gle between
this fact the
S reduced to
angle of inr.
and reflected
netrating the
cattered, ap.
louds, haze
sunlight and
t of the sun
effects. The
ed with the
S concerning
adiation and
osphere are
/ithin the ob-
radiation de-
ant reduction
s on the field
| and on the
endent from
r the spectral
Jour infrared
ellow filter to
the effective
corded with
Y 500nm and
photographic
intensity and
h, on ihe
urves of the
during which
| surface, the
yellow in the
cess) of the
e coloured in
um.
Besides the
t of radiome-
n the photo-
eflectance of
»flectance is
cture of sur
wacteristic oi
ription of the
ı be given by
s the relation
a differential
and the fe
)
[sr] (2
The reflectance function is dependent from the wave-
length of the electromagnetic spectrum and must be
defined as a spectral dimension. It can be determined
only with an enormous measuring and computation
efford.
By an integrated approach of interaction mechanisms
between objects and matching colouration on the film
emulsion following statesments can be made:
To predict the colouration of an object or parts of an
object on a photograph a lot of parameters (sun ra-
diation, atmosphere, object characteristic, camera
objective, camera filter, film emulsion etc.) must be
considered, determined or measured.
+ Due to the radiometric and spectral characteristics of
the above mentioned parameters it is impossible to
get a determined transformation between colours of
an object on a photograph with a true colour (TC)
emulsion and such one of a colour infrared (CIR)
emulsion.
+ The correlation of reflectance within similar objects
classes (e.g. coniferious forests) is very high.
+ Distributions of radiation intensities on TC and CIR
photographs (e.g. difference between sunny and sha-
dowed areas) are correlated to a high degree.
3. TRUE COLOUR VISUALIZATION
3.1 Field of tasks
Within a water management project in the district of Alt-
heim (Upper Austria) a land-use classification was
necessary in order to obtain draining and water mana-
gement parameters. The classification for the whole area
(approximately 300km?) was done by statistical methods
using colour-infrared aerial photographs (photo scale
appr. 1:15.000). The positions of - in the ground coordi-
nate system - regular sample plots were distorted for
each photograph to the photo coordinate system by the
parameters of the interior and exterior orientation, using
à digital terrain model (Bart! et al, 1996).
For the visualization of results, such as the location of
measured river profiles, potential areas for floods depen-
dent on different disaster levels, orthophotos were pro-
duced. Due to the limited financial resources of this pro-
fct the request of the customer to get true coloured
orthophotos by making an own photo flight had to be
cancelled and orthophotos were produced using the
colour-infrared photo material. To solve the problem that
the non photogrammetric or interpretation specialists
feels often confused and irritated by the ,false colours“ of
the CIR photographs true colour orthophotos were
Simulated by the algorithms described in this paper.
3.2 Production of orthoimages
Te en of the aerial photographs to orthophotos
N i one by means of digital photogrammetry. First of
a aerial (CIR) photographs - photographed by the
NE n „Bundesamt für Eich- und Vermessungswesen“
Y à private photo flight company (Fischer) - were
471
scanned on the photogrammetric scanner ZEISS-
INTERGRAPH PhotoScanner PS1 with a pixel resolution
of 30um. For saving disc space the image data were
compressed using a JPEG algorithm.
The parameters of the exterior orientation were deter-
mined by means of aerotriangulation (measured on the
analytical plotter ZEISS P3 and computed with the model
adjustment software ,PATM" of the Institute of Photo-
grammetry, University Stuttgart).
The rectification of the CIR aerial photographs was
computed on the soft copy station INTERGRAPH
ImageStation 6787 of the Institute of Surveying and
Remote Sensing, University of Agriculture, Forestry and
Renewable Natural Resources. The ground pixel size of
the orthophotos was 0.5m * 0.5m. Due to the modest
(geometric) accuracy requirements of the orthophoto
(only used for visualization purposes) the terrain
information was derived from the Austrian-wide digital
terrain model of the Bundesamt für Eich- und
Vermessungswesen with a ground resolution of 50m.
3.3 Production of Pseudo True Colour orthophotos
from Colour Infrared orthophotos
As described in Chapter 2 there is no unequivocal solu-
tion to transform colour-infrared (CIR) pixel values to
true-colour (TC) values. A best-fitting relation between
CIR values and TC values can be approached. The TC-
images simulated in this way are called ,pseudo true
colour” (PTC) images since the result naturally is not
identical with a TC-image.
3.3.1 General mathematical approach: The three
bands of a digital colour infrared image (red band cor-
responding with the natural infrared part of the electro-
magnetic spectrum, green band corresponding with the
natural red part of the electromagnetic spectrum, blue
band corresponding with the natural green part of the
electromagnetic spectrum) desribe a three dimensional
colour space (Rıp Gem B,,). The three bands of a digital
true colour image (red band corresponding with the natu-
ral red part of the electromagnetic spectrum, a.s.o.) also
describe a three dimensional colour space (R,., G,., B,.).
The relation between the CIR colour space and the TC
colour space is approximated by the following linear
transformation:
Rrc — dio aii- Rem t A12 - Gom + A13 - Ber
Grc — dao à21* Rein - à227 Gem - à23- Bin (3)
Brc= aso+ as - Rer+ asn- Ger + ass- Ber
3.3.2 Determination of transformation elements: To
estimate the transformation elements described in
equation (3) identical pixels (,spectral and radiometric
control points“) in both colour spaces (CIR- and PTC-
colour space) have to be found. For this purpose true
colour photo material of a part of the project area is
needed. This reference photo material can be obtained
by true colour photos from archives or by taking
photographs with amateur cameras along with the CIR-
photo flight. To get (geometrically) identical pixels also
the reference photo is rectified to an orthophoto with the
same ground pixel size and geometry as the CIR
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B7. Vienna 1996