IN-FLIGHT CAMERA CALIBRATION FOR DIRECT GEOREFERENCING
E. Honkavaara
Finnish Geodetic Institute, Department of Remote Sensing and Photogrammetry, Geodeetinrinne 2, 02430 Masala,
Finland — eija.honkavaara @ fgi.fi
KEY WORDS: Accuracy, Calibration, Camera, Georeferencing, GPS/INS, Orientation
ABSTRACT:
Direct georeferencing is increasingly applied in connection of the photogrammetric film cameras. The prerequisite for the use of
this technique is an airborne system calibration. The calibration issue was investigated by analysing 10 calibration blocks obtained
with four GPS/IMU/optics-combinations. The earlier studies already showed that in-flight interior orientation determination was
highly relevant with the data. The objectives of this study were to investigate the further extension of the system calibration model
with the typical image deformation parameters and to evaluate efficient calibration routines for daily use. The physical image
deformation parameters appeared to be quite problematic due to their high correlations with other parameters. Mathematical image
deformation model of Ebner appeared to have a consistent behaviour; with the examined data the maximum corrections for image
coordinates were 4-12 pum depending on the optics. It appeared that the accuracy of the direct orientation observations was the
limiting quality factor. A minimal block geometry with a single bi-directional flight line and no ground control points (GCPs)
allowed the determination of the principal point and boresight unknowns; the calibration with this minimal block structure was
clearly advantageous. Single GCP improved the reliability, but in order to obtain good accuracy, several GCPs were needed.
1. INTRODUCTION
‘The future of airborne image acquisition is direct georefe-
rencing (DG) by combining direct image orientation with
digital imaging. DG is also becoming an integral part of the
image production based on film cameras. A central step in DG
is the in-flight system calibration, which can be made by using
permanent test-fields or by using a calibration block photograp-
hed in the mapping area. Different approach is self-calibration,
which can be made if integrated sensor orientation approach is
taken (see Heipke et al. 2001).
The boresight parameters are the central parameters in the
system calibration. Practical results of DG have shown that the
sole boresight calibration is not sufficient in applications with
higher accuracy requirements. As the conclusion of the OEEPE
test Heipke et al. (2001) recommended to include the interior
orientation parameters to the system calibration whenever
possible. Wegmann (2002) and Jacobsen (2003) have made
similar conclusions based on the OEEPE data. Results of
Cramer et al. (2001, 2002) have also shown the importance of
the extension of the collinearity model by additional parame-
ters. Honkavaara et al. (2003) reported results of 11 practical
calibration blocks with 4 GPS/IMU/optics combinations. The
determination of the interior orientations was a necessity; with
all the systems appeared a significant (20-40 jum) correction in
the direction perpendicular to the flying direction (yO) and with
one optics appeared a significant (25-35 um) correction in the
principal distance. Due to the systematic errors and
instabilities, Hansa Luftbild German Air Surveys performs
routinely the airborne system calibration in the mapping area
(Schroth 2003). Also ICC of Barcelona determines the system
calibration frequently e.g. using a minimal block configuration
(Baron et al. 2003).
National Land Survey of Finland (NLS) photographed a large
number of calibration blocks with four GPS/IMU/optics-
166
combinations in summer 2002. The first results of the datasets
were reported by Honkavaara et al. (2003). In this study the
NLS calibration data has been further processed. The first
objective was to study the expansion of the above-mentioned
parameters with other standard additional parameters
modelling image deformations. The second objective concerned
cost effective daily calibration routines.
2. CALIBRATION
2.1 Mathematical model of calibration
2.1.1 Grouping of the parameters. The in-flight
calibration is performed by using standard bundle block
adjustment techniques. Honkavaara et al. (2003) used the
following grouping of feasible parameters:
1. Boresight misalignments (do, do, dk)
2. Flying direction dependent corrections
a. Constant position shifts dependent on flying direction
(e.g. lever arm (dX, dY, dZ)icver)
b. Camera interior orientation (dc, dx0, dy0)
3. Other image deformations: the available parameters model
physical distortions (e.g. radial and tangential distortions)
or try to compensate systematic image deformations using
mathematical polynomials.
4. Datum transformation: a full or a partial 7-parameter
similarity transformation (dX, dY, dZ, a, p, y, scale)auun.
Modes of image and GPS/IMU-position and -attitude observa-
tions used in this study are discussed below. Burman (2000),
Cramer et al. (2002), and Wegmann (2002) have reported
related work.
Int
the
mc
on
(1¢
M
kn
re
pe
m
ar
m
m
