oth,
ries a high
ScanSAR-
; required.
ntific and
capability
geometric
ed. A new
ly coarser
ng will be
ovided by
nt will be
ements the
gment for
he X-band
ent is the
ures of the
respect to
techniques
ATM and
al, 2003),
o the high
x rate of
ice will be
e-Doppler
geocoding
ortions in
and _right-
X terrain
onificantly
les DEMs
ing terrain
Corrected
jecause it
elto3m
correction
‘e.g. from
idjustment
s the SAR
uct is seen
ted (GTC)
Geocoded
1) can be
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004
2. SAR GEOCODING
2.1 Geometric Distortions in SAR Images
TerraSAR-X is a side-looking synthetic aperture radar (SAR). It
transmits pulses in X-Band to the Earth, which will be reflected
back to the instrument and received again by the radar. The
signal travel time is measured and therewith the range distance
between the antenna and the ground target. Due to the side
looking geometry of SAR-systems undulated terrain is
significantly distorted during the SAR mapping process.
The most important and well known local image distortions are
(Schreier, 1993) foreshortening, layover, and shadow. The area
on ground is not seen by the radar. But also the range
displacement effect needs to be considered that causes elevated
features to be mapped in false range positions — namely to
closely to near range. These effects as well as the varying
ground resolution caused by varying slopes can be corrected
using a digital elevation model.
Three approaches can be applied to geocode SAR images and
will be discussed in the following chapters. Common to all
cases is the backward geocoding also denoted as object-to-
image approach.
2.2 Rigorous Range Doppler Approach
For each output pixel, which defines a co-ordinate triple
(easting, northing, height) in the output map projection, the
corresponding azimuth and range positions in the input image
have to be determined. This is based on the Range-Doppler (1)
and range equations (2) applicable to SAR images. Due to the
dynamic imaging principle of SAR this is an iterative and hence
time consuming search procedure. The orbit position is varied
until the range and Doppler equations are simultaneously
fulfilled (Meier et al, 1993).
Lo.
15-5
Q) FG jyrn-cm, :d-|p- 5]
ay FC = fpr
(5i) are the pixel co-ordinates where i are the azimuth and j the
range positions. fpc is the Doppler reference function applied
during the SAR processing. p and s are the earth surface point
and sensor position vectors, A is the SAR sensor wave length
and ry and m, the slant range offset and the pixel spacing.
2.3 3D Interpolative Approach
The principle of this approach is to perform the rigorous
transformation for grid points and using an interpolation to fill
the grid cells (Raggam, 1988). The radar image range and time
co-ordinates are determined by interpolating between anchor
points. At first a three-dimensional grid of points (co-ordinates
in easting, northing, height) is generated and the corresponding
pixel co-ordinates (in azimuth and range) of the input image are
determined using the rigorous Range-Doppler approach. The
grid covers the output area and its height extension spans the
entire elevation range of the underlying DEM. Starting from the
azimuth and range co-ordinates at a reference elevation
correction terms in azimuth and range are interpolated using the
individual height values from the DEM. In order to correct non-
linear terrain effects a quadratic term for height interpolation is
considered.
The main purpose of the interpolative approach is to reduce the
computing time for generating a terrain corrected geocoded
product. Tests showed that the throughput can be improved by
at least a factor of 5 compared to the rigorous geocoding.
Geometric degradation strongly depends on the grid size.
(Raggam & Gutjahr, 2003) showed that the interpolative
approach is precise from the geometric point of view if the
mesh sizes is less or equivalent to 1000 m on ground. In
comparison errors caused by accuracy deficiencies of e.g. the
DEM and the sensor model parameters cause significantly
larger location errors.
2.4 Interpolative Ellipsoid Correction
The ellipsoid correction is a special case of the 3d interpolation
approach. A net of points is transformed applying the rigorous
range-Doppler-approach. As only constant elevation values
need to be considered the grid cells are filled using a bilinear
interpolation (Roth et al, 1993).
2.5 Implementation Issues
Rigorous, 3d- and interpolative ellipsoid correction are
parameteric geocoding approaches. They are independent from
the radar wavelength and can be applied to other space- or
airborne SAR data as well. The current version of the
geocoding system supports input from Envisat-ASAR, ERS, J-
ERS, Radarsat-1, SIR-C / X-SAR and DLR's airborne system
ESAR. The pixel spacing of the in- and output data as well as
the Doppler reference function are parameterised and are stored
in configuration files. Even though most SAR processors refer
to zero-Doppler the geocoding system is able to consider other
reference functions. Multi-polarised data are considered as
multi-layer images.
3. TERRASAR-X IMAGING MODES AND
POLARIZATION
This chapter summarises the main features of the TerraSAR-X
imaging modes. (Roth, 2003) provides more details.
3.1 SpotLight (SL) and High Resolution SpotLight Mode
(HS)
TerraSAR-X achieves the highest geometrical resolution in the
SpotLight modes. During the observation of a particular ground
scene the radar beam is steered like a spotlight so that the area
of interest is illuminated longer and hence the synthetic aperture
becomes larger. The maximum azimuth steering angle range is
+0.75°. The size of the observed area on ground is smaller than
the one in all other modes. HS and SL modes are very similar.
In SL mode the geometric azimuth resolution is reduced in
order to increase the azimuth scene coverage. Characteristic
parameters of the SpotLight and High Resolution SpotLight
modes are listed in Table 1.
Value SL
10 km x 10 km
Value HS
5kmx 10 km
Parameter
Scene extension (azimuth x
ground range)
Incidence angle range (full
- 20° - 55° 20° - 55°
performance)
Azimuth resolution 1m 2m
3r ange resolution
Ground range 1.5m-3.5m 1.5m-3.5m
(55?-20? incidence angle)
841
Table 1: Parameters of SpotLight and
High Resolution SpotLight Modes
