NetNews Usenet Archive 1993 #3/NN_1993_3.iso/spool/sci/virtual/4632

Internet Message Format | 1993-01-28 | 6.8 KB

Path: sparky!uunet!ogicse!news.u.washington.edu!stein.u.washington.edu!hlab
From: [e-mail] (David Caballero)
Newsgroups: sci.virtual-worlds
Subject: DESIGN: 3D Forests
Message-ID: <[e-mail]>
Date: 27 Jan 93 15:26:25 GMT
Article-I.D.: shelley.1k7bukINNod6
Organization: University of Washington
Lines: 112
Approved: [e-mail]
NNTP-Posting-Host: stein.u.washington.edu
Originator: [e-mail]



This article is to reply those asking me for a bit more of information
about 3D modelling and rendering of forests, i.e John Curtis, Chris
Hand, Diego Montefus co, Paul Simpson, Thomas Eskridge, Michael
Almquist and Thad E Starner:

    I have been involved as analyst and programmer in a research team
of Forest Fire Simulation from 1989. In such project I developed a
first approximation of what I thougt a burned forest should seem like.
It was entirely programmed on a PC with starndard VGA graphics and,
may I have to admit it, it was a severe limitation to my expectations.
The first module of the project was then presented as a paper in the
IXX World IUFRO Congress, at Montreal, Canada, in August o f 1990.
Then I showed it to Peter Kourtz, one of the greats on Forest Fire
Research (he is actually involved in some VR like projects with fire
fightning helicopter simulators and computer programming using voice
recognition) at Petawawa Forest Institute. He, then, said: "Oh, very
nice graphics". Yes, it was, but I was seeking a little bit further.

    At the end of that very year I went to the II World Congress on
Forest Fire Res earch, an then again, I had the opportunity to show my
work to Richard R. Rothe rmel, Forest Fire Research Leader at
Intermountain Research Laboratory (USDA) a t Montana, and Francis
Fujioka, at Riverside (CA, USDA). This last one told me: "Well, your
work sounds interesting but, what do you consider is the next step?
". And the next step, I told him, was the high end graphics and the
human inter facing with computers.

    All this facts lead us to the present. Nowadays I am working with
other har dware and software, but my goals are the same. I intend to
a) Represent high en d computer generated pictures of landscapes to
see what can be in the future wi th such or such activity taking place
on natural environments. The area of acti vities I am refering to is
civil engineering and, more close, forest and natura l environmental
engineering. b) To involve VR techniques for the forests and ge neral
natural environmental management, by means of developing new software
or adapting the existing one to VR environments. In this way I am
thinking about 3 D GIS whit data input and management via standard VR
devices (such as dataglove s etc.). (I have a lot to speak about this,
because I have several ideas that I want to be real). c) To develop a
VR Operating System for Forest Fire Crew tra ining. Supose all those
guys being involved in hazarous situations when a fire occurs and are
rarely expected. It will take artificial intelligence inputs of those
with experience and, virtually, give the other some hints as the
virtual forest fire simulation take place. d) One of my dreams is to
submit a work to I MAGINA and SIGGRAPH showing how can I represent
natural environments, focusing mostly on tree and forests modelling
and rendering. At this point I have got an idea called CYBERJACK, a
project I want to develop along 1993 year to be prese nted at 1994.
(Is a story about CYBERnetic timberJACKs, really amazing).

    Getting to the point I am nowadays working over a Silicon Graphics
IRIS CRIMSON XS24 (Is hard to find, buy or share such computers in
Spain) with TDI s oftware, i.e. TDI Professional Animator Explore. The
package counts with a modu le of 3D tree modelator, namely TDIAMAP,
developed by CIRAD at Montpellier (Fra nce). As forest engineer I have
to say that the species are very realistic ones but it has a drawback:
the huge facets databases it generates. Trying to modela te a whole
forest, with thousand of large, medium and small trees as well as ot
her plant levels such as shrubs and grass, lead us to millions of
facets to be computed for ONE final rendered frame.

    The firsts attempts to render it where unfructuous because only 16
trees co llapsed the swap disk space. So I decided to render the image
in several stages and then compose all of them with a Z buffering.
Because of the type of Z-compo se involved the same space disk and
time rendering, so nothing solved. But, and here is the idea, why to
Z-compose?.

    So I went to the root of the problem. Why takes so many disk space
and time the rendering of a forest?. Is it because of the complexities
of the tree model used? The answer was NOT. The rendering routines of
TDI Explore analizes each pixel of the screen tracing a rendering line
crossing trough ALL of the facets that are behind that very pixel.
This is the key. To ratify this hyphothesis I modelled 10 thousand of
non-complex objects, i.e. a single hexaedron, and place d it one in
front the other in a single row. Then I placed the virtual camera i n
front of the first one so only one object could be visualized. But
there were 10,000 x 2 = 20,000 facets in " front" of each pixel being
analized by the rend er routine. And Tah taaah Y "process stopped due
to no more swap space". The sa me limitation I had with a very complex
object.

     The solution then was obvius. I have to render the forest splited
in scrip ts with trees placed roughly in the same plane but with FEW
FACETS OVERLAPPING each other. Then a single compose OVER did the
work. I called it Multiple PLane s Rendering Compose. Some limitations
to the method are: 1) The camera is neede d to move rougly parallel to
the normal of the analisys planes. The rendering o f the quck turns
are rather complicated to set. 2) The method implies complex s cript
editing, precise object location and a very carefull order in the
final c omposition. 3) It involves a lot of scripts, one per analisys
plane. 3) How to render the shadow casting over the floor and over the
other trees (I think I've got the solution projecting the shadow along
a vector parallel to the analisys planes). 4) It is possible to render
turns of the camera by succesive composing of renders, not by planes,
but by concentric cilindric surfaces from the camer a location.

    At the moment I am programming in C-Shell all the process
involved. I will send you a copy of the code when I finish it.

    Please, send me any comments about this article that, be care of
it, is the only one being published at the date.

    Thanks all of you sharing knowledges. VR is rougly known here in
Spain and I can't say if there is projects on it. But it is
overwhelming how it is being accepted in the rest of the world. Don't
wait for the future, the future is her e right now.

BEST REGARDS
GNOMO