models, gives a point standard error of the control 
points equal to about s, = 0,70 m. This agrees quite 
well with a direct terrestrial control of signalized 
points. 
Assuming a standard error of the unsignalized 
points equal to 2,50 m, we will give a value which 
is 0,10 m too high when counting the control points 
as without error. 
As mentioned above, the field operators could find 
in the photopraphs only 950 of the 1400 signals. From 
these 950 the stereo operator could positively iden- 
tify 420 only. This unexpected fact was confirmed by 
letting one of the field operators identify signals under 
a mirror stereoscope, in an area where he did not 
make field work. 
This reduction of the control materials is partly 
avoided by assuming the same accuracy in these 
points without visible signal, as for series where no 
signals exist, and using the differences for computing 
standard deviation. 
STEPS OF THE PHOTOGRAMMETRIC WORK AND 
SOURCES OF ERRORS 
Pricking of points in the photographs 
The first error which can be done in this procedure, 
is wrong identification of the point to be pricked. 
This is an error which can be of any size, and has 
no normal distribution. It is probably the reason of 
the relatively great number of gross errors which we 
will find in the final result. 
The second error is the pricking of the identified 
point. Some investigations are made to find this 
precision: 
Pricks which refers to the boundary point itself, 
should be identical in two photographs, except for the 
pricking error. When having a group of such points, 
pricked by two different operators, and transform 
the image coordinates from one of the photographs 
over to the other, we will get the precision of the 
pricking, supposed that the two operators have the 
same precision and they are not correlated. 
This method is also used for accidentally chosen 
points. 
When reducing the error of the final map coordi- 
nates of points where identification errors not are 
involved, with the measuring error, we also get the 
precision of the pricking. 
Some tests gave standard deviation values between 
0.05 and 0.11 mm in the image, or between 0.75 and 
1.65 m in terrain when the photo scale is 1 : 15 000. 
The highest value is the most correct, because the 
materials which gave the lowest figures proved to be 
strongly correlated. However, due to other reasons 
the most probable value is set to s, — 0.09 mm or 
1.35 m. 
Absolute orientation of the models 
The standard error in planimetry for the points used 
for the numerical absolute orientation is given as 
average for three groups of photography: 
10 
A. Series used for photogrammetric passpoint de- 
termination, scale 1 : 25 000, terrestrially meas- 
ured passpoints, 6 models, s, — 0.52 m 
B. Series used for coordinate determination of 
signalized points, scale 1 : 15 000 and 1 : 17 500, 
photogrammetrically measured passpoints, 16 
models, s; — 0.44 m 
C. Series used for coordinate determination of un- 
signalized points, scale 1 : 10 000, 1 : 15 000 and 
1 : 20 000, phot. measured p.p., 54 models, sc = 
0.50 m 
sy and s; compared with s, shows that the relative 
accuracy of the photogrammetrically determined pass- 
points is of the same magnitude as the terrestrial ones. 
When the models are absolutely orientated on four 
corner points, the standard error for an arbitrary 
boundary point will be between s, | y/2 and sy /2, 
depending on if the points are near the corners or 
the centre. As an average the influence of the absolute 
orientation can be set equal to s,, = 0.30 m. 
All modells have got numerical absolute orientation 
in the planimetry by conform, linear transformation 
on to the passpoints. 
Stereoscopic measurement of the points 
This work includes two operations that-give errors 
which have to be considered. 
The first operation is to place the floating mark 
at the same place in the model as the prick has in 
the paper prints. Two different methods have been 
used: 
The most simple method is the one called *'visual 
transfer" without auxiliary equipment. That means 
that the stereo operator alternately has to look at the 
paperprints, monoculary or stereoscopically, and in 
the stereo restitution instrument. The method is slow 
and tedious, and the accuracy mainly depending on 
the operators patience. A test of the accuracy is done 
by a double plotting by two operators, which are sup- 
ry 
e (75 mM A M0 p M 9 9 = e—i deja mh an 
e 
4 m 71 M t 
bM AO | =~] HA