48
1. Comparison of different methods for analytical strip formation
1.1.
At the commencement of the groups activities no material was available for
practical tests, so the comparison of the calculation-procedures has been based on a
theoretical case. For this purpose the mathematical division of the I. T. C. computed
plate-coordinates of 26 photographs which together represent a strip of 25 models.
The computation of terrain-coordinates had to be done from those plate-coordi
nates, from the knowledge that photograph no 1 had a vertical axis and that in the
first model bx = 180 mm. The plate-coordinates, although originally computed with 9
decimal places have been rounded off to microns, this being the unit in which most
stereo- comparators give their registrations.
For the computation three different methods have been applied, viz:
1. A method developed by Prof. Inghilleri at the “Centro di Addestramento el Studi
Fotogrammetrici del Politecnico di Milano“ (ref. Technical paper no. 2 of the Milano
Institute, May 1959).
2. A method developed by Prof. Dr. K. Rinner and applied by the “Bundesamt für Eich-
und Yermessungswesen“ at Vienna in cooperation with the Mathematical Laboratory
of the Technical University (ref. Deutsche Geodätische Kommission, München 1957 —
Reihe A, Heft 25).
The result of these computations have been published already in “österreichische
Zeitschrift für Yermessungswesen, Special no. 1, Vienne 1962.
3. A method developed by Ir. C. M. A. van den Hout and applied by the International
Training Centre for Aerial Survey at Delft (ref. Bolletino di Geodesia etc., Firenze
1962. p. 418—427).
Rinner’s formulae-system differs in two points from the other two:
1. The condition for the intersection of two homologues rays is expressed in terms of
the shortest spacial distance between the two rays, whereas van den Hout and
Inghilleri use the ¿/-parallax.
2. The scale transfer from one model to the next is computed using all 3 common points
between adjacent models whereas the others use only the nadir point.
1.2. Results
A comparison of the computed strip coordinates with the known values shows
small differences which can be explained by the fact that the plate-coordinates have
been rounded off to microns. Due to the above-mentioned difference between Rinner’s
method and the other two, the influence of these rounding-off errors is also slightly
different, mainly in the X coordinates. All three methods are essentially iterative
procedures and their efficiency can be evaluated partly by comparing the number of
iterations per model. This comparison shows no appreciable difference for the first
21 models of the strip. With the Italian method, however, considerably more iterations
were required for the 22nd and 23rd models, and the 24th and 25th models did not
converge at all.