Seijing 2008 
The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part Bl. Beijing 2008 
1345 
:y GPS) a new 
the CartoSAT- 
igital elevation 
with accuracy 
ence DEM has 
JTM 31N, and 
ady; 
adiometry; 
at; undertaking 
s with varying 
k points; 
the DEM; and 
Number of GCP 
the four images of the study area. The corresponding entropy 
values are presented in Table 5. The entropy value of all the four 
images is around 5.6 bits/pixel; this value indicates good contrast 
of the image, which is essential for employing auto-correlation 
techniques while generating DEM. 
Image 
Entropy 
Left-aft 
5.51 
Left-fore 
5.45 
Right-aft 
5.73 
Right-fore 
5.62 
Table 5. Entropy Values 
4. DEM PROCESSING 
the positional 
Figure 2. Number of GCP’s and Resultant RMSE 
4.1 DEM Extraction 
With 4 control points, the rms error is around 5 pixels. With a 
minimum of 6 or 7 control points, the rms is of the order of 2 - 3 
pixels, and it was possible to correct the imagery to the specified 
accuracy. 
he coefficients 
wo stereo-pairs, 
liai adjustment, 
in both Easting 
ectification and 
e been ignored. 
30th the stereo 
nber of control 
its - 0 
its - 5 
its - 10 
its - 15 
3.2 Analysis of 2-D Geometry 
Using the individual GCP falling in each image a transformation 
model is generated to rectify both the images, and the 
corresponding rms values are presented in the Table 3. The high 
rms value in y-direction of the fore images indicate the relief 
displacement in the along track direction due to sensor orientation. 
Description 
RMS (X) 
RMS (Y) 
Left-fore 
3.00 
7.26 
Left-aft 
3.59 
2.24 
Right-fore 
1.19 
6.29 
Right-aft 
1.26 
1.95 
Table 3: RMS Values of the 2-D Transformation Model 
e Table 2 and 
dodel 
Model 
III 
IV 
.56 
5.35 
:.14 
3.09 
.41 
6.56 
1.69 
1.75 
1.93 
1.18 
¡.22 
10.74 
.84 
2.98 
'.62 
11.20 
:.52 
4.59 
1.82 
1.76 
¡.60 
5.63 
Image-to-Image tie down models are also developed to evaluate 
the relative geometry of fore-aft images and left-right images. The 
corresponding rms values are presented in the Table 4. The high 
rms value of the models is due to different viewing angles. 
Description 
RMS (X) 
RMS (Y) 
Left - Aft/fore 
1.00 
8.85 
Right- Aft/fore 
0.59 
7.57 
Left-fore / Right-fore 
5.48 
0.71 
Left-aft / Right-aft 
7.12 
1.19 
Table 4. RMS Values of the 2-D Relative Geometry 
3.3 Analysis of Radiometry 
CartoSAT-1 acquires data with radiometric resolution of 10-bit. 
The entropy, a measure of amount of information content 
measured in units of number of bits per pixel, is estimated for all 
The DEM has been extracted from both the stereo-pairs, using the 
LPS software and presented in Figure 3. The posting interval for 
the regular DEM is 10m, and generated for the full overlap of the 
image pairs. However, the assessment presented here is only for 
the area for which a reference DEM is available (Figure 1), i.e. 
approximately 50% of the acquired image area. During the 
creation of the DEM, all GCP’s and tie points are used as seed 
vertices. This input enhances the relative position of the DEM 
generated. Since the main objective of the study is generating a 
DEM and assess its quality, no manual methods are employed to 
enhance the quality of the DEM. 
Figure 3. DEM Mosaic of the study area 
Based on the internal software quality reporting, around 84% of 
vertices are classified as “Excellent” or “Good” or “Fair” and this 
value indicates the generated DEM is of good quality in nature. 
The results of the internal quality are presented in Table 6.