generated for a specific user view pyramid.
The ray tracing technique must be able to
decide which terrain or object polygons are
intersected by initial rays or reflected rays.
The search through all possible polygons for
each ray is very time consuming. Many
optimizations are used to reduce the number
of intersections that must be calculated.
A ray traced image may show very realistic
effects such as reflections off of shiny
surfaces, and generally is the type of image
that is used in movie sequences.
Inverse ray tracing is also popular in
rendering. In this case instead of tracing a
ray from the source until it intersects the
terrain and finds its way into the view
pyramid, the ray is cast from the viewer's
eye, through image pixels in the imaging
plane, and onto the terrain. This technique
must also look for the intersection of the ray
with the terrain polygons, but much fewer
rays have to be cast. The terrain reflectance
properties are assumed to be lambertian, in
which for a given source ray, reflection
occurs in all directions. Multi-bounce
scattering is not easily accommodated within
inverse ray tracing.
IV. GTRI RENDERING APPROACH
GTRI has developed a generic C rendering
approach using the methodology of data base
to screen rendering as described above. This
technique implements photo texture mapping
for both terrain and objects, can handle
arbitrarily large spatial databases,
incorporates antialiasing techniques for
motion rendering (simulation), and runs in
near real time (approximately 10 seconds) on
a Sun SPARC 1 class workstation. Dynamic
indexing into multiple levels of database
resolution creates appropriate background
fuzziness and antialiasing while reducing the
total number of polygons to process. Image
data or GIS feature layers may be draped
over the terrain, giving a user the capability
to dynamically see raw or analyzed
information in a natural perspective. Using
118
a graphical user interface (GUI) a user
may navigate throughout the database
interactively, evaluating dynamically the
visibility of certain features from his
autonomous path. The perspective images
of the spatial data set provide the ability
for the user of the system to totally
immerse himself within the spatial data
base with ultimate free movement. The
scenes that are created as the user moves
are high quality, detailed, views of the
spatial data set. The user is truly
immersed into a virtual spatial data set
using a Virtual GIS (VGIS).
As the speed of rendering and display
increases with the Joy’s Law increase in
computing power , and as more of the
process is implemented in parallel, full
real time immersion is possible. While
virtual immersion within the spatial data
set is practical in the near term (1 to 2
years), the speeds for typical GIS spatial
analysis functions should also be
increasing. Thus, a user might place
himself on a hillside and watch a
rainstorm drop water into a watershed (a
graphical illustration of a meteorology
model), watch as the water finds its way
into streams (as modeled by watershed
runoff models), and see the river rise and
create flood hazards. This is the very
beginning for the merger of modeling,
GIS, and visualization.
V. REALTIME GIS INTERFACE
In addition to the ability to immerse
oneself within spatial data, a user must
also be able to perform the natural GIS
query functions and dynamically see the
results of GIS analysis. The query
function is implemented with the dynamic
movement control structure of the view
model. All interaction occurs with the
rendered image. The user can place
himself high above the database and
essentially see a two dimensional view of
the spatial data, or he can zoom into the
data to view it from a specified view
point
point
retrie
coord
and t
GIS
Wher
avail:
origit
layer:
such
terrai
of the
the lz
imag:
with ‘
infori
persp
corre
river,
pick «
VI. T
C
Geor;
Comr
game
of Ge
Villas
plann
const
set ba
foot
maps.
parki
These
attrib
On the
were
build
of the
the b
road :
inforr
was |
of ev
buildi
were
could
obscu
