thoroughly dried out.
Although the flight altitude was only 500 m above ground-
level and the air was very dry an atmospheric correction of
an evening flight stripe and a morning flight stripe was
tested. To calculate the correction it is necessary to
integrate the temperature and humidity values of a
radiosonde (weather balloon) during flight time and to
combine them with the pressure and trace gases of a
standard summer atmospheric condition. The comparison
of the original data set with the atmospherically corrected
picture shows mostly constant temperature differences of
1 - 1,5 °C higher temperature in the corrected pictures. This
moderate difference can be explained with the little
difference between soil temperature and air temperature of
evening and morning flights. For that reason a costly
atmospheric correction is not necessary for the complete
data set.
2.3 Reference Data
For the presented steps the following data were used:
- topographic maps (scale 1 : 10 000)
- boundaries and sealing degrees of 40 terrestrial test
sites (total area of 145,3 ha)
For further analyses and as a reference for all test sites the
following data shall be integrated:
- map of biotopes (scale 1 : 10 000) including sealing
degree values divided in five classes
- layer of buildings taken from topographic maps (scale 1
: 10 000)
- maps of sealing degrees calculated from satellite data
(Landsat-TM and SPOT-XS)
3. DATA PREPROCESSING
3.1 Data Calibration
The scanner data have to be converted in true reflection
values and temperature values using the recorded flight
parameters. Each pixel value has to be calibrated with the
following formulas:
for the optical bands 1 - 10:
Calibrated Radiances L; = C * ( D/G - Dgg,) (1)
and for the thermal infrared band 12:
Calibrated Radiation Temperature
T; — Tag, + ( Togz~ Tass) / ( Dep - Dggi) * ( Di- Dggi) (2)
parameters:
D; - - Video Count
C - Calibration Factor
G - Gain
Des - Blackbody 1 Count
Dag - Blackbody 2 Count
Tg, .- Temperature of Blackbody 1
Tes, - Temperature of Blackbody 2
3.2 Geometric Rectification
488
Data are rectified using the rectification software PRESDO
(JANSA, 1983). Therefore about 30 ground control points
per km? were marked as points in the scanner data and in
scanned topographic maps at the same time (Scale
1 : 10 000), so altogether about 10 000 ground Control
points were inserted for all flight stripes. Data are
resampled with a resolution of 2 m. A possible resolution of
1 m is avoided, because the necessary storage Capacity
would be increased by factor 4 (for 2 m resolution 550 MB
are needed).
The limited precision of rectification is an argument not lo
resample down to 1 m. The mean rectification error
therefore 2,5 pixels (5 m), its maximum is 5 pixels (10 m)
The software PRESDO only allows a relative rectification vj
flight stripes on a topographic reference in file Coordinates,
so the rectified data have to be transformed into Gauss.
Krüger-Projection (Bessel ellipsoid) by editing the upper lef
corner co-ordinates of each stripe.
is
3.3 Mosaic of Stripes
To compose the rectified flight stripes into complete
pictures a mosaic process is needed, which is worked out
in the software ERDAS-IMAGINE (8.2). In order not to
modify the intensity value of each pixel, ie. real
temperature value, the process is performed without
contrast matching. Tests of mosaic with included contrast
matching over the whole overlap area partly show
unsharpness which can be explained with inaccuracy of
rectification. For that reason each flight stripe was cut outin
a certain way that the overlap area between two stripes siil
is only 5 - 10 pixels. A linear rectangular feathering is
applied on this remaining narrow unsharp overlap area
which connects two stripes.
3.4 Calculating Additional Information Layers
For the estimation of urban soil sealing the vegetation index
and the cooling characteristics of certain urban areas are
very important.
A cooling is calculated using the difference of evening ani
morning temperature and displayed in a "cooling picture"
(3)
ATzT T
evening ^ ' morning
The vegetation index NDVI (Normalized Difference
Vegetation Index) for the whole area is calculated on the
basis of the following formula:
NDVI = (Band 7 - Band 5) /( Band 7 + Band5) (4
Scanner
3.5_Composition of a Multispectral and Multitemporal Total
Data Set
For further steps it seems promising to composite al
available information layers in one total composition for
each flight area. Layer comparison and classification
procedures are now easier to handle. For software reasons
it is necessary to have all layers in the same format. So te
additional calculated auxiliary bands, which were in floating
point format, are now transformed to 8-bit unsigned inlet?
format. Table 3 shows the picture composition in detail:
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B7. Vienna 1996
-— er Xr Pf) qoc rTÀ 5
>