Michael Breuer
2.6 Roll distortion
The roll distortion yields to offsets across the flight track. The pixels are shifted, due to variations of the roll angle. The
amount of one offset per pixel remains not constant even for a constant roll angle. It grows with an increasing scan
angle of the line-of-sight (Albertz et al., 1995). An oscillating roll angle leads to a wavelike capturing of the terrain
surface.
2.7 Pitch distortion
Movements in pitch can be serious if they lead to gaps in the image data. This can arise due to sudden forward or
backward shocks of the sensor that cannot be compensated by a stabilized platform. Then there are some areas of the
terrain that are depicted repeatedly whereas other parts remain unscanned.
2.8 Flying height distortion
The image data are affected by changes in scale if the flying height varies. This can result in underscans. Therefore
variations of the flying height should be kept always within a certain tolerance because the scan rate and the along track
velocity are balanced on the background of a chosen scale (res. spatial resolution) (Albertz et al., 1995).
2.9 Relief displacement
Elevation differences of the terrain result in displacements that increase with growing scan angle. The displacements
occur in only a single direction (i.e. along the scan line, res. across the flight track). Its effect raises with increasing
object height, with increasing distance from the nadir line, and with decreasing flying height (Avery and Berlin, 1992;
Buiten, 1993).
2.10 Earth curvature
Scanning systems are not affected by earth curvature because of the low flying height of the aircraft (Richards, 1986).
3 AN APPROACH TO MEET THE USER'S NEEDS
The user needs the hyperspectral data as orthoimages, i.e. in the scale and the geometric properties of a map or another
orthoimage. This should be achieved easily with high precision in a considerable period of time (Toutin, 1995).
Therefore reliable software which allows for high quality radiometric and geometric correction must be readily
available. The radiometric correction however will not be discussed in this paper but should be considered as an integral
part of hyperspectral data preprocessing. To get precise
results in geometric correction based on a holistical ;
concept one agrees today that this can be done only by Hoa
using a parametric model (Pope and Scarpace, 2000; I IE
E
Schläpfer et al. 1998; Jacobs, 1988). Nevertheless the
non-parametric models do not seem to lose in value (Ji
Orientation
and Jensen, 2000; DeJong et al, 1999; McGwire, Bua mh
1998). But these approaches can only achieve good TTR ET Orthoimade
. * . *. t a e.
results under special conditions (e.g. flat terrain, good inen zz
flight conditions, etc). Another reason may be the fact Orientation image Control
that non-parametric methods are available in most Fov. IFOV, ——4 | Geometric | 4—— Points
. scan rate, a.Incr. :
remote sensing software systems whereas the access to Correction Ii x y
software based on parametric modeling is still limited.
> Ground Control
| Points
4 INITIAL DATA
Hyperspectral
The initial starting point when someone gets a raw NR,
hyperspectral data set differs from each project to Z— MÀ
LE
another. For this reason all possible initial auxiliary ==
informations that might come with the hyperspectral
data are investigated in the following. Figure 2 gives eu 2
an overview. ;
Figure 2. Different types of initial Data
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B3. Amsterdam 2000. 95