indicate the resi 
X; , = 
‘ ‘Nadir 
j JNadir 
(z x y - ej) 
(3) 
where Va<i) =i-coordinate of true nadir point, 1 ^^=]-coordinate of true nadir point, ^elevation of the test point at 
position (i,j), and z x v =altitude of ER-2 for the current x and y.To allow for a more precise selection of the corre 
sponding surface point, the DEM oversampled to a grid size of 6 m is used. To define the intersection point on the 
surface, the normalized dot product of x] and X, . is calculated. The vector x] ; best representing X is that for 
which the dot product DP has the smallest difference from 1. To represent the pixel size dependent on the topog 
raphy, the four comer points ( 6 m grid size) of every pixel are separately calculated. To prevent changes to the 
radiometric characteristics, the original value is selected by an improved extraction algorithm during the resam 
pling to 18m grid size, thereby eliminating the need to interpolate the values. 
2.4. Discussion of the Results 
There are currently no well-established methods of quantitatively assessing the success of a geocoding process. 
Visual inspection provides useful information, but cannot be used to intercompare methods. Statistical results 
based on residual calculation of single ground control points allow only a local error assessment 
For the discussion, the Rigi scene is selected because of the more challenging topography. Figure 4 shows the 
greyscale representation of the bands 13, 18, and 28 of the geocoded image overlaid by the scanned forested areas 
(solid bright line) 
Figure - 4: Greyscale representation of the southern part of the scene Rigi with bands 13,18, and 28 of the geo 
coded image overlaid by the scanned forest (solid line). 
S 
(pararti 
S 
(par am 
S 
(non-pan 
Table -1: 
No systematic 
was geocoded 
3 ) presents the 
solution. 
The whole prc 
language, Resi 
3 - ATMOSP1 
To utilize the ( 
reflectance us 
begins with tl 
number for a' 
Figure - 5: 
Through AVI 
In general, the result shows a good correspondence between the geocoded image and the map for all existing 
topographical locations. A few locations show minor miscorrespondence. These problems are almost always 
restricted to single pixels and no general tendency can be recognized. The errors may result from changes in real 
ity between the time the aerial photographs (which are the basis for the topographic maps) were acquired (1987) 
and the AVIRIS data acquisition, and on the fact that maps are the result of a generalization while AVIRIS dis 
plays every occurrence within its resolution characteristics. The additional verification calculates the average 
deviation of the geocoded image compared with the forest and lake map for checkpoints. Lines 1 and 2 of Table 1 
114