The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part Bl. Beijing 2008 
1250 
accuracy is ten times lower when only along track overlap is 
used. 
Estimated vertical accuracy of Medusa 
Figure 5: Effect of along track overlap 
Estimated vertical accuracy of Medusa 
Figure 6: Effect of across track overlap on generalized vertical 
accuracy 
Because of the distinct rectangular shape of the sensor, the 
image base is much smaller along track compared to across 
track. The extreme altitude at which the Medusa operates 
amplifies this. 
The same “normal case” approach also allows us to get a first 
impression on the effect of system temperature on the geometric 
performance of the instrument. Simulation studies have 
indicated that the focal length changes with 20 pm per °C, we 
can simulate the effect on the overall planar accuracy, assuming 
a flying height of 18000 meter (Figure 7): 
Estimated planar accuracy of Medusa 
Temperature Offset [°C] 
Figure 7: Effect of temperature on generalized planar accuracy 
Figure 7 indicates clearly that the variable focal length, caused 
by temperature differences, has a limited effect on the 
generalized photogrammetric performance of the Medusa 
instrument. 
When we compare this temperature related effect with 
variations in overlap, it becomes clear that the temperature 
effect on focal length is negligible (Figure 8): 
Estimated vertical accuracy for Medusa [stdev [m]] 
Figure 8: Effect of changes in focal length versus along and 
across track overlap. The vertical lines visible in the plot 
correspond to isolines of equal generalized vertical 
photogrammetric accuracy [stdev[m]] 
The highest gradient in generalized vertical photogrammetric 
accuracy is visible along the “% overlap” axis. Relative to this, 
changes in focal length due to expected temperature differences 
has no influence. 
In a second phase, sensitivity analysis is performed more in 
detail based on collinearity equations, a camera model and 
image distortion models. A refined (image) simulator is 
developed that allows us to simulate each characteristic of the 
Medusa instrument in detail and estimate the effect on 
measurement precision and accuracy. Figure 9 illustrates the 
concept. 
These simulations allow us to get a thorough understanding of 
the type of image distortions we can expect. Moreover, it allows 
us to estimate the effect on the geometric performance of the 
Medusa instrument if we cannot correct for the distortions. 
Figure 10 illustrates preliminary results of this approach: In 
case changes in focal length would not be taken into account 
when extracting object (real world) coordinates out of imagery 
using direct georeferencing, measurement errors of up to 6 
meters can be expected at the outer regions of the imagery. No 
additional error is simulated in the exterior orientation 
(theoretical GPS/IMU values) for these direct georeferencing 
simulations.