The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part Bl. Beijing 2008 
Aspect 
[“Л*«« 
~j North 
i North East 
■Bl East 
¡ SouthEast 
— South 
Ш Southwest 
ГП West 
■Ц Northwest 
Figure 6a. Aspect map 
Figure 6b. Aspect Categories 
Drainage network has been extracted from standard procedures 
(Jenson and Domingue 1988) from DEM having resolutions 5m 
and 10m. The procedure includes generation of depression less 
DEM, computing flow directions, flow accumulation and delta 
value. The effect of DEM resolution on the drainage order was 
compared by automatically deriving the drainage order from the 
extracted drainages. Drainage network was generated from 5 & 
10 m DEMs by taking similar area specific (5000 m 2 ) 
accumulation threshold for initiating the drainage on both DEM. 
The results are shown in figures 7a and 7b and Table 4. 
8. RESULTS AND DISCUSSION 
The RPCs provides very good solution of ground coordinates 
with the help of a few GCPs. The accuracies can be improved 
using increasing the number of GCPs and order of the 
polynomial. Table 2 show the improvement in the RMSE at 
both Control and check points for different combinations of 
number of GCPs and polynomial order. When no GCP was 
used, the Check point accuracy at longitude, latitude and height 
were 22.6m, 98.4m and 22.9m respectively. This provides an 
estimation of the location accuracies using only RPCs provided 
along with data. By using a single GCP, RMS Error was 
reduced to the order of 10m in both planimetry and elevation. 
Drainage Order 
/\y 1st order 
2nd order 
/\/ 3rd order 
/\y 4th order 
/ / 5th order 
Fig. 7a Drainage Network extracted from 5m DEM 
Drainage Order 
/\y 1st order 
2nd order 
/\y 3rd order 
/\y 4th order 
/V 5th order 
Fig. 7b Drainage Network extracted from 10m DEM 
Stream 
Order 
Stream Number Derived from 
5 m DEM 
10 m DEM 
1 
198 
190 
2 
34 
30 
3 
8 
7 
4 
2 
2 
5 
1 
1 
Table 4. Comparison of Drainage Network Derived from 5 m 
and 10m DEM. 
Polynomial order 1 was used for RPC refinement process using 
5, 8 and 10 GCPs. Similarly polynomial order 2 was used for 8 
and 10 GCPs. The model accuracies provide very good results 
by changing the polynomial order from 1 to 2. RMS Error at 
check points were also reduced while using polynomial order 
and also for the increase in GCP number. Effect of increase in 
GCPs for polynomial order 1 from 5 to 10 shows an oscillatory 
behaviour. The results are positive for order two when GCP 
were increased from 8 to 10, where total RMSE at Check points 
is reduced. When 10 GCPs and polynomial order 2 were used 
for refinement of RPCs, RMSE at GCPs was minimum. The 
same effect was not exactly reflected in Check points probably 
because of the exact point identification on hilly terrain.