984 
The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B4. Beijing 2008 
Figure 4: Control points (blue) of the DLR network in comparison with the Duxbury and Callahan control point network (letters). 
The density has increased by a factor of two while the uncertainty of the control point coordinates is approx, a 6 times 
lower. 
To successfully compute a block adjustment with Viking orbiter 
data, navigation information needed to be improved prior to the 
computation. Therefore, we improved the pointing information 
by fitting the predicted limb position in the image to the 
observed figure. For images with very high resolution - 
showing only parts of the surface but not the limb of Phobos - 
an overlay of the Duxbury and Callahan control points was 
produced and moved to fit to surface features in the image to 
improve the camera orientation. To further optimize results, 
accuracy values for orientation data were adjusted according to 
the computed normalized residuals. Overall, a bundle block 
adjustment with Viking orbiter data was very sensitive to 
variations in the stochastic model. Object points were 
determined with uncertainties Ox, o y , Oz of 111 m , 90 m, 101 
m when a measurement accuracy of 1 pixel was assumed. 
For the computation of a global control network both data sets 
were combined through conjugate points in overlapping areas. 
Since reduction of the Viking orbiter image measurements was 
very dependent on the stochastic model chosen Viking image 
orientations were introduced as unknown parameters rather than 
additional observed values. The computed spacecraft 
orientations for MEX and the Viking orbiters were then used to 
determine the object point coordinates of the control points in a 
second run. Furthermore, a Baarda gross error detection 
(Baarda, 1968) was applied to remove misidentified 
measurements. Thus, uncertainties of the 3D ground 
coordinates could significantly be reduced by a factor of 3 for 
points in SRC images and a factor of approx. 6 for points in 
Viking images (cf. Tabel 3). 
4.4 Global Shape 
Control points are irregularly distributed over the surface of 
Phobos (see Figure 4). To determine the shape of Phobos a 
triangulation between the points was computed. A few gross 
errors remain in the coordinates of the control points, 
represented by artifacts. Even though the distribution of the 660 
control points is quite irregular (Figure 4) the network is a good 
basis for a global shape model, which shows many details such 
as of medium sized craters. A preliminary version of the shape 
model is shown in Figure 5. 
4.5 Rotation Parameter 
For the reduction of the control point network, nominal values 
of unknown parameters of the bundle block adjustment are 
computed with respect to the body centered, body fixed, 
coordinate frame. Thus, a different orientation of the observed 
body may lead to a solution with reduced stresses in the 
network as the nominal camera orientations - acting as 
unknowns and observed parameters in the adjustment - do 
change with the orientation of the body. This situation is 
exploited to observe rotational parameters of Phobos. 
Even though Phobos is in a synchronous orbit around Mars, a 
free libration can be observed due to the varying angular speeds 
along an elliptical orbit. Furthermore, Phobos’ pronounced non- 
spherical body interacts with the gravitational field of Mars 
causing a forced libration, a superimposed sinusoidal oscillation. 
We concentrated on the observation of the forced libration 
amplitude from the control point network. The least-squares 
adjustment of the control point network was computed for a 
varying forced libration amplituds. Subsequent studies of the 
residuals of the control