anbul 2004 
  
  
of 
oimage 
  
  
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B2. Istanbul 2004 
Table 4. Root-Mean-Square error of orbitimage 
  
  
  
  
GCP CHKP 
Unit: meter GSD |RMSE E [RMSE N [RMSE E [RMSE N 
2.3 1.91 1.69 3.75 3.01 
QuickBird 0.6 0.60 0,51 0.83 0.90 
  
  
  
  
  
  
  
3.3 Computation Time 
We used a personal computer with 3GHz CPU for 
orthorectification. The equal grid patch backprojection method 
is applied in the orthorectification. Table 5 is the statistics of 
calculations for two images. We setup the model error should 
be less than 0.05 pixel in patch backprojection. Referring to 
Figure 11, for SPOTS and QuickBird data, the variance of 
terrain in single tile should be smaller than 500m. We used the 
actual DTM to do the terrain analysis for both sensors. The 
elevation range of SPOTS is from 0 to 2100 meter, when the 
terrain variance is smaller than 500 meter in single tile. The 
smaller patch size is 160 by 160 meter, so we used tile sized 
160m*160m to do the equal grid patch backprojection. The 
elevation range of QuickBird is in between 0 to 700 meter. We 
analyzed the DTM with respect to QuickBird. When the terrain 
variance is smaller than 500 meter in single tile, the smaller 
patch size is 160 by 160 meter. We spent 45 minutes in doing 
QuickBird's orthorectification. As for SPOTS, we spent 55 
minutes. 
Table 5. Orthorectification computation time 
  
  
  
  
  
  
  
  
  
SPOTS QuickBird 
Computation time (min) 55 45 
Patch Size (m*m) 160*160 160*160 
Patch Size (pixel*pixel) 64*64 266*266 
Orthoimage Size (pixel*pixel) |29480*30320 |30786*29186 
Orthoimage Size (mb) 874 1744 
  
  
We used SPOTS data to do the Patch Size Optimal 
Backprojection instead of QuickBird, because its terrain 
variations and patch size is small. Also, when using quadtree to 
do the terrain analysis, the result is same as equal grid. The 
result is shown in Table 6. When using Patch Size Optimal 
Backprojection, it takes only 28 minutes for orthorectification, 
and both the quality and computation time is satisfactory. We 
spent 55 minutes for equal grid patch backprojection, and more 
than 10 hours for point-by-point method. 
Table 6. Comparison of computation time for SPOTS 
  
  
  
  
  
  
  
  
  
  
  
Patch size optimal|Equal grid patch|Point-by- 
backprojection backprojection  |point 
Computation time|28 55 >10 hr 
(min) 
Terrain variation|0~2100 0~2100 0~2100 
(m) 
Patch size (m*m) |Max:1280*1280 |160*160 NULL 
Min: 160* 160 
Terrain allowance|500 NULL NULL 
in a patch(m) 
Number of patch — |6367 217580 Number of 
point 
5760000 
  
3.4 Summary 
The experimental results indicate that: (1) we proposed a 
scheme for patch size optimization while the model error is less 
than 0.05 pixels, (2) the orbit adjustment accuracy is better than 
2 pixels when 9 ground control points is applied, (3) the 
proposed “Patch Backprojection" reduced the computation time 
of orthorectification, and (4) the orthorectifation result is better 
than 2 pixels, which is almost identical to the one tested in orbit 
modeling. 
4. CONCLUSIONS 
In this study, we have proposed a procedure of fast 
orthorectification for satellite images. The proposed method 
used the patch backprojection in orthorectification. Patch 
backprojection method is a feasible way to improve the 
efficiency with respect to the point-by-point backprojection. In 
order to control the model error of patch backprojection, the 
model error analysis of the proposed method is also presented. 
Data sets including SPOTS and QuickBird have been tested in 
validating the proposed method. Experimental results indicated 
that the proposed scheme may minimize the orthorectification 
computation time, while the model error is insignificant. 
REFERENCES 
Orun, A.B. and K. Natarajan, 1994, A modified bundle 
adjustment software for SPOT imagery and photography: 
tradeoff, Photogrammetric Engineering and Remote Sensing, 
60(12): 1431-1437. 
Chen, L.C. and L.H. Lee, 1993. Rigorous generation of digital 
orthophotos from SPOT images, Photogrammetric Engineering 
and Remote Sensing, 59(3): 655-661. 
Chen, L.C., L.Y. Chang, 1998. Three dimensional positioning 
using SPOT stereo strips with sparse control, Journal of 
Surveying Engineering, ASCE, 124(2): 63-72. 
Chen, L. C., and T. A. Teo, 2002, Rigorous Generation of 
Digital Orthophotos from EROS A High Resolution Satellite 
Images, International Archives of Photogrammetry and Remote 
Sensing, 34(Part 4):.620-625, 2002. 
Kim T., D. Shin., and Y.R. Lee, 2001, Development of robust 
algorithm for transformation of a 3D object point onto a 2D 
image point for linear pushbroom imagery, Photogrammetric 
Engineering and Remote Sensing, 67(4): 449-452. 
Mather, P.M. 1999, Computer processing of remotely-sensed 
images, John Wiley & Sons. 292 pages. 
Konecny, G., P. Lohmann,, H. Engel, and E. Kruck,, 1987, 
Evaluation of SPOT imagery on analytical photogrammetric 
ilnstruments, Photogrammetric Engineering and Remote 
Sensing, 53(9): 1223-1230. 
Mikhail E.M and F.Ackermann, 1982, Observation and least 
squares, University Press of America, New York, pp.401. 
Thierry T., 2003, Error tracking in IKONOS geometric 
processing using a 3D parametric model, Photogrammetric 
Engineering and Remote Sensing, 69(1): 43-51. 
Westin T., 1990, Precision rectification of SPOT imagery, 
Photogrammetric Engineering and Remote Sensing, 56(2): 247- 
253