Local time: Friday, 05-Dec-2008 12:08:40 EST
Last update at /special/conference/apwww95,Friday, 21-May-2004 09:47:08 EST

Scientific Visualisation using VRML

Geoff Horne
horne@vislab.su.edu.au, http://www.usyd.edu.au/su/vislab/

Bernard Pailthorpe
bap@vislab.su.edu.au

Ben Simons
ben@vislab.su.edu.au
Matthew Arnison

matthewa@vislab.su.edu.au

Nicole Bordes
nicole@vislab.su.edu.au
Abstract
The Sydney Regional Visualisation Laboratory, Sydney VisLab, provides a user interface for researchers from both universities and industry to effectively process and visualise data as well as providing a front end to Australia's supercomputers and large scientific instruments.

The use of the image capable WWW browsers has allowed these researchers to publish their work to a wide audience, both to their peers and to people with a general interest in the field of research. However the WWW allows these researchers to only provide `snapshots' of their work.

The recent introduction of the Virtual Reality Markup language (VRML) now means that entire `environments' can be created, centred around a particular area of research. This could range from something simple such as visualising some simple geometry through to a virtual museum or a complex virtual laboratorie that allows the user to manipulate and visualise a database of Three dimensional computer generated information.

Keywords
VRML Visualisation Webspace Science VisLab Publishing

HTML -> VRML

Limitations of HTML

HTML, while providing a more enhanced method of publishing information, still has its limitations. These limitationa are tied up in the conventional rules of typesetting and publishing. HTML was, and in some ways still is, a TEXT only language. Until recently images could not be merged cleanly with text. While the use of hyper links allows for the creation of more dynamic documents, with better cross referencing, the presentation of detailed and complex scientific information can be hampered by the linear restrictions of a page layout.

What is VRML ?

The Virtual Reality Markup Language (VRML) is a an addition to HTML for defining a 3 Dimensional object or space. Much like an inline image or movie, a link to a VRML document will point to an environment than can be explored and manipulated in 3 dimensions. For the purposes of publishing information it provides a completely new way to view information and creates `documents' that are in some ways more intuitive to the reader than plaint text or images. Some of the uses for VRML include:

3D indexes

Information that is better arranged in 3 dimensions can be laid out as objects in space. the relation of an index entry is thus spatial, with related references arranged around the object.

Navigation

An environment can be created for exploration, such as a virtual world or a model of the human body. instead of reading a description of the environment, it is possible to navigate through it and see the information

Object manipulation

Individual objects, such as engine parts or building models can be created and then manipulated. instead of producing elevation views of the object it is possible to manipulate the object in real time and to view it from any desired angle.

Scientific Visualisation

An inherent problem with any computational simulation is finding a simple way to analyse the results. With a physical experiment the results may come in a multitude of forms, from energy fluctuations and changes in radioactive decay to spectroscopic phenomena. The common factor with these results, however, is that they are observed, usually with the aid of an instrument. By producing information in this manner we enable the data to be quickly and easily understood. A computational simulation is not, without modification, as capable of producing such results. At best it can produce output in the form of numerical data. This is a significant impediment to most areas of research as a method is required to transfer numerical results into coherent visual images that allow us to absorb the information in a more natural manner. To date, the most significant and effective technique for interpreting, and communicating the results of a computational simulation is with the aid of 3D computer graphics and animation.

Visualisation serves a vital role in the understanding of information. As humans we interpret data in a slightly obtuse manner, we don't see specifics, we observe differences. We describe a conceptual idea rather than a specific incident.

Thus, when analysing data, if we want to see errors in a single field of a multi-dimensional dataset, it might be possible just to look for the abnormalities in the data. If, however, we wanted to see how these abnormalities were changing over time, or what the contributing factors were, then the results could not be obtained as easily. We would instead need to express this information, with the aid of visualisation, in such a manner as to make these trends visible. Plotting a graph provides us with instant information, visualisation is just the next step to this analysis. Rather than being restricted to simple 2 dimensional diagrams, it is possible to plot positional data, scalar fields and vector information all in the one space. Ideally all the information that we would wish to obtain from the data should be obtainable from a single viewing of the data .

The addition of colour and depth allows the observation of data that has many dimensions. We can use spatial position to define three dimensions but by adding colour we add another (4th) dimension. If we change the size of our point in space this becomes a 5th dimension, with the inclusion of vector maps we can add a further 3, 4, or 5 dimensions per vector, depending on how we display our information. With fine tuning of the display method (such as colouring all values green or red depending on whether they are above or below a specific value) minute changes, that could take days to find by looking at the numbers, can be seen instantly in the visualisation.

All this information, However, is very hard to publish. It is not possible to look `behind' an image and we cannot control the navigation of an MPEG movie. And in many circumstances all the value of a visualisation is lost when the project goes to print.

Visualisation Methods.

Current applications

Visualisation is, generally, performed through the use of a graphics work station via either custom written software or third party applications such as AVS, alias, wavefront, PRISMS, etc. These systems suffer from two problems, they are not particularly portable, and they're expensive. This creates a hurdle in any efforts to present the information to a wider audience, or, as is usually the case, an audience in another room. Thus compromises have to be made

Current output

The presentation of information that has been produced as a result of a visualisation is currently limited to the following forms:

None of these methods allow direct manipulation of the data, and are a poor substitute for the original visualisation system.

Collaborative Research.

The original ideal behind the URL was to allow researchers who were physically separated to retain access to their data. this allowed them to co-produce papers and generally collaborate on projects despite limited resources at some or all of the sites. The current methods of publication available today, while allowing basic information to be distributed, do not provide a medium to truly allow collaborative research involving scientific visualisation. Any one involved in such a project, if they are to remain functional, must have access to a graphics work station and the required (expensive) software.

Future output

Future publication methods may in some way hope to overcome these problems. The essential problem to note, however, is that the method of publication has to be one that in some way is able to reproduce the original visuaisation system as much as possible.

VRML, while not providing the original visualisation interface, at least reproduced the `final stage' in the process through a geometry viewer that will render databases that can be accessed via a URL and a common geometry viewer

Geometry Standards

A further limitation to collaborative research is the problem of the `standard'. As anyone who has attempted to distribute information will testify, the format in which it is published will go a long way to determining it's final audience. Text while having a common printed form, in its electronic form has a myriad of different formats, none of which are supported by more than a few applications. Images, while all looking the same, electronically can be produced in at least 42 different formats, and their view ability is thus dependant on the application being used.

The universal support of The World Wide Web and HTML has taken a major step towards the consolidation of these problems. Text can now be published in a standard format (HTML 2.0). It is now possible to obtain a web browser for any hardware platform, and GIF/JPEG have become defacto image standards for publishing.

The production of geometry has not reached this point yet, but if it is to be universally understood a geometry standard needs to be defined

VRML will hopefully be the solution to this problem. Currently a product's success can be greatly influenced by its support on the web, so if researchers are going to produce VRML environments then they are going to have to produce them in VRML. As the growth in this area increases, more geometry converters will be produced and visualisation systems will eventually directly support the VRML geometry standard.

While, currently, being an ascii based format, VRML has one strong advantage in that it is a hierarchical format. This can reduce both rendering and downloading times. Multiple occurrences of an object can be defined through a hierarchical reference to a single object and parennting of properties can firther rduce the overall size of the file needed to describe the database. It also allows the rendering of bounding boxes or object place holders on systems that are incapable of rendering the complete environment. It is this a compact, efficient and portable method of defining geometry.

Examples

weather

Visualisation of the Bureau of Meterology's McIDAS data initially using AVS. A topographic map of australia has been overlaied with false colour representing the temperature of the country.

Stonehenge

A model of the Stonehenge site, using a hierarchical VRML dataset (Courtsey Silicon Graphics Inc.)

chemistry

Visulaisation of Aspartic acid from the NYU MathMol library

cfd

Isosurfase of the x momentum of airflow around an object

architecture

Architectural modelling, using texture maps and VRML

Current limitations

Unfortunately, the currently limitations of both the WWW and VRML make it a limited system:

Network

Despite efforts to optimise a dataset it is often hard to keep the database at sizes of less that 0.5-1 Mb. The current network bandwidth makes the downloading of this amount of data without being directly connected to the source network unfeasible. Further increases in both network bandwith and optimisation of network routings are required before the regular trannsmissio of data this size can occur.

BIG datasets

Visualisation is usually required in situations where the datasets are of such a size that their information cannot be examined by traditional tabular or graphical techniques. This implies that the final geometry will involve a large number of points. Even after optimisation, most scientific datasets will be megabytes in size.

limited hardware

In order to examine a 3D model you need either a very fast CPU, or dedicated graphics hardware. Currently only Silicon graphics (SGI) workstations are capable of rendering shaded VRML datasets, making the distribution of information to people without access to such machines virtually impossible

advanced rendering

The addition to a dataset of such things as texturing or transparency, produces such a load on a CPU that without dedicated hardware, it is impossible to properly browse (ie in real time) such a dataset. Realistically, such a dataset can only be examined (at present) with a high end SGI workstation.

The future

database structures

The current trend in the production of VRML databases is to produce a single object that contains all the information. this is both inefficient and unwieldy , and usually leads to big datasets. VRML is designed to make use of its hierarchical nature, not only to allow for multiple copies of objects, but also to break the geometry into a number of smaller files that can be downloaded piece by piece.

In this way a large database can be loaded in parts like an HTML page and the viewer does not have to wait for the entire object to get to the local host.

Geometry should also be optimised with respect to points and polygons. Many datasets will have concurrent points that can increase the size of the file. buy first sorting and consolidating these points the overall size )and thus download time) of the geometry can be reduced

VRML tools

As yet there are no available tools for the creation of VRML databases, but the nature of the language, and the number of third party converters, makes this a minor hurdle. Like HTML, it is possible to create a simple database with a standard text editor.

wrl converters

Currently there are number of converters for the various geometry formats, they include, but are not limited to:

push/pull

With the continued support of server push/client pull and other mime enhancements for dymamic uptating of files will greatly enhance the power of VRML. It will eventually be possible to produce VRML documents that are not only dynamically updated, but via an HTML page it will be possible to reproduce a simplified version of the original visualisation system at the users end.
[Return to Table of Contents]
COPYRIGHT © 1995 by AUUG95 and APWWW95 Charles Sturt University. ALL RIGHTS RESERVED.