PHOTOGRAMMETRIC ENGINEERING
rotation thus calculated would be slightly
different from that which would be obtained
from the first method—besides, it would be
directly affected by the errors in identification
of the points common to the two models, and
must therefore be considered as less reliable.
The basic concept of the method of inde
pendent models is that it is always better to
make a measurement and derive a correction
for it mathematically than to apply the cor
rection manually. One is then justified in
asking if it is really necessary to obtain a per
fect orientation, and if it is not preferable to
simply measure the residual parallaxes, the
relative orientation being then refined by the
calculations in the course of treatment of the
data. To this concept, proposed by Inghilleri,
one can object that the currently available
instruments, even if they do permit measure
ment of the parallax, are not generally
equipped for automatic recording of this
parameter so that it must be done manually,
with consequent risk of error and waste of
time.
Pursuing this direction still further, one can
eliminate relative orientation completely, and
reduce the parallaxes, by means, for example,
of By ,which can be recorded at the same time
as the coordinates x, y, and z. These quantities
define the directions of the two corresponding
perspective rays and are equivalent to mea
surements made on the stereocomparator: it
is by this method (mentioned as a matter of
interest) that the first trials of a so-called
analytical method were carried out by I.G.N.
in 1957.
The Problem of Transferring Points in
Analytical Methods
The logical end result of this evolution is
the measurement of photographic coordinates
x,y for each control point on all the photo
graphs upon wihch it appears, with the aid of
a comparator. But it is not enough to make
these measurements with the greatest pos
sible precision; it is also necessary to be cer
tain that exactly the same point is observed
on all the photographs, because any error in
identification will have exactly the same effect
as a measurement error. It appears, then,
that the transferring of control points is a
critical problem in all instrumental proce
dures and that this transferring should be
done with a precision equal to that of the
measurements.
When they were first introduced, analytical
methods were designed to duplicate mathe
matically the operations that were normally
done by plotting instruments, that is to say,
the formation of the spatial model; this idea
led to the use of stereocomparators which, by
stereoscopic fusion, assured correct identifica
tion of corresponding points in the stereo
pair. But it is also necessary to be certain of
the identification of points common to neigh
boring stereopairs in the strip; stereocom
parators having three plateholders have been
designed for this purpose. But there is a less
expensive solution, applicable to instruments
having only two plateholders. In certain of
these instruments, the movements of the
plateholders are independent and the co
ordinates x, y, and x 2 , y 2 are recorded directly.
In passing to the next stereopair, it is easy to
recover the previous coordinates on the com
mon plateholder to assure identification, the
stereoscopic setting then being accomplished
by adjusting the other plateholder. However,
experience shows that the execution of a
setting with four independent movements is
not very convenient. Other stereocomparators
are provided with an xy-movement acting on
both plateholders together and a differential
movement dx, dy between them; it permits
convenient setting but does not allow recov
ery of a point already observed in the preced
ing stereopair by simple resetting of the coor
dinates. The IGN uses a stereocomparator
designed especially to solve this problem; the
recorded coordinates are indeed x, y, and x 2 ,
y 2 , but there is a differential control capable of
adjusting the left or the right plateholder
at will.
There remains still the problem of identi
fication of points on different strips, a prob
lem which has taken on greater importance
with the generalization of treatment in blocks
by purely analytical methods.
The simplest solution is evidently to mark
all control points on the ground before taking
the photographs, with targets visible on the
photos. But this solution is not always possi
ble; besides, it entails considerable work
which is scarcely justified except for large-
scale surveys, and it eliminates, at least par
tially, the principal advantage ol aerotriangu-
lation which is that it reduces the amount of
field work. Another inconvenience, not neg
ligible, is that it imposes difficult require
ments of precision on the taking of the photo
graphs if it is desired to have the points fall
exactly at preplanned locations. Finally,
experience indicates that pretargeted points
are sometimes difficult to recover.
On the other hand, pretargeting permits
the use of a simple monocomparator, a com
paratively low-cost instrument which gives
perfectly independent measurements.
4