Fos.prd.uth.gr

On Developing Standards for the Creation of VR
City Models
BOURDAKIS, VassilisLaboratory of Environmental Communication and Audiovisual Documentation, University ofThessaly, Greecehttp://fos.prd.uth.gr/vas/ The paper is an inclusive summary of research work on creating VR city models carried
out over the last six years in the UK and Greece aiming to put into discussion the guidelines/
rules developed by the author.
The paper is structured in three sections referring to the main stages in terms of either
technical expertise and problem solving or conceptual structuring of information: creation
of 3D city models, CAAD versus VR in digital city modelling and finally utilizing digital
city models.
The expected outcome of the work presented is the establishment of a body of knowledge
that will facilitate the development of standards and guidelines for the creation of city
models. There are obvious advantages in having a compatible set of city 3D models. On
the other hand, there are different rules to be followed and issues to be solved, according
to the scale of the model, level of detail that is needed—all these rules relate to the projected
use of the model.
Keywords: digital city models, 3D modelling, virtual reality, urban planning
Background
application, due to lack of architectural / urban Architectural and urban design is full of assumptions planning experience in the field—with the exception and conventions to an extent not applicable to many of work carried out in the UCLA Urban Simulation other disciplines. This has always been the case with paper based drawings considering the fact that 3D (www.artcom.de/contacts/city-and-architecture/ information, the urban structure, had to be expressed welcome.en.old.shtml:May, 2001) and others—VR and communicated on a 2D medium, the drawing.
conventions and methodologies are adopted.
This approach has worked quite well with paper Furthermore, a close examination of the techniques drawings since professionals are trained to read the above mentioned groups use, reveals that VR drawings quickly and with minimum effort, though methodologies are taken for granted and therefore average performance is obtained using a computer.
carefully tuned and appropriately built models are The main reason being that one has to decide and employed rather than custom tools for urban scale either imitate the paper based approach and thus do everything in 2D or advance to 3D modelling. Opting for the second means that a new set of rules and experience obtained in constructing the VRML models conventions have to be invented/developed together of Bath city, and London’s West End together with with an even better understanding of 3D space. In the work he is currently involved in Volos, Greece to creating a fully 3D interactive Virtual Reality (VR) develop a set of standards for the creation of CAAD urban models and, mainly, their conversion to VR 2D modellers in general tend to over-design, model models. Due to the nature of the topic and the size in deep detail, usually irrelevant for a VR application.
limitation of the paper, the work carried out so far by This well established trend must be avoided. The only the author is not analytically documented. It is solution is prototyping a representative area of the recommended to refer to a series of papers by the proposed model, converting to the final VR author addressing VR city modelling in depth. Most environment and testing before establishing the detail papers are also available online (fos.prd.uth.gr/vas/ Data source scale and projected use are also important in establishing the accuracy level of the Creating 3D digital city models
model. 2D source data should be at a scale close to The starting point for any digital model is the data 1:1000 and definitely better than 1:5000 if the resulting source. At an urban level, source could be the existing model is to be of an accuracy level of less than half a 2D city plans and stereo pair aerial photographs, metre and close to twenty centimetres. The aerial bearing in mind that most 2D city plans are based on photographs scale is vital at achieving a high level of photogrammetric data. However it is recommended detail and should be close to 1:3000–anything over that, in order to have maximum compatibility with that results in 3D models with an accuracy of more existing city plans, creating the whole 3D model from than a metre a value unacceptable for planning scratch should be avoided. Established work in terms oriented city models, but may well be suited for other of defining the 2D data for each properly, element in general is not replicated thus model construction is a hybrid one using existing plan data and aerial constructs expanding over kilometres in length and photographs for the extraction of the elevation width. It is not practical to work on such a project information of the buildings, height of trees, chimneys, unless a subdivision strategy for the digital model is roof geometry information, dormer windows etc.
employed. The aim of such a subdivision is twofold; Furthermore, discrepancies on plan data are avoided to be able to store the source data in manageable simplifying the job of engineers that may use these sized files and to enable engineering professionals obtain the needed parts of the city in order to base Choosing the appropriate software platform is their work on a well-established dataset and keep the essential. Using a widely available commercial main 3D database of the city up to date. The most software platform warranties compatibility, ease of appropriate subdivision unit is that of an urban block access to prime data source and conversion to other furthermore simplifying naming conventions by data formats. Depending on the projected use of the following the current urban block numbering system.
model, in-house tools may be needed. However, careshould be taken in order to be able to export the Making the step from CAAD to VR in
geometry to one of the widely accepted formats city modelling
namely Data Exchange Format (DXF) and Virtual Moving from a CAAD 3D model to a VR application Reality Modelling Language (VRML). Similarly, the involves the complex stage of data conversion. Issues software used should be compatible with data common to all VR applications will be analysed.
obtained from foreign sources–engineers working on Urban models are very often constructed using a particular project within the area modelled, modular the country’s unit origin in metres. For example in the UK, Ordnance Survey origin is used; that is Lands The next issue that has to be tackled is that of the End, Cornwall the most south-western tip of Britain.
level of abstraction in modelling. CAAD operators and CAAD software have no problems dealing with large Modeling & City Planning – 15 3D City Modeling numbers (a typical city modelled maybe a million units Having converted the dataset the success of a away from origin point) or with the units; a metre with VR application is judged by the degree of geometry three decimal points. However, VR applications often use integer mathematics for certain geometric VR applications have a fairly clearly defined upper calculations. This leads to a series of rounding errors limit of amount of geometry they can handle and great problems with Z buffering. Furthermore, successfully which is quite low and unsuitable for trying to rotate or spin the model is impossible since urban scale modelling. In general, software browsers rotate about the co-ordinate origin; 0,0,0.
developers and VR designers/artists recommend All navigation, rotation, and most Z buffering problems replacing geometry with simplified repetitive texture are solved once the origin point is translated to the mapped shapes. However, this approach can only be centre of the urban model. A side effect is the file size successfully implemented in certain types of models reduction since all numbers dealt with are much and it seems to be producing acceptable results in American towns, skyscraper filled city centres high- CAAD software use a world based co-ordinate rise office blocks and generally highly repetitive system; X and Y for the plane definition and Z for environments. As an example, the author in the elevation. VR software typically utilise a screen based process of converting the CAAD model of Bath in VR, co-ordinate system–X and Y across the computer classified the existing types of properties, roof screen and Z out of the screen. This effectively means geometry, and other urban entities in a list of over that Y and Z have to be swapped from a CAAD model 200 elements (Bourdakis, 1996). Bearing in mind the to a VR. Depending on the CAAD software, the output generalisations that took place in creating this VR format, and the process followed, this is dealt with “restricted” list, the projected utilisation of the model varying degrees of competence and correctness.
and the inevitable downgrading of the available data, Sometimes the conversion process introduces it was decided to discard this approach. Repetitive transformation matrices making the VR output file elements indeed exist in urban scale models but due difficult, if not impossible, to comprehend and edit to the level of detail of the available data it was manually. Failing to exchange the Y and Z results in restricted to building elements: windows, doors, models that cannot be “walked” through, since the cornices and chimney stacks as well as street viewer-perceived walking is carried in X and Z meaning the viewer comes flying from the sky to the Architects and planners have a concept of Levels of Detail (LOD) based on the paper scaled A serious problem of model translations is the representation of real space (1:5000, 1:1250, 1:500 structure of the geometrical description itself. In most etc). Projects very often are seen in scales up to 1:100 CAAD packages, the operator can define surfaces or 1:50 for a building, which means a building is that are perceived as double sided. Indeed, the isolated and examined at a higher level of detail.
standard on hand modelling surfaces is double sided.
However, there is no such concept as spatial structure This means, considering three points in 3D space, within different levels of detail. The whole area is the surface defined by triangle (A, B, C) is the same sequentially “worked” at different levels but it is never as the one by triangle (A, C, B). VR applications visualised with varying levels of detail at the same usually define surfaces as being single sided and time (maybe VR’s way of “seeing” the environment is anticlockwise. Some renders have the option of going to be accepted and approved by engineers - it rendering double sided faces - at the penalty of a is simply the way the paper based representations considerable speed reduction usually by a factor of are structured that causes this behaviour). The closest concept to organising information in building practice is to use BLOCKS/ INSTANCES/GROUPS (naming be noted that not all windows, doors etc. of an conventions used on different modelling software) urban block are under one LOD node. LOD nodes where a set of objects are joined together, defined as are created on the basis of keeping concise, more one, and used in various other places without the need or less square (in plan) areas together. This usually means organising them per street facade.
Implementing LODs on VR urban models needs a completely different approach. A highly detailed Level 4 architectural detail such as chimney pots,
urban block may be over 20,000 triangles. One cannot string courses and pilasters are added. At this expect that more than a few such blocks will be easily level, some photographic texture maps are also navigable considering current graphics hardware included on windows and shop fronts. The Level capabilities. A low polygon count representation (30- 3 structure is kept and Level 4 is visible at 50 triangles) of each urban block should be used instead, when the camera is a few hundred metres Landscape modelling is an issue addressed differently away. However care must be taken as there is a in 2D modelling and VR applications. The main threshold on how many LOD calculations are problem is the continuity of the landscape and the acceptable, versus geometry / texture use. For inability to use LODs as described previously–having example, deciding to add textures on building facades, different resolution tessellated models of the and switching them on and off per building using LOD landscape to exchange at set distances is a very nodes will bring the application to a halt, not because resource consuming exercise. GIS companies have of the burden of loading all these textures, but due to researched this issue and there are terrain the need to do all the LOD checks for each building on visualisation solutions available (Terravision by SRI each camera movement! It is better for the browser International www.ai.sri.com/TerraVision/:May 2001) to do tests for 4 LODs per urban block than 200. This that will have to be integrated with the urban model.
leads to a sub-structuring of the model into streets Having converted and optimised the dataset to within each main urban block. Long streets may be VR the final issue that has to be considered is the process of updating the city database. Typically, there The following structure has been developed by will be alterations to the city model (either because of the author and successfully tested at the Bath city new developments or via the use of the VR model in model (Bourdakis & Day, 1997). The four levels of planning evaluation stages etc). Since most VR platforms available are not suitable for real-time editing Level 1 a simple volumetric description of each
of the underlying geometry, the CAAD database must terrace with a flat roof at the average height for be updated and the relevant alterations re-exported that terrace (typically under 200 triangles per to VR–effectively establishing the CAAD database as urban block). Roads, pavements, and landscape Level 2 each building is modelled with accurate
Utilizing digital city models
wall and roof geometry and tagged as a separate Implementational scope for city models
object in the model. This means that each property Urban VR models can be broadly classified under in the city can be identified and used for data three main categories (Bourdakis, 1998a): design and linking. Trees that are within the urban block are planning, education and general research and finally also visible. Visible at 150 metres.
commercial and entertainment. It should be noted that Level 3 windows, doors, parapets, party walls and
implementational directions vary greatly according to free-standing garden walls are added. It should Modeling & City Planning – 15 3D City Modeling the application scope with certain tasks being clearly of information is not an issue, quality of the image is fairly poor and not appealing or even attracting attention, questioning their financial and commercial demonstrate (Shiffer, 1995) how and to what extend computers will be used in the near future by engineers Concluding with the potential uses of urban VR as part of their everyday practice, creating, modifying systems, it should be stressed that in many of the and improving our cities online using centrally stored examples discussed above, the notion of a city space sharable databases (Day et al, 1996). Due to the is used loosely, denoting the various ways people nature of the proposed use of such models, the low perceive cities. It may vary greatly from one building polygon count fully texture mapped model approach complex, to a High Street, a neighbourhood or indeed adopted by more commercially oriented projects (Virtual Soma www.planet9.com/vrsoma.htm:May Limitations of VR
2001) is not feasible lacking severely in accuracy and An analysis on creating standards for VR urban models would be incomplete without some warnings Over the last decade, the potential of VR as a and a future work section criticising design or teaching aid has been under investigation. Research implementation limitations of VR applications as they however has focused on primary and secondary education and as such, urban environments have not VR applications’ ability to handle the sheer size been employed. The closest to urban scale projects of the models involved. Even with careful planning are investigation studies on spatial ability via “you are and use of LODs rendering frame rate is suffering.
here” type of maps for pre-adolescents (Phillips, Large triangle based models are not supported very 1997). Research work, investigating abstract data efficiently since software engineers focus on primitive representations based on architectural notions of based models. Furthermore, accuracy is not highly space such as Vector Zero and CASA’s own “Map of valued amongst VR developers and the need for three the Future” where the digital city becomes the front decimal points is often seen as excessive and wasting end facilitating navigation, should be noted.
The lack of copyright protection or digital signature commercially viable fields is quite different to research stamping in the geometry together with the ease of and development. The implications and return of transferring files across the Internet is hindering the investment is one issue that is extremely difficult to development and availability of urban models.
assess and thus persuade the client. Consequently, VR applications are largely customised for the “wow” factor and the hi-tech issue is the main repetitive geometry, tasks, behaviours etc. Once that selling point of the technology at it currently stands.
fails to be the case, VR apps tend to be inefficient This is clearly demonstrated by the various urban VRs and slow. Furthermore, hardware seems to be badly (in the loose sense of the term) created to enhance, tuned for VR related tasks—typically available improve virtual shopping mall applications which grow hardware provide either good CPU performance and is a fashion similar to that of shopping TV channels low graphics (PCs), or the opposite (graphics last decade. Typical examples are North American city centres, focusing on the prominent city landmarks Navigation is another issue that needs serious (mainly skyscrapers, large office blocks and distinctive consideration. In paper based environments, routes), which are pasted on a 2D map of the area.
everything is right in front the engineer’s eyes—the Detail is scarce, accuracy is questionable, conveying only tool needed is a large drawing board. On the computer screen, in non-VR approaches, the screen is viewed as a “window” to a much larger drawing, Communication Tool for Urban Planning in A.
plan view is the one used most of the time. In a VE, the screen size or HMD limitations force us to an Towards New Design Conventions TechnicalUniversity of Bialystok, pp.45-59.
approach closer to the CAAD one. However, a plan Bourdakis, V.: 1997b The Future of VRML on Large only view is unacceptable, walking on ground level Urban Models in R.Bowden (ed) UKVRSig’97 confusing and generally disorienting (Bourdakis, Bourdakis, V. and Day, A.: 1997 A VRML model of Bath in R.Coyne, M.Ramscar, J.Lee & K.Zreik Results – Proposals for VR city
(eds) Design and the Net, europIA Productions, models standards’ development
Summarising the issues presented in this paper, the Bourdakis, V.: 1996 From CAAD to VR; Building a author proposes a set of rules for future urban scale VR models: Plan compatibility by modelling based on Third UK Virtual Reality Special Interest GroupConference; Full Paper Proceedings. De existing 2D plans, Standardization on units, Origin– Montfort University, Leicester, pp.5-13.
relative coordinates, five LOD construction (the four Day, A.K., Bourdakis, V. and Robson, J.M.: 1996 described plus textured landscape form), VRML 97 Living with a virtual city in Architectural format for delivering the 3D geometry, library of Research Quarterly, Vol.2: Autumn 1996, pp.84- reusable complimentary elements (street furniture, trees, etc) work on metadata–ways of mapping Jepson, W., Liggett, R. and Friedman, S.: 1996, information on the model. This is by no means an Virtual Modeling of Urban Environments in exhaustive set of rules but a starting point for further Phillips, P.C.: 1996 Portraying the City. In: P.C.Phillips (ed) City Speculations Princeton References
Bourdakis, V.: 1998a, Utilising Urban Virtual Shiffer, M.J.: 1995, Multimedia Representational Aids in Urban Planning Support Systems in F.T.
Marchese (ed) Understanding Images; Finding Bourdakis, V.: 1998b, Navigation in Large VR Urban Meaning in Digital Imagery Telos, Springer- Virtualworlds,Springer-Verlag Berlin,pp.345-356.
Modeling & City Planning – 15 3D City Modeling

Source: http://fos.prd.uth.gr/vas/papers/ecaade01/15_01_bourdakis.pdf