The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part Bl. Beijing 2008 
1118 
RPC model refinement (Tao et. al. 2004). Ritesh et. al (2006) 
evaluated various algorithms for generation of drainage network 
from Cartosat-1 DEM. Li et. al. (2007) studied the 3D geo 
positioning accuracies by integrating IKONOS and QuickBird 
stereo images using rational polynomial coefficients. 
On the basis of these demonstrative studies, an attempt has been 
made to find out the planimetric and elevation accuracies of the 
DEM and Ortho image from Cartosat-1 stereo data. The RMS 
error was computed at GCPs and Check points by varying the 
number of GCPs and polynomial order for refinement of RPCs. 
Terrain parameters such as slope, aspect, and drainage network 
has been extracted from DEM. 
2. OBJECTIVES 
The prime objective of the study is to generate DEM using 
Cartosat-1 data and derive terrain parameters. The detailed 
objectives of the study are as follows. 
• Experimental design and execution of DGPS survey 
and establishment of Ground Control Points (GCPs). 
• To study the effect of number of GCPs and order of 
polynomial for RPC refinement for the generation of 
DEM 
• Generation of DEM and Ortho image from Cartosat-1 
stereo data 
• Effect of DEM resolution / accuracy on ortho image 
generation 
• To retrieve the terrain parameters such as slope, 
aspect and drainage network. Comparative evaluation 
of the drainage order derived from different DEM 
resolutions. 
3. STUDY AREA 
Part of Alwar District, Rajasthan state, India was taken up for 
the study. The area falls between 27°30’ and 27° 50’ in latitude 
and 76°30’ and 76°50’ in longitude respectively. The study area 
has heterogeneous terrain with elevation ranging from 200 to 
600 meters in WGS 84 datum approximately. The major 
cultural features include Arravali range, Alwar city, Shyamaka 
Reserve Forest etc. 
4. DATA USED 
Cartosat-1 stereo data acquired on 4 th November, 2005 was 
used for the study. The details of the data are given in Table 1. 
Sensor 
Path 
Row 
Orbit 
No 
Sun 
Elevation 
Sun 
Azimuth 
PAN AFT/ 
PAN FORE 
0523 
0270 
2713 
44.5740° 
159.6077° 
Table 1. Details of the Cartosat-1 data used 
DGPS observations at Control points, IGS data observations, 
ancillary files and GPS satellite precise orbit file. 
5. RATIONAL FUNCTIONS 
A sensor model relates 3D object point positions to their 
corresponding image positions through the collinearity 
condition equations. The RFM relates object space coordinates 
to the image space coordinates. The image pixel coordinates (x, 
y) are expressed as ratios of polynomials of ground coordinates 
(X, Y, Z). Generally they are represented as third order 
polynomials. Ratios have a forward form: 
x = P! (X,Y,Z) / P 2 (X,Y,Z) (1) 
y - P 3 (X,Y,Z) / P 4 (X,Y,Z) 
This equation is called upward RF. Usually RF model is 
generated based on a rigorous sensor model. 
Pi ( i =1,2,3 and 4) are the polynomial functions with the 
following general form: 
Pi = an + a 2 i X + a 3i Y + a 4 ;Z + a 3 jXY + a&XZ + a 7 iYZ + 
a 8i X + a9i Y 2 + a 10 i Z 2 + am XYZ + a 32 i X Y + am 
XZ 2 +a 14i YX 2 + a 15i YZ 2 +a 16i ZX 2 + a 17i Z Y 2 
+ a 18i X 3 + a 19i Y 3 + a 20 i Z 3 (2) 
The order of the terms is trivial and may vary in different 
literature. The number of coefficients in the polynomial can be 
reduced gradually by applying different conditions (P 2 =P 4 ) & 
(P 2 =P 4 =1). The first coefficient in the denominator is known 
(a 2 =a 4 =l). A minimum of 7, 19, and 39 GCPs are required to 
resolve the first, second and third-order RFM having 14, 38 and 
78 number of RPCs respectively. 
Refinement with polynomials corrects the remaining error and 
refines the mathematical solution. Values between 0 and 3 can 
be selected to correct the original rational function model. The 
0 th order results in a simple shift to both image x and y 
coordinates. The 1st order is an affine transformation. The 2 nd 
order results in a second order transformation; the 3 rd order a 
third order transformation. In general, a 0 th or 1 st polynomial 
order is sufficient to reduce error not addressed by the rational 
function model. A higher order polynomial requires more GCPs. 
6. METHODOLOGY 
A standard methodology has been adopted for the generation of 
DEM, ortho Image and terrain parameter retrieval and is shown 
in Figure 1. It comprises of reconnaissance survey and DGPS 
survey, establishment of reference station by network 
adjustment with IGS stations, establishment of a sub reference 
station with respect to reference station, establishment of GCPs 
with respect to sub-reference station, stereo data analysis using 
RPCs and updation of RPCs using GCPs, generation of DEM, 
Ortho image, retrieval terrain parameter from DEM, accuracy 
assessment of DEM and ortho image, generation of DEM by 
refining Rational Polynomial Coefficients (RPCs) with different 
number and distribution of GCPs, validating the DEM, 
generating the DEM, Ortho images at the best check point 
RMSE and extraction of terrain parameters. 
7. DATA ANALYSIS 
Establishment of GCPs 
One of the most important parameter for DEM generation and 
validation is to establish the coordinates of the GCP’s. The GCP 
coordinates are required for geo-referencing the satellite 
imagery and also in the bundle adjustment for generation of 
DEM. In addition to this, it is also required to have some 
GCP’s for validation of DEM. These were established using