Table 4 collects all the anomalous points of the 11 tests. The name of each
point is followed by the indication of the coordinates with an anomalous difference A.
Table 5 shows for each test the values of the differences A computed for the 14
points believed to be anomalous after a first judgement and eliminated from the
computation of the means. The values A on these 14 points have not been considered
for the computation of m A .
This table, in which one can note a good accordance of sign and value of the
differences for each point in all the 11 tests, is sufficient to justify the elimination
of these points from those on which the results of a block are estimated. In chapter
II, the I.G.N. of Paris examined each one of these points and investigated the reason
for their anomalous behaviour in the preparatory operations of the polygon and in
the distributed documentation.
It is more difficult to establish which ones of the anomalous points of table 4
are anomalous for non photogrammetric reasons, particularly for uncertain identifi
cation of the point itself. Tables 6 and 7 can help us in this investigation.
Table 6 subdivides the anomalous points in several classes. The first one includes
points anomalous in only one test. The second one includes points anomalous in
two tests. The third one includes points anomalous in three tests, and so on.
Table 7 contains the list of the values of A anomalous in 4 or more tests at the
same time. To these points anomalous in 4 or more tests, we added other points,
anomalous in only 3 tests, but that had been eliminated in some of the remaining
tests, during measuration or computation, for various and unspecified reasons. By
means of a line in the centre of the corresponding space, we have indicated that the
photogrammetric coordinates of the point have not been communicated.
Since the probability that the same point is anomalous in 4 tests at the same time
is very small, we can say that the points of table 7 are anomalous for non photogram
metric reasons and are to be added to those of table 5.
For some of these points, this supposition is confirmed by the agreement of the
sign of the difference of the same coordinate and by the order of magnitude of the
difference itself.
In this connection, we must consider the following points: 24, 40, 44, 53, 54,
91, 95, 105, 130, 137, 146, 164, 170, 175, 184, 231, 252, 272, 291.
The procedure of statistical investigation on the anomalous points, the principle
of their isolation from the other points and the probabilistic elements on whose basis
we thought to ascribe to them non photogrammetric causes, is not free from criticism.
Other types of more qualitative research, taking into account also eventually the
remarks contained in the reports of the different centres, could have arrived at more
results. However, as the notes of the reports are not complete and homogeneous, the
method we adopted maintains the merit of complete objectivity in respect to each
test, all of which were treated by the same procedure.
We should still pose the question whether the dispersion of the errors of points
around the average behaviour of the errors themselves, from which the values of
w 2 a of table 3 have been computed, is to be ascribed to real photogrammetric causes
(as poor bridging between models, internal deformation in the models, discontinuities
between adjacent strips), or to the wrong identification of the ground points and to
their uncertain individuation by the operators.
Unfortunately, on the basis of the material in our possession, it is impossible
to reply to this question: in fact, it would be necessary to have other experimental
data obtained under different operative conditions.
1G9