-46- 
A summary of the results published sofar is shown in table 1 
Author, Method of no. of no. of Channel 
Year restitution control check 
points points 
  
  
Colvo- nil 
  
coresses 
1973 similarity 
  
Báhr & affine 
  
polynomial 
2nd degree 
  
nil 
similarity 
affine 
20term polynomial 
Helmert transfor- 
mation 
2nd degree 
polynomial 
least squares pre- 
diction filtered 
5 =0,7, K 70,1 
Le. 
© 
Forrest polynomial 16 un- 
1975 knowns after correc- 
tion of syst. effets 
polynomial 21 un- 
knowns after correc- 
tion of syst.effects 
parametric 21 
unknowns 
Bern- 
stein «C: parametric 
1976 
  
Trinder parametric 
1976 12term polynomial 
affine 
  
Table 1 : Accuracy tests for Landsat MSS Imagery 
The results seem to indicate that the use of C.C.T!s combined with proper indentification of 
control points via digital image correlation offers the same promise as careful vusual selection 
of control combined with a high order polynomial fit (berstein, Wong). The early results are 
most likely caused by bad identification (and choice) of control. 
An accuracy test for satellite images other than of Landsat has been performed by Schuhr /77/ 
for Nimbus - III photos. 
The approach was to calculate the geographic coordinates of an image point using orbital data 
with the aid of the reconstucted subsattelite path and the scan angle by use of sperical trigo- 
nometric relations. The known ground coordinates could then be compared. While the image grid, 
superimposed onto the image at latitude and longitude crossings of even 19° by orbital tracking 
data in NASA's eledtron beam recorder was to be in error by * 60 km, the use of 4 control points 
for a transformation yielded standard deviations of + 7 km (or + O,13 mm at averag image Scale) 
for the check points afterwards.