ACQUISITION OF DATA IN AEROTRIANGULATION
coordinates to better than 5 microns.
It is not a vain exercise to emphasize the
importance of the plan for taking the photo
graphs. One is usually satisfied with parallel
strips with the longitudinal and lateral over
lap as small as possible for reasons of econ
omy, but this small overlap makes it neces
sary to select the pass points at the extreme
edges of the format, in the areas where the
resolution is poorest. Besides, this plan has
the disadvantage of not permitting the cal
culation of a block in the air, without using
any known points: the transmission of lateral
strip tilt transversely from one strip to an
other is not really assured, since the tie points
between the strips are nearly alined. The
vertical control network on the ground should
then necessarily include, at the extremities of
each strip, the points necessary to define the
lateral tilt, which would be an oppressive
requirement, whereas an essentially rigid
block can be supported by a small number of
points distributed in random fashion. The
adoption of a lateral overlap greater than 50
percent provides this rigidity; if also the
camera stations were alined laterally per
pendicular to the direction of the strips, one
would obtain an ideal arrangement (this
correspondence can be assured easily enough
by adopting a longitudinal overlap of 90 per
cent and by choosing afterwards the photos
which are suitable). But such an arrangement
for the taking of the photographs has not been
used very much, except for very limited
tests, because of the excess of work which it
imposes, notably in the instrumental phases
as a consequence of the increase in the num
ber of photographs to be treated. This in
crease in work can, however, be maintained at
an acceptable level by the use of super-wide-
angle cameras, and it would be largely com
pensated by the near-elimination of field work
and by the gain in accuracy due to the redun
dancy of the system.
A less laborious solution for assuring the
internal rigidity of the block is to use cross
strips, long known and successfully tested in
the research of the ISP on the Massif Central
test area. It is truly surprising that such a
simple and efficient technique is not used
more often,
It may be well to note that a judiciously
conceived plan for taking photographs will
effectively reduce the propagation of certain
systematic errors: the taking of photographs
in parallel strips with alternately reversed
headings minimizes the effect of horizontal
curvature and the effect of progressive varia
tion in scale; if there are transverse strips,
systematic torsion can also be eliminated.
But no flight scheme will eliminate the ver
tical curvature; this must be taken care of by
elevation measurements on the ground or by
the use of statoscopes.
Instrumental Procedures
From the earliest days, there has been a
general tendency toward simplification of
instrumental procedures, achieved at the cost
of an increase in the volume of calculations.
The earliest methods required the use of
universal plotting instruments, with which
the models of each strip were formed succes
sively, the scale and the absolute orientation
being effectively transmitted from model to
model. But as early as 1948 there appeared at
IGN the method called constant altitude in
which no attempt was made to transmit
longitudinal tilt, the Bz component being
maintained equal to zero. Very soon the
models were formed with the By component
equal to zero and with a constant Bx compo
nent in order to maintain as constant as
possible the effect of the instrumental errors.
Still later, the transmission of transverse tilt
was also abandoned, ending in the method
called independent models.
The transmission of absolute orientation
from model n— 1, n to model n, re + 1 can be
done in two ways:
(1) . By measuring the parameters defining the
orientation of the camera n in each of the two
models. These parameters can, for example,
be the angles of inclination of the principal rays,
referred to the reference system of the instru
ment. If the instrument does not have graduated
circles or levels for making this measurement,
the orientation can be defined by the parameters
governing two particular perspective rays,
corresponding, for example, to two marks on the
edge of the camera frame or to points marked on
the photographs. It is sufficient to align these
points monocularly, at an arbitrary distance z
and to record the x and y instrumental coordi
nates as if they were ordinary control points.
(2) . The second procedure does not require
any special measurements: the transmission of
the orientation can, in effect, be calculated from
the spatial coordinates x, y, z of at least two
points in the two models, referred to the com
mon perspective center.
The two methods are not in all respects
equally rigorous. The first involves only
directions, and is not affected by errors of
identification: the transmission of orientation
is, in effect, based entirely on the reduction of
the transverse parallax. The second, on the
contrary, is concerned with points, intersec
tions of two corresponding rays; but these
rays will not intersect exactly because there is
always some residual parallax, so that the
3