International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B6. Istanbul 2004 
metric cameras, the well-known shape and size of fiducial 
marks, simplify the problem. These marks are composed by 
several symmetrical elements (the centre of symmetry coincides 
with the centre of the mark) and they are easily differentiated 
from the image background. These characteristics allow the use 
of several measurement techniques: binarization and centroid 
localization, area-based matching or feature-based matching. 
The fiducial marks in most present aerial cameras are composed 
by lines and circles (concentric rings) that define the centre of 
the mark. For this reason, it is possible use the capabilities of 
the Hough transformation in order to define which pixels 
belong to the same circle. The method provides the parametric 
characteristics of the extracted circles (centre coordinates and 
radius), basic information for the results debugging. The 
obtained maxima in the accumulated table define the 
characteristics of the circles, and therefore the position of the 
fiducial marks. The measurements of the marks will be used in 
order to obtain the transformation between the image space 
(pixel) and the calibrated photocoordinates (obtained from the 
calibration certificate). This methodology has been 
implemented in a software program developed under C++ 
Builder language programming (Figure 5). 
The program input informations are: calibrated coordinates of 
the fiducial marks obtained from the camera calibration 
certificate (this information can be introducing using the 
keyboard or imported from a camera file); digital image where 
we want to apply the interior orientation process (using this 
image it is necessary obtain a gaussian image pyramid); 
direction of the +x coordinate axis (that can be provided 
indicating the marginal information position) (Figure 4). 
With this information (and information that the system obtain 
directly from the header of the digital image file such as 
resolution) it is possible to obtain an initial approximation to the 
fiducial marks position in the digital images. 
u (columns) 
v 
(rows) 
  
Y 
  
Figure 4. Image space coordinates (u,v in pixels) 
and calibrated photocoordinates (x,y in milimeters) 
For each fiducial mark, the system will go to the previously 
selected zone and will search the corresponding circle. The 
process begin using the initial available approximations in the 
higher level of the image pyramid (with low resolution, low 
accuracy and a small number of pixels) and this results will be 
used as approximations for the next levels —lower levels— (with 
higher resolution, higher accuracy and larger number of pixels). 
Once, all the measurements are finished, an affine 
transformation (6 parameters transformation) is carried out. 
The obtained results (measurements) have been compared with 
the obtained ones using a commercial digital photogrammetric 
workstation (SOCET SET —in manual and automatic modes—). 
The obtained results are compared in Table 1. 
  
  
  
  
  
  
  
  
  
  
  
Mark OIA SOCET SET SOCET SET 
manual automatic 
X y X y X y 
1 15247 7684 | 15248 7684 | 15248 7684 
2 188 7682 188 7683 187 7683 
3 7716 153 7716 153 7716 153 
4 7717 | 13213 7717 | 13213 72712 | 13213 
5 15245 153 | 15246 154 | 15246 153 
6 188 | 15213 188 | 15212 188] 13213 
7 186 152 187 153 186 152 
8 15246 |. 15215 | 15247 | 15214 | 15247 | 15214 
  
  
  
  
  
  
  
Table 1. Comparison of the mark measurent working 
with OIA and SOCET SET (values in pixels) 
If there is any problem in the automatic measurement it is 
possible use the manual measurement option for solve any 
measurement error. 
  
170 
Figure 5. OIA program. Main display and results windows 
3.2 Non-metric camera images. 
In the non-metric cameras digital imagery, one of the most 
frequent methodologies is based in the location of the frame 
image corners by the frame borders line intersection. These 
lines can be extracted using the Hough transformation for 
straight lines. This type of automatic inner orientation consists 
in the calculation of the format centre position (indicated 
principal point, IPP) (Figure 6). 
     
u (columns) 
    
       
pd 
  
  
v 
(rows) 
LÉ 
I FF 
“= 
Pa 
Figure 6. Non-metric camera inner orientation