164 
Fig. 3: Distortion of the scanning on the ground 
(detector exposure ti*e t B (( Scanning period t^ << angular eotion period T) 
a) under siaultaneous influence of roiling and pitching eotion ¿and «p 
b) under siiultaneous influence of yawing and rolling eotion ¿e and 
Since the orientation data are acquired with a considerably 
higher accuracy than the on-line values to be realized 
according to Section 2, it is also possible to carry out a 
highly accurate off-line post-rectification and a geometric 
allocation o-f the pixel to the terrain. If there exist 
topographic data o-f the recorded area or already a digital 
terrain model, the distortion (caused by differences in the 
terrain height) can be corrected. Otherwise a mean terrain 
height is defined. 
The output quantities x,y,z of a pixel in the terrain are 
functions of the coordinates x c , y c , z c (coordinates of the 
projection centre) and of the angles y 5, , to , ae (Fi g. 4). 
The accuracy requirements on the measured orientation 
parameters for the coordinate determination lie in the 
range of some dm for x, y, z and some mgon for V s , uj , <*£ . 
A pixel size of 10 jum, a flying height of 6 km, and a 
calibrated focal length of 10 cm require a < 0.3 m 
and a (J2 „ ** %!> mqon« 
Flight tests with push-broom scanners entailed the 
necessity of a series of radiometric preparations. It 
should be noted that a geometric on-line correction which 
shift the pixels in a row (e.g. on-line correction of the 
roll movement) is reasonable only up to a point. This 
derives from the fact that radiometric error influences in 
a scanner line (as for example the different sensitivity of 
the individual sensors) can only be corrected with 
difficulty, i.e. the individual correction values would 
have to be stored in the computer aboard the aircraft and