Istanbul 2004
stimation and
itobahn exit
1 phenomena
multipath-
led layover
tion gradient
alls. Because
e the same
bund in front
; same range
jved from a
of different
sing e.g. full
combination
ear bright in
lignment of
esponses by
between the
unce signals
ne of bright
footprint.
tly occluded
SAR image,
ildings) may
interest, like
v areas s, On
| the viewing
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004
h Ss l. h s;
ground, ew
c
Figure 7. Shadow and Layover from buildings displaced in
range direction.
The viewing angle increases in range direction over the swath.
Assuming a range of the viewing angle 0 between 40° and 60°,
the shadow length of a certain building is more than doubled
from near to far range. In Fig. 7 such a situation is depicted
(shadow length s,, 5,). A worst case will arise if a road
between two building rows is oriented parallel to the sensor
trajectory. The street is partly occluded from shadow and partly
covered with layover. An object on the road can only be sensed
undistorted, if a condition for the road width wg holds:
W,>s, +1=h-(tan(0, ) + cot(0,)). (1)
In the sketch (Fig. 7) the angles Gs and 6sf vary remarkably
between the two buildings. In reality the angle changes only
slightly over the width of a road. Hence, for the estimate of the
size of the problem areas a constant angle (6s 6s/) is assumed.
An angle of 55? in both cases and a building height of 20m give
a minimum Ws of 40m.
44 Simulation of shadow and layover
It is essential to determine a-priori the optimal SAR acquisition
parameters in order to minimize the influence of layover and
shadow for a selected area of interest or an object class. For this
purpose, maps and high-resolution elevation data from a GIS
are required [Soergel et al., 2003b].
Based on a DEM, layover and shadow regions are simulated
[Meier et al, 1993]. From the sensor position, the DEM grid is
sampled in range direction. Layover and shadow regions are
detected analyzing the distance and the viewing angle. By
intersection of these results with the map data, the affected
areas of buildings and roads are identified
Figure 8. a) LIDAR DEM superimposed with the layers of
buildings (yellow) and roads (red). b) SAR image
.A simulation of shadow and layover areas with the given SAR
parameters is shown in Fig. 9a. The results are given in Table
1. The large viewing angle suggests a larger portion of shadow
compared to layover. This would be the case if all objects in the
scene were detached, their signals not interfering with each
other. However, in the test scene, the ground distance in range
direction between the objects is often small. This results in
many mixed pixels where shadow and layover are both present.
Less than 20% of the road area can be sensed undisturbed.
Ep m
e
Figure 9. Simulation results for given SAR parameters from Fig
8b. (layover: white, shadow: black,
layover+shadow: dark gray, reliable: bright gray). a)
illumination from north (top), b) illumination from
west (left).
4.5 Determination of optimal aspect and viewing angles
In order to estimate the optimal aspect and viewing angles for
an arbitrary SAR measurement a large number of simulations
have been carried out. Fig. 9b shows the result for a simulated
illumination of westward aspect direction. The aspect angle was
altered systematically in steps of 5°. For each of the 72 aspect
directions the layover and shadow areas were detected for 9
different viewing angles. The viewing angle 0 was chosen
between 30° and 70° with 5? increment. This results in 648
simulations. From this set, the best single SAR illumination
direction was determined. The results are shown in Table 1.
Parameters given optimal
Area | Complete | Roads | Roofs | Roads | Roofs
Shadow 28 39 16 27 14
Layover 25 19 34 16 28
Mixed 19 24 75 19
Reliable 28 20 43 39 52
Table 1. Results (in 96) of shadow/layover detection for given
(see Fig 8b) and optimal SAR parameters
The optimal illumination direction with respect of the visibility
of buildings was from the east with a viewing angle of 60°.
More than half of the roof area would be visible in SAR data
acquired with these parameters. In the case of the roads, the
best result is achievable for an illumination from exactly north
with a viewing angle of 45?. This direction coincides with one
main road orientation.