nbul 2004 
according 
) 
—(1) 
v2 
- Ha) 
  
| window 
ndow 
nding part 
ng part of 
reference 
orrelation. 
n can be 
1 between 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004 
e the estimation process is efficient, 
e the evaluation tools from least squares estimation are 
available. 
Feature based matching uses symbolic descriptions of the 
images for establishing correspondence. It is assumed that such 
a symbolic description can replace the original image suitably 
well and all information that is necessary for matching is 
contained in the attributes of the features and possibly their 
relations. Using features instead of the original intensities 
permits to select a representation that is much more invariant 
with respect to distortions such as illumination, reflectance or 
geometry (Lang and Fórstner 1998). 
4. EXPERIMENTS AND RESULTS 
Ten photographs, which belonged to Selcuk University Campus 
area, were taken by Wild RC 10 camera. Fig.5 shows 
subimages and locations of fiducial marks for Wild RC 10 
camera. Table 1 shows calibrated data of fiducial marks for 
Wild RC 10 camera. The focal length is 153.29 mm and 
photographs scale is 1:10000. The films were scanned by Zeiss 
Scai Photogrammetric Scanner. They were scanned with a pixel 
size of 21 pm, 8 bits per pixel, yielding a ground resolution 
21.0 cm. Total scanned image size is 10956 pixel*10955 pixel. 
  
  
  
  
  
Figure 5. Subimage and locations of fiducial marks for Wild RC 
10 camera 
Table 1. Calibrated data of fiducial marks for Wild RC 10 
camera 
  
  
  
  
  
  
  
Fiducial Calibrated data of fiducial marks 
marks 
x(mm) y(mm) 
1 -106.007 106.004 
2 106.009 106.005 
3 106.005 -106.002 
E- 4 -106.007 -106.004 
  
  
  
  
A computer program in Borland C++ Builder software is 
developed to find out the image fiducial marks, particularly for 
this project. The program has manual and automatic interior 
orientation module, bundle adjustment module and automatic 
digital elevation module (Karabork 2002). 
Cross correlation matching method is used for measuring of 
fiducial marks (Fig.6). Four fiducial marks on each image in 
this project are automatically measured by cross correlation. 
Both residuals of fiducial marks on each image (Table.2) and 
root mean square error of transformation between pixel and 
image coordinate system are calculated. 
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
Input Reference Search 
P» Window P Window 
Generation Generation 
Transformation Determination Cross 
«| ofbest fit pixel |g Correlation 
  
  
  
  
  
  
  
  
Figure 6. Workflow of interior orientation 
Table.2 Residuals of fiducial marks on each image 
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
Fiducial No 
Image No 1 2 3 4 
9061 v, (pixel)| 0.30 -0.30 0.30 -0.30 
vy (pixel)| -0.01 0.01 -0.01 0.01 
9062 v. (pixel)| 0.30 -0.30 0.30 -0.30 
vy (pixel)| -0.01 0.01 -0.01 0.01 
9063 V, (pixel)| 0.30 -0.30 0.30 -0.30 
v, (pixel)| -0.01 0.01 -0.01 0.01 
9064 v, (pixel) | 0.30 -0.30 0.30 -0.30 
vy (pixel)| -0.01 0.01 -0.01 0.01 
9065 v, (pixel)| 0.30 -0.30 0.30 -0.30 
v, (pixel)| 0.24 -0.24 0.24 -0.24 
9066 v, (pixel)| 0.55 -0.55 0.55 -0.55 
v, (pixel)| -0.26 0.26 -0.26 0.26 
9067 v, (pixel)| 0.30 -0.30 0.30 -0.30 
vy (pixel) 0.49 -0.49 0.49 -0.49 
9068 v, (pixel)| 0.30 -0.30 0.30 -0.30 
vy (pixel)| 0.24 -0.24 0.24 | -0.24 
9069 v, (pixel)| 0.05 0.05 0.05 0.05 
vy (pixel)| -0.01 0.01 -0.01 0.01 
9070 v, (pixel)| 0.54 -0.54 0.54 -0.54 
v, (pixel)! 0.49 -0.49 0.49 -0.49 
  
  
  
  
  
  
813 
In addition, interior orientation of 10 images are manually 
achieved by Erdas Imagine Orthobase software. Both residuals 
of fiducial marks on each image and root mean square error of 
transformation between pixel and image coordinate system are 
manually calculated. Finally, both manual and automatic 
measurement techniques are compared with each other in terms 
of interior orientation accuracy (Table 3).