3. THE INDIAN EARTH OBSERVATION SATELLITE
IRS-1C
The "Indian Remote Sensing Satellite" (IRS-1C) was laun-
ched on 28 December 1995 and, following satellites 1A
(launched 1988) and 1B (launched 1991), is the third and
most powerful earth observation satellite in the Indian pro-
gramme. IRS-1C has been continuously transmitting data
since the summer of 1996. It is fitted with three main in-
struments (CCD detectors operating in pushbroom mode).
The panchromatic line camera (PAN) can be swivelled
when recording stereo images or else to achieve enhanced
time coverage. The structurally identical twin satellite IRS-
1D was launched on 29 September 1997. It will halve the
repetition time of currently 24 days as well as simplifying
operational acquisition planning. The sensor parameters
are shown in Table 2.
PAN LISS-II WiFS
Spectral 0.5-0.75 0.52 (green) 0.62-0.68 (red)
0.62 (red) 0.77-0.86 (NIR)
0.77 (NIR)
1.55-1.75 (SWIR)
Spatial 5.5-5.8 23.5 and 70.8 188
resolution (m) (SWIR)
Strip width 70.5 142 770
(km)
Radiometric 64 shades of |128 shades of 128 shades of
resolution grey (6 bit) grey (7 bit) grey (7 bit)
Repetition rate 124 24 5
(days)
Table 2: Parameters of IRS-1C sensors
The European data provider GAF/Euromap has in the me-
antime set up a production system for the generation of
ortho-imagery and natural colour products.
4. IMAGE QUALITY AND PREPROCESSING
OF IRS-1C DATA
4.1 Data Search and Procurement
Searching for IRS-1C data via the ISIS satellite image ar-
chive of the German Remote Sensing Data Centre (DFD)
and using the search parameters Area, Sensor, and Rec-
ording Date is comfortably straightforward. Besides sear-
ches for stored image data and important recording varia-
bles (recording date, degree of cloud, covered areas etc.),
it also facilitates the display and downloading of "quick-
looks" measuring 500*500 pixels. These only allow the
degree of cloud and any hazy areas to be assessed, howe-
ver. What cannot be verified using either a 1:1,000,000-
scale representation of multispectral imagery 140*140 km
in size or panchromatic imagery shown at 1:500,000 scale
(70*70 km) is the presence and nature of image defects.
Data procurement via Euromap GmbH was similarly
straightforward. At the same time, data distribution could be
further simplified in some respects, since several inquiries
by phone were required when scenes were to be relocated
within a strip for example. The fact that the data of the
SWIR channel are only part of the product as long as no
data shift is included only became apparent once the data
had been procured. Neither is it possible at present to mesh
image data with the neighbouring strip. An improvement in
the locational stability of scenes would also be desirable.
The areas covered tend to vary rather a lot while still being
accorded the same scene designation.
4.2 Ima uali
The image quality of the IRS-1C LISS quarter-scene pur-
chased and of the PAN sub-scene can be rated as being
very good overall. That applies in particular to the inner
image geometry, which facilitates high-precision rectifica-
tion. Image defects in the form of blobs and faulty strips
were present in the PAN scene, however. When recording
extremely bright and smooth surfaces, the sensor is over-
set, especially when the sun is very low, and it requires
quite a bit of time to reproduce correct reflection values.
Defects are difficult if not impossible to remove. Whilst indi-
vidual lines or points can be easily rectified by generating
mean values, it is not generally possible to repair a multiline
omission without visible image impairment. The area then
has to be filled with data from another image set. In the
case of a faulty strip approx. 10 image lines wide, the satel-
lite image provider (Euromap GmbH) generously made part
of another panchromatic scene available free of charge.
Slight banding in the orbiting direction was discernible in
the panchromatic image scene (approx. 6 % intensity fluc-
tuation at a distance of approx. 4 pixels), this being troub-
lingly noticeable in some colour composites on very dark
surfaces such as bodies of water. This may be a result of
the sensor's CCD line not yet being 100 % calibrated.
4.3 Georeferencing
Panchromatic imagery was georeferenced using scanned-
in topographical maps of 1:25,000 scale with the multispec-
tral image (LISS) subsequently being rectified onto the
panchromatic image. This ensured a high locational accu-
racy between the two image products, a prerequisite for the
generation of merge products. Table 3 reproduces the recti-
fication error.
Given a mean error of 0.78 pixels, a mean accuracy of
« 4 m for the PAN scene and of « 8 m for the LISS scene
was achieved. The independent monitoring points confirm
these findings. Very high locational accuracy is of great
importance with regard to the overlaying of high-precision
geodata in later processing or interpretation stages. Over-
laying a georeferenced IRS image product with, in turn, an
urban biotope chart (survey scale 1:10,000) and an ATKIS
data record (survey scale 1:10,000) impressively underli-
ned the high locational accuracy and hence also the fine
inner geometry of the IRS data (see III. 1).
For an extra charge, the data distributor Euromap is now
offering georeferenced data with a locational accuracy of
5 m for panchromatic and 15 - 20 m for multispectral ima-
gery.
252 Intemational Archives of Photogrammetry and Remote Sensing. Vol. XXXII, Part 7, Budapest, 1998
Recor
—-
Scene
Rectifi
—
Numb
Pass |
Monitc
Table 3
4.4 At
Given
atmos
future.
absolt
both r
very h
ration:
mount
relief-
1997).
reseal
recour
Cy apr
by R.
intens
sphere
easier
illumin
visual
adjust
the m
atmos
have |
cesa
4.5 IR
Image
produ
al res
this. H
cing tl
by co
multis
the re
partici
produ
1C d
1997)
