375
The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B4. Beijing 2008
oriented perpendicular to each other. The effects of de
viations from these requirements are discussed in section 4.
2. For the identification of corresponding image points
effective matching techniques must be applied. This is
discussed in section 5.
For both aspects practical experiments have been carried out.
The test data have been provided by the Institute for Planetary
Research, German Aerospace Center (DLR). They have been
acquired in an aerial survey over an urban area with the HRSC-
AX in combination with high-precision DGPS and INS. This
camera system, developed for aircraft operation, has a scan line
of 12,000 pixels (Scholten et al., 2002). The area was flown
from south to north and from west to east (marked with arrows
in Figure 3) in an altitude of approx. 3650 m above the ground.
The ground sampling distance was approx. 15 cm. The data
were already corrected radiometrically, and the orientation
parameters were provided for each individual image line.
♦
ilifiBi
plane (Figure 3 lower images). Such a rectification process was
first established by Wewel et al. (1998).
In this example the WGS84 ellipsoid was selected as the
reference plane and the conformal UTM-projection with a
locally defined middle meridian was applied. Due to the local
UTM-projection and the small area, the influence of meridian
convergence can be ignored.
The deviations from the ideal configuration are significant. It is
evident from Figure 3, that the simple image geometry gets lost.
This means that the flight direction must not coincide any more
exactly with the direction, in which the objects are shown in
vertical parallel projection. This has a direct effect on the new
process for the generation of true orthoimages. The influences
of the aircraft motions can be estimated with the sensor rotation
angles roll, pitch and yaw. For the calculation the boresight
angles between the INS and camera system were used (see
Sandau, 2005). Figure 4 shows the calculated orientation angles
as well as the variations in both flight strips, containing 13000
image lines.
flight 1
Ï#*«
flight 2
N..
Figure 3. The pushbroom scanner data of a test area with the
flight directions shown on top, in the left image from south to
north, and in the right image from west to east. Below, the
flight motion corrected image data.
4. GEOMETRICAL INVESTIGATIONS
4.1 First Geometrical Correction
From the images in Figure 3 it is clear that some geometric
corrections are necessary before further processing. Due to the
significant distortions in the raw data through aircraft motion
and due to the deviations in position of the individual scan lines,
the identification of corresponding points and the combination
of the resulting ground positions could not be effective. Thus,
the geometric distortions by aircraft motion must be corrected
in a first step.
This correction can be achieved utilizing the internal and ex
ternal orientation data of the camera, which is available for any
scan line. The correction of the image data refer to a reference
Figure 4. Sensor movements roll, pitch and yaw for both flight
lines. Roll in the upper left, pitch in the upper right diagram.
Yaw for flight 1 in the lower left, for flight 2 in the lower right.
4.2 Effects of Orientation Angles
It is obvious from Figure 4 that the orientation angles change
from one image line to the next. In order to study the influences
of these changes the effects are discussed individually.
The pitch angle is of special interest. It indicates the deviation
out of the nadir direction, which has a constant effect on the
complete line. The vertical parallel projection is changed to an
oblique parallel projection. Thus, objects above the reference
plane are displaced in adjacent lines dependent from the height
of the objects. This relief displacement also yields an incorrect
ground coordinate of an image point. As an example for both
image strips: if an object height of 50 m is assumed and the
pitch angle is 0,4 degrees the displacement will be 2 pixels or
0,3 m.