical abstraction to deﬁne houses. This formal def-

inition allows to encode houses using two structures:

PLA (points, lines and areas) and PLAV (PLA + vol-

umes).

Finally, Xu et al. (Xu et al., 2010) propose a

model including geometric, semantic and topologi-

cal aspects of 3D City Models. To achieve this, a 3D

city model is enriched with a thematic module which

contains semantic and topological information; then

the items of the thematic module are mapped onto

the geometric model. This work also introduces a

semi-automatic integration tool for the semantic en-

richment process.

6 COMPARATIVE ANALYSIS

AND DISCUSSION

In the above sections a number of papers from the lit-

erature have been reviewed. In this section we will

propose a model which intends to serve as a linking

point between low-level geometry, topology and se-

mantics. Later, we will compare the reviewed works

and discuss about the applicability of the cited models

to different areas.

6.1 Our Proposal: A Three-module

Framework

Our goal is to propose an architecture to represent

building indoors accomplishing with the following re-

quirements:

1. The model should cover different views of the

same model: from low-level geometry of archi-

tectural drawings to high-level semantic informa-

tion about the structural distribution and the con-

nectivity among physical spaces (e.g. adjacency

between rooms, between a room and an exterior

area of the building, etc.).

2. All the views should be related, so that infor-

mation from different levels can be mapped efﬁ-

ciently.

3. Other information models should be easily de-

rived from our model.

In order to achieve the described requirements, we

introduce a three-level architecture containing infor-

mation about CAD drawings and topological struc-

ture of building indoor (ﬁgure 1).

The ﬁrst module represents the input data of

the framework and contains CAD architectural ﬂoor

plans. They may include information about a vari-

ety of aspects such as structure, furnishing, plumb-

ing, electricity, etc. together with meta-elements such

as measurements and annotations. Due to the huge

variety of possible models represented in CAD ﬂoor

plans, we will proposea set of constraints on the range

of input data. The second module contains the infor-

mation relevantto represent the structure of a building

interior, i.e., walls and openings, obtained as the re-

sult of semiautomatically ﬁltering the information of

the ﬁrst module. The available low-level geometry el-

ements (wall lines and opening blocks), are processed

in order to obtain high-level information (walls, wall

intersections and openings). The third module con-

tains a topology graph derived from walls and open-

ings. Its associated dual graph represents the subdi-

vision of a building into closed spaces. Finally, the

SEMANTIC AND TOPOLOGICAL REPRESENTATION OF BUILDING INTERIORS - An Overview

171

room structure gathers structural and semantic infor-

mation (topology graph) and geometric information

(wall lines).

An initial approach of our proposed model has

been tested with real CAD ﬂoor plans with promis-

ing results:

• CAD ﬂoor plans have been processed semiauto-

matically in order to get semantic and topological

features like walls, rooms (including their inner

polygons and their contour).

• Geometric information from CAD ﬂoor plans is

linked to semantic information obtained semiau-

tomatically.

• We have successfully implemented modules

which automatically export instances from our

model to CityGML and COLLADA.

Figure 1: Framework based in a three-level architecture to

represent building information models.

6.2 Comparative Analysis

Before presenting a summary of the main features of

the reviewed papers (table 1), we introduce some pre-

vious considerations related with geometry, topolog-

ical connectivity, topological adjacency and seman-

tics.

6.2.1 Geometry

The majority of the reviewed papers include geomet-

rical information as the basis of their models. How-

ever, they do not deal with geometry at the same

depth. We distinguish among the following items:

1. Works that do not mention anything regarding ge-

ometry because they only focus on semantic is-

sues. They appear in table 1 with a dash (-).

2. Works mentioning geometric elements without

giving details of the underlying representation.

They appear as implicit.

3. Geometric elements like vertices, edges, faces, re-

gions, discrete cells or volumes appear abbrevi-

ated respectively as V, E, F, R, DC, VO.

4. Other works use spread models like IFC,

CityGML, Geographic Markup Language (GML)

or BIM’s

6.2.2 Topology

We distinguish among connectivity and adjacency,

when applicable.

1. Two spaces in a building model are topologically

connected if there exists a door or a window be-

tween them. Thus, models with information about

openings have topological connectivity. Connec-

tivity is explicit if it appears represented in the

model, and implicit if it can be deduced by ana-

lyzing the model.

2. Two spaces in a building model are topologically

adjacent if they share at least one item (e.g. rooms

sharing one wall). Adjacency is explicit if the

model contains information about relationship be-

tween spaces, or implicit if the model does not

contains this information, but it can be deduced

analyzing the geometry.

6.3 Semantics

For each reviewed work, we specify which semantical

items it contains, according to the legend: RO, O, PA,

CR, S, T, W, L, C, CO represent respectively rooms,

openings, passages, crossroads, stories, tags, walls,

lifts, ceilings and corridors.

7 CONCLUSIONS AND FUTURE

WORK

In this work we have analyzed a number of exist-

ing approaches to the problem of the representation

of building indoor models. Initially, a set of criteria

to classify the existing models have been introduced.

According to these criteria, some works have been

classiﬁed into three main groups: (1) 2D (2) 2.5D and

(3) 3D.

Another topic covered in this paper has been the

analysis of the wide range of building models accord-

ing to geometric, semantic and topological features.

This analysis allows us to state some conclusions:

• Due to the huge variety of existing representation

models for building indoors, and the wide range of

ﬁelds of application, it is quite complex to achieve

GRAPP 2012 - International Conference on Computer Graphics Theory and Applications

172

Table 1: Comparison between the overviewed works.

Group Work Geometry Topology Semantics

Connectivity Adjacency

2D models

(Franz et al., 2005) - Implicit Explicit RO, O

(Lamarche and Donikian, 2004) Implicit Explicit - PA, CR

(Pluemer and Groeger, 1996) V, E - Implicit -

(Stoffel et al., 2007) V, E, R Explicit Implicit RO, O, S

(Li et al., 2010) DC - - T

(Zhi et al., 2003) V, E, R Explicit Implicit R, O

(Hahn et al., 2006) Implicit Explicit Implicit RO, O, S

(Merrell et al., 2010) Implicit Explicit Implicit typed-RO, O

2.5D models

(Slingsby and Raper, 2007) Implicit - Implicit W, L, O

(Tutenel et al., 2009) Implicit - - RO

(Germer and Schwarz, 2009) Implicit - - RO

(Dongen, 2008) Cubes - - W, C

(Choi et al., 2007) - Explicit Implicit RO, O, S, W

3D models

(Choi and Lee, 2009) R Explicit Explicit RO, O, CO

(Clemen and Frank, 2008) V, E, F, P Explicit Explicit -

(Van Berlo and de Laat, 2010) IFC, CityGML Explicit Implicit RO, O, S, W

(Paul and Bradley, 2003) V, E, F, VO Explicit Explicit W, C

(Billen and Zlatanova, 2003) V, E, F, VO Implicit Explicit Buildings

(Hagedorn et al., 2009) GML Explicit Explicit RO, O, S, W

(Van Treeck and Rank, 2007) B-Rep Explicit Explicit RO, W

(Borrmann and Rank, 2009) VO - - Buildings

(Isikdag et al., 2008) Implicit Implicit Implicit O, S, W

(Boguslawski and Gold, 2010) V, E, F - Explicit -

(Yan et al., 2010) BIM BIM BIM BIM

(Xu et al., 2010) F, VO Explicit Explicit RO, O, S, W, C

Our proposal V, E Explicit Explicit RO, O, PA, S, W, C

Legend: V=vertices, VO=volumes, E=edges, F=faces, P = planes, R=regions

DC=Discrete cells, RO=rooms, O=openings, PA=passages, CR=crossroads

S=stories, T=tags, W=walls, C=ceilings, L=lifts, CO=corridors

uniﬁed models. Thus, we consider unavoidable a

thorough design of the problem to be solved, in-

stead of the application of general approaches.

• In most of the reviewed works, there exists a

lack of automation for getting building models,

without regard of the ﬁeld (BIM, GIS, Spatial

Databases or custom models). Therefore, algo-

rithms for the extraction of semantic information

from CAD ﬂoor plans need to be developed. In

this area, we have proposed some methods to

recognize rooms semi-automatically from vector

ﬂoor plans in AutoDesk



DXF format, obtaining

promising initial results.

• The use of formal approaches is recommendable,

since it allows to take advantage of demonstrated

results. For example, a lot of works formulate a

problem in terms of graphs. Thus, existing algo-

rithms from graph theory can be applied without

need to demonstrate the validity of the solution.

Some of our work (in progress) tries to solve the

wall automatic recognition using graph theory.

REFERENCES

Billen, R. and Zlatanova, S. (2003). 3d spatial relationships

model: a useful concept for 3d cadastre? Computers,

Environment and Urban Systems, 27(4):411 – 425. 3D

Cadastres.

Bjork, B.-C. (1992). A conceptual model of spaces, space

boundaries and enclosing structures. Automation in

Construction, 1(3):193 – 214.

Boguslawski, P. and Gold, C. (2010). Rapid modelling of

complex building interiors. In Proceedings of the 5th

International 3D GeoInfo Conference.

Borrmann, A. and Rank, E. (2009). Speciﬁcation and im-

plementation of directional operators in a 3d spatial

query language for building information models. Adv.

Eng. Inform., 23:32–44.

Cerovsek, T. (2010). A review and outlook for a [‘]building

SEMANTIC AND TOPOLOGICAL REPRESENTATION OF BUILDING INTERIORS - An Overview

173

information model’ (bim): A multi-standpoint frame-

work for technological development. Advanced Engi-

neering Informatics, In Press, Corrected Proof:–.

Choi, J. and Lee, J. (2009). 3d geo-network for agent-

based building evacuation simulation. In Lee, J. and

Zlatanova, S., editors, 3D Geo-Information Sciences,

Lecture Notes in Geoinformation and Cartography,

pages 283–299. Springer Berlin Heidelberg.

Choi, J. W., Kwon, D. Y., Hwang, J. E., and Lert-

lakkhanakul, J. (2007). Real-time management of spa-

tial information of design: A space-based ﬂoor plan

representation of buildings. Automation in Construc-

tion, 16(4):449–459.

Clemen, C. and Frank, G. (2008). Architectural indoor sur-

veying. an information model for 3d data capture and

adjustment. In Proceedings of the American Congress

on Surveying and Mapping.

Dongen, J. V. (2008). Interior mapping. In CGI 2008 Con-

ference Proceedings.

Franz, G., Mallot, H. A., and Wiener, J. M. (2005). Graph-

based models of space in architecture and cognitive

science - a comparative analysis. In Proceedings of the

17th International Conference on Systems Research,

Informatics and Cybernetics.

Germer, T. and Schwarz, M. (2009). Procedural arrange-

ment of furniture for real-time walkthroughs. Com-

puter Graphics Forum, 28(8):2068–2078.

Hagedorn, B., Trapp, M., Glander, T., and Dollner, J.

(2009). Towards an indoor level-of-detail model for

route visualization. In Proceedings of the 2009 Tenth

International Conference on Mobile Data Manage-

ment: Systems, Services and Middleware, MDM ’09,

pages 692–697, Washington, DC, USA. IEEE Com-

puter Society.

Hahn, E., Bose, P., and Whitehead, A. (2006). Persis-

tent realtime building interior generation. In Pro-

ceedings of the 2006 ACM SIGGRAPH symposium on

Videogames, Sandbox ’06, pages 179–186, New York,

NY, USA. ACM.

Isikdag, U., Underwood, J., and Aouad, G. (2008). An

investigation into the applicability of building infor-

mation models in geospatial environment in support

of site selection and ﬁre response management pro-

cesses. Advanced Engineering Informatics, 22(4):504

– 519. PLM Challenges.

Lamarche, F. and Donikian, S. (2004). Crowd of virtual hu-

mans: a new approach for real time navigation in com-

plex and structured environments. Computer Graph-

ics Forum, 23:509–518.

Lee, J. (2001). 3d data model for representing topologi-

cal relations of urban features. In Proceedings of 21st

ESRI International User Conference.

Li, X., Claramunt, C., and Ray, C. (2010). A grid

graph-based model for the analysis of 2d indoor

spaces. Computers, Environment and Urban Systems,

34(6):532 – 540. GeoVisualization and the Digital

City - Special issue of the International Cartographic

Association Commission on GeoVisualization.

Merrell, P., Schkufza, E., and Koltun, V. (2010). Computer-

generated residential building layouts. ACM Transac-

tions on Graphics, 29(6):181:1–181:12.

Paul, N. and Bradley, P. E. (2003). Topological houses. In

Proceedings of the 16th International Conference of

Computer Science and Mathematics in Architecture

and Civil Engineering.

Pluemer, L. and Groeger, G. (1996). Nested maps - a for-

mal, provably correct object model for spatial aggre-

gates. In Proceedings of the 4th ACM international

workshop on Advances in geographic information sys-

tems, GIS ’96, pages 76–83, New York, NY, USA.

ACM.

Slingsby, A. and Raper, J. (2007). Navigable space in 3d

city models for pedestrians.

Stoffel, E.-P., Lorenz, B., and Ohlbach, H. (2007). Towards

a semantic spatial model for pedestrian indoor navi-

gation. In Hainaut, J.-L., Rundensteiner, E., Kirch-

berg, M., Bertolotto, M., Brochhausen, M., Chen, Y.-

P., Cherﬁ, S., Doerr, M., Han, H., Hartmann, S., Par-

sons, J., Poels, G., Rolland, C., Trujillo, J., Yu, E., and

Zim´anyie, E., editors, Advances in Conceptual Mod-

eling Foundations and Applications, volume 4802 of

Lecture Notes in Computer Science, pages 328–337.

Springer Berlin / Heidelberg.

Tutenel, T., Bidarra, R., Smelik, R. M., and de Kraker, K. J.

(2009). Rule-based layout solving and its application

to procedural interior generation. In Proceedings of

the CASA workshop on 3D advanced media in gaming

and simulation (3AMIGAS).

Van Berlo, L. and de Laat, R. (2010). Integration of bim

and gis: The development of the citygml geobim ex-

tension. In Proceedings of the 5th International 3D

GeoInfo Conference.

Van Treeck, C. and Rank, E. (2007). Dimensional reduc-

tion of 3d building models using graph theory and its

application in building energy simulation. Eng. with

Comput., 23:109–122.

Weiler, K. (1988). The radial-edge structure: a topologi-

cal representation for non-manifold geometric bound-

ary modeling. Geometric Modeling for CAD Applica-

tions.

Xu, W., Zhu, Q., and Zhang, Y. (2010). Semantic modeling

approach of 3d city models and applications in visual

exploration. International Jorunal of Virtual Reality,

9(3):67–74.

Yan, W., Culp, C., and Graf, R. (2010). Integrating bim and

gaming for real-time interactive architectural visual-

ization. Automation in Construction, In Press, Cor-

rected Proof:–.

Zhi, G., Lo, S., and Fang, Z. (2003). A graph-based al-

gorithm for extracting units and loops from archi-

tectural ﬂoor plans for a building evacuation model.

Computer-Aided Design, 35(1):1–14.

GRAPP 2012 - International Conference on Computer Graphics Theory and Applications

174