ieces of data 
50 per coordinate, The 9 000/per model will together form an empirical error distribution large enough to be 
considered as the general empirical photogrammetric three -dimensional error distribution, Let us call it the 
Empirical Photogrammetric Error Distribution, 
  
It is important to state that the x y z errors could be considered as samples of the corresponding standard devia- 
tions, Their squares can easily be summed. 
€) € D 
X *y -p , where p isthe radial error in plane 
Qi 
p +z =r , where r isthe radial error in space. 
the 
r will be used as an indicator of total photogrammetric error, There exists an unlimited number of pz com- 
binations all giving the same r. 
o 
8.1 The Empirical Photogrammetric Error Distribution and X= Distribution 
  
The Empirical Photogrammetric Distribution was calculated in the following way. The experimental correspon - 
ding xyz errors were sorted in pz classes with one fifth of the general p and s as class interval. The number of paired 
Z 
p $, in each pz class gave the absolute frequency of the class. 
Z 
Such frequency distributions were calculated for the three negative scales, A mean frequermcy distribution of all ne- 
  
gative scales was also calculated. 
They will not be published as tables. Only the mean frequency distribution will be presented, figure 8. 1. The 
corresponding cumulative distributions were calculated. Only the mean cumulative distribution will be presented, 
  
  
figure 8.2, We underline that all errors larger than three standard deviations are included in the values for three 
stand.dev. 
  
Dividing the absolute frequencies by the total number of pz errors we calculated the relative frequencies or probabi - 0 7^ 
lities for the empirical frequency as well as cumulative error distributions, Only the means for the negative scales 
  
will be published. 
Table 8.11. The Empirical Photogrammetric Frequency pz Error Distribution. 
Table 8,12, The Empirical Photogrammetric Cumulative pz Error Distribution, 
The same method of calculating and sorting error was used in order to find the empirical maximum pz error. 
  
We used a larger unit, namely three times the standard deviation. That should give errors larger than three stand. dev. 
and less than our criteria in section 5. The result is presented in 
Table 8.13 Maximum pz error in the Empirical Photogrammetric Error Distribution, within chosen limits. 
  
  
  
Negative scale p cm Z cm T cm 
1:3 500 0<108 < 141 -100 < -90 <100 110 
1:6 000 0<179 <282 -200 < +192 « 200 207 
1:12 000 04255 <424 -300 « 470 «300 264