ıl 2004
Sensor
:elona,
Direct sensor orientation based on GPS network solutions
Helge Wegmann®, Christian Heipke®, Karsten Jacobsen”
“Ingenieurbüro Wegmann,
Leo-Rosenblatt-Weg 6, 30453 Hannover, Germany
info(@ib-wegmann de
t . ^ = . . : ; x
"Institute for Photogrammetry and GeoInformation, University of Hannover,
Nienburger Str. 1, 30167 Hannover, Germany
heipke, jacobsen/@ipi.uni-hannover.de
Commission I, WG V
KEY WORDS: GPS, IMU, Direct image orientation
ABSTRACT:
Direct sensor orientation, i.e. the determination of exterior orientation based on GPS and inertial measurements without the need for
photogrammetric tie points, has gained considerable popularity over the last years. One pre-condition for direct sensor orientation is a
correct sensor and system calibration. The calibration can only be carried out by a combination of a photogrammetric solution and a
GPS/inertial solution, which is equivalent to the concept of integrated sensor orientation.
In the work carried out so far, GPS has been identified as the most critical part in terms of achievable accuracy. Strategies for
improving differential GPS results are available for terrestrial applications, but have not yet been used in direct and integrated sensor
orientation. One of these solutions consists in using a network of reference stations rather than a single station only.
In this paper we present our work on direct sensor orientation using a GPS network. After describing the related mathematical
models we report the results of an experimental test. The test data were drawn from the OEEPE test “Integrated sensor orientation”.
The results show, that while for many applications a network may not be necessary in case of short baselines and good GPS data, it
still improves the accuracy of direct sensor orientation to some degree. More important is the fact, that our approach is able to detect
gross errors in the reference station data and therefore has the potential to improve also the reliability of the results.
Î Introduction
The development of sensors and related processing methods
for the economic, accurate and reliable collection of 3D geo-
spatial information has been a topic of intense
photogrammetric research in recent years. Besides the new
digital aerial cameras, sensors for directly determining the
exterior orientation based on GPS and inertial measurement
units (IMU) have found large interest. The integration of GPS
and the inertial measurement system has been strongly
promoted at the University of Calgary for a long time already
(Schwarz et al. 1984; 1993) and in the meantime a series of
tests and pilot projects has been conducted demonstrating the
potential of these methods (e.g. Skaloud 1999, Cramer 1999,
Heipke et al. 2002b). The current situation is that GPS has
been identified as the most critical part in terms of achievable
accuracy.
When using GPS and IMU observations to determine the
exterior orientation of photogrammetric images, one can
differentiate between the so called integrated sensor
orientation, in which all available information including tie
points is processed simultaneously to achieve highest
accuracy, and direct sensor orientation, in which the exterior
orientation is computed based on GPS/IMU observations
only, and object space coordinates are derived in a separate
step (Heipke et al, 20022).
Direct sensor orientation consists of three steps - a sensor
calibration step to be carried in advance, as well as GPS/IMU
pre-processing and the determination of the exterior
orientation for the actual mission.
During sensor calibration the parameters describing each
sensor individually and those describing the relationship
between the different sensors need to be determined. These
153
parameters include the interior orientation of the camera, the
angular differences between the IMU and the image
coordinate system (boresight misalignment), and additional
parameters modelling e.g. GPS errors. The system calibration
parameters, and in particular the boresight misalignment, can
only be determined by comparing a photogrammetric solution
based on image coordinates of ground control and tie points
with the pre-processed GPS/IMU results.
GPS/IMU pre-processing includes the transformation of the
raw GPS signal and IMU measurements into trajectories in
object space for the camera projection centres and roll, pitch
and yaw values at a high frequency (usually 50 — 200 Hz).
The common method of integrating GPS and IMU
observations is via Kalman filtering. It provides the optimum
estimation of the system based on all past and present
information (for details see Grewal et al. 2001).
The determination of the exterior orientation parameters then
consists in applying the sensor calibration to the pre-
processed GPS/IMU values. One of the applications of direct
sensor orientation is 3D point determination via spatial
forward intersection based on the refined exterior orientation
parameters.
In this paper, we deal with direct sensor orientation and
present as a novel aspect of our work a GPS network solution
for photogrammetric point determination. In the next section
we describe our model for sensor calibration based on pre-
processed GPS/IMU data. We do not deal with GPS/IMU
pre-processing itself. We then introduce our test data which
are drawn from the OEEPE test “Integrated Sensor
Orientation". We derive sensor calibration parameters, and
compute 3D coordinates of independent check points which
we compare to the known values, both with single reference
stations, and using the GPS network. Finally, we comment
our results and draw some conclusions for future work.
