9 -
ee also [50] - [54]).
d in [54] and [55].
eing abandoned in
t been intended to
Tectronic compu -
Ling situation for
it electronic compu-
puters ( [55] ). Thus
le computation with
the theoretical ba-
's are equivalent
y will, for quite so-
omputation is not yet
~y with simple elec-
[56] ). In view of
mputers it must, ho-
r ith slight modifica -
riment a new analo-
cross strips (after
ent of heights with
> yield weight coef-
Dnic computation
n on strip-adjust-
f aerial triangula -
putation. It seems
ik-adjustment is an
ical approach, prin-
edures, program -
the considerable
e not yet led to prae
ure ( [60] - [71] and
ite where the practi-
iow a matter of rou-
Lth plate coordinates
hmid [39] and De
'ammetric applica -
is been published as
puters are used in
idjustment procedu -
; and yields simulta-
y Proctor and Amer
([56], [72] - [74]). Extensive experiments about the convergence of the iterations have been
carried out. By application of overcorrection factors the originally encountered number of about
80 iterations per block has been brought down to the order of 20 iterations for blocks of about
100 - 200 models (see also Muller [70]). A high speed computer is needed in order to keep the
procedure economical.
ITC in Delft is using a block adjustment procedure developed by Van den Hout ( see
[75] ). Up to now it is programmed only for planimetric adjustment. It is also working with in
dependent stereo-models as basic units. The numerical solution of the normal equations is di
rect, non-iterative. It is achieved with a computer (Stantec Zebra) ranging at the lower end of
middle class computers. At present blocks up to about 500 models can be treated.
From a third group of block-adjustment procedures, which use strips as adjustment
units, the procedures of Schut and Bervoets are in routine application. These procedures ad
just up to now planimetry and height separately. They correct the strips in a block by polyno -
mial formulae of 2 n< ^ or 3 rc * degree which are determined together in the block connection. The
numerical solution is iterative. Procedures and test results are described in [15], [76] - [8l] ;
see also [82].
The majority of adjustment procedures which are at present either in practical use or
in an experimental stage of development have been and are developed practically independent
from each other .Up to now no comparative evaluation, neither theoretical nor experimental, has
been attempted although a number of distinctions are obvious, and are generally known. In de
veloping and evaluating block-adjustment procedures one is concerned with a number of pro -
blems and considerations of different nature. Considering the items listed under 1. 2 as a basis
of discussion may help to bring forward the most important points of view.
3. 4 The theoretical approach to block adjustment has been classified (see [1] and [7 3] ) accor
ding to the elementary units which have to be tied together by the adjustment. One has distin
guished between procedures working with single photographs, stereomodels, sections, or strips.
The list might be completed by mentioning triplets (3 photographs) and multiplets (i. e. 9 pho
tographs). Generally speaking any form of subblock may be used.
Regarding this search for the most suitable unit of adjustment a few general remarks
may be made : The most general and hence the most detailed and most accurate approach to
block-adjustment is at present given by the use of single photographs (bundles of rays) as basic
adjustment units, with plate coordinates as uncorrelated observations. Given this system as a
reference one can consider any other system as being derived from it, as any other unit can be
derived from single bundles of rays. According to the rules of least squares adjustment any ad
justment procedure working with derived elements remains completely rigorous and thus iden
tical with the direct solution of the single plates approach in case the error properties (i. e.
the correlations) of the derived elements are taken into account. It is even possible to achieve
a rigorous solution when working with units which have been formed by allowing approximations
and simplifications. In such a case the subsequent adjustment can still rectify these approxima
tions. Thus as far as basic least squares theory is’concerned the search for various adjust
ment units is not substantial. It is even not substantial - still referring to a completely rigo -
rous solution - with regard to the problem of numerical treatment. The total amount of nume -
rical work in an adjustment in phases is generally not less than that in the direct approach.
It should be pointed out, however, that a sophisticated combination of simplified deri
ved units and relaxation procedures might give advantages in the numerical treatment and still
achieve results which are equivalent with a rigorous solution. It is conceivable that ultimately
the most efficient numerical solution of the general task of block-adjustment would be such a
" hybride system ", which would also incorporate the problem of obtaining approximate values
of the unknowns.
Regarding the present situation, the use of various adjustment units other than single
plates becomes important, however, at a second stage of thought, namely when certain appro-