163
the control points, a condition not always achieved in the experimental blocks, partic
ularly if they are of great dimensions.
The means M are poorly determined on the borders of the block and particularly
for the angles, where a part of the circle around the points is outer.
Finally, the means M are influenced by the effects of the identification errors
of the control points which are sometimes sizable. These errors are not to be ascribed
to the photogrammetric proceedings and do not have a great importance during the
application.
However, notwithstanding the uncertainties arising from the above mentioned
facts, we think that the three indices M m , a z m and m 2 A represent a significant element
as a basis for judgement concerning the accuracy of the results of a test.
In fact, the choice of the diameter D may be established by basing it on objective
data, thus making it much less arbitrary.
Evidently, the dimensions of the circle must not be less than the dimensions of
a single model, since just the variations and the oscillations on the orientation points
of the single models are of interest. Moreover, B must be such as to also include the
discontinuities between adjacent strips without masking the overall systematic be
haviour of the strips. It follows that the dimensions of D must not be larger than two
adjacent strips.
To these objective reasons, we must add the statistical condition that in each
circle, such a number of points must be included to make practically négligeable the
presence of moderately anomalous points.
Within these limits, the choice of the radius of the circle is established with suffi
cient approximation.
The other causes of uncertainty about the worth of these indices become unim
portant if one considers their mean total significance.
3 - SYNTHETIC MEASUREMENT OF ACCURACY OF A BLOCK! RESULTS
The principle of synthetic measurement of the accuracy of a block, illustrated in
the preceding paragraph, was applied to the results of tests obtained by the different
centres. Some of these tests are not independent, since they represent the ana
lytical solutions using the same numerical data, according to different proceedings
of adjustment. Accordingly, we chose among those tests, only those which resulted
from independent measurements and are adjusted with as uniform a number of points
as possible.
The computation of the indices M m , a 2 m , m 2 A was carried out at first for tests 1,
5, 8, 13, 14, 15, 16, 17, 18, 19, 20.
Tests 1, 5, 8, executed by the N.R.C. of Ottawa, have been adjusted on 7, 8, 8
points respectively; tests 13, 14, 15, 16, executed by the I.G.N. of Paris, have been
adjusted on 9 points. The other 4 tests make use of a completely different number
of points.
But afterwards, on explicit request of Mr. Schut of the N.R.C. of Ottawa, the
same calculations were also executed for tests 3, 6 and 10. These three tests are
obtained by means of the same measures of tests 1, 5 and 8 but are adjusted on
16 points distributed in pairs along the perimeter of the block. In Mr. Schut’s opinion
they allow more valid comparisons with the other results.
Moreover, we also complied with the request of Mr. Bonneval of the I.G.N. of
Paris and we made the comparison between the results of tests 11 and 13 pos-