3s,, when this limit 3s, is set equal to 9 m. 
The gross errors are separated in the two other 
classes, one for errors between 9 m and 15 m, and one 
for larger than 15 m. 
Nearly 100 points are omitted due to mistakes in 
the procedure, either by the field operators or the 
later treatment of the materials. 
Table II gives the main results of the investigations 
at Radoy. Here are also the figures for the use of 
plastic base copies given, as this material here is 
relatively larger than from Idd. 
Only one class of gross errors is given, as the 
number of these is small. 
Within the test field only ‘‘open vegeation” exists. 
Table II. Testfield Radoy 
  
Number of 
Standard error 
£TOSS errors, 
e> 9m 
Sp— ( Ze 3 i 
n + 
Number, 
Number of points|  % of total 
  
  
  
  
  
(open veg. only) number 
Series 1 : 15 000 VA 
Paper copies 2.60 2 
185 1.1% 
Series 1 : 15 000 VA 2.36 4 
Copies on half- 
transparent plastic base 178 2,2% 
  
The number of points where the standard devia- 
tion is computed from two mappings without use of 
visible signal, is somewhat higher than for visible 
signals. The precision found from this material con- 
firms the standard error result in Table I. The 
standard deviation is regularly less than the standard 
error, and in average 0.19 m. This difference can be 
explained by possible correlation, due to similar pro- 
cedures. 
The contribution of standard error from the dif- 
ferent operations dealt with, is: 
Signalized control points, s =0.70m 
Pricking of points, s =135m 
Absolute or. of the models, Sao = 0.30m 
Visual transfer of points, S4, = 0.80 m 
Height setting. $, = 100m 
The value for visual transfer is given here, as this 
method was used more than the one with stereo- 
scopic transfer. 
When taking the root of the quadratic sum of 
this figures, we get the expected final value with the 
identification error not included. For “open vegeta- 
tion" we have to exclude s;. For *dense vegetation", 
however, most of the points refer to treetops. 
12 
We get: 
i = 
Sov = (se? + Sp? + Sao? + Svt?) ?= 1.75m. 
x 
Sav = (Se? + sp? + Sao? + Svt? + sn?) ? = 2.00m. 
According to these values, point errors larger than 
5—6 m should appear very seldom. But on the con- 
trary, they appear frequently. This implies that the 
error in identifying the points is considerable and is 
characteristic by numerous cases of gross errors. 
Another possible source of gross errors is the 
correction for points pricked in details outside the 
boundary. Mixing up these data is easily done, both 
in the field and by the mapping procedure, and will 
be destructive to the points concerned. But as the 
number of this kind of points is only 10 %—20 % 
of the total number (the highest percent for smallest 
scale), is the identification probably the main source 
of gross errors. 
for the series 1: 15000 VA, we will have the stan- 
dard error for the operations not included above. 
respectively for open and dense vegetations = (2,73°— 
1.752) 122.10 m, and s — (3.132—2.002) $—2.40 m. 
These values, which mainly are caused by errors in 
identification, represent certainly not normally distri- 
buted errors. Most of these errors are near to zero, 
but a relatively large number have values between s, 
Using the values s, — 2.73 m and s,, — 3.13 m 
and 3s,. Also errors greater than 3s, are numerous, 
but they are not included either in the values of s, 
or s above. 
The most surprising fact which can be drawn from 
the accuracy figures, is the small difference between 
the scales. 1 : 10 000 and 1 : 15 000 show no difference, 
and 1 : 20000 lies only about 15 % higher. As the 
errors from the photogrammetric operations is ex- 
pected to increase more than that, the conclusion has 
to be that the error in identification is least for the 
scale 1: 15000, and has not increased much down 
to the scale 1 : 20 000. 
Surprising is it also to find that the result for the 
series 1 : 15000 normal angle was not better than 
for the wide angle series, especially because the field 
operators express that they found this series most 
easy. One reason may be that just this feeling of less 
difficulties has dimished their attention and care- 
fulness. 
x 
However, s, has regularly a slightly higher value than 
s.. This can be explained by the regular greater model 
distances from the photo centres in this direction, 
which is near vertical to the base direction, combined 
with the height setting error. 
As found above, the identification error contributes 
s, and s, are not given separately in the table.