197
Parameter
Approximate Value
Estimated Value
Standard Deviation
Ro [m]
825185.4208
825185.4208
10' 8 (fixed)
AR [m]
7.89809
7.89769 m
0.00083
To [s]
97.7800
97.7800
10' 8 (fixed)
AT [s]
0.0023828
0.0023828
10' 8 (fixed)
foo [Hz]
398.573
398.573
10' 8 (fixed)
foi [Hz/col]
0.0
- 0.0057
0.00102
ÎD2 [Hz/lin]
0.0
- 0.0087
0.00759
f D3 [Hz/ col lin]
0.0
0.0000028
0.0000046
do [m]
33.7
33.8218
0.00434
di [m/col]
0.0
-0.0001480
0.000009
d2 [m/lin]
0.0
0.0000289
0.000015
d3 [m/coMin]
0.0
0.0000000093
0.0000000031
d 4 [m/col 2 ]
0.0
- 0.0000000273
0.0000000092
ds [m/lin 2 ]
0.0
0.00000004445
0.0000000035
Table 1: Estimated parameters and their standard deviations (first set of parameters).
(ascending and descending ERS-1 SAR images, SPOT
images, orthophotos, reference DTM, land-use map, etc.),
has been made available to ORFEAS participants. The
results analysed in the following were obtained through
the ORFEAS data set.
Two ascending ERS-1 images of the ORFEAS data set
were chosen for the processing. They were acquired at
September 12 and 15, 1991. The baseline length is 161.5
m, which is about optimal for DEM generation. From the
original images two sub-images of 1500 pixels in range by
5000 pixels in azimuth were extracted and processed with
the ISAR software. The mean coherence of the filtered
images equals 0.57. The considered area has an
extension of approximately 25 by 35 km. The maximum
height difference within the area is about 1150 m. This
area includes many portions affected by foreshortening,
layover and even shadow that make difficult the phase
unwrapping.
The images were processed in batch mode with the ISAR
software. Before phase unwrapping, the interferogram
was compressed 4 times in azimuth (complex average).
The unwrapping generated four major zones of
integration. The zones were manually "welded” and the
unwrapped phases were checked and corrected for
aliasing errors.
5.1 InSAR Calibration
Input data for the InSAR calibration are the precise master
and slave orbits and the GCPs. In order to derive accurate
orbits, the precise state vectors calculated by the
GeoForschungZentrum (Germany) were used. The orbits
are described by polynomials of the 5 th order; the
polynomial coefficients were estimated by LS adjustment
using the 7 state vectors closest to the image acquisition
interval.
The GCP identification was carried out using the
orthophotos (scale 1:25000) of the ORFEAS data set and
the SAR amplitude and coherence images. The
identification in image space was realised using both the
SAR amplitude and coherence images. In fact, depending
on the land cover type and topography, such images bring
quite complementary information.
An example of GCP measured on the amplitude image is
shown in Figure 2. In this case the river and the bridge
can be easily recognised in both images and the point
was chosen nearby the bridgehead. The example of
Figure 3 regards homologous points measured on the
coherence image. In this case a little stream is the linear
feature which can be recognised in both images. In the
coherence image it appears as decorrelated (low
coherence) with respect to the background.
Being very difficult to recover a sufficient set of full GCPs,
we decided to measure height GCPs, i.e. points not well
identified in planimetry whose height can be determined
accurately. Such points were chosen in the centre of very
flat fields. The identification of a suitable set of GCPs was
very demanding. For the entire scene 20 GCPs were
collected: 13 of them are full GCPs and 7 are only height
GCPs.
The InSAR calibration was performed with two different
sets of parameters. They differ for the parameterization
employed for the interferometric constant Die. In the first
set a second order polynomial was used (with do, di, d2,
d3, d4 and ds as parameters, see equation (6)), while in
the second one, a bilinear polynomial (with do, di, d2 and
d3 as parameters) was adopted.
5.2 Analysis of the First Set of Parameters
The calibration requires the selection of the suited set of
parameters. In order to avoid a very ill-conditioned normal
matrix, from the original parameter set (Ro, AR, T 0 , AT, foo,
f D i, 1d2, fD3, do, d-i, d2, d3, d 4 and ds), the parameters Ro
and foo were excluded from the adjustment.
