I suspect there are valuable lessons to be learnt from the computer
scientists on these techniques, who use them for rendering 3D environments.
I know Sanjay has already collaborated with Yiorgos Crysanthou of the
University of Cyprus (previously UCL), but there is also work by Mel
Slater at UCL, who was working on 'virtual light' -- essentially
precalculating what you can see from everywhere within a 3D environment.
There are two real problems with 3D isovists / visibility graphs, one
is speed of computation (e.g., ray tracing is very slow), and the other
is the size of the representation. Computer scientists use B-trees to
speed up the process of discovery, but as far as I know the compact
representation does not exist yet (at one stage a few years ago I
believe Mel was using an array of computers with 8GB of memory).
On quick and dirty methods to retrieve the information for non-computer
scientists, when we first started depth mapping, David Chapman, Alan
Penn and I started by using the depth buffer from a 3D rendered scene
using OpenGL (this is where "Depthmap" gets its name from). At this
point we threw away the 3D information and just took a 2D isovist slice!
However, I'm sure this would probably be the simplest way to start out
making 3D isovist tools for generic 3D models -- perhaps not for GIS,
but I am sure the two should converge at some point.
The reasons we didn't keep the 3D information were complexity of
representation, and the belief that the use of 3D was limited if we were
dealing with the interaction of people with space, who are essentially
constrained to a 2D plane for movement. Alan has mentioned this before,
but there is a paper by Marcos Llobera, called "Extending GIS-based
visual analysis: the concept of visualscapes" (IJGIS, 2003) which
contains a good discussion of the issues.
Alasdair
Sanjay Rana wrote:
> Prof. Kyushik Oh (http://www.casa.ucl.ac.uk/news/pages/Kyushik.html) did
> show me an AUTOCAD solution last year which did visibility analysis for
> Planning Process, but it didn't make a visibility graph but more sort of
> rendering 3D vistas.
>
> I think I have found a link to the lab of Prof. Lee but its all in Korean..
>
> http://lacomi.uos.ac.kr/gislab/index.htm
>
> We need a translation here :)
>
> Sanjay.
>
>
--
Alasdair Turner
Lecturer in Architectural Computing
Bartlett School of Graduate Studies
UCL Gower Street London WC1E 6BT
|