GERY
iting planet
ocessing of
1ages in the
etric bundle
athematical
stment with
the MOLA
quisition of
concept of
lata as con-
ection 3. In
ing are pre-
camera con-
nsors for si-
multi-phase
ur panchro-
The conver-
s arc 21 and
rranged per-
ne. The im-
;quired line-
. One image
yn ground of
sntre and in-
] Hoffmann,
ibout 60 km
th of about
image from
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B4. Istanbul 2004
Figure 1: High Resolution Stereo Camera ((ODLR, Berlin)
Figure 2: Part of image from orbit 68
In addition to the nine sensors of HRSC, there is another sensor
called Super Resolution Channel (SRC). The SRC delivers frame
images with 1024 x 1024 Pixel with a ground resolution of 2 -
3m. This sensor delivers no stereo images. Therefore, a pho-
togrammetric point determination is not possible and the sensor
will be not longer considered in this paper.
The three-dimensional position of the spacecraft is constantly de-
termined by the European Space Agency (ESA) applying a com-
bination of doppler shift measurements, acquisition of ranging
data, triangulation measurements, and orbit analysis. The orbit
accuracy at the pericentre is given as an interval of maximum and
minimum accuracy for the whole mission duration (Hechler and
Yafiez, 2000). Table 1 shows the accuracy interval for X (direc-
«^ of flight), Y (perpendicular to the direction of flight), and Z
radial).
X Y Z
[10-2120m | 25-795m | 1-80m
Table 1: Orbit accuracy at pericentre (Local frame)
The attitude of the spacecraft is derived from measurements of
à Star tracker camera and from an Inertial Measurement Unit
(IMU). The accuracy of the nadir pointing results from a com-
853
bination of attitude errors and navigation errors. The values for
accuracies are 28 mgon for all three angles ¢ (pitch), w (roll), and
k (yaw). They are supposed to be valid for the whole mission
(Astrium, 2001).
These measurements result in an observed three-dimensional po-
sition and attitude of the spacecraft which can be considered as
the approximated exterior orientation in classical photogramme-
try. However, these observations are not consistent enough for
high accuracy photogrammetric point determination.
The interior orientation of the HRSC has been calibrated by
Dornier at Friedrichshafen (Carsenty et al., 1997). During the
six month journey to Mars the interior orientation has been veri-
fied by the means of star observations. So far no deviations from
the calibration have been experienced and the interior orientation
parameters of the HRSC is considered to be stable.
2.2 Mars Observer Laser Altimeter (MOLA)
In February 1999 the Mars Global Surveyor (MGS) spacecraft
entered the mapping orbit at Mars. During the recording time
(February 1999 to June 2001) the MOLA instrument acquired
more than 640 million observations by measuring the distances
between the orbiter and the surface of Mars. In combination with
orbit and attitude information these altimeter measurements have
been processed to object coordinates of points on the ground.
Each orbit results in one track of MOLA points (see Fig. 3).
\
A A
AN \/ Vy ;
A SA X
V
2 #1 RE X
77/7 RS
Figure 3: Part of MOLA tracks
The along track resolution is about 330 m with a vertical neigh-
boring precision of 37.5cm from shot to shot, i.e., from laser
point to laser point. The absolute vertical accuracy is in the or-
der of 10 m. The surface spot size is about 168 m in a 400-km-
elevation mapping orbit (Smith et al., 2001). The across-track
shot to shot spacing depends on the orbit and varies with latitude.
In general, the distance between neighboring tracks on the ground
is up to more than 1 km (Kirk et al., 2002).
In addition to the surface described by the original, irregularly
spaced MOLA track points NASA (Neumann et al., 2003) dis-
tributed a grid-based global Digital Terrain Model (DTM) which
is derived from these MOLA points (see Fig. 4). The accuracy of
DTM is 200 m in planimetry and 10 m in height. As mentioned
before, the special thing about the laser points is, that they can not
be identified in the images in an easy way. l.e., image coordinates
of most of these points can not be measured, and therefore, it is
not possible to treat them as normal GCPs in a bundle adjustment.
