M 
International Archives of the Photogrammetry, Remote Sensing 
and Spatial Information Sciences, Vol XXXV, Part B7. Istanbul 2004 
  
Figure 2a. Image ITAI-el 123941.1B 
(©2000-2004 IPT Informatica Per il Territorio srl. 
©2000-2004 ISI Imagesat International Ltd) 
2.3 The QuickBird imagery 
The orthorectification of a panchromatic QuickBird (Basic 
roduct) image using the terrain data provided by the EROS 
liscussed in the next sections. The dataset, 
libration, was acquired on Sept. 29, 
p 
stereo-pair will be c 
corrected by the internal ce 
2002 (10:16 a.m) with a significant off-nadir angle (20.8 
The image swath is 16.5x16.5 km, the nominal spatial 
degree). 
resolution 0,61m and a significant cloud coverage (16%) 
partially compromises its quality. 
2.4 Ground Control Points (GCPs) and CheckPoints (CPs) 
collection 
A large set of points (57) have been surveyed through a 
differential GPS survey (see figure 3). A subset of the whole 
dataset will be used for geocoding the QuickBird image 
(GroundControlPoints), the remaining for validating the final 
accuracy achieved (Indipendent CheckPoints). 
For the DSM extraction from the stereo pair the control points 
for geocoding each image have been derived from 1:5000 maps. 
In Figure 3 the red dots represent the distribution of GPS 
surveyed points, while the white rectangle illustrates 
approximately the limits of the EROS scenes. 
3. DSM EXTRACTION 
For both processes of DSM extraction and QuickBird image 
ication PCI Geomatica Orthoengine v9.1 software 
d. It implements a rigorous model for EROS and 
and a Rational Function Model for 
orthorectif 
has been use 
QuickBird images 
orthorectification of QuickBird images. 
In order to correct geometrical distorsions of the EROS images 
and produce a DSM, the software uses a 3D physical (rigorous) 
model based on collinearity equations and complanarity 
equations for the stereo-model computation. Though the EROS 
model is designed for 1A images, we imported 1B images as 
generic files with average parameters derived from metadata. 
The DSM extraction procedure allows the generation of relative 
or absolute DSM. The first does not require ground control 
collected from maps), but tie-pointing 
between the pair allows to obtain a model of elevation not tied 
to a cartographic reference system. Conversely, when an 
absolute (geocoded) elevation model had to be sought, a certain 
number of GCPs are needed. 
points (surveyed or 
Figure 2b. Image ITAT-el 123943.1B 
(©2000-2004 IPT Informatica Per il Territorio srl. 
(92000-2004 ISI Imagesat International Ltd) 
As the results of the present work have to be related to a 
national grid, the National Gauss-Boaga Reference System, the 
DSM has to be geocoded through ground true. 
First step of the DSM generation is the collection of GCPs over 
the two images. In this work 16 GCPs (collected from 1:5000 
maps) for each image have been used. 
  
Figure 3. QuickBird image illustrating the point dataset 
acquired with the DGPS survey and, in the white rectangle, the 
EROS image limits 
enly distributed over the scene and cover all of the 
elevation range of the relieves. The availability of tie points 
recognizable over both images is also useful for the DSM 
extraction procedure because they improve the matching 
(stereo-correlation) between the two images. 
3CPs the geometric model is computed for both 
arately with a rigorous and specific model and the 
ast square adjustment 
They are ev 
Through 
images sep 
stereo model geometry is refined with a le 
process. 
As soon as the matching is performed, a quasi-epipolar 
geometry for the stereo-pair has to be created. Using epipolar 
1050