The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B7. Beijing 2008
1073
However, other features have the potential to provide additional
information about the imaged scene. The three images in Figure
2 show three different bridges in Southern Germany. A
relatively short bridge crossing a freeway is displayed in Figure
2 (top left). Most of the radar signal on top of the bridge is
reflected away due to the smooth surface of the tarred road.
Therefore, the road appears dark in the image and almost no
insight can be deduced from direct backscatter. However,
although the bridge surface itself cannot be seen nicely, the
width of the bridge may be derived from two parallel bright
lines. They are due to double-bounce effects from the guardrails
on top of the highway bridge. Such guardrails are typical
features of highway and freeway bridges. They also appear in
the top right image and the bottom image in Figure 2. The top
right image shows a freeway bridge crossing a deep valley close
to the city of Schwäbisch Hall. In addition to two guardrails on
each side of the bridge, a third guardrail can be seen in the
middle of the bridge, separating the two opposed driving
directions. A typical SAR effect, useful for modelling bridges
over land, is occlusion. A shadow of the bridge body and the
pillars is observable in the top right image of Figure 2. Its shape
depends on the bridge body, the height of the bridge above the
terrain and the undulation of the terrain. The greater the
distance between the bridge body and the bottom of the valley
becomes, the further away is the shadow from the bridge
(assuming a constant height in the occluded area). This shadow
distinguishes objects elevated above the ground from non
elevated objects. It describes the bridge as a three-dimensional
object, discriminating it from roads.
3. BRIDGE HEIGHT ESTIMATION FROM OPTICAL
AND SARIMAGERY
3.1 Displacement of elevated objects in optical imagery
One database containing optical imagery covering the entire
Earth with reasonably high resolution is GoogleEarth. Ortho-
rectified optical images from both space borne and airborne
sensors are available free of charge. In case rapid change
detection is needed during the occurrence of a natural hazard,
imagery from GoogleEarth can be used as a reference. Since the
optical images are ortho-rectified, distortions due to terrain
undulation are significantly decreased. However, elevated man
made objects stay distorted because they are not included
within the digital elevation models (DEM) that are used for the
rectification process. Such DEMs only contain points on the
bare ground. This effect is shown in Figure 5 while an overview
of the entire bridge is given in Figure 4. Figure 5 shows a
section of a long railroad bridge, crossing a valley. The terrain
height of the valley bottom is varying as it can be seen in Figure
4 (left) and from the distance between the shadow and the
bridge in Figure 5. The further away the shadow is situated
from the bridge, the deeper is the valley.
The three-dimensional modelling of a scene using one optical
image and one SAR image is possible due to the different
perspectives of the sensors (Figure 3). Optical sensors usually
acquire images with off-nadir angles smaller than 30°. SAR
sensors are side looking and e.g., military airborne sensors
acquire imagery with off-nadir angles up to 70°. Hence, two
different perspectives of the same object on the ground allow
for the estimation of height values. Additionally, SAR sensors
measure distances whereas optical sensors measure angles.
Hence, the points A, B, and C in Figure 3 are projected to Aq,
Ground
A B C
Imaged Ground Area
Figure 3: Comparison of optical and SAR sensor geometrie
B 0 , and C 0 by the optical sensor while they are imaged to A R ,
B R , and C R by the SAR sensor.
Figure 4: Images of a railroad bridge made of concrete-steel
near the city of Zellingen in southern Germany, (left) oblique
photo taken out of the aircraft with a consumer grade camera,
(right) optical image from GoogleEarth.
Figure 5: Blow-up of a GoogleEarth from Figure 4; compared
to the two parallel red lines, the distortion of the bridge in the
ortho-rectified image can be observed.
The distortion of the railroad bridge in Figure 5 becomes
obvious by comparing the alignment of the two parallel red
lines with the bridge body. The greatest distortion occurs at the
highest elevation of the bridge above the valley bottom. Since
the image was taken from the south-west, maximum distortion
occurs towards the north-east.