5. A Web3D-orientated Predictive Model

After the predictive model was set up, it was optimised in order to be communicated through a Virtual Reality system. This work was done over a variety of stages, carried out by using different software platforms. Firstly, in a GIS environment, the previously collected data (and so the predictive model), were converted to VRML files; then, all of these were adjusted using 3D modelling software and, finally, exported into an innovative Web3D-orientated development platform.

5.1 From GIS to VRML

Starting from a TIN-based terrain model, obtained by interpolating contour values and elevation points, a Z-value parameter was added to all the cartographic data, both in vector and raster format. In this way, a GIS 3D Analyst tool enabled us to turn multiple layers such as lithology, land use, topographic and ortho-photographic maps into 3D models which then have been geo-referenced at the same coordinate system.

As for the representation of the known sites, since one of the main purposes of this virtual representation was to provide users with an effective way to visualise and interact with the archaeological dataset derived from a GIS, the symbology of the archaeological record was conceived in order to provide an immediate perception of the chronology. As end-users are not necessarily going to be content experts, we chose to simplify the archaeological record by splitting it into three main symbol categories, representing as many chronological macro-areas. Therefore, we started by distinguishing a Prehistoric, a Classical and a Medieval period, in which the 104 known sites (pertaining to a very broad range of chronological phases) have been classified. For that reason, we decided to put into the Prehistoric period category all the sites dating to the Palaeolithic (middle and late), the Eneolithic, the Neolithic, the Bronze and Iron Ages. We put the Archaic, Hellenistic and Roman period sites into the Classical category, while all later ones were assigned to the Medieval period.

After this database allocation, it was necessary to choose some representative, symbolic 3D models to indicate which chronological phase they related to. To this end, a very fast and effective solution was provided by looking at the Google 3D warehouse, a Web repository where anyone can search for and download specific models, based upon specific keywords.

Figure 6: 3D icons illustrating different site tipology and chronology

Figure 6

We used different three-dimensional icons to distinguish each site according to its own original function (Figure 6). In order to make users easily understand the chronological distribution and the typology of the archaeological sites all over the study area, we adopted two main criteria in their representation: firstly, we chose to use different colours to distinguish the chronological phase each site pertained to, using green for Prehistoric sites, red for Classical and blue for the Middle Ages. Further information about the original site function is then provided by the icon shape. While some symbols represent a specific chronology (e.g. a lithic artefact-shaped icon marks only a Prehistoric site), others can define more than one time period (e.g. an anchor identifies both a Roman and a Medieval harbour).

Afterwards, each model was exported from GIS as a VRML file. This is a 3D file format, originally designed as a web standard for VR applications, which currently serves as a de facto standard for 3D data interchange, as almost every software application dealing with 3D is able to import and export models coded in this format.

Thus we produced different VRML models, one for the terrain representation, made up of overlapped aerial ortho-photos, one for the toponomastics of the study area, expressed by means of 3D text icons, and other models for the thematic layers concerning the land use, the lithology and, most of all, the map of the archaeological risk derived from the predictive model. Moreover, a terrain model along with 3D symbolic icons representing the distribution of the known sites enabled us to obtain a complete representation of the archaeological landscape of the study area. In this manner, such a virtual landscape integrates superimposed information related to the toponomastics, geological and land use aspects as well as the archaeological record represented by intuitive 3D symbolic icons expressing both the different chronological phases and the functions of the human settlement in this territory.

5.1.1 3D Modelling

In a further stage of the work, VRML files were imported into 3D modelling software in order to be processed and optimised for a better outcome in a web-orientated environment.

One of the problems of exporting models from GIS3D extension is the incorrect representation of some geometrical features, because of different conventions between different software applications: the normal vectors used to determine the orientation of the polygons, are always inverted and because of this the final outcome is a model in which the texture is not visible. Additionally, GIS3D extension tends to duplicate vertices in the final VRML model, and we found many overlaps, resulting in data redundancy and shading problems. In order to fix these problems, prior to proceeding to the simplification of meshes (in order to make them usable for real-time usage), we had to perform simple post-process steps, such as normal flipping and vertex welding. Finally, the model was converted into the AAM geometry format (which is the XVR native format for the description of triangular meshes) in order to be visualised with the XVR technology (Carrozzino et al. 2005). For this purpose, it is necessary to have previously plugged in a specific exporting tool.

5.1.2 The Web-orientated model

XVR allowed us to prototype a 3D real-time application quickly, where the different models previously obtained have been integrated in order to set up a Web3D-orientated model. Once the VRML models had been converted to AAM files, we obtained different, independent 3D thematic representations of the Pisa coastal plain landscape i.e. land use, lithology, known sites and archaeological predictive model. At this point, all of the previously obtained data had to be integrated in a virtual environment in order to be visualised with a typical web-browser (it should be stressed that so far only Internet Explorer is supported via ActiveX Control, although Firefox and Chrome can run it via the 'IE View' NPAPI extensions and more extensive support to other browsers/platform is currently under development). For this reason, all the data were imported inside XVR Studio, the Integrated Development Environment which enables the building of a VR application and then embedded in a web page, with dedicated HTML code automatically added by the IDE.

Series of screenshots from XVR model

Select the image above to view the model in a new window [Instructions]. A small activeX plug-in is required. For non-IE users, use the 'IE View' NPAPI extensions (available for Firefox and Chrome users).

Once the functions related to the interactive visualisation of the model had been defined, by means of a VR-orientated scripting language, the application was compiled in a bytecode that will subsequently run on the XVR Virtual Machine and visualised in the 3D graphics context embedded, as previously described, inside the web page. As for the Pisa coastal plain model, we decided to visualise and experience it in two different ways: a passive and an interactive mode. As a matter of fact, the purpose was to test the process by presenting the model at an international conference, where an off-line installation was set up in order to let people experience such a model either by passively watching it or by navigating and moving into the virtual environment almost without any limitation, just by dragging the mouse. Users were also able to switch through different thematic layers, such as land use and archaeological risk maps (via function keys).


© Internet Archaeology/Author(s)
University of York legal statements | Terms and Conditions | File last updated: Thu Dec 1 2011