itional
:anical
analog
e very
, While
impro-
en the
iomical
perties
' being
1 This
o far.
' flexi-
e fully
o meet
control
1omical
uld be
° total
SUCCESS
ppears
special
Several
1doubt-
means
ethods.
mputer
curacy
though
nviron-
je elec-
'ercome
'e close
ogram-
nciples.
making
hat the
plotter
ctically
e digits
It also
gid nu-
THE DESIGN OF PHOTOGRAMMETRIC PLOTTERS, HELAVA 123
merical methods may be used for determining the values of various parameters by the
method of least squares. The most important advantage, however, is that digital prin-
ciple is most suitable for the solution of the data processing problem of photogrammetric
plotting “in toto”. This refers particularly to the problem of automation. A digital in-
strument may be made to follow a comprehensive and versatile program, make deci-
sions, and manipulate the data accordingly. For these reasons the use of digital computa-
tion as the basic computation method appears imperative in most advanced plotters and
mapping systems.
The most serious limitation of digital methods has been their speed. This speed is
very high, but still not necessarily high enough for fast and “continuous” plotting. The
formulas to be solved are quite complicated, and for an adequate degree of fidelity in
tracing the details these formulas must be solved tens or hundreds of times in a second.
The seriousness of the speed problem depends partly upon the magnitudes of the para-
meters and mainly upon the intended speed of plotting. The solution of this problem may
be found by employing suitable computational techniques and by accepting a favourable
design “logic” for the plotter. The problem of the design will be considered in a later
section of this paper. In the following paragraphs a few of the most important general
features of computation techniques will be discussed.
The digital computers may be divided into two groups according to their internal
method of handling the number or word transfers. These groups are:
1. the “entire-word-transfer” computers and
,
2. the “incremental-transfer” computers.
The second group may be divided again into two sections:
2a. fixed increment computers;
2b. variable increment computers.
The entire-word-transfer computers are best known. They are also often called gen-
eral purpose computers, although a “general purpose” machine could also be based on the
incremental principle. The entire-word-transfer computer handles the data words or
numbers as units of a certain length of digits. Thus the inputs to each computer opera-
tion are the complete values of numbers, and whenever changes of the input occur the
entire calculation cycle must be repeated. In general application this method is very
favorable because it lends itself well to the programming of various problems. In special
applications, however, it has definite drawbacks, e.g., if the input data is of such a nature
that it changes relatively slowly without discontinuities, a considerable amount of time
is lost because complete numbers must be handled although only their least significant
digit has changed. Nevertheless, even computers of this type are now sufficiently fast for
the photogrammetric plotting problem, provided that the plotting speed is not excessive.
At the same time, being versatile as far as programming is concerned, they offer con-
venient means for the solution of fringe problems, such as least squares adjustments, etc.
The incremental computers are based on “partial-word-transfer”. Their arithmetic
accepts incremental changes in the values of the variables as basic information. This
leads to a computer that on the one hand is less flexible from a programming point of
view, but on the other hand loses no time in repeating calculations which have already
been done. This type of computer is very suitable for process control operations where
variables change continuously. In this sense the photogrammetric plotting process offers
a most suitable application for incremental computation.
The incremental computers are usually based on the principle of fixed increments.
This simplifies the computer design, but at the same time brings about some difficulties
if the input values are changing faster than in a normal operation. The “slewing rate”
can be greatly improved by using a parallel mode of operation. However, this means that
the computer must become larger and more expensive. The use of a computer of this kind
has been considered in more detail in reference [5].