1211 
The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part Bl. Beijing 2008 
image coordinates at an image scale of 1:14.300 and a GSD of 
0.08 m. The accuracy in ground coordinates at the GCPs is 
± 0.04 m in X and Y and with ± 0.03 m in Z within the expected 
values. 
Figure 6: Orthophoto mosaik Merklingsen 2 (M2) 
o(AZ) 3.44 2.41 146.89 
Attitude angles [ 0 
a ( A to) 
8.37 
8.33 
- 
a( A(p) 
8.58 
5.95 
- 
a ( A k) 
26.15 
20.97 
- 
Table 3: Standard deviations of the differences between the 
approximate exterior orientation parameters of the aircrafts and 
final values of the aerotriangulation 
The high deviations reveal that the approximate values of 
UAVs are only a little help in the photogrammetric processing 
chain. Reasons for the differences are of systematic and 
unsystematic nature. Unsystematic sources are related to a 
missing /incorrect time synchronisation between the camera 
exposure and the associated GPS-position. Also the accuracy of 
the GPS itself (no DGPS), wind gusts and other factors may be 
sources of errors. Likewise there is the strong drift of the 
miniaturised inertial sensors. Another current problem is the 
axes of the acceleration sensors do not coincide with the 
coordinate axes of the camera, because the integrated GPS/INS 
solution has been developed for the autopilot functions of the 
model plane and not for photogrammetric applications. 
4.2.3 Accuracy assessment 
The determination of the residuals at GCPs generally provides 
optimistic accuracy values. Therefore some of the GCPs were 
declared as checkpoints. For the Flight M 2 the average 
accuracy at the checkpoints were -0.13 m in X, 0.01 m in Y and 
0.00 m in Z. For the flight W1 the average residuals 
were -0.04 m in X, -0.01 m in Y and -0.08 m in Z. The values 
show that the high stability of the block W1 and the high 
quality camera allow for a block with reliable subpixel quality. 
The perspective centres become well determined by the 
aerotriangulation. A comparison to the a priori GPS / INS 
values measured during the flight is given in table 3. 
M 1 
M2 
W 1 
Coordinates [m] 
a (AX) 
16.92 
4.47 
7.53 
g(AY) 
16.64 
2.33 
21.82 
5. PHOTOGRAMMETRIC POTENTAIL OF MICRO- 
UAV’S 
Despite the poor results of the empirical tests there is high 
photogrammetric potential for direct georeferenciong of Micro- 
UAVs. This potential is limited by a number of factors. The 
theoretical optimum of the direct georeferencing is determined 
by the accuracy of the GPS/INS and the flying height. For 
instance using the new MINC autopilot system of Mavioncis 
with an attitude accuracy of 0.6 - 1.2° and a flying height of 
300 m results in a theoretical positional accuracy of 3.15 - 6.3 
m. However the theoretical accuracy level is not achievable, 
even for highly sophisticated solutions, Grenzdorffer and Zuev, 
2007. The following Table 4 gives an overview of the relevant 
factors, which may influence the accuracy of the direct 
georeferencing of an UAV: 
Source of error 
Problem 
Possibility of correction 
Impact on accuracy * 
Interior orientation 
Changes of centre point 
With zoom lenses of consumer 
cameras the centre point may 
change 
Given: camera calibration before the 
flight or simultaneous calibration 
Systematical error, > 1 m 
Changes of focal length 
With zoom lenses of consumer 
cameras the focal length may 
change 
Given: camera calibration before the 
flight or simultaneous calibration 
Systematical error, > 1 m 
Radial distortion 
Nearly constant over time 
Generally not necessary, but correction 
available with every calibration 
Systematic, increase 
toward image comer 
Exterior Orientation 
Time synchronisation 
between GPS and camera 
Unknown exposure time of 
camera, related to GPS time 
stamp 
Given: Synchronisation of camera and 
GPS 
Non systematic error, > 4 
m 
GPS 
No DGPS 
Given: use of DGPS-Logger / 
Large (< 4 m)