International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004
4.1 Related work
Approaches for monitoring traffic have been proposed for both
spaceborne systems, e.g. RADARSAT-2 [Livingstone et al.,
2002], Shuttle Radar Topography Mission SRTM [Breit et al.,
2003] and airborne systems, e.g. [Ender, 1999], [Eckart et al.,
2000]. The moving target indication (MTI) requires vehicle
detection in the SAR data and parameter estimation. Target
detection and estimation can be performed either incoherently
with a single SAR sensor, e.g. [Kirscht, 1998], [Livingstone et
al, 2002], or coherently (along-track Interferometry), with
much higher fidelity, with two [Gierull & Sikaneta, 2003] or
more apertures [Ender, 1998].
For processing a SAR image from a single aperture all scene
objects will be assumed stationary. The relative motion between
sensor and scene causes a Doppler frequency shift, which is
exploited to achieve a high azimuth resolution and to determine
the correct azimuthal position of objects. The movement of
objects causes artefacts. The radial velocity component of the
movement leads to an azimuthal displacement (because of the
additional Doppler frequency shift), the parallel component
leads to blurring. This first effect can be observed in Fig. 5
comparing a SAR image and an aerial image acquired
simultaneously. A cargo ship entering a lock (motion from left
to right = in range direction) is mapped on the ground in the
SAR image.
Figure 5. Azimuth displacement of a cargo ship entering a lock.
a) SAR image, b) aerial image
4.2 Detection of single vehicles
Some image based approaches exploit this azimuth
displacement of detected vehicles from the expected position to
estimate the radial velocity component relative to the carrier.
Using the information about geometry of the route they try to
derive the velocity of a vehicle. One-channel implementations
may register only fast transverse or parallel movements.
Other approaches exploit SAR data from two or more channels
resulting from several antennas or subapertures oriented in the
direction of flight which observe the same scene at different
times. A coherent processing of the data allows calculation of
along-track interferograms showing moving objects by phase
differences. Introducing knowledge, (e.g. typical size or shape)
moving objects can be detected in the interferogram by model-
based image analysis [Schulz et al., 2003].
SAR-MTI (moving target indication) processing has shown
promising results in rural and suburban area. Figure 6 shows an
example of a SAR-MTI result which depicts an Autobahn exit
and automatically determined positions of vehicles after
complete MTI processing including clutter suppression, moving
target detection, high-precision azimuth position estimation and
target tracking [Ender, 1999]
Figure 6. SAR-MTI result after target tracking (Autobahn exit
Ingolstadt/Nord) [Ender, 1999]
4.3 Exploiting context
However, in urban areas SAR specific illumination phenomena
like foreshortening, layover, shadow, and multipath-
propagation burden the interpretation. The so-called layover
phenomenon occurs at locations with a steep elevation gradient
facing towards the sensor, like vertical building walls. Because
object areas located at different positions have the same
distance to the sensor, like roofs, walls and the ground in front
of buildings, the backscatter is integrated into the same range
cell. In general, the signal mixture cannot be resolved from a
single SAR acquisition. The signal contributions of different
objects inside a resolution cell can be separated using e.g. full
polarimetric or interferometric SAR data or even a combination
of both [Guillaso et al. 2003]. Layover areas appear bright in
the SAR image (see Fig. 8b). Perpendicular alignment of
buildings to the sensor leads to strong signal responses by
double-bounce at the dihedral corner reflector between the
ground and the building wall. All these double-bounce signals
have the same time-of-flight. This results in a line of bright
scattering in the azimuthal direction at the building footprint.
At the opposite building side, the ground is partly occluded
from the building. This region appears dark in the SAR image,
because no signal returns into the related range bins.
Layover and radar shadow (e.g. caused by tall buildings) may
hinder the visibility of neighboured objects of interest, like
roads. The sizes of the layover areas /, and shadow areas sg On
the ground in range direction (see Fig. 7) depend on the viewing
angle 0 and the building height A.
Interi
————
gro
Figur
The
Assu
the s
from
(shac
betw
traje
covei
undi:
In th
betw
sligh
size |
Ana
amit
4.4
it is |
paraı
shad
purp
are ri
Base
[Mei
samy
detec
inter:
areas
Figu