a Digital Terrain Model (DTM). The geometric correction
makes use of a ray tracing algorithm in an iteration pro-
cedure (Zhang et al., 1994).
Attitude data
À
Computation of
orientation parameters
by least squares adjustment
|
Calculation of
temporary terrain coordinates
for ground control points
using equations of collinearity
|
Calculation of
polynomial coefficients for the transformation
from temporary to geodetic coordinates
by least squares adjustment
using ground control points
(and tie points for mosaics)
]
Loop of all input pixels
|
Calculation of
temporary terrain coordinates
using equations of collinearity
and actual DTM information
|
Transformation into a
geodetic coordinate system
using polynomial equations
|
Interpolate new DTM value
(actual - new DTM value)
>eps?
Final geodetic coordinates
Image data
Y Y
Resampling
of the final orthoimage
À
Orthoimage
Figure 3: The »GASIS« approach
for the geometric correction of airborne scanner data
The result of these calculations is a set of data where all
image points (pixels) are converted to temporary ground
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B3. Vienna 1996
points. They are transformed to the »true surface« using
polynomial equations, which can be derived by only a few
ground control points. These polynomial equations are
necessary in order to correct the low frequency trends
that may be inherent in the original measurements of the
flight attitude parameters. However, the resulting »true«
positions form an irregular pattern. In order to derive a
regular grid of pixels, as it is necessary to generate the
final orthoimage, appropriate interpolation algorithms are
applied for resampling.
In many cases it is necessary to combine adjacent strips
to an image mosaic. This is achieved by the simultaneous
determination of the coefficients of all polynomial equa-
tions in a least squares adjustment. This procedure uses
not only ground control but also additional tie points, thus
reducing again the number of necessary ground control
point measurements.
The principle of the approach is outlined in Fig. 3. Al-
though up to now only data of the DAEDALUS scanner have
been processed with GASIS. The reason for this is, that
only for this system, which is operated by the German
Aerospace Research Establishment (DLR) in
Oberpfaffenhofen and equipped with a special flight
attitude measurement system, image data and precisely
specified attitude data have been provided. This data was
flown in the context of several research projects. But in
principle the developed approach and its proved functions
can be applied to any other type of airborne line-scanner
data.
The only modules, which have to be adapted to new types
of data, are the interfaces to the image and attitude data
formats as well as the derivation of the classical orien-
tation parameters from these recorded flight attitude data.
This procedure, however, is in any case one of the most
important processing steps.
5. FURTHER DEVELOPMENTS
In order to fulfil common requirements for the processing
of airborne line-scanner imagery, the main extentions of
GASIS will be:
* the generation of interfaces to data formats of other
line-scanner data and their associated flight attitude
data, especially GPS and DGPS,
e the development of a general module for the
derivation of the classical orientation parameters from
any recorded flight attitude data,
* the integration of powerful radiometric mosaicking
techniques,
* the generation of standard file exchange formats.
Allthough a standard of input data can not be expected in
the near future in scanner technologies it should be dis-
cussed to define an experimental data standard for line-
scanner raw-data which includes the full performance of
image and attitude data. This aspect will become still
more important with respect to the developments in digital
aerial photogrammetry.
In addition to the desribed approach an other solution of
the geometric correction problem is under discussion.
While at the moment orientation parameters are calcu-
lated from
carried out,
data as ob:
a result of
could be d
be improve
Zhang, W.
from Airboi
meters. In
Remote Se
945-950.
Zhang, W.
borne Line
In: Procee
Sensing C
pp. 11-447 -