PHOrOGRAMMETRIC ENGINEERING
lation than from a model-by-model triangu
lation. However, for the latter he employs
the modified Herget method in which the strip
coordinates of points in the preceding model
are enforced during the orientation of a photo
graph. Besides, his conclusion is based upon
the size of systematic errors and not upon the
residuals after polynomial strip adjustments.
It is claimed as an advantage of triplet
orientation that it is here possible to recog
nize and eliminate errors in the x-coordinates
of points in the triplet overlap. In the two-
photo orientation recognition is possible by
comparing the heights of these points in the
two models in which they occur. Errors are
more evident than in the triplet orientation
because here the least squares adjustment
does not minimize them. Points with such
errors are here simply eliminated from use
in the scale transfer and, if the x-error causes
F-parallax, also from the relative orienta
tion.
On balance, the triangulation by means of
independent relative orientation of each two
successive photographs and scaling of the
resulting models has the advantage that it
uses simpler formulas and requires only half
the computation time [69] of the triplet tri
angulation. The latter, in the version used by
Tewinkel and in the version recommended
by McNair, has the advantage that a
smoother fit between models is obtained
automatically.
3. RADIAL TRIANGULATION
Numerical radial triangulation is discussed
in ref. [70] to [76]. Roelofs and Timmerman
describe results and the influence of errors if
the classical rhomboid triangulation is used.
Van den Hout describes the use of triangles
in a block adjustment by the Anblock method.
The lack of general availability of the exten
sive literature on the rhomboid triangulation
may explain recent interest in the numerical
formulation of the old graphical radial tri
angulation by means of alternate resection
and intersection (ref. [74] to [76]).
Polynomial Adjustment of Strips
References [17], [36], [38], [39], [69], and
[77] to [94] deal with the adjustment of strips,
individually or in blocks, by means of poly
nomial transformations. The adjustment
serves to correct the strips for deformation
and to obtain a reasonable fit of the trans
formed strips at the ground-control points
and at the tie points.
Reference [84 a to f] have made it gradually
clear that conformal transformations by
means of polynomials of higher than the first
degree do not exist. Therefore, in practice the
adjustment of a strip is performed by means
of various nearly conformal transformations.
It can be performed as a sequence of two-
dimensional transformations (see e.g. [85],
[86]) or of transformations of planimetry and
of heights (see e.g. [82], [88]), in each case
with appropriate correction of the remain
ing coordinate or coordinates. Alternatively,
after an initial positioning, it can be per
formed as one three-dimensional transforma
tion (see e.g. [36] and [84a]). A proper correc
tion for strip deformation cannot be guaran
teed if independent polynomials are used for
all three coordinates [99] or for planimetry
and heights [69].
Before a block adjustment is performed,
the strips must first be positioned approxi
mately. In this positioning, a second-degree
correction for longitudinal curvature should
be included. This makes it possible to perform
the block adjustment of planimetry and that
of heights separately.
The transformation of planimetrie ground
coordinates of control points to the axis-of-
flight system of a strip is rather common
(see e.g. [82]) but is rather awkward if a
block adjustment is performed. It can be
avoided either by using the known param
eters in the formulas for the initial position
ing directly in the correction equations and
applying the polynomial transformations to
axis-of-flight coordinates [36] or by applying
a conformal transformation to the coordinates
obtained after the initial positioning [85, 86].
As in the cases of block adjustment of
photographs and of models, normal equations
can be formed for the simultaneous adjust
ment of all strips. The direct solution of the
equations is here rather simple [85], [87]. Al
ternatively, the iterative procedure can be
used in which strip after strip is transformed
and tie points from overlapping strips are
used as additional control points [36], [78],
[85]. The iterative procedure is simpler to
program than the simultaneous solution but
it consumes more computer time. With the
amount of control that is commonly available
in topographic mapping, as a rule a sufficient
convergence of the iterative procedure is ob
tained after about ten iterations of the plani
metrie adjustment and five iterations of the
height adjustment [35], [85]. Tewinkel [9]
and Jacobs [80] even state that after careful
positioning of the individual strips there is
often little need for a block adjustment.
Practice as well as theory have long since
shown that it is advisable to keep the degrees