son
get
| of
age
ms
zed
es,
ole
deo
to
ing
ach
Its.
ing
pth
ery
the
pint
‘ain
1 à
ffer
As
ject
; of
Was
put
ess
thin
e in
ach
10St
eve
lel
jare
ffer
the
cial
2 of
ical
Py.
A
£e
Was
to
)ase
needed high resolution, and which parts
could get away with mathematical texture.
Techniques such as Delauney triangulation
were used to optimize the number of
polygons necessary to adequately represent a
given area's terrain. Areas that had a
complex terrain required more and smaller
polygons (generally triangles or rectangles)
while areas that were flatter could be
adequately represented by fewer, larger
polygons. When using mathematical texture,
this technique could result in a reasonable
representation of the terrain, and more of the
compute time could be spent on rendering
high resolution versions of the natural and
manmade objects.
Normally, a real time system can process a
given number of polygons in 1/30th of a
second. If more detail is given to the
terrain, then less detail will be available to
the object. The size of the individual polygon
may not be a major time factor. This
tradeoff between realism and speed has
always existed in simulation.
In the late 1980's photo texture became the
trend in simulation systems. Instead of
generic or mathematical texture functions to
determine the color of a specific polygon, an
image of a real world object or terrain was
used. | When the terrain polygon was
projected to the screen, an interpolation or
mathematical function was not used to
determine how to fill the projected polygon.
Instead, the real world image was
transformed through the same perspective
transformation and pixel by pixel used to
populate the polygon. Photo texture is
especially effective when it is necessary to
portray a very complex setting such as an
urban landscape. Buildings within this urban
environment might take very many polygons
to describe in a way that allowed the viewer
to see a realistic view. Instead, a simple
rectangular polygon could be projected and a
digitized photograph of a real building could
be mapped onto that projected polygon. A
simulation system that takes advantage of
photo texture might spend less time on the
polygon projection and concentrate on the
117
relatively straightforward image
processing functions to remap the image
to the polygon.
The above techniques use a data base to
screen type of rendering in which within
a limited view area, all polygons are
rendered using the depth buffer to mediate
visibility. It can be seen that this
technique is very time dependent on the
size of the terrain and object database in
terms of the number of polygons to be
projected and filled. For databases using
real imagery as the terrain texture
information, it is not uncommon for over
100,000,000 polygons to be in a spatial
database. For example, a merge of SPOT
satellite information (60 km x 60 km)
with a 10 meter pixel size with a portion
of a Landsat Thematic Mapper scene (185
km x 185 km) with a 30 meter pixel size
would easily give over 70,000,000
triangles to be rendered if the whole data
base could be seen. Several other
methods are often used in the generation
of perspective images, ray tracing and
inverse ray tracing.
Ray tracing is a very straightforward
procedure which often takes significant
computer resources, but that generally
results in high quality rendered images.
Ray tracing assumes that there is at least
one light source radiating light rays onto
the spatial database including terrain and
objects. Parallel rays are cast from the
source toward the terrain. As each ray
intersects the terrain it is either absorbed,
reflected, or transmitted through the
terrain material. If it is reflected, it may
intersect another part of the data base or
it may be reflected away from the data
base. If it intersects another part of the
data base, another calculation of
absorption, reflection, or transmission
must be performed. Ray tracing usually
limits the number of multiple bounces that
are performed.
Those rays which are reflected toward the
viewer's eye represent the image that is