2. GENERATION OF ORTHOIMAGES
In general line scanner imagery has its own variable ab-
solute orientation for each scan line. This is significant
especially if airborne scanner imagery is processed
(Zhang et al., 1994). In this case parametric rectification
methods using the measured absolute orientation has to
be applied. Non-parametric rectification where three-
dimensional ground control points are used to decribe the
continuous orientation can only be applied to imaging line
scanners of spaceborne missions like SPOT (Albertz et
al., 1990). Within these approximately circular orbits the
sensor orientation does not contain high frequent and
non-continuous variations of position and pointing. Plane-
tary missions are mostly flown on elliptical orbits which
can also be considered to be continuous but the varying
observation distance does not allow the application of
non-parametric approaches. Too many ground control in-
formation would be necessary, but such information is not
available in sufficient quality and quantity for Mars and for
most of the other planets.
Therefore and because of the decribed technical capabi-
lities of both cameras orthoimages of HRSC and WAOSS
data will be derived using a combination of parametric and
non-parametric approaches (Fig. 2). Grid points defining a
patch pattern within the image are projected to the sur-
face using the absolute orientation while all positions
within the patches are described by projective transfor-
mation using the patch edges as identical points (anchor
points).
The patchwise transformation of image data of HRSC and
WAOSS is based on the assumption that position and
pointing of the sensor varies continuously within shorter
periods. The length of these periods can be expected in
the range of up to 50 lines while changes within the sensor
commanding might appear in a frequence of up to 8 lines,
thus adequate patch sizes seem to be between 8 and 50
pixels.
Proper position and pointing information will be acquired
during the mission in form of so called SPICE kernels
(including also information about planetary constants and
instrument parameters). The orientation data will be im-
proved by photogrammetric bundle block adjustment (Ohl-
hof, 1996).
A ray tracing algorithm (Jahn et al., 1992) is applied to
compute the intersection of the line of sight vector with
the surface given by the DTM which is defined above a
triaxial ellipsoid as the height reference body. Patches will
be rectified if the DTM (Uebbing, 1996, Wewel, 1996) con-
tains reliable height information for all patch edges.
The intersection points of the patch edges will then be
transformed to a given map projection defining not only
the projection type but also the scale of the final ortho-
image. All pixel positions within the map projected pat-
ches will then be transformed back (indirect rectification)
to the patch in the input image using projective trans-
formation with the patch edges as identical points. This
process assumes that surface variations within these
patches can be regarded to be constant. However, if the
terrain information, given by the DTM, shows higher fre-
quencies the patch size has to be decreased, if neces-
sary up to a size of 1 pixel.
Image sequence
of HRSC/WAOSS
Y
Check for areas with
constant macropixel format,
constant exposure time and
without gaps
Commanding information
of each image line
Define anchor point patches
with a given size
within these areas
| Position and pointing
information
Define the line of sight vectors
of the patch edges
| Geometric calibration
information
Intersect the rays
with the reference body
(defined by DTM)
| Raster-DTM
Transform the intersection points
into a map projection
|
Define projective transformation
between patch edges
in the image and
in the map projection
Transform all pixel positions
within the patch
from map projection to the image
using projective transformation
|
Interpolate grey values
for all pixel positions
within the patch
Y
Orthoimage
Figure 2: Orthoimage generation of HRSC/WAOSS data
In addition to this some map projection types describe
distortions with respect to the surface which do not allow
the application of projective transformation in large pat-
ches. In these cases patch sizes larger than 5 again are
not appropriate.
As an example for such an extreme map projection Fig. 3a
shows an image acquired during the GALILEO mission
from the north pole region of the Moon (partly in shadow),
Fig. 3b displays this image map projected to the Sinu-
soidal projection where the pole is represented by a point.
Fig. 3c shows the distortion of the input patches.
352
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B2. Vienna 1996
In
tio
(Ne
luti
ant
loc
be
duc
The
dat
dat
pro
Mis
Ortl
info
CLE
