2004 
  
y to 
hing 
were 
in of 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B1. Istanbul 2004 
  
Figure 6. Automatic extracted tie points 
2.4 Model calculation 
In this stage, the adjustment calculations will be done on the 
rigorous model to solve the parameters. Some tests have been 
done to find out how many GCPs are needed and are logical for 
solving model to achieve a reasonable accuracy. 
The test has been done with changing the GCPs to CPs and vice 
versa. Also, removing and contributing tie points in the 
"calculations have been tested. The criterion for the test is based 
on the RMSE of the GCPs and CPs. The results of the test for 
PCI OrthoEngine using the Toutin's model for SPOT 5 data 
are: 
1- When the number of GCPs goes up and becomes 
more than 7 points, the role of tie points in the 
calculations will be small. 
2- The minimum number of GCPs for solving the model 
with logical error in each image is six. Also, it is 
logical to have six GCPs in each image in practical 
projects. Thus, six GCPs per each image are used in 
this study. 
3- When the user uses minimum number of GCPs, 6 
points, the tie points makes a normal error 
distribution in the entire image. 
The result for 6 GCPs for each image and 13 tie points is: 
Residual Info for 2 Images (Residual Units: Image Pixels) 
No. of GCPs: 12, XRMS=0.49, Y RMS=0.43 
No. of CPs: 22, XRMS=067 YRMS- 304 
No. of Tie Points: 13 , XRMS=0.15, Y RMS = 0.04 
Residual Info for 2 Images (Residual Units: Metres) 
GCPs: X RMS = 4.85, Y RMS =2.21 
CPs: X RMS = 8.90, Y RMS = 13.99 
Tic Points: | XRMS= 1.40, Y RMS = 0.37 
Please see the Appendix I for more information about the used 
points. 
2.5 Creating Epipolar Images 
After solving the orbit modelling parameters, the images will be 
resampled in epipolar lines. In the epipolar images, Y 
parallaxes are minimized and X parallaxes are remained. This 
makes the search area for matching process to be narrow and it 
makes the matching computation to be simpler and faster. 
2.6 Automatic DEM Extraction 
The next step is to extract DTM automatically from epipolar 
images. This software uses correlation function for image 
matching. The algorithm of DTM extraction asks from the user 
to give the minimum and maximum height in the region. Also, 
the correlation coefficient for each DTM cell could be saved in 
another image. 
2.7 Geocoding Extracted DEM 
This process projects the generated DTM from epipolar images 
to the ground coordinate system. 
In the whole process, no edit has been done on the extracted 
DTM because the goal is to find the accuracy of the automatic 
extracted DTM. 
2.8 The DTM result 
Table 1 and Table 2 show the error analysis on GCPs and CPs 
in the generated DTM. 
  
  
  
  
  
  
  
GCP Elevation | GCP calculated | Difference 
ID (m) Elevation (m) (m) 
G0001 -25.8 -34.9 9.] 
G0002 566.9 583.1 -16.2 
G0005 5.4 8.1 -2.7 
G0007 499.7 459.6 40.1 
G0010 -22.5 -34.3 11.7 
G0011 -22.3 -24.9 2.7 
  
  
  
  
  
  
Table 1. Error analysis on GCPs in generated DTM by PCI