from the point of view of the contemplated investigation, since the best that can
be made of the non-reseau block would be to try out the use of double minor
control in the absence of the reseau. This purpose can be well served by the reseau
block, simply by ignoring the presence of the reseau. It appears therefore that the
whole investigation could be carried out using only the reseau block. I am not
going to enter here in a discussion of the criteria to be used in the analysis. These
will be deferred to another communication. For the present, we are only concerned
with general approaches.
A possible design of the experiment which might, at first sight, appeal to the
exacting people is to compute the reseau block so that we can obtain the errors
corresponding to the following combinations of conditions:
(
1) Single minor control, using collimating marks for the estimation of distortion;
(2) Double minor control, using collimating marks for the estimation of distort-
ions;
(3) Single minor control, using the reseau to eliminate distortion and lack of
flatness;
(4) Double minor control, using the reseau to eliminate the distortions and lack
of flatness.
The contribution of the reseau would be revealed from comparison of the
results of (1) and (3), and also of (2) and (4); while comparison of (1) and (2)
and also of (3) and (4) would reveal any improvement due to the increase in the
number of minor control points.
This plan of the experiment can, however, still be criticized on the ground
that it would involve a four timerepetition of all calculations which would be
costly and would also interfere with production.
An alternative which suggests itself is to divide up the block of photographs
into four plots (the terminology is suggested by an analogy with agricultural
experiments), and to allocate the four treatments enumerated above to these four
plots at random, say
where C refers to the use of collimating marks, R to the reseau and S and D to
single and double minor control respectively. As before, the contribution of the
reseau appears from the comparison of (C-S) and (R-S) as well as from (C-D) and
(R-D); the improvement due to doubling the minor control will be seen from
(C-S) and (C-D) on one hand, and also from (R-S) and (R-D) on the other. The
comparisons can of course be based on the internal consistency with no reference
to ground control, but for direct evaluation of the final improvements in terms
of the magnitudes of the errors in the map coordinates, ground control becomes
necessary. In this block the amount of ground control averages one point per 12
photographs. The problem is thus resolved into weighing the advantages of gaining
a deeper insight into the roles of the factors against some additional ground work.
25