International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B-YF. Istanbul 2004 
photographs have not been used, as it is expected, that these are 
the areas of grossest distortion. 
       
  
     
  
  
i o o 
: es © Os sera 
  
  
T5 RE eh 
  
  
Figure 2. Point location 
2.3. Results 
The co-ordinates of 33 geodetically determined points have 
been measured from the stereo model. The differences between 
geodetical co-ordinates and the ones, defined by 
photogrammteric method, have been calculated, as well as the 
RMS of photogrammetrically defined co-ordinates (see 
Ttable 1). 
RMS [cm 
+1.2 
+1.4 
+1.9 
Axes 
along axis X 
along axis Z 
lane XZ 
in debt (along axis Y) +6.1 
Table 1. Accuracy assessment 
in picture 
  
The co-ordinates, established by photogrammertric method have 
been examined for systematic errors, whereas the theoretical 
and empirical values of the mean absolte errors and median 
errors: 
Theoretical = 0.80 m (4) 
Viheoretical = 0.67 m 
where; r is the mean absolte error, 
m is the RMS, 
V is the median error. 
The comparison shows standard distribution of errors of 
photogrammetrically defined co-ordinates (see Table 2 and 
Table 3). 
  
  
  
  
  
  
  
  
  
  
Mean absolte X [em] Z [em] Y [em] 
Empirical +1.0 xi. XS. 1 
Theoretical +10 £1.1 t4.9 
Table 2. Mean absolte error 
Median X [cm] Z [cm] Y [cm] 
Empirical +0.8 +0.9 +4.7 
Theoretical +0.8 +1.0 +4. 1 
  
  
  
  
  
  
Table 3. Median error 
The frontage of the building has been vectorised in stereo mode, 
the three dimensional model (DTM) of its surface and ortho- 
image have been created. The co-ordinates of 30 points have 
been recognised from the ortho-image and have been compared 
"to the ones, obtained from the geodetic measurements. The 
RMS for the location of points from ortho-image has been 
calculated to be + 2.2 cm. 
3. CONCLUSIONS 
e The shortcomings of non-metric camera can be eliminated 
by usage of special software. 
e In order to make adjustment by means of self-calibrating at 
least 5 control points or 9 control distances must be 
measured. Having known camera parameters, the number 
of control data is reduced. 
e The accuracy of the ortho-image directly depends on the 
quality of the digital three dimensional model (DTM) of 
the object and does not exceed the one of the stereo model. 
e The unknown lens distortion and the inconsistency of the 
parameters of internal orientation did not render systematic 
influence on the results of the experiment. 
e The results from the experiment show, that this method 
completely meets the requirements of the assigned task and 
could successfully be applied, being an inexpensive and 
efficient technology for application in various fields, such 
as preservation of building frontages of cultural 
monuments. 
REFERENCE: 
l. Non-topographic photogrammetry, 2nd edition. 1989, 
American Society for Photogrammetry and Remote 
Sensing, Falls Church, Virginia. 
138