a of the
1g as no
the data
| to mesh
ement in
lesirable.
still being
ene pur-
as being
the inner
rectifica-
lity strips
recording
r is over-
requires
n values.
'hilst indi-
enerating
x multiline
area then
et. In the
the satel-
nade part
harge.
ernible in
nsity fluc-
ing troub-
very dark
a result of
ited.
scanned-
multispec-
onto the
nal accu-
site for the
5 the recti-
curacy of
ISS scene
ts confirm
s of great
\-precision
ges. Over-
in turn, an
an ATKIS
ly underli-
so the fine
ap is now
ccuracy of
ectral ima-
IRS-1C PAN full scene
IRS-1C LISS lll quarter scene
Recording date 16.09.96 16.09.97
Scene size 70*70 km 70*70 km
Rectification base TK25 georeferences PAN scene
Number of pass/monitoring points 160/21 95/35
Pass point accuracy RMSE, 0.45 pixels / 2.6 m 0.20 pixels / 4.6 m
RMSE, 0.64 pixels / 3.2 m 0.26 pixels / 6.0 m
RMSE... 0.78 pixels / 3.9 m 0.33 pixels / 7.6 m
Monitoring point accuracy RMSE, 0.49 pixels / 2.5 m 0.12 pixels / 2.8 m
RMSE, 0.45 pixels / 2.3 m 0.14 pixels / 3.2 m
RMSE... 0.67 pixels / 3.4 m 0.18 pixels / 4.1 m
Table 3: Rectification error of IRS-1C data
4.4 Atmospheric Adjustment
Given the ever improving resolution of satellite imagery,
atmospheric adjustment will assume great significance in
future. Change detection methods and the application of
absolute spectral signatures for land-use classifications
both necessitate atmospheric adjustment. There are now
very high-performance program packages with good ope-
rational attributes available which can also be deployed in
mountainous terrain, being capable of compensating for
relief-related lighting discrepancies (ATCORS, Richter,
1997). Atmospheric adjustment of data was effected in the
research project using the ATCOR3 program and having
recourse to a relief model (grid width 25 m, altitude accura-
cy approx. 15 m in mountains). (This section was covered
by R. Richter [DLR], to whom many thanks!) By converting
intensity values into relative reflection values, the atmo-
sphere and lighting-adjusted image products become far
easier to interpret and classify, shady slopes being better
illuminated for example. Whereas the intensity values of the
visual green and red channels are reduced by atmospheric
adjustment, those in the near-infrared channel are raised in
the middle and more effectively differentiated. Following
atmospheric adjustment in NIR, surface waters for instance
have reflection values of 0 % whilst lush vegetation produ-
ces a 60 %-plus reading.
4.5 IRS-1C PAN and LISS Image Merge Products
Image merging has the objective of producing the image
products best suited to a given task of interpretation. Spati-
al resolution and colour rendition play a decisive part in
this. High spatial resolution can be achieved without sacrifi-
cing the colour information required for surface evaluation
by combining high-resolution panchromatic images with
multispectral data records. Important colour composites are
the real-colour and the infrared representations (the latter
particularly so for assessments of vegetation). Real-colour
products can be calculated even if, as in multispectral IRS-
1C data, the blue channel is missing (cf. Schumacher,
1997).
Calculating colour composites from IRS-1C data proved
more difficult than had been expected. The end-results
were unconvincing as regards both their sharpness of detail
(incomplete imaging of high-resolution panchromatic infor-
mation) and their colour fidelity (incomplete retention of
multispectral information). The reason for this may be that
the panchromatic image only covers two spectral bands of
the multispectral image (visible green and red), near-in-
frared lies beyond the panchromatic spectral range, and
hence congruency of imaging between the panchromatic
and the multispectral does not obtain.
IHS transformation (IHS), principal component transforma-
tion (PC), and procedures originated by Brovey (B) were
tested for their capacity to produce colour composites. The
best visual results were yielded by the PC process. The
IHS and B methods produced significantly less well-focu-
sed results. The quality of the PC results, however, de-
pends on the image section selected. At the time of writing,
therefore, a need for further research can be discerned
where the generation of optimum colour composites based
on IRS-1C imagery is concerned.
4.6 Overlaying IRS Data with other Geodata
An overlay with geographical base data, data from the Offi-
cial Topographic-Cartographic Information System (ATKIS)
for instance, can facilitate the creation of digital or analogue
cartographic bases that are ideal for visual interpretation.
The superimposition of vectoral data records is also neces-
sary for the assessment of segregated surface units or
even for their up-dating. Visually estimating the dimensions
of structural utilisation within a construction block, for ex-
ample, is eminently feasible by superimposing the construc-
tion-block boundaries. Furthermore, the vector data records
can themselves be up-dated by means of image interpreta-
tion (e.g. up-dating of land-use planning, regional planning,
ATKIS, biotope mapping). Figure 1 shows the IRS PAN
image with construction-block vector data overlaid.
Intemational Archives of Photogrammetry and Remote Sensing. Vol. XXXII, Part 7, Budapest, 1998 253
