9
measures
sxpansion
j that the
ourse, on
Dry work,
e a priori
■ficient of
Ly at our
ble.
it not for
mgulation
ice of the
nal block
be found
the more
tie smaller
very best
t the best
measures;
determine
towns and
a «better
jr answer,
ise of the
work, it
is a better
ions: what
is to stop
• processes
speaking,
eatures of
and not a
s of maps
5 with the
3 or down
o walk up
¡round, its
te valuable
this shape
[ is surely
ted result,
¡eometrical
alone but
rdinates in
and polar
coordinates (heights). The idea that photogrammetric triangulation has errors
due to earth curvature comes simply from the failure to accept the obvious fact
that a height is not a rectangular coordinate and should not be treated as if it were.
3. THE RÉSEAU
A word or two must be said about the use of the réseau in analytical photogram-
metry. It is unfortunate that while this device has been in use for many years
in Great Britain, it is only relatively recently that it has been incorporated in
cameras having modern high performance lenses.^ It has therefore been hitherto
impossible to carry out realistic tests to decide whether and by what amount the
use of the réseau improves aerial triangulation with good photography. The
opportunity occurred to do such a test in Southern Rhodesia. The Surveyor
General was anxious to see whether analytical techniques could be used for
providing coordinated property boundaries; and he initiated an experiment with
the cooperation of University College London and Messrs. Fairey Surveys who
undertook the photography and provided an observer for making the observations
on the Hilger and Watts stereo-comparator at University College. The camera
used was the Hilger and Watts 105 fitted with the Aviogon lens and réseau; and
the computations were carried out by Mr. P. Boniface on the University of London
Ferranti-Mercury Computer.
For a number of reasons it has been impossible to present a complete analysis of
the results for this conference, but one set of results is available and is instructive.
The photographs for this set were taken at a height of 7500 feet giving a scale of
1/15,000 with the six inch lens. There were 12 photographs in two strips of six
pictures forming a roughly rectangular area with a perimeter control of six points.
All control and check points were pre-marked. The adjustment was carried out
by the Amer method.
Each photograph to be observed was placed on a stageplate carrying an accurate
centimetre grid so that the lines of this grid were displaced by a small amount
from the réseau crosses appearing on the pictures. For every photographic point
observed, the operator measured the difference of coordinates to the nearest réseau
intersection and also to the nearest grid intersection. Pointing errors to photo
graphic points were thus common to the réseau and non-réseau observations and
differences in the results could thus be ascribed absolutely to the use or non-use
of the réseau.
The results, on an adequately large number of check points, were very good
indeed, the root mean square vector errors at plate scale being + 13 ( u with the
réseau and + 12/x without. The difference may be statistically significant but it
is clearly unimportant; but what is interesting and requires explanation is that
the réseau gives no improvement. The same diapositives used in the Thompson-
- Watts plotter gave a vector error of + 18 u which, while larger, is still very good
for an analogue instrument. Unfortunately, as we have said, a full analysis has
not been done, and explanation of these results must be to some extent conjecture.