all 
  
g- 
le 
19 
er 
cy 
  
  
Pasajes data set is rather poor compared to that of Dubai. 
  
  
  
  
  
  
— 
Variables Result: s 
t site 
fes Channels Pixel es) d 
(scale factors) size pie groun 
BW (¢1) | FW (¢1) | 13.5 m| 0.94 12.7 m 
Pasajes 
HR (°1) | FW (°3) | 45m| 0.87 3.9m 
Dubai-A | HR (e1) [FW (¢3) | 4.5m| 0.31 1.4m 
  
  
  
  
  
  
  
  
Tabelle 5: Influence of channel combinations on relative 
orientation via interactive stereo measurement 
In order to take full advantage of the digital type of MOMS- 
02 data an automatic matching procedure - at least in 
order to derive conjugate points as input for the computa- 
tion of an exterior orientation or a Digital Elevation Model 
(DEM) - is ingenious instead of interactive, manual measu- 
rements. The used matching method - including a point 
search based on a modified Fórstner-operator with a follo- 
wing least-squares adjustment - is described in more detail 
by Wang (1994). 
Table 6 elucidates that the combination of the FW- and 
BW-looking channels results in a significantly better rela- 
tive orientation than the multi-resolution approaches. With 
the latter methods information is lost due to the interpola- 
ted scaling of one of the channels - either by enlarging one 
oblique channel to the size of the HR-channel or by redu- 
cing the HR-channel to the size of an oblique channel. 
  
  
  
  
  
  
  
  
  
  
  
  
Variables Result: s 
Teststto Channels Pixel ol round 
(scale factors) size p 9 
BW (e1) | FW (e1) |13.5m| 0.11 1.5m 
HR (*'/3) | FW («1) | 13.5 m| 0.16 22m 
HR (e1) | FW (e3) | 4.5m| 0.21 0.9m 
Dubai-A 
HR (*1/3) | FW (e1) |13.5m| 0.08 1.1m 
  
  
  
  
Tabelle 6: Influence of channel combinations on relative 
orientation via automatic matching 
By comparing corresponding values of tables 5 and 6 it is 
evident that automatic matching procedures are superior 
to interactive measurements in terms of the number of 
derived conjugate points and the achieved accuracy (at 
least by factor 1.5). Nevertheless, interactive measure- 
ments are still a necessary task for determining the abso- 
lute orientation. 
3.3.4 Transformation models: As exspected, applying 
the strict transformation model of the collinearity equation 
within the scan line yields slightly better results than using 
just a polynomial rectification (table 7). But if no stereo 
745 
imagery is available, polynomials have to be taken into 
account. This is the case for the test site Harare, which 
also shows relatively flat terrain and where we have found 
planimetric residuals in the range of 5 to 6 m whereby the 
degree of the applied polynomials was without importance. 
  
  
  
  
  
  
  
  
  
  
  
  
; 4 - : -resi | 
Variable: Pixel Mea- | Result: ICP-residuals 
: sure- 
Transformation | sizes 
ment sx sy sz 
Collinearity 4.5m/ |Stereo | 4.2m| 3.6m| 4.5m 
13.5 m 
Polynomial 45m |Mono | 4.8m| 57m - 
Constants: Test site = Dubai-A / Reference = 1:10 000 / No. of GCPs 
(ICPs) = 25 (30) 
  
  
  
Table 7: Influence of different transformation models 
3.3.5 Summary: Like for standard evaluations of aerial 
photography the object point accuracies obtained from 
MOMS-02 imagery depend on a variety of factors. To 
examine the potential of the sensor itself the best values 
can be inspected which have been obtained under "opti- 
mal” conditions. 
Comparing these values with the demands as pointed out 
in section 3.1 it is evident that the horizontal accuracy of 
MOMS-02 data is suitable for target scales of 1 : 25 000 or 
even larger. Because no suitable reference material was 
available for test sites with varying terrain, only the proba- 
bly too optimistic values for flat urban areas of Dubai and 
Harare can be taken into consideration. Nevertheless, 
even these values cannot always satisfy the high demands 
so that for some applications the quality standards have to 
be reduced. 
It was shown that using multi-resolution imagery for man- 
ual stereo measurements can improve the accuracy while 
automatic matching methods prefer images of the same 
resolution (i.e., FW- and BW-channels). A matching solu- 
tion which incorporates all three channels simultaneously 
could not be realized yet. 
Further tests will be carried out during the MOMS-02/P 
mission for test sites in developed countries with highly 
accurate reference material. 
4 INFORMATION CONTENT ASPECTS 
The suitability of a system for topographical mapping with 
respect to the information content depends on a variety of 
ground-side factors like relief and contrast of the topogra- 
phy and of scanner-side factors like the spatial and spec- 
tral resolution or the capability to identify three- 
dimensional objects. Due to this bundle of factors and the 
different demands for different applications no general 
measures but only a variety of recommendations for a suit- 
able input source can be defined, e.g. by Konecny (1995). 
Practical investigations have been carried out by visually 
comparing radiometrically enhanced MOMS-02 imagery 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B4. Vienna 1996