ISPRS Commission III, Vol.34, Part 3A „Photogrammetric Computer Vision“, Graz, 2002
3. TEST FLIGHT CONFIGURATION
The presented data from different calibration flights are part of
a big production project in Saudi Arabia flown by Hansa
Luftbild German Air Surveys. Within this project more than
9000 images (scale 1:5500) were captured at 12 flight days
from January, 29" — March, 25™ 2001, covering a time span of
approximately 2 months. Parallel to the image data recording,
GPS/inertial positions and attitude data were provided by the
IGI AEROcontrol IId system, whose IMU was rigidly mounted
at the camera body. For each mission day the same fully
signalised flight line was normally flown twice with opposite
flight directions for system calibration — typically once in the
morning before and once in the evening after mission flight.
This calibration strip consists of altogether 21 signalised ground
control points (GCP) located in the standard or Gruber positions
of each image resulting in 7 captured images per calibration
line. Since almost all images were taken with the same Z/I-
Imaging RMK Top30 — GPS/inertial installation (calibration
flights 1-19, January, 29" — March, 15“), the results from the
multiple calibration flight data allow for first investigations on
the long term stability of system calibration. Only the last two
missions were flown with a wide-angle RMK Topl5, therefore
the inertial unit had to be demounted and fixed to the new
camera body for this last two mission days (calibration flights
20-23, March, 24™ and 25"). These wide-angle flights are non
considered in more detail in the following.
The input data for the system calibration were provided by IGI
and Hansa Luftbild, respectively. The GPS/inertial data were
processed using the AEROoffice software, afterwards the
integrated GPS/inertial positions and attitudes are interpolated
on the camera exposure times. The pre-surveyed translation
offsets are already considered during GPS/inertial data
integration. The image coordinates were obtained from
MATCH-AT automatic aerial triangulation, where the GCP
image coordinates were measured manually.
4. TEST RESULTS
Based on the integrated GPS/inertial-AT described in Section 2
the calibration of system parameters was done for each
calibration flight based on the given 21 GCPs and the exterior
orientation results from the integrated GPS/inertial system.
Since no quality measures for the GPS/inertial positions and
attitudes were available from GPS/inertial data integration an
assumed accuracy of 0.1m and 0.005gon was introduced for the
stochastic model. This empirical accuracy should be expected
from such high quality integrated GPS/inertial system if its
accuracy potential is fully exploited.
Within system calibration the inevitable angle offset and
position shifts (if significantly present) are estimated in
combination with the Ebner self-calibration parameters. In
order to separate between global and strip-dependent shift
parameters, the two flight lines per flight day were considered
as one calibration block. Since the automatic AT was done for
the different flight lines separately, the two strips are tied
together only via the identical GCPs. For two of the normal-
angle flight days only one complete calibration strip was flown
due to weather conditions. These non-complete calibration
flights are not considered in the further processing. Overall,
eight complete calibration flight days are available for the
GPS/inertial normal-angle camera configuration.
4.1 Quality of GPS/inertial exterior orientations
As one first result the directly measured exterior orientations
from GPS/inertial are compared to the estimated values from
AT. The remaining differences serve as first indication of the
quality of GPS/inertial position and attitude determination. In
Figures 3-6 the particular position and attitude differences are
shown for the distinct camera stations from four representative
calibration flight lines handled as two calibration blocks flown
on January 29" and February 18. The statistical analysis from
all considered normal-angle calibration flight blocks is given in
Tables 2 and 3, respectively.
4: e eer EU
Aa TETE TE
075. à
os "v
E 3
E A $
e
$ oo es 7 s
2 =
© -0.25H + = =
amt :
e—e dEast e—e dOmega j|
-0.75 +-+ dNorth -0.03 +--+ dPhi
A ae 4-4 dKappa
=. El : la LT
1-007 Up GE UT à oe. 8. 10. 125 5
# Image of flight day
Figure 4. Attitude variations
(Flights 1+2, Jan 29).
# d o of flight day
Figure 3. Position variations
(Flights 142, Jan 29).
1.001 ——1L——1L——-1— 719—174 0.04 [——1——1————T————771——
Position [m]
Attitude [gon]
e
8
T
E e-edEast | F À
-0.75 +-+ dNorth -0.03r- 4
È 4-4 dVertical 3 E
«1.00 E——L LL EL DAL ce
0 2 4 6 8106 12/14 18 0 2 4 6 8
# Image of flight day
Figure 5. Position variations
(Flights 10+11, Feb 18).
ee dOmega
+--+ dPhi
4--A dKappa
10 128. 14 6
# Image of flight day
Figure 6. Attitude variations
(Flights 10+11, Feb 18).
As it can be seen from Table 2 the variations (STD) in the
GPS/inertial positions are quite consistent and mostly in the
range of 2dm which coincides with the typical GPS positioning
performance after differential phase processing. Nevertheless,
significant offsets or even drift effects are present, which can be
clearly seen from Figures 3 and 4. Additional systematic errors
are seen in attitude determination (Table 3). Although the mean
variation in o— and q-angle is within the 15” level for the
presented normal angle flights, the differences in « show larger
systematic effects for some calibration blocks resulting in large
STD values >0.01gon (>30”). As illustrated in Figure 6 for the
calibration block flown on February 18" the x-angle shows a
clear strip dependent systematic offset, which might be due to
non optimal system alignment or insufficiently damping of
inherent inertial error behaviour during GPS/inertial data
processing. Errors in the estimated gyro scale factor will result
in such a jump between two flight lines with opposite flight
directions. Any uncompensated error will deteriorate the quality
of object point determination after direct georeferencing.
Former airborne tests using the AEROcontrol IId system have
shown consistently higher quality results indicating that for
some of the mission days the accuracy potential is not fully
reached with the GPS/inertial data investigated so far. This fact
reconfirms the high demands for careful processing of
GPS/inertial data and well defined test flight conditions
especially when data are used in system calibration and for later
production projects.