EXPERIMENTAL RESEARCH ON SEVERAL TYPES, ETC.
5
j — i, 2 .... m
1 — I, 2 ... . fl
where the expressions of parallax and height appears as functions of the base com
ponents, of the height of the first photogram taking point, and of the direction
tangents of the projecting rays on the z x and zy planes.
Equations i) impose the vanishing of the parallax in a point of the model
and contain, as unknowns, the five parameters of relative orientation ; equations 2)
impose a known height to a point of the model, contain, among the unknowns,
the component b x of the base and are used for the transfer of scale.
We wish to underline the fact that, while in analogical photogrammetry the
two operations of scale transfer and relative orientation are executed one inde
pendently on the other, in the analytical program we used, both types of equations
are solved at the same time. Moreover, there is no limit to the number of equations
that can be used, on condition that proper instructions are given to the computer.
The program we used foresees a maximum of 18 equations at the parallaxes and of
6 height equations. The discrepancies obtained solving the set of equations with
the least squares method will be either residual parallaxes if they refer to equations
1), or deviations in height if they refer to equations 2).
It can be demonstrated that, if the photograms were perfectly nadiral and only
one equation of type 2) was used for scale transfer, this last equation would be bound
by the preceding parallax equations, but would not bind them even when, being
n > 5, a procedure of least squares should be resorted to. In other words, the va
lues of the 6 unknowns that we obtain with the simultaneous solution of n equa
tions 1) and one equation 2) are the same that we would get by saparately sol
ving the parallax equations (as in the instrumental method) and then determining
b x as a function of the unknowns cp, co, y, b y , b z previously obtained. This is no
longer true when the number m of equations 2) which are introduced in the set is
larger than 1. In this case there exists a mutual influence between the formation
of the model and the scale transfer, which we will further investigate. Each one of
the five strips has been triangulated 4 times, using different combinations of equa
tions 1) and 2) ; namely the types of computation we performed are the following :
6-1 — 6 parallax equations -f- 1 scale transfer equation
6-3 = 6 parallax equations -j- 3 scale transfer equations
18-1 = 18 parallax equations + 1 scale transfer equation
18-6 = 18 parallax equations + 6 scale transfer equations
Furthemore, the measures on the plates have been carried out independetly
two times (measures A and B) ; for each one of the two measures the above mentio
ned computations have been repeated.
The instrument used for the measure of plate coordinates is a stereocompara
tor O.M.I. TA3. The position of the orientation points, which have been pointed at,
is shown in fig. 1, that represents three consecutive photograms. On each one of