ctors 
actor 
| the 
y p. 
(7) 
1 Fig- 
  
  
based 
1e fea- 
| same 
nation 
metric 
f esti- 
in the 
[1ikhail 
cas are 
r map- 
s have 
5; block 
consisting of at least two flight strips with side- 
lap of 20%-60% and endlap of 60%-80%. Also 
number of control points and their distribution 
is determined according to standards. This kind 
of approach requires quite a lot of work and time 
especially in the case of map revision. The most 
time consuming phases are the field surveys and 
the targeting of the ground points (Figure 4.). 
  
  
Field surveys 
and targeting of 
ground points 
  
  
  
  
Photography 
- Aerial blocks 
- 30%-60% sidelaps 
- 60% 80% endlaps 
  
  
  
  
  
  
Photogrammetric 
triangulation 
  
  
  
Stereomapping 
- pointwise mapping 
  
  
  
  
Figure 4 
In our procedure the standard methods of point- 
wise photogrammetry are not necessary. By 
using linear features we can deal with different 
types of images which is not typical with tradi- 
tional methods. Also the use of images produced 
by different kind of sensors is possible. Actually, 
the problem with different exterior orientations 
of the satellite image rows, censored by row de- 
tectors, can be handled with the use of linear 
features. 
Because we are not using targeted ground points 
as the control datum but the existing features of 
area in question, geodetic measurements on field 
are not needed. This approach depends on the 
assumption of having a priori knowledge of the 
mapping area. The numerical information of ob- 
ject coordinates can be stored in any kind of 
database. The only requirement is that coordi- 
nates are stored with some logic. In our applica- 
tion we are using a GIS/LIS-system as a storage 
device. The system we are using, ARC/INFO, 
does not support 3D objects, so the third coordi- 
nate is handled as an attribute information. For 
data acquisition we are using routines provided 
by the system. Perhaps the severest drawback 
in this method is that information about the 
accuracy of object coordinates is rarely available. 
When all suitable object curves have been found, 
their images on photographs have to be found. 
The criteria of suitability is not very implicit. It 
503 
The flow of a traditional mapping process. 
can be based on the criteria of accuracy as well 
as the position and orientation of curves respect 
to the area of the photograph. Anyhow, it will, 
at least partly, be dealt in a heuristic manner. 
From 3D point coordinates of curves the parame- 
ters of linear features will be constructed. At 
this stage some estimates for the accuracy of 
parameters can be received. As we know the im- 
age of a linear feature is also a curve in a two 
dimensional photograph. In some case, image of 
a line could be a point, but that is very rare. 
With analog photographs observations can be 
done in the same way as it has been done before 
with analytical plotters. To find the curves from 
digital photographs we can use two different ap- 
proaches. First, we can try to fit some 2D spli- 
nes into the image and estimate visually when a 
curve fits in with a trace of the feature. In many 
graphic libraries there are many different draw- 
ing types of curves available. Another approach 
is to imply edge detection methods to find the lo- 
cation of edgels according to the grey level values 
and link them to a whole curve. The latter ap- 
proach is more accurate, but because it is based 
on automatic algorithms some robust estimation 
techniques should be implied. To develop this 
kind of system will probably take many years. 
The next problem is to find the correspondence 
between the object features and their images on 
photographs. A manual procedure with an oper- 
ator pointing out the correct images of the fea- 
tures on images is perhaps most reliable at the 
moment. For this purpose several LIS/GIS-sys-