The precision requirements on navigation data are more 
demanding if direct setting of orientation elements in an 
analytical plotter is considered, for the purpose of stereo- 
restitution of individual pairs of photographs, without 
further updating by relative orientation. Some preliminary 
simulations have shown that attitude data and position data 
with standard errors of 5 mgon and of 10 um (3 0,6 m for 
1:60 000), respectively, give coordinate errors of model 
points of 
M. y = 17 um and i, 42 um. 
In this case the precision of attitude and of position data, 
respectively, is balanced. The resulting accuracy of model 
points, as defined by the (near) intersection of rays, is 
quite acceptable in planimetry. The vertical accuracy, 
however, is rather poor. The external orientation data would 
have to be more precise in order to meet mapping standards 
for vertical accuracy and contouring. The above values can 
be transferred to other precision assumptions, because of 
proportionality. The results could still be improved, if 
after the absolute setting of the orientation data relative 
orientation would be performed. 
With regard to blocks several cases can be distinguished. 
Using directly the orientation data, as derived from naviga- 
tion data, the intersections of rays of all block points can 
be computed, without any further tying operations by block- 
adjustment. Assuming standard errors of 10 um (in photoscale) 
and 4,2 mgon for position and attitude data, respectively, a 
simulation gave the following r.m.s. coordinate errors of all 
block points: 
Uu = 15 um and HH. = 27 um. 
X,Y 
The values for points inside the models (2 rays per point) 
are larger by 33 % and 55 %, respectively. 
The most extreme case of a block is given when the orienta- 
tion data are used directly and points are determined in 
pairs of photographs only, without any consideration of mul- 
tiple intersection of rays or tying operation. It means that 
the photographs could be set in an analytical plotter and 
plotting would commence without any additional operation at 
all. This case was also studied by simulation assuming stan- 
dard errors of 4,2 mgon and 10 um for attitude and position 
data, respectively. The resulting accuracies for points in 
the overlap areas of adjoining models are: 
= 17 um and u_ = 30 yum. 
Hx Ly Z 
The remaining y-parallaxes have Wpy * 31 um. 
In this case the gaps between models are of special interest, 
as no tie-conditions are imposed. We can distinguish gaps 
between models within a strip,across strips and diagonally 
across strips. The results of the simulation are summarized 
in the following table: 
- 14 -