The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part Bl. Beijing 2008 
1370 
system level correction, Level 1) (Krishnaswamy and plan and better than 2.5m in elevation (Gachet and Fave, 2006). 
Kalyanaraman, 2002; National Remote Sensing Agency, 2007). Figure 2 shows the GCPs’ distribution on the study area. 
2.3 Standard accuracy of references DTMs 
The Root Mean Square Error (RMSE) is a measure often used 
to assess the accuracy of elevation data and is defined as 
follows: 
RMSE = 
(1) 
3. DATA PROCESSING 
The Cartosat-1 data processing was done using ENVI 4.3®. A 
commercial off-the-shelf software was selected for 
investigating the capabilities and limits of the system in the 
DTM’s generation using standard image processing tools, so 
from the point of view of a typical remote sensing user. The 
data processing involved the following aspects: i) pre 
processing, ii) optimization of the DTM’s extraction procedure 
and, iii) analysis of the influence of GCPs in the modelled 
DTMs. 
where: 
AZj are the elevation residuals (i.e., the differences of the 
elevation measures with respect to reference data) 
n is the number of measures 
Another statistics often used to evaluate the overall accuracy of 
elevation data at a fixed confidence level (a) is the Linear Error 
(LEa). LEa performs a comparison in the elevation data 
towards reference measures: an LE90=2.5m implies that 90% of 
the measures to be tested vary from the reference measures by 
2.5m or less. 
Accuracies of references DTMs is shown as follows: 
■ MNT DBTOPO® DTM: RMSE=lm in elevation 
over the whole France, as from PI IGN specifications; 
■ SRTM DTM: absolute LE90=6.2m in elevation for 
Eurasia (Rodriguez et al., 2006). 
2.4 Standard accuracy of references GCPs 
The GCPs supplied by the PI IGN were derived from the 
French BD ORTHO® and had an accuracy better than 1,5m in 
(a) 
3.1 Data pre-processing 
All the dataset were first converted into the UTM-WGS84 
F31N reference system. The MNT DBTOPO® DTM was 
resampled from its original 19m x 26m cell resolution to 25m x 
25m and the SRTM DTM from its original 60m x 90m cell 
resolution to 90m x 90m. 
3.2 Optimization of the DTM’s extraction procedure 
Before generating the DTMs from the Cartosat-1 images it was 
investigated the influence of the parameters involved in the 
generation process (i.e., number of tie points used, search 
window and moving window sizes, correlation coefficient and 
terrain detail). A sensitivity analysis led to the final optimal 
configuration as shown in Table 1. 
Sixty tie points were used in all the tests. They have been 
automatically detected using a regular grid scheme covering the 
entire images, obtaining an Y parallax of 1.27 pixel 
(corresponding to 3.18m). By increasing the number of tie 
points, it was not observed any improvement in the Y parallax. 
Figure 1. Cartosat-1 stereo orthokit collected over Salon de Provence (France) on February 6, 2006. (a) 2.5m Aft panchromatic 
image, (b) 2.5m Fore panchromatic image.