'anbul 2004
cal city of
past while
istitute the
lels used to
> could deal
it. of data it
itional data
he problem
curate and
of “Virtual
possible to
/e continue
nformation
1 we intend
s including
as well as
hrough the
Automatic
plications,
imation of
3 V/6, CD-
. Virtual
ising VR
" "Digital
ze in the
html
p
to
ml
he City of
allowed us
ect (2002-
Research”
ience and
ftware for
Trade and
marks or
PHOTO-REALISTIC SCENE GENERATION FOR PC-BASED REAL-TIME
OUTDOOR VIRTUAL REALITY APPLICATIONS
E. Yilmaz^ , H.H. Maras?, Y.C. Yardimci *
* GCM, General Command of Mapping, 06100 Cebeci, Ankara, Turkey - (eyilmaz, hmaras)@hgk.mil.tr
? Informatics Institute, Middle East Technical University, 06531 inónü Bulvari, Ankara, Turkey -
yardimy@ii.metu.edu.tr
KEY WORDS: Outdoor Virtual Environment, Synthetic Scene, Crowd Animation
ABSTRACT:
In this study we developed a 3D Virtual Reality application to render real-time photo-realistic outdoor scenes by using present-day
mid-class PCs. High quality textures are used for photo-realism. Crowds in the virtual environment are successfully animated.
Ability of handling thousands of simultaneously moving objects is an interesting feature since that amount of dynamic objects is not
common in similar applications. Realistic rendering of the Sun and the Moon, visualization of time of day and atmospheric effects
are other features of the study. Text to speech is used to inform user while experiencing the visual sensation of moving in the virtual
scene. Integration with GIS helped rapid and realistic creation of virtual environments. The overall rendering performance is deemed
satisfactory when we consider the achieved interactive frame rates.
1. INTRODUCTION
Virtual Environments (VE) where we can pay a visit are no
longer far away from us. Developments in rendering
capabilities of graphics hardware and decrease in the prices can
easily turn an average PC that we use in our daily life into a
cyberspace. PC-based VR applications are used in many fields
such as computer games, simulators, entertainment, education,
medical applications or real-estate presentations. Higher realism
level of the rendered scenes effects the users positively. Rough
terrain models, thousands of trees, moving human beings,
animals, vehicles, and buildings are some of the common
elements of typical outdoor scenes. The scene needs to be
refreshed as the user moves or any element of the virtual world
changes position. These renderings must be performed timely
so that the user is not annoyed. In a previous paper we
addressed scene generation for low altitude flights (Yilmaz et
al, 2004) Here we developed a cost-effective real-time VR
application that is capable of rendering realistic crowded
outdoor scenes from the eyes of first-person. Competition in the
video game industry lead to development of algorithms that can
be used to render faster and better VEs. Most of the algorithms
or rendering techniques that we used in this study originated
from entertainment industry.
1.1 Outdoor Virtual Reality
We focused on outdoor VR to be potentially used as a military
application such as visualization of battlefield in which we plan
to conduct further study. Outdoor VR has many challenging
issues that take place between real-time and realism boundaries.
A typical outdoor scene contains terrain, vegetation, culture,
sky, clouds, people, animals etc. The total polygonal cost of
these items is often beyond the rendering capabilities of many
graphics cards.
Corresponding author.
1.2 Type of Virtual Reality Systems
We can categorize VR systems into two main groups: non-
immersive and immersive. Non-immersive systems let the user
observe virtual world through conventional display devices.
Such systems are also called as desktop virtual reality.
Immersive systems totally replace real world scenes with virtual
ones. Head Mounted Display (HMD) is a typical immersive
system tool. In this study we tested desktop virtual reality
system but our study is ready to be use with immersive system
tools.
1.3 Real-Time Image Generation
One of the primary requirements of VR application is the ability
to update images at high speed. Ideally, this should be no
slower than conventional video frame refresh rates, which are
25 Hz for PAL and 30 Hz for NTSC. Human eye is able to
integrate a rapid succession of discrete images into a visual
continuum that takes effect at the Critical Fusion Frequency
(CFF), which can be as low as 20 Hz. Image size and brightness
are two important factors that determine CFF (Vince, 1995).
Commercial Image Generators used for flight simulators
provide 60 Hz for daylight and 30 Hz for night scenes. In our
application considering the capabilities of our target
configuration and the complexity of the scenes we accepted 25
Hz and upper frame rates as real-time. Frame rates over 10 are
considered as interactive. Frame rate is also strongly related
with the type of VR system. While lower frame rates annoy the
user in a non-immersive system it may cause motion sickness in
an immersive system.
1.4 Object Database
Real-time VR applications mostly provide an object database
that contains 3D models, 2D textures etc. to help creation of life
like scenes. This database mostly covers different versions of
the same model at different level of details (LOD). Low