■
The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B7. Beijing 2008
bridges over water. The same two bridges are depicted again in
InSAR allows for obtaining directly a three-dimensional model
of the scene. However, calm water surfaces act like a mirror and
hence the signal to noise ratio (SNR) is poor. Thus, water
surfaces show poor coherence values leading to artefacts in the
final Digital Surface Model (DSM). One possibility to more
precisely estimate the water level is to determine the distance
between the bottom of the bridge and the water surface.
Knowing the absolute height of the bridge body, the height of
the water surface can then be derived.
an optical aerial image in Figure 1 (centre image). Usually
is due to double-bounce reflection between the bridge and the
water surface or vice versa. Triple reflection leads to the third
stripe where the radar signal first hits the water surface, then
lower parts of the bridge body, and finally the water surface
again. The height h of the bridge over the water surface can be
estimated by measuring the slant range distance As either
between the first and the second or the second and the third line.
Knowing the off-nadir angle 0 of the SAR sensor, h is estimated:
h = As/ cos (0)
(1)
However, the exact horizontal position and outlines of the
bridges cannot be easily derived due to the multiple bounce
effects. This is done by fusing the three-dimensional InSAR
scene with an aerial photo taken in nadir view. The double
bounce stripe in the InSAR data is replaced with the optical
data and the water surface is replaced with the newly computed
value (Soergel et al.,2007).
2.2 Bridges over land
Bridges over land appear differently in SAR imagery compared
to bridges over water (Wegner & Soergel,2008). Multiple
parallel lines at a bridge, typical for bridges over water, do not
occur in SAR images of bridges over land (compare Figure 1
(top) and Figure 2). This is due to the different reflectivity
properties of water and land. Calm water strongly reflects the
radar signal whereas inhomogeneous terrain and vegetation
leads to almost diffuse reflection on the ground under the bridge
body. In case of bridges over land, multiple bounce effects only
occur at dihedral reflectors on the bridge body and where bridge
pillars meet the ground. Hence, the bridge height cannot be
determined directly from the distance of parallel lines and the
off-nadir angle.
Figure 1: Test images of bridges near the city of Dorsten in
southern Germany, (top) InSAR magnitude image pair taken in
X-Band (illumination direction from right to left), (centre)
optical aerial image, (bottom) 3D visualization of optical and
InSAR data after fusion.
This technique becomes possible because the very calm surface
of the canal leads to strong signal due to multiple reflections. In
particular, double and triple-bounce effects are typical for
bridges over water, illuminated perpendicularly by the sensor.
Three lines can be observed for each bridge in SAR data of
coarser resolution. In high resolution imagery such lines widen
to stripes. The first stripe in illumination direction represents
the height of the bridge body because most of the radar signal is
backscattered directly from the bridge body. The second stripe
Figure 2: SAR test images of highway bridges acquired by the
MEMPHIS sensor in ka-Band, (top left) SAR image of a bridge
near Manching (illumination direction from left to right), (top
right) SAR image of a bridge near Schwäbisch Hall
(illumination direction from left to right), (bottom) SAR image
of a bridge near Wolnzach (illumination direction from top to
bottom).
1072