tuelle-
iduelle
multa-
les re-
les in-
‘ochées
re ces
soudre.
permis
ui exi-
s to be con-
neval gives a
d methods in
ct of a block
ngulation, es-
gulation, but
ormal slotted-
ulation strips.
ck adjustment
. and the most
lent, and pos-
it gives only
t be used for
do not think
meeting here.
and analytical
1sidered them
tical air trian-
according to
stance, by Mr
ime and some
ording to ap-
then basically
rial triangula-
angulation, we
cal solutions,
ent strips and
nate obtained
hod of cross-
| of Professor
yr third order
e co-ordinates
for the whole
LA COMPENSATION DES BLOCS DE BANDES, DISCUSSION 61
strip. Finally, there is the special method, such
as the use of independent bases or the use of
auxiliary data for the block adjustment.
A second group of solutions will be all those
methods which do not deal with the strip itself
but divide the strip in sections and find certain
corrections for these sections for the whole
block together. These sections could be single
models, of course, or they could be a certain
number of single models together.
Another block of adjustment methods would
be mechanical methods, such as for instance the
stereo template method which has been devel-
oped to a high standard by the United States
Geological Survey. Then there is the mechanical
height adjustment method, as shown in the last
Congress of the IGN. Finally, there is the anal-
ogue computer adjustment method at Delft,
International Training Centre. There are two
different kinds of analogue computers, one for
the height, which is basically similar to the IGN
method, and a second one which is based on
dividing the strips in sections.
This is what I should like to mention out of
the features of Mr Bonneval's report. I would
like to add some personal remarks, namely on
the problem of using rigid solutions for block
adjustment on the one hand, or of using approx-
imative solutions on the other. It is not thought
that there are finally rigid solutions which can
easily be obtained, even if it includes the solution
of very large equation systems or including the
finding of the very large number of unknowns.
But I do think there will still always remain the
fact that a big number of unknowns will always
cost much more in computing time than a
smaller number of unknowns, and so I would
like again to stress the fact which has already
been emphasised by Professor Thompson: that
we do have to investigate very seriously whether
a careful choice of unknowns — that is, reducing
the number of unknowns in the block — will
not essentially decrease the precision that we
can obtain, but essentially decrease the work
and the cost of such a block adjustment.
With regard to the presented papers, I
would only like to make some remarks to Mr
Schut on the paper that was published in the
National Report of Canada for the use of second
order polynomes, and the results which have
been shown. I do think that a second order
polynome will never give a good correction for
long strips. This was theoretically also shown by
Professor Fórstners paper where he demon-
strates that after a certain strip length a parabola
cannot at all follow the real course of error in
aerial triangulation. Of course, such methods as
published there can meet the required specifica-
tions, if we have only a small block and we just
do not ask for higher accuracy. But as compared
with the block adjustment method the errors are
about four times as large.
That is all I have to say here, and I would
like to ask Dr Schmid or Professor Fórstner to
say something.
Discussion
Prof A. BRANDENBERGER: I have to make
only a few very short remarks. I would like to
point out certain specific problems we might
find in block triangulation.
Basically, we have two approaches: the first
one uses first stereo plotting instruments; and
the second one is a complete analytical solution
in one unit. The first approach, using stereo
plotting instruments, is based on triangulation of
individual models, and afterwards the discrep-
ancies between individual strips will be eliminat-
ed by block adjustment procedures. It is, of
course, necessary that the discrepancy between
adjacent strips is not too large. If the discrep-
ancy is too large there is no sense in proceeding
with block adjustment. Such a block adjustment
of individual strips, which have been triangulat-
ed by means of a first order stereo plotting
instrument, has to be done in such a way that
the individual models are not distorted too
much. If, for instance, the discrepancies be-
Archives 5
tween adjacent strips are too large, it is possible
that by the block adjustment you introduce
deformations in an individual model, you will
feel these deformations afterwards when you
plot the model. This is a point to which we have
to pay attention.
In regard to a new approach, namely block
adjustment using an analytical solution where
we measure picture corners by means of stereo
comparators and where we compute co-ordi-
nates, there is also a special problem there:
namely, that we have to adjust the block by
means of a conformal system, or we have to use
a non-conformal system, and this with regard to
the fact that probably error propagation across
the parallel strips might not be the same as the
error propagation in the direction of the strip.
These are some items which I have to men-
tion, and actually I would like to point out, of
course, that from the practical standpoint all
these block adjustment procedures have to be