than 
con- 
the 
id is 
the 
. the 
both 
will 
; the 
0% 
allest 
yurce 
13 m 
stan- 
bove, 
,73?— 
0 m. 
rs in 
listri- 
Zero, 
en Sp 
TOUS, 
of s, 
from 
ween 
'ence, 
s the 
s ex- 
n has 
r the 
down 
r the 
than 
field 
most 
f less 
care- 
table. 
> than 
model 
ction, 
bined 
ibutes 
more to the final standard error than the different 
photogrammetric operations do together. This leads 
to the suspicion that it should be possible to improve 
the result without changing the photogrammetric pro- 
cedure, by using the best possible field operators and, 
which probably is the most important, not letting them 
hurry up in the work but using patience. It is also 
important, seen from this point of view, to get the 
best possible photo-material. 
Another source of errors of the same kind as the 
identification error, is the visual transfer of the points 
from the paper copy to the diapositive. Hurrying up 
the operator will also here easily cause errors greater 
than the standard error, while patient work will give 
sufficient accuracy. A stereoscopic method for point 
transfer outside the restitution instrument is therefore 
recommended. 
The identification errors may be divided in two 
kinds: First the cases when the field operator has 
found the place in the photograph where he is 
standing, but prick in another detail than intended. 
Most of the errors between s, and 15 m is supposed 
to have that reason. This group will also include the 
cases when the stereo operator could not decide 
whether the prick was intended to refer to a treetop 
or a detail on the ground. No informations followed 
from the field operator to help the stereo operator in 
such cases. The second kind is when the operator has 
failed in localizing himself. Most of the errors ex- 
ceeding 15 m belongs to this category, and occur 
mainly in dense and homogeneous forest, which also 
the table confirms. 
The value of s, for scale 1 : 15 000 VA from Table 
I and Table II is 2.73 m and 2.60 m respectively. The 
errors of the control coordinates are included in both, 
but as the control for the latter one is terrestrially 
measured, this error is here smaller than for the first 
one. Hence, by reducing the first one with s, — 0.70 m, 
they can directly be compared: 
1 
From Idd: SPoy (2.73? — 0.70?) 2.64m 
From Radoy: 5poy 2.60m 
Thus the results from two independent tests coin- 
cide very well, which gives higher significanse to the 
figures. The percentage of gross errors are, however, 
2.8 96 and 1.1 % respectively, but this difference 
may be accidental as the total number in this group 
is relatively small. 
Even if the standard error found in these investiga- 
tions could be accepted for a mapping purpose, seems 
the method, owing to a number of errors larger than 
3s of about 6 % of the total number, hardly ac- 
ceptable. But before giving up the method because 
of that, should it be compared with the other possible 
methods and tried to be improved. 
Correlation 
Investigation into the correlation between the results 
for the different scale series with the same field 
operator, and between the results for different opera- 
tors for the same series, shows that large identifica- 
tions errors are not pure accidental. 
Comparing the different series, correlation coef- 
ficients from 0.21 to 0.46 is found. This correlation 
is not very strong, and may partly have been caused 
by the field operators remembrance from series to 
series. 
Comparing different field operators using the same 
series of photography, correlation coefficients up to 
0.81 are found. This is a remarkably high value, which 
shows that a photography may be systematically inter- 
preted in a wrong way. As the correlation coefficient 
is dominated by the relatively large errors, the main 
reason for it is probably the identification in the field. 
But due to the fact that the coefficient always is 
somewhat higher in y-direction (direction near ver- 
tical to the base) than x-direction, the systematic 
error in height setting has to be one of the reasons of 
the correlation. As one stereo operator did all the 
restitution work for the Idd test, this is very well 
possible. This should be an evidence to the assump- 
tion that information to the stereo operator about 
which height the prick refers to, would improve the 
result. 
Other results 
A simplification of the method, by always pricking 
the position of the boundary point itself, without 
regard to whether the point can be recognized as a 
detail or not, is triel. The result showed a decrease in 
accuracy of about 30 % from the ordinary method. 
The reason of that may be that the field operator is 
not forced to use time to find a visible detail, and 
therefore has more freedom to make a bad result. 
The method of pricking unsignalized points could 
be considerably simplified if the pricking could be 
done in the same diapositives which is going to be 
used for the restitution. This is tried using a half- 
transparent plastic base for the copies, called Crona- 
paque. The material itself gave some difficulties, and 
the photographic quality of the copies was very bad. 
Hence the result was expected to be bad, but never- 
theless no significant difference from the ordinary 
method appeared. Without the difficulties mentioned, 
the use of this material probably would be an im- 
provement. The number of gross errors, however, 
would hardly decrease. 
The results from Radoy (Table II) show an in- 
creased accuracy by using plastic base copies. This 
indicates that the error in point transfer from paper 
copies to the model has been relatively high in this 
case.