3.Types of Model

The distinction between perception and structural models was first made, to my knowledge, by Mark Lauden Crosley (1988, 124) who talked of 'perceived' and 'built' models in architecture. A perceived wall is represented as a single, rectangular object, while a built wall is represented in terms of the actual units of construction. I shall deal first with models of perception, as they are the sort which have been most produced in archaeology, and then move on to models of structure, which is an area where, I argue, there is much scope for development.

3.1 Perception Models

Although data visualization (including virtual reality) is an important part of three-dimensional modelling, I argue that it should be explicitly divorced from the related field of photo-realism at a research level. Perception modelling can be performed by surface or solid modellers.

Data Visualization

It would be possible to store (and where applicable process) all spatial data without any attention to presentation (as for example is possible with CAE - see above). However, a series of algorithms can be used to present the data (Fig 4) in a way more amenable to visual interpretation (e.g. hidden line removal (Fig 5), shading surfaces (Fig 6), or rendering (Fig 7) (rendering is the addition of a visible surface. Rendering algorithms rest upon calculating the way in which light and shadow fall upon surfaces in a picture).

Conventionally, archaeological spatial data are recorded and interpreted in two dimensions. Visualizing these data in three dimensions is of great assistance in understanding and interpreting spatial relations. The modelling of the Bath Temple Precinct enabled Barry Cunliffe to use a 'synthetic camera' to view the precinct to attempt to understand, among other things, the architect's conception of the relation between power and space (Reilly 1992, 150). The computer model for the baths at the Roman fortress at Caerleon was the only part of the exhibition to give a real impression of how the spaces in the building related to one other (Chapman 1990, 33). The problems of conceptualising the formation of the tell site of Toumba Thessaloniki (occupying 1.7 ha at the base, rising 23 m above the modern ground level and encompassing two and a half millennia of occupation) were reduced with the aid of a CAD surface model (Kotsakis et al. 1995).


In addition to representing what is present, it is straightforward to work with reconstructions. Such reconstructions should start with a model of what has been recorded in situ, with the construction of a model in mind. This applies to almost all reconstructions to which I have found reference. A great advantage of computer reconstructions over pen and paper drawings is the flexibility in making a large number of small changes to a drawing without having to re-draw the whole. Such alterations are perhaps easier to make on mainframe solid modellers such as WINSOM (Winchester Solid Modeller) whose data are expressed as programming language code (in WINSOM's case the Extensible Solid Modelling Editor) rather than through a CAD graphical interface, where the variables to be altered are not always explicit (Burridge et al. 1989, 550; Reilly 1992, 155).

Projects chosen for reconstruction work have tended to concern sites characterized by the repetition of a few elements, or a simple design, for example the Bath Temple Precinct, the legionary baths at Caerleon and the Hoffman limekiln (Reilly 1992, 150-152; Chapman 1990, 44-70). Structural elements can be defined and repeated in the drawing with considerably less effort than defining a series of unique elements. This was one of the initial strengths of computer modelling which led to its adoption by designers (see Section 2.1). Work has also tended to be with sites of simple phasing. Reconstruction work is in many ways more similar to the design process than archaeological recording, as it is allowed to simplify to a far greater extent, and it is not constrained by the individual detail of change to one of a set of identical elements, and the detail of degradation after abandonment. The recording of this detail can increase the work involved by an order of magnitude.


Photo-realism works with the same core of representational algorithms as data visualization, but is concerned with producing a model which looks as much as possible like the original, rather than aiming at helping to conceptualise something about the way that the elements in the building connect together. There is a thin dividing line between these two approaches. Such is the degree of merger in graphical modelling that some solid modellers are now developed with photo-realism as a feature (Burridge et al. 1989, 548).

The interior of the Dresden Frauenkirche
Fig.8 The interior of the Dresden Frauenkirche (Schneider 1994)
IBM T. J. Watson Research Centre, Visualisation, Interaction, and Graphics department

The most prestigious recent example of photo-realism is the IBM Frauenkirche project, in Dresden (Collins 1993), intended largely to raise funds for the rebuilding programme. Apart from heritage work, the other main sphere in which photorealism is practiced is historic buildings recording; for example the Holy Sepulchre in Jerusalem (Chapman 1990, 37). Because these models are almost always produced by photogrammetry, which has had limited functionality within buildings, they have tended to be hollow shells with an enormous quantity of surface detail. These models are simply the continuation of traditional standing building survey work on a computer platform. The emphasis has become, however, the recording of every detail, without regard for its perceived significance. It is an attempt to remove conscious simplification from the modelling process by the production of a total record.

Virtual Reality

As computational power has increased, reconstructions have moved from a series of carefully chosen stills, as at Bath, to 'walk-throughs' or 'fly-throughs', where an animation has been produced allowing the viewer to follow a pre-selected course through the model. With improvements in visualization techniques and processor power The use of parallel processors significantly reduces the time taken to produce photo-realistic images as has been shown by the INSITE project at the University of Bristol (Chalmers et al. 1995).

Fig 9 Reconstructed interior, Ggantija Temple, Malta

It should be possible to develop virtual reality tours of buildings or sites. The obstacle is that the animation must be developed in real time rather than as a batch process in advance. A limited virtual reality tour has already been achieved with the GIS package Grass running on a Silicon Graphics UNIX workstation, a type of computer designed for graphical work (Forte & Sarti 1995). There must in any case be a pay off between the complexity of the model, computational speed and the quality of the animation (measured, for example, in frames per second).


Early three-dimensional work in buildings archaeology (e.g. at Bath, Caerleon and Furness Abbey) was very commonly done using solid modellers. This was because of collaboration with software engineers, both in university departments and IBM's UK Scientific Centre, who wished to develop the visualization capabilities of their solid modellers, and test their robusticity on a more complex problem than design (c.f. Chapman 1990). The collaboration between archaeology and engineering was not a 'free lunch' for archaeologists; they had to agree to be guinea pigs for the research of computer programmers. As solid modellers were both more difficult and more expensive (in both labour and CPU time) to use than the surface modellers which were becoming available at the same time, it is difficult to justify the use of solid modellers for perception modelling.


© Internet Archaeology
Last updated: Thu May 1 1997