ul 2004
ck
s during
the fact
to shad-
; due to
; a com-
id GPS:
her with
d route.
ime due
t can be
ombina-
luced in
'rzinger,
s differ-
n all di-
ons (see
jns with
be used
tterns in
, 1999),
in situ.
vertical longitudinal
SQ
O-
7
3 t
T : :
1
cA
I: T 1 r + T + r + Y r Y f 7 -— T T i
10. 14 18 2 26 30 34 10: 14. 185. 22 26 30 34
Figure 3: Accelerations during human walking
Different types of hard or soft underground cause differ-
ent step pattern. This patterns change completely with a
slope > 10°. Taking the wrong time intervals will re-
sult in wrong number of steps. This can cause travelled
distance errors of 10m or more meters in dead reckon-
ing mode (Ladetto et al., 2000). (Ladetto and Merminod,
2002) present a combination of GPS, INS, electronic com-
pass and barometer in a small wearable system that is used
in a pedestrian navigation system for the blind. The system
is capable of detecting several walking patterns including
up and down stairs movement. It can track the person’s
path even indoors with several navigation algorithms that
compare the information of the output of all sensors and
using different strategies in case of sensor outages or con-
flicting information.
A wealth of research deals with fusion of several sensors in
order to overcome the weaknesses. Vision based methods
e.g. are used in combination with INS to improve head mo-
tion tracking accuracy (Ribo et al., 2002, You et al., 1999)
with the computer vision algorithms providing information
of low frequency movements and INS for fast movements.
3.1 Occlusion
An unprocessed overlay of virtual objects in a video stream
or on a head mounted display system will not result in a re-
alistic impression of the fusion of real and virtual scene.
Without any further processing virtual objects behind a
real object like a building will be displayed in front of that
building instead of being occluded by it.
In order to solve the occlusion problem in computer graph-
ics we can use depth information (z-buffer) of objects to
be displayed. A matrix is used to store distances from the
projection center to the object models for each pixel. The
object with a smaller distance to the projection center oc-
cludes the one with a greater distance.
In augmented reality the depth information of the augmented,
virtual objects can be calculated because the geometry about
these objects is accessible. But for the real objects on the
video stream there is in principle no information about the
geometry. Additional data acquisition is necessary in order
to generate the depth information for real objects. Other
1051
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004
information of depth |
of virtual scene |
information of depth
of real scene
virtual water surface
>
comparison
: AS ;
real > virtual ^ ^. virtual > real
EE -
color of
virtual objects video image
t
re combination ^^
Figure 4: Schema of occlusion processing
possibilities are the use of image processing (Simon and
Berger, 1999) or probabilistic models of form and posi-
tion (Fuhrmann et al., 1999), The virtual objects that are
used to provide the depth information about the real objects
are not displayed. They are called "phantoms" (Grohs and
Maestri, 2002, Fuhrmann et al., 1999). The used occlu-
sion solution in this work is shown in figure 4. Here the
depth information (phantoms) is extracted from a digital
elevation model derived from laser scanning or, if avail-
able, from building models. These models are used to con-
tinuously calculate the depth information according to the
camera movements. The phantoms used for the depth in-
formation are not visible, they are replaced by pixels from
the video image. The depth information for the real and
the virtual scene are compared, pixel by pixel. A real ob-
ject that is more distant from the projection center than a
virtual object at a distinct pixel will occlude the real one,
and vice versa.
4 ARFOR DISASTER RELIEF: HIGH WATER
4.1 Motivation
In 2002, 30% of all damaging events, 42% fatalities, 50%
economic and 37% insured losses are due to high water
worldwide (Rückversicherungs-Gesellschaft, 2002). Apart
from storms, high water causes the most damage in com-
parison to other natural disasters. Disaster management
