4l 2004
ticated
:|lite"s
es the
targets
ye are
| more
nately
e final
1ave a
m the
If the
cover
ıld be
dle of
'y are
tually
rin a
justed
dark
oured
fixed
nding
to be
ns for
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004
2.4 Linear motion system
The satellite is moved in two mutually perpendicular directions
using a servo-controlled linear motion system. The positioning
and repeatability of the servo system needs not to be the most
accurate because the satellite position and orientation are
determined solely using photogrammetry. The satellite cannot
be allowed to move when images are captured, so the stability
of the servo system is very important. All dead motion and
mechanical gaps should be eliminated in order to get reliable
final results.
A calibrated servo system, with an accuracy of a few
millimetres, makes it possible to automatically measure the
satellite targets from their predicted positions in images.
3. CALIBRATION
The measurement system requires careful system calibration to
get final measurements in a correctly scaled, orthogonal 3-D
coordinate system. Calibration of the whole system consists of
the camera calibration, laser calibration, and the calibration of
the servo system.
3.1 Camera calibration
The outer and inner camera systems can be calibrated
simultaneously using a free-network bundle adjustment
(Haggrén and Heikkild, 1989). Approximate values for the
unknown parameters can be obtained using the projective
adjustment method developed by I. Niini (2000). Normally, the
absolute scale of the coordinate system is fixed by observing
the end-points of a known scale bar and by using the known
distance as a constraint in the adjustment. More information
about the calibration method can be found in (Niini, 2002).
Here, due to the scale difference between the two systems (the
inner system size is less than one tenth of the outer system)
only one scale bar or distance cannot be used easily. A
calibration tool containing targets for both systems is
appropriate. À scale bar with four collinear targets can be used.
It contains two larger targets in the ends of the bar, and two
smaller ones close to each others in the middle of the bar. The
first large target and the first small target are made
recognisable by adding a smaller index target close to them.
All targets are white circles in a black background, and they
can be searched and measured automatically from the images.
Since the mutual distances of the calibration targets are known
precisely, e.g., from a coordinate measurement machine, the
bar can be used to get the systems into a common coordinate
system and scale. During this part of calibration, the satellite is
kept at a fixed position, preferably in the middle of the
measurement area. The targets in the calibration tool are then
observed from the images so that both systems can see their
corresponding targets simultaneously. This is made in several
different positions and orientations. The image coordinates of
the satellite cover targets are also measured in this fixed
position, and taken in the adjustment, so that the satellite base
position also becomes calibrated and can be used later as a
reference.
485
To strengthen the calibration of the outer system, additional
observations (say 50-100) of the calibration bar are made well
distributed in the entire measurement volume. The calibration
of the inner space is also strengthened using another small
calibration tool, scale of which may be known but needs not to
be known. It is then only constrained to have a constant, but
unknown length.
The overall calibration can be further strengthened with a
precise calibration plane that contains several unknown
circular targets for both systems. This plane can be placed in
several different positions and orientations below the satellite,
and both systems can observe their dedicated targets.
Corresponding points can be determined using the epipolar
geometry obtained from previously determined calibration
parameters. The image observations of the plane points enter
the adjustment, and the corresponding 3-D points are
constrained to lie on the same plane. The calibration also gives
the relative positions of the plane targets. Then this calibration
plane can be used to recover an accidentally lost calibration
between the two systems.
3.2 Satellite reference position
The calibration volume has to include the area where the
satellite targets can move. Additional calibration points are
generated easily by moving the satellite over the entire linear
motion area, simultaneously registering the image coordinates
of the satellite targets. These image coordinates are added into
the adjustment, along with the constraints that keep the
relative geometry of the satellite targets constant in all
measured satellite positions. Thus, the relative geometry of the
satellite targets will be adjusted using image observations over
the entire calibration space.
3.3 Laser calibration
The laser light planes are calibrated in a simple manner after
the satellite camera system has been calibrated. It is only
required that the visible 3-D laser lines are measured from at
least two distinct object planes. The laser planes are then
obtained from these 3-D lines. The intersection of the two laser
planes is the 3-D line that is projected down as the laser cross
in the object surface.
After the laser calibration, it is possible to calculate where the
laser cross locates in any known plane in the satellite
coordinate system. Using the similarity transformation between
the satellite base position and current position, the position of
the laser cross is also known in the outer coordinate system.
Optionally, the information from the laser planes can be used
to constrain the measurements, because the visible laser line
segments have to lie on the same known plane of the
corresponding laser.
3.4 Servo system calibration
The servo system has its own coordinate system and scale. The
axes may not be exactly straight, or may not have the same
scale. There may also be minor discontinuities in certain
positions along the axes, especially when the axes are made
from several racks. The calibration of the servo system is based
on a polynomial model for the servo coordinates.