(a) Visible signals in the control series. 
It is this main group which must be used as the basis for estimating accuracy, 
where boundary points are not signalled. Table 2 shows the results for the 
calculation of accuracy for the different series (scales). It can be seen from the 
Table that m p in all the series in considerably higher for areas of dense vegetation 
than for sparse vegetation, but there is comparatively little difference between 
the series. 
The scale of the photographs in the actual scale-area seems to be of little importance. 
Normal angle has given poorer results than wide angle in the same scale, but the 
number of large errors is smallest for normal angle, see class III, Table I. 
Table 3 shows a comparison of the results of all the series, added together for 
each field worker, and in Table 4 the same is done for all the series and workers 
grouped for each survey. 
(b) Signals not seen in the control series. 
As regards this main group, the mean error is calculated from the difference between 
the control series and each of the experimental series. The series are presumed 
to be of approximately the same value. The control series is here not more 
accurately decided than the other series, and m p is accordingly lower throughout 
than the equivalent for the visible signals in the control series. The results can be 
used as a basis for comparison of each of the series, the workers and the surveys, 
but one gets no other impression than what is obtainable from the visible signal 
group in the control series. There is therefore no reason to investigate this more 
closely here. 
The connection between errors in each of the series. 
There must be certain qualities about a point which ensures that both field worker 
and plotter identify it in the same way in two series of photos, even though ghe 
point is differently photographed. 
A correlating calculation for worker A for errors in x and y between series 1:15.000 
wide angle and the other series confirms that there is to some extent strong positive 
correlation between the errors in each of the series. There are many reasons for 
this, among them the uncertain height adjustment in the Autograf where pricks 
are on treetops, the same direction of flight in photographing, errors in the con 
trol base, etc. 
Conclusion. 
The experiments done in the Idd area show that one should be able to keep a 
mean error of ± 2.5-3.0 m, without signalling in sparsely wooded areas, fields 
and other unwooded areas, if (a) the quality of the photographs is good, (b) many 
measurements from clear details in the photographs are used, (c) boundary points 
are transferred to, for example, glass diapositives, for plotting, and (d) precise 
constructing work is carried out in the machine.