23
correlation factor
square side
milli meters
Fig. 1.63:1. Graphical presentation of the correlation factors between points in the corners of squares with varying sides.
The curve was determined from tests of aerial photographs and additional points from multicollimator tests of an aerial
camera.
1.7 The Accuracy of the Elements of the Interior Orientation of the Photo
graphs
The bundle of rays is reconstructed from the image coordinates of the photographs and the elements
of the interior orientation. The accuracy of the reconstructed rays is therefore dependent upon the accu
racy of the image coordinates and the elements. These elements, however, should refer not only to the
camera, but to the photograph taken with it, and the accuracy should be determined from a calibration
of the photograph at the time of reconstruction. For this purpose the accuracy of the image coordinates
of the photograph expressed as a standard error of unit weight is a basic factor, as the accuracy of the
separate elements can be expressed as standard errors which are products of the basic factor and the
square roots of the corresponding weight numbers.
The clearest determination of this accuracy is obtained by calibration and adjustment according to
the grid method as applied in the multicollimator calibration procedure. Assume a calibration as describ
ed in reference 1.61:2 and section 1.61. Through the adjustment procedure corrections are determined
to preliminary data for the position of the principal point (x' 0 ) and (y' Q ) and for the principal distance
(c). The standard error of the corrections is the basic factor in determinations of the accuracy. From the
general solution of the normal equations as shown in reference 1.61:2 the accuracy of the corrections can
be determined. For an adjustment based upon discrepancies in the center point and in four points on
a circle around it the corrections of the preliminary coordinates of the principal point are:
dx' 0 = dx o — cdcp
dy' 0 — dy 0 + cd ( <)
The weight numbers of the corrections are:
Q x o x o = Q x o x o + C "Q<Pf 2cQx o9 £,
Qy 0 y o = Qy 0 y 0 T - c2 Q<^>+ 2 c Qy 0 w
After substitution of the weight and correlation numbers from the solution of the normal equations
it is found that:
Q x o x o = Qy 0 y 0 == Vs
The standard error of the corrections is then found as: