the need for a new implementation should cause de
signers to think again about the best way to do things. It
should not be forgotten that relative and absolute orien
tation as two separate operations were conceived for
analogue instruments. Lohman reports in [14] problems
of using relative orientation with narrow angle photo
graphy such as the Russian KFA 1000 images. With the
advent of automatic selection of conjugate points, such
points will not be in the classical (von Gruber) positions,
but on the other hand different cameras may not always
have suitable geometry for bundle solutions and we still
may not have control points. Platform (aerial and space-
borne) positions may be well known with the use of GPS
or refined orbit determination techniques as used with
ERS-1. Tilt is now found accurately with INS and it may
be necessary for only a few control points over large
areas. The fundamental methods of orientation in a
plotting instrument should be re-evaluated to consider a
combination of techniques, based perhaps on a bundle
adjustment as the general case, but also providing for
single models to be set up without control points. The
operations of inner orientation and relative orientation will
be automatic and even ground control point identification
may become automated.
The key to the use of digital systems is automation.
When the goal of full automation is reached, workstations
will no longer be required! In the meantime however
human interaction is necessary. Leberl states in [13] that
,,No digital elevation models can be generated with high
quality by automated image matching: attempts to
automate are typically contaminated by bad matches so
that manual fixes are needed, thereby destroying any
throughput advantage the automation may have prom
ised". This is true only up to a point. It is not true for close
range work and all advantage will not be destroyed.
Farrow and Murray report an interest in automatic DEM
production at Ordnance Survey [6], but for specific tasks
such as revision of existing DEMs and to use a digital
workstation to help solve problems caused by vegetation
and man made features. Rosenholm reports (from per
sonal communication) that very reliable DEMs can be
derived from SPOT data in the production environment
of the Swedish Space Corporation, with accuracies of
6 - 7 m but that half of the production time will be spent
in post processing and manual editing. Considerable
attention is being paid to validation techniques which
make use of consistency checks (ridges and water
courses) and visualisation.
Feature extraction is still far from automatic although
work by Förstner, discussed in this issue of NGT Geo-
desia (p. 372), Schenk [19] and McKeown [15] are laying
the foundation for advances in this area.
Users
Digital systems are now used for map production in DMA
and systems are being developed at other major agen
cies. Some of the issues concerning this have been
discussed above. One of the key changes which photo-
grammetrists must face is the establishment of infor
mation systems and their use by professionals in a range
of disciplines. The dominant of these is GIS. Photogram-
metry provides a means of creating and revising the data
in a GIS. Satellite data is also an important component of
a GIS, and that comes in raster form and needs cor
recting to be registered with other data. However GIS
data is fluid, it is by its nature changing, as the features
it describes change, so the revision of the data becomes
the most important operation. The operation of data revi
sion is continuous and must take place on the data itself,
thus the revision operation must be part of the system.
Photogrammetry thus becomes a small part of a large
system in which all data is in digital form and in which
raster data plays an important part. It is therefore almost
inevitable that photogrammetry will be used to provide
digital data and that many of the operations will be auto
matic and not require the traditional skills of the photo-
grammetric operator.
In the area of CAD/CAM, photogrammetry is the front end
of an end to end system in which robot cameras obtain
the images which are measured with interactive tools and
compared with the design or as built surveys. Chapman
and others have developed HAZMAP for modelling of
nuclear plant [2]. This type of activity has a great potential
in construction, manufacturing industry and utilities and
will also remove the photogrammetrist from the data
acquisition loop. A similar system has been designed for
highway surveys. Novak has described a system for a
mobile mapping system [17] in which various data from
sensors such as GPS, inertial navigation devices and
video cameras, is integrated to produce 3D data for a
GIS.
Where are we going?
This section can be divided into two; first a discussion of
the technical issues and second a consideration of the
future role of photogrammetrists.
There can be no doubt that a technical revolution has
taken place. Within a space of 20 years photogrammetry
has developed from a science predominated by the pro
duction of topographic maps on analogue instruments to
one applying itself to a much greater range of problems,
most of which require spatial data in digital form. The
technology exists to permit the analytical stereo plotter to
be superseded by fully digital systems. However the tra
ditional map producer, particularly the national mapping
organizations do not yet see significant advantages in
taking the next step away from the traditional user inter
face. It is only an interface which is different: the orien
tation software and data capture techniques are basically
the same. Kolbusz in a review [11] of the proceedings
of the ISPRS Symposium on Digital Photogrammetric
Systems [5] may have identified the reason for this. He
concluded by notingYour reviewer failed to find in this
collection of papers sufficient critical assessment of what
the universities and vendors are now offering the users.
Do the systems work and are the user interfaces and
reliability acceptable?" In reality the economic case has
to made and the keys to this are reliability of new tech
nology and efficiency compared to the current equip
ment. Organisations still remember the teething problems
of introducing analytical plotters, often these were seem
ingly minor problems with computer interfaces or coping
with the infrequent, but awkward, problem. New users,
those primarily concerned with acquiring data to solve
there own problems are likely to be more receptive. To a
GIS-user who needs three dimensional data to correct or
help analysis of his data, a digital workstation into which
digital images are read and which outputs 3D raster data,
such a system is ideal. To the engineer coping with the
370
NGT GEODESIA 93 - 8
