The rest of the exercise is devoted to geometric correction and 
re-sampling. In order to illustrate the main principles, an 
artificial image with 10 lines by 10 pixels is used initially. A 
ground control point file is established from given pairs of 
control points by GCP. (Keyboard mode is used for file- as 
well as map-coordinates). The transformation equation is 
determined by COORDN. LRECTIFY is used to produce one 
image-file for each of the three resampling methods Nearest 
Neighbour (N), Bilinear Interpolation (B) and Cubic 
Convolution (C) as shown in Figure 4b. The students also 
perform, based on the transformation equation, a "manual 
resampling-calculation" for some selected pixels. In this 
connection it can be mentioned that the way COORDN shows 
the coeffisients of the transformation equation is somewhat 
confusing. 
  
Figure 4b. Transformation of a synthetic image with different 
resampling algorithms. 
An artificially distorted version of an uncorrected Landsat MSS 
image (denoted "AR=6:1" in Figure 4c) is the basis for the part 
of the exercise which deals with capture of ground control 
points using the digitizing tablet. The result of this rectification 
is denoted "GCP" in Figure 4c. 
In the next part of the exercise, GCPs are measured on a map 
and entered into ERDAS with the keyboard. The map is a large 
scale topographic map (scale 1:5000) and the image is a SPOT 
level 1B panchromatic subscene. The result of this rectification 
process is denotetd "SPOT-MAP" in Figure 4c. When rectifying 
the image, the students are asked to define corners 
corresponding to one of the 1:5000 map sheets covering the 
actual area. 
In the last part of the exercise the goal is to perform a "system 
correction" of the same distorted Landsat MSS image as is 
mentioned above. Based on the "aspect ratio", the earth rotation 
and the angle between the orbital- and the equatorial plane, the 
transformation equation is calculated. The result of the 
subsequent resampling is denoted "MOD" in Figure 4c. 
385 
  
Figure 4c. Image rectification with various methods. 
SATK2. Exercise 2. Determination of terrain elevation 
from a single oblique SPOT scene. 
The main emphasis in this exercise is rather on the under- 
standing of the satellite image geometry than on the use of the 
ERDAS system. By measuring distances (expressed in pixels) 
by CURSES in a level 1A (not geometrically corrected) pan- 
chromatic SPOT subscene, the students shall calculate the 
terrain elevation for selected pixels in the image from known 
true horisontal distances. 
In addition to the subscene in which the measurements are 
carried out, another obliqe SPOT subcene with a different view 
angle covering the same area is also available. Together these 
two subscenes form a stereo model By intelligent use of 
READ, an image which can be stereoscopically viewed through 
red/green glasses (the anaglyph method) is produced on the 
screen. As this paper is in B/W, the effect cannot be 
demonstrated here, but on the display screen of the ERDAS- 
system, the effect is striking. Since the two images in the 
stereo-model are taken from different orbits, the orientations of 
the two images are slightly different when they are registered 
to a common reference system, as shown in Figure 5. 
  
Figure 5. Stereo model from two oblique panchromatic SPOT 
subscenes.