Michael Breuer
measurement means are available there are always noise effects which have to be modeled for error estimation and
quality management. Often combined GPS/INS is not available because these systems are very expensive if they reach
high precision. Therefore low-cost navigation systems or simple GPS receivers providing a positioning accuracy of 30
meters or more are widely used. Such data serve only for approximation. Sometimes they turn out to be useless to
achieve the desired correction accuracy.
2 THE PROBLEM
No remote sensing acquisition system can render
a true geometric reproduction of the earth’s
surface without corrections (Binnenkade, 1993).
This is why methods for geometric processing of
the data are needed. The discussed problem in
this paper deals with the elimination of all dis-
turbing effects from the image data. These effects
are superimposed (see Fig. 1) and some of them
are correlated (Rose, 1984). The main effects are
briefly described in the following paragraphs.
flight track
A B C
2.1 Panoramic Distortion
Figure 1. Schematic diagram of distortions
The cause of panoramic distortion is the
cylindrical projection (see 1) across the flight track. For that reason the scale as well as the spatial resolution varies
within one scanline. From the nadir (the place of maximal resolution) it decreases towards the sides. Furthermore the
changes are different in x and y directions. A pixel is quadratic at the nadir and becomes a trapezoid at the sides. The
panoramic distortion is sometimes also referred as tangential scale distortion (Buiten, 1993, Richards, 1986).
2.2 Underscan
The aim of hyperspectral scanning is to capture the terrain surface without any gap. The rotation velocity of the mirror
(res. the scan rate) must be consistent with the velocity of the airplane along the flight track to meet this demand. It
seems imaginable to pitch the scan rate constantly with the along track velocity during a flight mission. However the
scan rate is constant in most practical applications. Therefore the pilot has to take care of keeping the along track
velocity constant so that no gap could occur in the image data. If the gaps could not be avoided the resulting effect is
referred to as an underscan (sometimes also referred as undersampling). In this case the information about the terrain is
lost (see Fig. 1, section C).
2.3 Overscan
An overscan (res. oversampling) is the opposite of an underscan. It occurs if the along track velocity of the airplane is
smaller than the allowed threshold. An overscan results in an increase of the along track scale of the image data. This is
less dangerous than an underscan because no information is lost (see Fig. 1, section A). In the ideal case all nadir pixels
are in touch (see Fig. 1, section B). But even then the margin pixels overlap especially if the airplane moves straight
ahead and if the field of view is large.
2.4 *S-shape" distortion
This distortion results from the fact that the airplane moves ahead during the time one line is captured. This results in an
S-shaped curvature (see Fig. 1). This effect is highly correlated with the yaw angle and the crab distortion. Each scan is
mostly approximated to be a straight line because the rotation velocity of the mirror should be high compared with the
along track velocity of the airplane. The "S-shape" distortion is sometimes also referred as sensor scan nonlinearities
(Richards, 1986).
2.5 Crab distortion
The crab distortion results in a skewed image and occurs when extreme crosswind is encountered during data
acquisition. In this case the axis of the airplane must be oriented slightly away from the flight axis to counteract the
wind (Lillesand and Kiefer, 1994).
94 International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B3. Amsterdam 2000.