424 
where 1 = 22.0 mm and A = the difference between a quantity in 
the reference picture and the same quantity in another picture. The 
second alternative is to determine the difference of the longitudinal 
tilt by means of the same horizons according to the following formula: 
(ARi — Ah x + AR-2 — Ah 2 + AR3 — Ah 3 ) p 
A = Vi (2) 
where the focal distance f = 34.2 mm. 
If formula (2) is used the standard error theoretically is 3.8 times 
smaller than that of formula (1), therefore only formula (2) has been 
used in this investigation. 
4. Procedure. 
The horizon pictures were measured in the stereo comparator Zeiss 
No. 72613, where it is possible to estimate 0.001 mm in parallaxmea- 
surements. Two reference horizons were used in two of the strips and 
three in the third (longest) strip. 
The diapositives of the aerial photographs were measured using the 
autograph Wild A8 No. 712, the accuracy of which can be characteriz 
ed by the standard error of unit weight 0.005 mm in the image scale 
according to test measurements. The scale of the diapositives was 
about 1: 30 000 and the model scale was 1: 15 000. After relative and 
absolute orientation the heights of the control points, the vertical paral 
laxes of 15 points and the orientation elements were observed. The 
autograph was tested several times during the measurements using the 
grid method for control of the accuracy and for determination of index 
errors etc. There was sufficient elevation control in every model while 
there were fewer horizontal control points. The procedure was begun 
therefore with completely oriented models and thereafter the scale was 
transferred to adjacent models by numerical and graphic methods. The 
proportions of the scale errors have in general been 0.3—-1.5 % 0 . 
5. The computations. 
The measurements of the horizon pictures were computed according 
to the formula (2). 
The measurements in the autograph A8 were corrected with respect 
to the index errors in the <p- and «-scales, the radial distortion, the y- 
parallax measurements at 9 points, the residual errors in the elevations, 
the curvature of the earth and the refraction. In addition, the inclina 
tion angles were corrected taking into consideration the fact that the 
coordinate and axis systems of the horizon camera and the stereo 
plotting machine are not identical. According to the definition mention 
ed earlier in this investigation k — zero (or n = 50g) in the camera 
system. On the contrary u in the plotting machine is dependent among