2 DERIVATION OF THE PRINCIPAL DISTANCE FROM THREE PHOTOGRAPHS OF A
PLANE OBJECT
2.1 Direct Relative Orientation of two Metric Photographs
If four object points forming a plane can be measured in two metric
photographs, a direct rectification of the object plane is possible
without iterations. This problem vas verified by Wunderlich /7/ using
matrices, and Killian /3/ using trigonometric relations. Both methods
solve a cubic equation and deliver two results. The correct result
has to be pointed out by analysis of the rectified data and
plausibility tests.
These algorithms may be modified for calculating the relative
orientation /2/. The derived parameters (coordinates of the
projection center, rotation angles, local object coordinates) can be
applied as approximate values in bundle adjustment. To get the scale
of the model the distance between two object points has to be
measured.
2.2 Orientation of three non-metric Photographs
Three photographs of a plane object are necessary to calculate the
principal distance (c ) if the principal point (x ,y ) lies in the
center of the image. ^ 9.4.9
By combining two photographs at one time, three relative orientations
can be calculated with 2.1 /2/. In the first step any value that is
smaller than c may be taken as principal distance Cc,. Each
combination defivers four rectified object points (Fig. 1). The
difference between the coordinates of: points 315: 3", 3" and 4!, 4", 4"
may be used as an indicator for the descripancy of the principal
distance e from the real value e.
aa
—
1'=1"=1" 21-2422" X
Fig. 1: Three rectifications in one model coordinate system
Ax; . (k) = x. (k) - x. (k) i-2,3,1
Ay: (k) = y! (k) - y; (k) j=1,2,3 .. image (combination)
J J k=3,4 .... point number
9 2 Xy: eis «is rectified object coordinates
dx (k) = 1/3 1 AX; (kK) dx,dy .... mean discrepancy in
dy” (k) = 1/3L y, (k) the x- and y-coordinate
= 531.7