The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part Bl. Beijing 2008 
The geometric calibration approach for the MEDUSA 
instrument needs to be able to cope with this dynamic 
behaviour. 
Second challenge is that the instrument cannot be calibrated at 
room temperature and atmospheric pressure. Therefore a ground 
calibration of the MEDUSA instrument would require the 
operation to be performed in a thermal vacuum chamber. 
Although this is possible in principle, it is difficult to realize the 
exact thermal conditions (radial and axial thermal gradients) in 
this simulated environment. 
A third aspect to be taken into account is the presence of the 
folding mirror. Further investigation is necessary to investigate 
how possible deformations of the reflecting surface, induced by 
temperature variations of the mirror material and its interface 
pieces to the athermal carbon fibre structure, need to be 
modelled. 
4. CALIBRATION STRATEGY 
The MEDUSA instrument is designed for large scale mapping 
and disaster monitoring. Therefore the primary focus of the 
calibration is on geometric correction of the imagery, which is 
the topic of this paper. Radiometric calibration of the 
instrument shall be considered in a second phase. 
Ground geometric calibration of imaging platforms remains a 
most complex operation (Zeitler, 2002). A direct drawback of 
ground calibration is that the results are only valid for similar 
operational conditions. 
In case of the MEDUSA instrument, this assumption is not 
valid (see above). For this reason we have opted for a full in 
flight geometric calibration approach based on block bundle 
adjustment. Such a geometric calibration strategy has proved to 
be successful for other more complex imaging systems such as 
the ASD40 from Leica (Tempelmann, 2003). 
4.1 Geometric Ground calibration 
During the performance test of the optical system in a thermal 
vacuum chamber a rough determination of the focal length and 
principal point will be performed. This will be used as starting 
value for the in-flight calibration method. 
4.2 Geometric In-flight calibration 
The in-flight calibration is prepared by a two-phase sensitivity 
analysis for various aspects of the instrument based on expected 
ranges of environmental parameters. 
In a first phase, the generalized photogrammetric accuracy of 
the instrument is explored for the “normal case” (Kraus, 2007). 
This provides a first insight into the significance of different 
parameters on the potential photogrammetric accuracy of the 
MEDUSA instrument. 
According to the normal case equations: 
cr x =a Y =m b -<J x 
Z 
°z= m b-° x ‘- 
with 
Z 
m b ~ 
FocalLength 
Z = elevation 
a x = accuracy of tie point identification in the image 
B = distance between observations 
For the expected accuracy at which tie points can be identified 
(o x ) a value of 1/3 of the pixel size is taken as a first estimation. 
In best case scenario’s, this accuracy can reach 14 of a pixel or 
better. 
The distance between observations can be estimated based on 
the image characteristics and planned along track image overlap: 
B = W-(l-R) 
with 
TTT Pixels • Pixelsize 
W = Z 
FocalLength 
R = overlap 
When we assume the Medusa instrument will operate between 
15000 and 21000 meters above ground, the expected overall 
planar accuracy can be estimated to be within the decimetre 
range (Figure 3). 
Estimated planar accuracy of Medusa 
Figure 3: Planar accuracy of Medusa versus altitude 
The overall vertical accuracy is expected to be approximately 
1.5 meter, when using 60% side overlap (Figure 4). 
Estimated vertical accuracy of Medusa 
Figure 4: Vertical accuracy of Medusa versus altitude. 
By comparing the expected vertical accuracy as extracted out of 
along track (Figure 5) versus across track (Figure 6) 
overlapping imagery, it is highly advisable not to use along 
track overlap for the extraction of elevation data. The vertical