The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part R4. Beijing 2008 
982 
one sequence are dedicated to long exposures to observe faint 
background stars with visual magnitudes of m = 4 to 9, which 
are used to control the camera pointing. 
3.2 Camera Pointing 
Camera pointing uncertainties have a major influence on the 
accuracy of astrometric measurements. Long-time exposure 
images containing background stars were taken during almost 
all Phobos flybys since mid 2005. Stars are usually observed 
just before and after a Phobos encounter, giving us the 
opportunity to control/correct predicted pointing information 
and check for possible pointing drifts during the flyby. In cases 
in which just one star was observed, we assumed that a twist 
angle with respect to the nominal orientation information is 
negligible. This was later confirmed by the reduction of the 
observations (cf. Section 3.4 ). To control the camera pointing 
we compute image coordinates of stars from star catalogs (ESA, 
1997, Roser and Bastian, 1992) applying nominal pointing 
information and compare these with the observed line/sample 
coordinates of the stars. Differences between the star positions 
are interpreted as pointing displacement. Hence, all further 
observations of the flyby are corrected accordingly. In general 
the nominal pointing information for the SRC is in good 
agreement with the observed positions of the stars and within 
the accuracy constraints set by the initial specifications for 
spacecraft orientation. Nevertheless, four outliers were found in 
orbits 682, 2706, 2739 and 4568 where discrepancies between 
predicted and observed star positions of up to 450 SRC pixels 
were observed. During one image sequence the pointing 
remains very stable. The average difference between the first 
and the last image of a flyby sequence is 8 pixels which relates 
to 0.004 degree. 
Figure 1: Examples of recent flyby images obtained by the SRC with different flyby distances 
3.3 Astrometric Measurements 
While our previous analysis (Oberst et al., 2006) was based on 
fitting an ellipsoidal model to the observed limb of Phobos, we 
further improved our method by making use of a control point 
network (Duxbury and Callahan, 1989a, Duxbury, 1991). The 
3-D Cartesian coordinates of 315 control points, all of them 
craters, some very large, are given with respect to the center of 
mass of Phobos. The control points are identified in the SRC 
images and line/sample coordinate pairs are measured. These 
points are defined as the center of a crater on the mean local 
surface (Duxbury and Callahan, 1989b). As craters are typically 
observed from quite different viewing angles, much effort was 
made to consistently measure their image coordinates. Points on 
the crater rim were collected and an ellipse fit method was 
applied to reconstruct the center of the crater on the level of the 
crater rim. On average 9, but a minimum of 3, surface features 
were identified and measured in each image. Using the Phobos 
orbit model, the corrected camera pointing and the Phobos-fixed 
coordinates of the craters, we computed the predicted image 
coordinates of the control points. To fit the observed image 
coordinates, the predicted line/sample pairs of the control points 
were transformed. An iterative least-squares adjustment was 
used to solve for the unknown transformation parameters, 
rotation, scale and 2 translations. To derive information on the 
position of the COM (center of mass) of Phobos, we then 
computed the predicted line/sample coordinates of the the COM 
for the specific image and applied the previously determined 
transformation parameter. 
3.4 Results 
By April 2008, over 280 SRC images were obtained during 92 
successful flyby maneuvers. As some images suffer from smear, 
lack of identifiable control points, insufficient surface coverage, 
or lack of background star observations, we considered 120 
images from 51 orbits for the analysis. For these images the 
iterative least-squares analysis to determine the transformation 
parameters converged rapidly. Results indicated that both, the 
scale factor and rotation parameter, are negligible for the 
transformation leaving translations in line and sample direction 
between the predicted and observed image coordinate sets. We 
studied the translation vectors in the object space, especially in 
along-track and across-track direction (radial to the central body 
or out of the orbit plane) for the two recently released orbit 
models (Jacobson and Rush, 2006, Lainey et al., 2007). While 
no significant across-track offsets were found, small along-track 
offsets between orbit prediction models and our observations 
remain, ranging from 1.5 km to 2.6 km, depending on the orbit 
model used (Willner et al., submitted). Phobos’ positions were 
observed with an estimated accuracy of ±0.1km to ±0.5km 
depending on the flyby distance. Comparing observed Phobos’ 
position with orbit predictions, both models show the same 
discrepancy pattern, whereas differences to the orbit prediction 
by Jacobson and Rush, (2006) are larger by approx 35%