International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B1. Istanbul 2004 
  
accurate aerotriangulation. AP coverages flown in the 
scope of IGN Topographic data base with a RMS error 
of 40 cm on the aerotriangulation are identified as a 
good candidate. 
* Altimetry should be controlled through an elevation 
model (including superstructures) with an accuracy 
better than 1 m. No such data existing over France, the 
APs were preprocessed and the reference DTM 
provided by IGN was refined to produce a terrain 
model suited to our objectives. In order to minimize 
residual errors, the site should not have extreme slopes 
and stable superstructures, forbiding montainous as 
well as over urbanized areas. 
2.1.3 Completeness of the measurement : In order to 
completely qualify and measure the SPOT 5 look angles, the 
reference data must cover the full swath which is, for HRG 
instruments, 60 km wide roughly in the east-west direction. 
The reference data, in effect, cover that swath but HRS 
images, twice as wide, could not be entirely covered once. 
Furthermore, when the HRGs improved steering mirrors allow 
to accurately overlap the reference site across track, no such 
device is available on HRS instruments reducing the 
opportunities to acquire the site during the in-flight 
commissioning. 
2.1.4 Quality of the statistics : the final estimation of the 
inner orientation of the instruments is obtained through the 
average of several measurements, say 200 per km of image for 
the modes with best resolution. Therfore a set of reference 
data 5 km wide with 80% overlap ensures a redundancy of 
1000 measurements per look angle. 
2.1.5 Accessibility of the site : the relative shortness of the 
in-flight commissioning phase together with the need to 
minimize the diachrony between the reference data and the 
SPOTS data implied the choice of a site with low cloud 
coverage statistics. 
2.2 The initial reference site 
A first attempt to apply an absolute method for inner 
orientation of SPOT images was led by L.FALALA and P. 
GIGORD in 1998 [Falala 1998]. The site of Yvetot in French 
Normandy region was selected without considering all the 
above characteristics and led to an unsatisfactory result 
although the main ideas of the method were implemented at 
that time. In order to improve the results, a site .more suited 
was then looked for taking into account most of the 
constraints. 
This drove to the choice of a site located in the South East of 
France between the Luberon chain and the Durance river, 
ensuring an acceptable meteorology. The landscape is mainly 
covered with scrubland and orchards and the topography is 
mostly made of gentle hills except in the eastern area. 
Within the scope of the regular BDTopo ® , IGN has flown a 
1/30000 scale APs coverage of the site in March 1997. From 
this coverage we acquired the 48 APs of a strip roughly 60 km 
wide spreading over the SPOT swath, the width of the strip 
being rougthly 8 km, together with the associated orientation 
models obtained through regular bundle adjustment ; these 
APs were scanned at a pixel size of 50 cm and further 
resampled to a 1.5 metre resolution. 
Following is a sample of the 1.5 metre resolution data. 
  
Along with the APs came a DTM from the IGN BDZ ; this 
DTM was improved to an estimated 1 metre RMS error over 
the site through the use of the scanned set of APs. 
The following Figure shows the refined DTM over the whole 
swath of the dataset. 
  
  
  
Digital Elevation Model 
The method was then experimented on SPOT limages with 
this set of data. 
2.3 The second generation site 
The time lap between the reference data and the SPOTS in- 
flight commissioning where most acquisitions occured in July 
2002 introduced both seasonal and long term diachrony in the 
process. It was therefore decided to acquire a complete new 
set of high resolution aerial imagery with the new generation 
4000x4000 digital camera PELICAN at a resolution of 50 cm 
and 80% overlap. The reference site is covered with 5 tracks 
of 200 images each, flown on July 2002. 
Following is a sample of the PELICAN data 
     
  
  
  
  
  
   
    
  
  
  
  
  
  
  
  
  
   
   
   
     
     
   
     
     
  
  
  
  
   
   
    
   
  
   
    
   
    
    
    
   
   
   
   
   
   
   
   
    
   
Internati 
  
The inn 
ensured 
Further « 
with RM 
Instead © 
on this 
orientatic 
through 
correspo 
elevation 
An aver: 
RMS er 
evolutior 
(typically 
distribute 
For data 
finally : 
compatit 
The evol 
justify to 
3.1 Desc 
The met 
based on 
the refer 
is associ 
and an ac 
geometri 
the accu 
initialize 
measurer 
measurer 
instrume 
Finally tj 
describec 
represent