86 RADIAL TRIANGULATION, DISCUSSION
we had to deal not with simple rhomboids but
with rhomboids which had variable numbers of
wing points; I think it went up to six or eight.
It would have been hardly possible to match this
case with the classical method of adjusting
rhomboids.
Finally, I want to mention that some of the
ideas developed here we have found also in a
paper presented by Mr Dmochowski of Poland.
However, there are some points of disagreement
because in computing the rhomboid directly
with the observations he gets also the closing
error at the end of the rhomboid, but instead of
taking the average of the two points which he
gets there, he considers the centre of gravity
of this aerial triangulation as the best possible
adjustment. This is not in accordance with the
least squares solutions. A second objection we
have against his method of adjusting the strip
of rhomboids, is that his method comes down
to what you can call an interpolation method of
the closing error at the strip, but instead of
distributing this closing error more or less
regularly he makes it a function of the individual
closing error of the rhomboid. That means his
corrections are not a smooth function of the
length of the strip, but jumped.
Mr J. Visser: I will base my comments on
the investigation of Professor Roelofs and the ac-
curacy of numerical radial triangulation. Again,
a radial triangulator — and this time one of a
much improved design — is readily available.
Every person charged with the planning of
procedures in photogrammetric activities should
think about possible applications of this numer-
ical radial triangulation, in particular if this
method may profitably replace spatial triangula-
tion on the one hand and template triangulation
on the other. Of course, this only applies to
projects where the planimetric co-ordinates ex-
clusively are required. In practice, I think this
concerns mainly relatively flat areas. It is ob-
vious that the operation of a radial triangulator
such as the Wild RTI is much simpler than that
of a universal first order instrument, and that
consequently inexperienced operators can be
used. Further, from a simple cross-calculation it
will be clear that the cost of the numerical radial
triangulation will be about $42 lower per trian-
gulated model than spatial triangulation. This
is due to the higher speed of the observations in
radial triangulation, and to the much lower cost
of amortisation of the instrument.
Now about the attainable accuracies. When
block adjustment methods, such as the ITC-
Jerie are applied, it is only the deformation with-
in the units formed by two successive models
which are of importance, and not the propaga-
tion of errors through the triangulated strips. Ip
his 1956 publication, Professor Roelofs derived
the formula for a systematic error in transfer of
scale and azimuth from one model to the next in
numerical radial triangulation. This, indeed,
gives a good insight into the deformations per
unit which are to be expected and which deter.
mine the so-called relative accuracy of the block
adjustment.
When considering the possibility of applica-
tion of numerical radial triangulation in the
National Cartographic Centre in Teheran, we
used a set of photographs which were taken
from actual flights in Iran. The tilts were larger
than those used by Professor Roelofs in his
numerical example. We used a series with
maximum tilt component of 1.7 grades and
averages of 0.48 grades, and 0.57 grades.
Assuming the ground to be relatively flat, dis-
regarding the influence of ground elevation, we
found that in principal point triangulation sys-
tematic errors of the scale transfer of the order
of 0.15 promille were no exception; and also
that errors of 0.30 promille could be expected.
In nadir point triangulation these errors, how-
ever, practically never were larger than 0.12
promille. The systematic errors in the azimuth
transfer are practically negligible in both cases.
As Professor Roelofs did, we chose as com-
parison the errors and scale and azimuth
transfer in spatial air triangulation, caused by
observational errors only. We found that for
rightangle photography the mean square error
in scale and azimuth transfer was about 0.12
promille, so if we may assume that the errors
in scale and azimuth transfer were caused by
observational errors in radial triangulation to be
equal to those in spatial triangulation — and it
is Mr Van der Weele who can give an answer to
this, and who will do so in his I T C-publication
which will appear in the very near future —
then indeed nadir point triangulation may
replace the more expensive spatial triangulation.
This, however, is not entirely true for principal
point triangulation.
Prof B. HALLERT: We have practised a
method for radial triangulation founded on
image co-ordinate methods only, and this has
worked very well. We have quite a large ex-
perience from it, and it seems particularly from
the point of view of theory of errors that in this
case we get very clear information about the
propagation of from the image co-
ordinates to the final results. Perhaps this could
Crrors
TET
e» — mM o
t M CO —
-—-— €)