'ests,
STOW
el. It
ikko
e for
logy
errer
es by
orest
g for
agen,
nsing
logy
1etric
1 3-
an in
arch
mote
ng of
, vol.
ellite
1suu,
ta —
r the
, PP-
Karsten Jacobsen
POTENTIAL AND LIMITATION OF DIRECT SENSOR ORIENTATION
Karsten Jacobsen
Institute for Photogrammetry and Surveying Engineering
University of Hannover, Germany
karsten @ipi.uni-hannover.de
IC-11 Sensor orientation
KEY WORDS: INS, GPS, Integration, Sensors, Orientation
ABSTRACT
Based on combined use of an inertial measurement unit (IMU) and GPS, sensor orientations can be determined directly.
Problems are included in the determination of the misalignment between the camera and the IMU, the long and short
term stability of the physical connection of the IMU to the camera lens cone, the separation of the GPS-shifts from the
misalignment and the reliability of the sensor orientations. In addition, the focal length has to be checked under flight
conditions if different flying heights above ground are used.
The misalignment of the IMU from the camera axis can only be determined with sufficient accuracy by a reference
flight strip flown in opposite directions and controlled with a standard bundle block adjustment. But also under this
condition, some limitations are caused by the very high correlation of the sensor orientations, especially for normal or
even smaller angle images.
For an operational project the sensor orientations of approximately 3000 photos, flown at different days, have been
determined by IMU + GPS. The misalignment was determined every day by means of a small reference strip, flown
before and after the images of the main block.
Limited, but still significant differences of the misalignment occurred. A check of the orientations against ground
coordinates with 252 photos measured manually and 460 photos used in an automatic aero triangulation has shown
sufficient results. Nevertheless larger discrepancies in the heading (kappa) have been shown — this also may be caused
by the limited stability of the IMU mount in relation to the lens cone of the used LMK 2000. The lens cone is fixed only
by one pin to the camera body. Also the other cameras do have similar mechanical problems.
In the case of Kodak DCS images, supported by IMU and GPS, a component calibration (misalignment and GPS shift)
was not possible because of the very small view angle. After a general shift of the GPS-data, these values have been
fixed to allow a sufficient system calibration. For the resulting ground accuracy a separation of the different error
sources was not necessary.
A general problem of the directly determined sensor orientations is the missing reliability. Even simple errors like in the
image numbers can only be seen by a model setup or the matching of created orthophotos. The model setup often is
disturbed by not acceptable y-parallaxes. This can be solved by a combined bundle block adjustment. Based on the
precise approximate image orientations an automatic aero triangulation can be handled much more easy, so the
additional effort is limited.
1 INTRODUCTION
The image orientation for photogrammetric data acquisition is a very important, but time consuming procedure. With a
block adjustment, the number of required control points has been
reduced drastically. The next step was the combined bundle block
adjustment with projection
center coordinates
determined by relative
kinematic GPS-positioning.
In the case of a real block
structure, attitude data are
not required, they can be
determined by the combined
block adjustment with GPS- Figure 2: block configuration of
data (figure 1), if at least 2 linear objects - IMU-data not
parallel flight strips are required
available. The flight strips
Figure 1: block configuration for combined
adjustment with GPS - crossing flight strips
every 20 — 30 base length or control points
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B3. Amsterdam 2000. 429