437 
1 
The weight coefficients are: 
Ü a 0 a 0 
Q m a m a 
n 
4 
[(n + l-2i)2] 
(31) 
The influences or the linear systematic errors have been corrected 
in accordance with the formulas (29) — (31). The results are given 
in the following table. 
0 
Q 
Strip 
so = 
= s 0 
A8+hk 
helm 
a 
s bdm a 
s 0 hk 
so = 
= S D 
A8+hk 
bdm 
a 
s bdm tt 
S Q hk 
6010a 
2.63 c 
— 0.39 c 
0.08 t: 
i.97° 
2.29 c 
— 0.14 c 
0.07° 
2.00 c 
6010b 
2.99° 
— 0.23 c 
0.14 c 
2.11 c 
2.62 c 
+ 0.29 c 
0.12 c 
1.30 c 
6011 
2.59 c 
+ 0.38 c 
0.28 c 
2.27 c 
1.65 c 
+ 0.55 c 
0.11 c 
2.35 c 
rms 
2.75 c 
0.34 c 
0.16 c 
2.08 c 
2.41 c 
0.31 c 
0.10 c 
2.02 c 
Three dm«-values of six deviate signifioantly from zero and all are 
greater than their standard errors. Thus there are systematic errors. 
In this case it seems to be proable that the reason is the systematic 
elevation variations of the horizon. The accuracy is after the adjust 
ment 2.05 c as the mean square value of all the values. It corresponds 
to 0.027 mm in the image coordinates. The theoretically estimated 
value 1.8 C does not deviate significantly from the obtained values 
according to the F-distribution on the 2.5 per cent level. The diffe 
rences have been tested using the chi 2 -test (see histogram 2) and 
they are well normally distributed. 
Summary. 
The accuracy and the precision of the horizon camera were investig 
ated by means of three strips, which were photographed in Northern 
Finland at a height of 4 500 in. Panchromatic film and infrafilter (670 
m/x) were used in the horizon camera. The differences in inclinations 
between a horizon image and other images in the strip were determined 
by stereoscopical measurements. The root mean square value of all the 
standard deviations of a measured inclination angle was l. c 9 (centesi 
mal minutes) which were determined by means of two or three repeat 
ed measurements. The inclinations from the horizon pictures were 
compared with those from double point resections in the autograph