The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part Bl. Beijing 2008 
1306 
No of 
GCPs 
No of Tie 
Points 
S 0 
9 
9 
0.393 
9 
34 
0.368 
9 
149 
0.284 
9 
494 
0.217 
34 
34 
0.303 
34 
194 
0.272 
34 
494 
0.227 
Table 5. Reference standard deviation using UCL along track 
Kepler model based on the collinearity equations and 
coplanarity equation with various number of tie 
points. 
7. DEM GENERATION 
7.1. Introduction. 
This section reports on the generation of a DEM using the 
Erdas Photogrammetry Suite (EPS) version 9.2. The DEM 
generation has a pixel size of 1 Om and is based on area-based 
matching which is also called signal based matching. The 
EPS is used because the UCL sensor model has yet to be 
linked to stereo matching software. 
In order to check the accuracy of the produced DEM the 
DEM provided by the PI (15m pixel size) is used as a 
reference. The area covered is a hilly area where the 
difference in heights within the whole area is 120m. 
7.2. DEM quality. 
For the CARTOSAT data the following strategy of the 
software is used as it produced quite good results. This 
strategy is the following: • 
• Search Size: 19 x 3 
• Correlation Size: 7x7 
• Coefficient Limit: 0.80 
7.3. DEM accuracy. 
The accuracy of the DEM is described in table 6. The 
produced DEM has not been edited for blunders (manually or 
automatically) and it is evaluated as it is extracted by EPS. 
No of 
GCPs 
MIN 
(m) 
MAX 
(m) 
Absolute 
linear error 
LE90(m) 
1 
-39.20 
138.62 
13.30 
2 
-41.87 
231.73 
13.45 
3 
-40.26 
220.94 
4.13 
4 
-36.51 
222.85 
4.07 
5 
-32.64 
89.12 
4.04 
6 
-32.63 
104.10 
4.10 
9 
-31.23 
89.07 
3.88 
34 
-39.25 
87.80 
3.97 
Table 6. Accuracy of DEM with different number of GCPs 
used in the orientation 
From the results in table 6 it seems that at least 3 GCPs should 
be involved in the orientation in order to reach Absolute Linear 
error LE90, close to 4m. However because of the min and max 
high errors it is needed to edit the produced DEM manually or 
automatically. 
As a general conclusion, in the DEM generation from Cartosat 
data when RPCs model is used, 4 GCPs should be measured in 
order to reach the accuracy as described in table 6 which is not 
improved, in reality, if the number of GCPs is increased. 
8. COPLANARITY IN DEM GENERATION AS A 
GEOMETRIC CONSTRAIN 
As this UCL sensor model is not linked with DEM generation 
software the importance of the coplanarity equation as a 
geometric constrain after the correlation process is tested in the 
similar correlation procedure of auto- tie point generation in 
EPS software. Under Auto-Tie generation procedure 4096 tie 
points are produced which are checked for their accuracy 
manually, where the wrong correlated points are found. On the 
other hand the coplanarity equation is applied on all tie points 
and it is found that all these ‘wrong’ points give values much 
higher than the expected value (close to zero). The above 
procedure denotes the important role that the coplanarity 
equation could have in the DEM generation procedure, which 
should be approved in the near future. 
9. CONCLUSIONS 
This paper has described the testing of the UCL Kepler Along 
Track Sensor model and the RPCs model on Cartosat data. 
Also a DEM is generated using Erdas Photogrammetry Suite 
software. The results that are introduced within the paper leads 
us to the following conclusions: 
• RPCs model reaches close to pixel accuracy when at least 
4 GCPs are used. 
• UCL model sufficient (subpixel) accuracy is achieved 
even in case of five GCPs, better than the RPCs model in 
all cases. 
• For DEM generation it shown again, as in case of SPOT5- 
HRS, (Michalis and Dowman, 2004) that the use of the 
along track stereo sensors is a very promising for DEM 
generation, as the image matching quality and the 
achieved accuracy is very high. 
However the most important achievement in this study is the 
development of the coplanarity equation. This has the 
following benefits: 
• When the coplanarity equation is involved in the solution 
one more equation per point (GCP, Tie Point) is provided, 
giving the opportunity to have a solution with less GCPs, 
with sub-pixel accuracy. 
• The coplanarity equation increases the precision of the 
solution significantly. 
• The coplanarity equation is a robust and rigorous equation 
which can be used easily and straightforwardly as a