International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part BI. Istanbul 2004
collected in a variety of scan directions that included: north-
south, south-north, east-west and west-east scan directions.
The footprints of the imagery used in the initial geometric
camera calibration of OV-3 can be seen in Figure 4 shown
below.
All of the image measurements were collected by auto
correlation methods. There were a total of 3,875 ground
control points from the geometrie calibration range. A total
of 33,093 image rays were observed on the satellite images.
This means that each ground control point was observed on
an average of 8.5 images. Some of the ground control points
at the center of the geometric calibration range were
observed on all of the images. Figure 4 shows the location of
the ground control points generated from the geometric
calibration range shown as blue triangles.
Figure 4: Imagery Footprints and Control Points Used in the
Initial Geometric Camera Calibration of OV-3
An advantage to using controlled aerial photographs is that
a large number of control points can be used for the
characterization and calibration of the camera. Figures 5-8
show the image residuals from the panchromatic arrays.
Each dot in the graph represents an observation of a ground
control point. It is clear that the ground control densely
covers the entire arrays.
To show apparent distortions at the focal plane, an
adjustment solving for only the focal length and camera
alignment parameters was performed. The resulting image
residuals show the remaining optical and focal plane
distortions. The distortion is divided into two separate
directions: line and sample residuals. The distortion in the
line direction is sometimes called the camera smile
distortion because of the characteristic shape. This
distortion is primarily due to radial distortion of the optical
system and is estimated in the design process of the camera.
The predicted smile distortion from the camera design
process agrees with the on-orbit observed distortion.
The distortion in the sample direction is parallel to the
direction of the arrays. Distortions in this direction can be
thought of as scale distortions along the arrays. The major
contributor to this distortion is due to radial distortion of
the telescope and is estimated in the camera design process.
The predicted scale distortion from the camera design
process agreed with the on-orbit observed distortion.
20
10
[pixels]
Aline (
-10
0 2000 3000 6000 8000
detector
Figure 5: Image Residuals in Line Direction Showing Smile
Distortion Before Camera Calibration
20
8 Asample
...
[pixel]
=
-20
0 2000 4000 6000 8000
detector
Figure 6: Image Residuals in Sample Direction Showing
Scale Distortion Before Camera Calibration
20
10
ij
à Aline 0
[pix
zio
=20
0 2000 4000 6000 8000
detector
Figure 7: Image Residuals in Line Direction
After Camera Calibration
20
10
Æ Asumple 0 pm
pixels)
—10
—20
0 2000 4000 6000 8000
detector
Figure 8: Image Residuals in Sample Direction
After Camera Calibration
Figures 5 and 6 show the observed distortion before the
initial geometric camera calibration. Figures 7 and 8 show
the image residuals after the initial calibration all of the
systematic distortion has been modelled in the calibration.
Internation
The stand
adjustmen
posteriori
micromete
at the end
estimate fc
seconds.
The comp
the initial
data used
section.
UA
The valida
performar
geolocatic
evaluation
control da
This secti
accuracy à
comparisc
The chec
assessmer
points in «
over a per
the calibr:
Generally,
a small di
mono ima,
is not usu
stereo pair
said to be
the figures
5.1 Mon
The mon:
adjusted
points an
the syste
measured
elevation
longitude
latitude a
point are
such as C
directions
and west
this asses
The mono
and a me:
CE histog
this CE sa
random v
which me
of the sy
somewhat
do not ad
The 90-tl
ranked ac
meters an
