International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004 
  
‚No.of | Mean | RMS | <2.0 | 2.0- | >5.0 | Max. 
Area Compared m) (m) m 5.0 m m m) 
Points ( : 
  
O+C+V+A | 29210494 | -1.21 | 4.80 | 60.7% | 16.8% | 21.3% | 424.2 
O+C+A 17,610,588 | -1.11 291. 177.096 | 13.096 |. 10.196 |: 358.9 
  
  
  
  
  
  
  
  
O+A 14,891,390 | -1.24 | 2.77 | 79.8% | 12.2% | 8.0% | 358.9 
Oo 11.795.795 .]..-1.00 1.28.1] 90.395 | 8.599 | 1.298 | 37.33 
  
Table 9. Accuracy measures and error classes for the triplet. O-Open areas; C-City àreas; V-Tree areas; A-Alpine areas. 
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
No.of | | Mean | RMS | <2.0 | 20- | >=50 | Max. 
Area Compared 
Points (m) (m) m 5.0 m m (m) 
O+C+V 20,336,024 | 0.45 4.78 |57.7%0121.3% | 209% | 125.2 
O+C 13,496,226 | -0.33 | 3.38 | 68.7% | 20.8% | 10.3% | 47.34 K 
o 3,969,734 -0.97 1.54 83% | 15.0% | 2.0% 39.4 
Table 10. Accuracy measures and error classes for the stereopair. O-Open areas; C-City areas; V-Tree areas. 
A 
5. CONCLUSIONS Fraser, C., Yamakawa, T., 2004. Insights into the affine model 
for high-resolution satellite sensor orientation. ISPRS Ir 
The presented results verify that 3D points can be determined Journal of Photogrammetry and Remote Sensing 58(5-6) (in m 
with a submeter accuracy which for the planimetry can be 0.5 print). cl 
m or less, if accurate GCPs are used. This was achieved also Fraser, C.S, Baltsavias, E., Grün, A., 2002. Processing of s 
with non-GPS GCPs and in mountainous areas with not very IKONOS images for submetre 3D positioning and building 
well defined GCPs. The number of GCPs can be small, their extraction. ISPRS Journal of Photogrammetry & Remote p 
accuracy being the main point. GCPs can cover only a portion Sensing 56 (3), 177 — 194. el 
of the image, although caution should be paid in areas with Grodecki, J., Dial, G., 2003. Block Adjustment of High- e 
large height differences. QB is not as linear as IKONOS and to Resolution Satellite Images Described by Rational bi 
achieve equal accuracy, needs an affine transformation after Functions. Photogrammetric Engineering and Remote 
employment of RPCs. The simple models (3D and 2D affine) Sensing 69(1), 59-70. 
do not always perform well, thus use of RPCs should be Gruen, A., 1985, Adaptive Least Squares Correlation: A 
generally preferred. IKONOS and QB orthoimages have been powerful Image Matching Technique. South Africa Journal 
generated with an accuracy of 0.5 — 0.8 m, for typical sensor of Photogrammetry, Remote Sensing and Cartography 14 
elevations of 65-75 deg. This requires, however, an accurate 3), 175-187. In 
DSM/DTM. Sophisticated matching algorithms have derived a Heller, O., Gut, O., 2004. Auswertung hochauflösender d 
5 m DSM with an accuracy of 1-5 m without editing and under _Satellitenbildern | IKONOS/QUICKBIRD. Report for in 
very difficult conditions. In spite of that, accuracy in open Vertiefungsblock, ETH Zürich, March. 
textured areas was Im or below. This potential has been very Otto, G. P., Chau, T. K. W., 1988. A “Region-Growing” qu 
little exploited up to now, especially with IKONOS which is Algorithm for Matching of Terrain Images. Proc. 4" Alvey m 
more stereo capable than QB, and presents an interesting Vision Club, University of Manchester, UK, 31 Aug. - 2 sp 
alternative technology for deriving DSMs. Future work will Sept. an 
focus on refinement of these investigations and possibly Zhang, L., Gruen, A., 2003. Automatic DSM Generation from be 
processing of new better quality images in the Thun testfield. StarImager (SI) Data. Proc. 6" Conference on Optical 3-D mc 
Measurement Techniques, Gruen, A., Kahmen, H. (Eds.), Hu 
ACKNOWLEDGEMENTS September 22-25, Zurich, Vol. I, pp. 93-105. am 
Zhang, L., Gruen, A., 2004. Automatic DSM Generation from e 
We thank the Swiss Federal Office of Topography and the Linear Array Imagery Data. IAPRS, Vol. XXXV, Part B3 
NPOC, Bern for providing in Thun the laser DSM and in (proc. of this Congress). Th 
Geneva the HRS images, Swissimage and the DHM25, the in 
Canton Geneva for providing the 25-cm orthoimages and the Mu 
laser DTM and Space Imaging USA for the IKONOS images in Th 
Thun and the RPCs of the Geneva IKONOS images. We also 
thank Oliver Heller and Oliver Gut, students at ETH Zurich, for dec 
contributing in this work, especially for the Geneva data. Wi 
resi 
REFERENCES DW 
Baltsavias, E. P., 1991. Multiphoto Geometrically Constrained Ye 
Matching. Ph.D. thesis, IGP, ETH Zürich, Mitteilungen No. Ms 
49, 221 p. fusi 
Baltsavias, E., Pateraki, M., Zhang, L., 2001. Radiometric and Kin 
e 
Geometric Evaluation of IKONOS Geo Images and their 
use for 3D building modelling. Proc. Joint ISPRS 
Workshop "High Resolution Mapping from Space 2001", 
Hannover, Germany, 19-21 September (on CD-ROM). 
528