ımber of cycle slips, or
satellite geometry, was
quent kinematic OTF
ncies for widelaning, a
'stablished, which then
1cy processing. During
ellites were visible, the
/cle slip detection. The
y interpolated to obtain
e time.
| sensors, a comparison
derived from GPS/INS
stment could be made.
5 have RMS values of
. Since the separation
rers was within 10 km,
ue to the small number
configuration. In order
tion satellites had to be
evation mask as low as
:eously tracked by both
50 60
)
. height
Camera Position
SM Mae
Li
*
" *
3 A Y
?
3"
azimuth
'amera Orientation
rameters are depicted in
roll, pitch and azimuth.
rm due to the fact that
short time periods when
1a 1996
flying over the test field. The errors in azimuth are randomly
distributed with standard deviations (STD) of 0.015? over the whole
mission. This implies that the dominant systematic errors in
INS-derived attitude have been almost completely eliminated by
frequent GPS updates. These updates should be theoretically more
effective in roll and pitch, see Skaloud 1995 for a theoretical
explanation, but this does not seem to be the case for either pitch or
roll. Both are at a level of about 0.03? (STD), and show a small drift
in pitch and some rather systematic features in roll. This could be due
to several reasons: unmodelled errors due to aircraft dynamics; time
synchronization errors between GPS/INS and camera; large gyro
noise due to aircraft vibration; residual errors in the
photogrammetrically derived parameters of exterior orientation due
to the strong correlation between position and attitude. Most likely,
a mixture of all these sources contributed to the larger than expected
attitude errors. Further analysis is planned to better explain some of
the unexpected features.
5. DIRECT GEOREFERENCING
Target points in the imagery can be directly georeferenced, once all
unknowns in the modified collinearity Equation 2 have been
determined. In this case, the parameters of interior orientation as
well as the offsets between on board sensors are known from
calibration and the parameters of exterior orientation are resolved via
GPS/INS integration. To evaluate the overall performance of direct
georeferencing, coordinates of 50 control points were recomputed by
means of Equation 2 and compared to their reference values.
Practically, the computation was done by using the calibration
parameters of interior orientation to correct the measured image
coordinates and running the bundle adjustment with no ground
control and fixed parameters of exterior orientation as derived from
GPS/INS. The three dimensional position residuals on all 50 check
points are depicted in Figure 7. The error distribution has a standard
deviation of 30 em horizontally and 50 cm vertically. Additionally,
the points are shifted towards South by a mean value of 35 cm.
Since the camera position and orientation were considered as known
parameters in the adjustment, errors in their determination are
directly propagated into the derived ground coordinates. Considering
the errors shown in Figures 5 and 6 at a flight height of 900 m, the
resulting position errors on the ground could even be larger. This
might indicate that the reference values derived by inverse
photogrammetry are not as reliable as indicated by their standard
deviations, due to the high correlations between them.
GCP ERROR (m)
1.5.9 19 17.21 25 29 33 37 41 45 49
POINT ID
€ east ii north à height
Figure 7: Check Point Residuals-Direct Georeferencing
6. CONCLUSIONS
The performance of an airborne data acquisition system for precise
attitude and position determination in support of airborne remote
sensing has been evaluated by means of aerotriangulation. Due to
poor satellite geometry, the positioning errors along the aircraft
trajectory were larger than usual, and reached several decimetres.
The attitude errors from the integrated GPS/INS showed some
unexpected features. While the azimuth error behaves randomly, as
expected, with a standard deviation of 0.01 degree, the errors in pitch
and roll are larger, about 0.03 degree, and seem to be systematic in
nature. They need to be further investigated. Direct image
georeferencing at 1:6000 using the measured GPS/INS attitude was
better than 1 metre (absolute), when compared to 50 presurveyed
control points. This accuracy is sufficient for many of the intended
mapping applications.
ACKNOWLEDGEMENTS
Flight time and expertise was provided by Rheinbraun
AG-Department of Photogrammetry, Cologne. This support is
gratefully acknowledged. The authors wish to thank Mr. Werner
Schneider for his effort in performing high quality photogrammetric
measurements and Mr. Darren Cosandier for providing the bundle
adjustment software.
REFERENCES
Ackermann, F. (1995) Sensor and Data Integration - The New
Challenge, in Colomina/Navarro (eds.) Integrated Sensor
Orientation, Wichmann, Heidelberg, Germany.
Cosandier, D., T.A. Ivanco, M.A. Chapman and M. Dylke
(1994) The Integration of a Digital Elevation Model in
casi Image Geocorrection, Presented at the First
International Airborne Remote Sensing Conference and
Exhibition, Strasbourg, France, September 11-15.
Friess, P. (1991) GPS-supported Aerial Triangulation - Empirical
Results, in Linkwitz/Hangleiter (eds.) High Precision
Navigation 91, Dümmler, Bonn, Germany.
Hofmann, O. (1988) A Digital Three Line Stereo Scanner System,
Archives of the Int. Society of Photgrammetry and
Remote Sensing, Commission II, WGII/6, pp. 206-213.
Hofmann, O., Kaltenecker, A., and Müller, F. (1993) Das
flugzeuggestützte, digitale Dreizeilenaufnahme und
Auswertesystem DPA - erste Erprobungsergebnisse, in
Fritsch/Hobbie (eds. Photogrammetric Week 93,
Wichmann, Karlsruhe, Germany.
Schwarz, K.P., M.A. Chapman, M.E. Cannon and P. Gong
(1993) An Integrated INS/GPS Approach to the
Georeferencing of Remotely — Sensed Data,
Photogrammetric Engineering & Remote Sensing, Vol.
59, No. 11, pp. 1667-1674.
Schwarz, K.P. and M. Wei (1994) ENSU 623 Lecture Notes,
Department of Geomatics Engineering, The University of
Calgary.
129
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B3. Vienna 1996
