International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part Bl. Istanbul 2004 
  
  
stereo coverage being mostly cloud covered in the left half of 
the image. Even though the test site is cloud free, DEM could 
not be generated. The DEM generated by automated Saphire 
software also contains large anomalies due to the presence of 
dense urban features in the test site where manual interaction is 
unavoidable, which are inadequate in the current software 
version. Hence this result has not been included. 
6. CONCLUSIONS 
6.1 Observations on Results: The results are available 
essentially for Montmirail data set only. Vertical bias of 30 m 
for Saphire DEM and 100 m for PCI DEM are observed while 
comparing them with the reference BDTOPO DEM in 10 m 
resolution. See Figure-1. However there is no vertical bias 
observed while comparing the Saphire DEM generated with Sm 
grid size with the reference Laser DEM. Also the results are 
improved while making the comparison with the Laser DEM. It 
is to be noted that the Laser DEM depicts the top canopy of the 
surface as seen by HRS-1 and 2 unlike the BDTOPO DEM, 
which represent the ground (bare earth) heights. Algorithmic 
and implementation differences do exist between Saphire and 
PCI Geomatica software packages. Figure-3 depicts the 
difference map of Laser DEM and Saphire DEM in four classes 
for a good reproduction capability in black and white print. 
Closer scrutiny reveals that there is lack of Interest points 
automatically identified during the stereo image matching step 
at those locations where there are large differences observed. It 
is also observed that these locations occur in high-slope areas, 
where occlusions might occur due to the different view angles 
of HRS-1 and 2 resulting in poor automatic digital correlation 
during stereo image matching. These results have been obtained 
without the use of any real ground control points. The corner 
and centre pixel coordinates supplied by SPOT which are 
accurate to about 50 m (25 m as claimed in Vadon, 2003) only 
have been used, both for Saphire and PCI Geomatica. 
6.2 Conclusions: The quality of SPOT5 HRS along-track stereo 
data sets is very good for the generation of digital elevation 
models. The quality of the height data is commensurate with the 
quality of the input data available. It is observed that without 
using any control points, Saphire, the modified IRS-1C/ID 
stereo photogrammetric software could achieve an accuracy of 
20 mat 89% (10 m at 75% and 5 m at 58%) of the DEM 
points in the case of Montmirail. See Table-3. The quality of 
corner coordinates provided with the HRS scenes (after revision 
for Montmirail scene) is sufficient to achieve a planimetric 
accuracy of about 10 m as against the 25 m localization 
precision claimed in (Vadon, 2003). The DIMAP format used to 
deliver the HRS image data sets is very user friendly. Saphire 
software provides results closer to the reference DEM, as 
compared to Geomatica. 
The along-track stereo data acquisition of SPOT 5 provides 
near-real-time stereo as well as superior modelling possibility 
with a single set of orbital parameters, as against the across- 
track stereo options available with SPOT 1-4 or IRS-IC/ID. 
The results could be improved if manual interaction for 
interactively increasing the density of conjugate points and 
identifying break-lines are introduced in Saphire. Availability of 
a near-nadir image would also help reduce the occlusions that 
occur in high slope areas. 
A detailed comparative evaluation of each internationally 
available DEM generation software for SPOT 5 on all test sites 
covering multiple terrain types would be ideal. 
Figure-3: Difference Map of Laser DEM & Saphire DEM 
   
   
   
t à 25 d 
vadam 
# js + > 
Feet u 
    
Legend: 
Color Difference 
Less Than 10 m 
11 mto 20 m 
2] mto 30 m 
More Than 30 m 
   
7. ACKNOWLEDGEMENTS 
Authors gratefully acknowledge useful discussions held with 
Mr. Kannan V Iyer and Ms. Deepa Padmanabhan of SAC. 
Authors thank Dr. KL. Majumder, Deputy Director, SAC for 
permitting to take up this work at SAC and Dr. KN Shankara, 
Director, SAC for permitting the first author to be a co- 
investigator of the CNES-ISPRS SPOTS HRS Study Team. The 
first author thanks the HRS Study Team Coordinators Mr. 
Baudoin Alain of CNES, France and Prof. Manfred Schroeder 
of DLR, Germany for admitting him to be part of the team. He 
also thanks Mr. Marc Bernard of SPOT Image, France, Mr. 
Sylvain Airault of IGN, France and Prof. Clive Fraser, 
University of Melbourne, Australia for providing the data sets. 
8. REFERENCES 
Baudoin, A., Schroeder, M., Valorge, C., Bernard, M., and 
Rudowski, V. (2003), The HRS-SAP initiative: A scientific 
assessment of the High Resolution Stereoscopic instrument on 
board of SPOT 5 by ISPRS investigators, High resolution 
mapping from space 2003, Hannover, 6-8 October, 2003. 
Fraser, Clive (2003), Personal Communication, July 17, 2003 
PCI Geomatics (2001), “OrthoEngine SE 3D” Reference 
Manual, October 19, 2001. 
Srivastava, P.K., Ramakrishnan, R., Nandakumar, R., Gopala 
Krishna, B., and Majumder, K.L. (1996) Cartographic potential 
of IRS-1C data products, ISPRS 18^ Congress, Vienna, Austria, 
July 1996, IntArPRS, Vol. 31 Part B, Commission IV, WG 
IV/3, pp IV-823. 
Vadon, Hélène (2003) 3D Navigation over merged 
panchromatic-multispectral high resolution SPOTS images, 
International Archives of the Photogrammetry, Remote Sensing 
and Spatial Information Sciences, Vol. XXXIV-5/WI0, 
International Workshop on Visualization and Animation of 
reality based 3D models, ISPRS Commission V, WG 6, Tarasp, 
Switzerland, 24-28 Feb. 2003. 
  
  
  
     
  
  
  
  
  
   
   
   
  
  
  
  
  
  
   
   
  
  
   
   
  
  
   
  
   
   
  
   
  
   
  
  
   
  
   
  
  
  
   
    
  
   
  
   
  
  
  
  
  
   
  
   
   
   
     
  
    
  
    
KE 
AB 
Pro 
DE 
ran 
are 
infi 
pre 
bea 
pre 
(ar 
cha 
On 
dat 
test 
No 
dat: 
CD 
fori 
test 
sofi 
zon 
Bar 
Ori 
Inci