time tags and exposure time data for each image lines. No
divergence between changes of exposure times, as indicated
in the housekeeping data, and drops of DN values could be
found.
1.90 -2.34 —
: observed 08
3 1.88 N -2.36 0
= : "d
S 1.86 -2.38 2
Z d
Q 1.84 -2.40 =
20 commanded 3
= 1.32 -2.42 9,
1.80 nw
0 2 4 6 8 12x10
line
Figure 8: In order to verify proper changes in exposure
time during imaging, pixel greyvalues are plotted along a
sky profile in line direction. Indeed, each change in the
integration time (solid line) is clearly visible as a decrease
in brightness by the corresponding magnitude in this
logarithmic plot. The prominent feature at line =7,400 is an
artifact.
4. GEOMETRY
4.1. Camera Coordinate System
For a geometric analysis of all HRSC channels, we first
thoroughly analyzed the setup of the experiment. Initially,
we assumed that the origin of the camera coordinate system
(Fig. 9) was located in the center of the rotational axis and
that the camera was somewhat offset from the origin of this
system, and in addition, tilted towards the observed scenery.
A À S di ah of"
LA w
Q4, [^
first pixel
mi U UL Yi
5F 4F 3F 2F1N 2A 3A 4A 5A
Figure 9: Camera coordinate system of HRSC as used
in this paper. Flight direction is u, the optical axis is w,
and v is parallel to the orientation of the CCD lines (5F
stero 2, 4F red, 3F photometry 2, 2F blue, 1N nadir, 2A
green, 3A photometry 1, 4A infrared, 5A stereo 1).
We collected large numbers of tiepoints in the nadir and the
two stereo channels and performed a least squares adjustment
to determine the offset and the angle. While the offset was
found to be small and insignificant for the analysis to
follow (0.028 + 0.0023mm), the angle was found to be
within -0.09039° + 0.0007°.
4.2. Camera Metric Properties
HRSC is a metric camera (Fig. 10). This implies that during
operation from orbit, all pixels —even though on different
352
CCD lines— will have the same areal coverage on the
planet’s surface, assuming that the surface is planar. As a
consequence of camera metric properties, each pixel's field-
of-view will differ according to the pixel's position on the
focal plane. It was one of the goals of ET3 to verify this
property of the camera, which is an important hardware
requirement for photogrammetric processing of imagery.
P
stereo angle
image space stereo angle
flight altitude
focal length of the lens
size of a sensor element
on focal plate
hf-p
size of a sensor element
on ground
u
>; Stereo w
Figure 10: Metric properties of HRSC: Relationship
between sensor element field of view and ground pixel
size
TIT [Ws T=.
In order to verify this characteristics, we took advantage of
the camera rotation during the ET3 experiment. Unlike
during a flight above a planar surface, ground pixel sizes for
sensor elements of the stereo and nadir channel,
respectively, will differ substantially (Fig. 11) in this
geometry. In the following, we concentrate on the stereo
and nadir channels. Howewer, the same considerations
apply to the other channels as well.
Y stereo angle
p = size ofa sensor element À
A
Maur T1 25 1FOVnadir Y N
stereo= | 05" IFOVstereo / x
IFOV = instant field of view
Figure 11: Relationship between ground pixel sizes
in CCD channels for a rotating HRSC.
From Figure 11, we determined that the field of view of
stereo pixels is about 10% less than that of the nadir pixels.
This results in darker images from the stereo channels
relative to the nadir channel. This offset in the brightness
level of the off-nadir channels, including the photometry
channels, is clearly observed in the histograms of the
images (cf. Fig. 3).
Likewise, due to smaller fields of view of pixels located on
the stereo CCD line (Fig. 11), an object will cover a larger
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B4. Vienna 1996
nur
CCI
{ El
Pre pede > (CS 65 med
"T
Hen
"elo
(Fig
ellip
para
We
this
of ai
Witl
elon
to n
agre
