eration of analogue plotters displayed very little basic
difference to the early instruments. The development of
cameras followed a similar pattern and today's aerial
cameras are highly refined, computerised versions of
Messter's automatic film camera of 1915.
The development of computational techniques is closely
tied to the development of the computing equipment.
Early methods for aerial triangulation in the 1930s only
developed rapidly during the 1960s when the electronic
computer made their use for production work a practical
possibility. The analytical plotter invented by Helava in
1958 also developed relatively slowly and it was not until
1976 that the instrument became fully accepted and it
took a further decade and a half to be accepted as the
standard plotting system. The developments from the
nineteenth century to the 1980's can be considered to be
evolutionary. Since photography was invented there has
been no great, sudden change, and certainly no revo
lution. This view has also been put forward by Leberl [13]
who notes that analytical photogrammetry is being made
obsolete after only 20 years of accepted application.
Development has speeded up over the last 30 years, as
indeed have most aspects of our lives, but the evolution
has been steady. I believe that we now face a much more
drastic change. For many years photogrammetry was a
branch of surveying which used photographs, and al
though the technique was used for mapping or measuring
features other than the terrain, these tended to be
exceptions. In 1966 the editor of The Photogrammetric
Record bemoaned the fact that useful non-topographic
photogrammetric techniques were not being adopted. He
quotes Deville: There is such a fascinating simplicity
about the method that it is at first difficult to understand
the reasons which prevent its adoption". Deville con
cluded that the apparent simplicity was a delusion. This
has now changed because photogrammetry can be used
almost as a black box technique and the mathematics are
hidden and old skills of the operator not required. There
have been cases of 'photogrammetry' being reinvented
because of the need in a new field of application. Geo
graphical Information Systems (GISs) need data and
environmental scientists need terrain models on local,
regional and global scales. Manufacturers and software
houses, previously not interested in the relatively small
photogrammetric market, are designing systems to pro
duce three dimensional data from stereo pairs and soft
ware to produce digital ortho-images. In order to see
where these developments are leading, let us look more
objectively at the current state of photogrammetry.
Where are we now?
Data
The data which is available is coming from a much wider
range of sources than previously and is of more variable
quality [21]. Aerial cameras with forward motion compen
sation (FMC) and tilt stabilisation (AMC) are producing
very high quality images with very low distortion which
extend the range of use of photogrammetry. This high
quality of photography should be matched by suitable
plotting equipment. At the same time much less expen
sive cameras are producing images which can be used
photogrammetrically because of the software available to
correct such images. Digital cameras are also producing
data at close ranges either directly over small formats or
by scanning over a longer period of time, for example the
Zeiss UMK Linescan. It will not be long before aerial
cameras are available with large CCD arrays; cameras
with up to 5000 x 5000 pixels are available but at a very
high cost.
Photographic cameras from space provide high quality
images at a low cost. Konecny quotes US$ 0.13 per km2
for mapping with KFA-2000 photography [12], (compared
to US$ 0.62 for SPOT and US$ 10 per km2 for aerial
photography,) but the shortage of suitable platforms limits
their use. The availability of Russian photography, now
with 2 m resolution, may result in an increase in use of
this type of data.
Digital data from satellite platforms is one of the major
catalysts for change. It is almost universally recognised
that it is possible to obtain more information from SPOT
data in its original digital form than in a hard copy. There
may still be good reasons for using hardcopy such as
non-availability of digital instruments with stereo viewing,
or a desire to use conventional methods such as the
analytical plotting instruments. A number of tests failed to
show clear improvements in accuracy of standard photo
grammetric techniques such as aerial triangulation, by
using digital data, (for example, in [16] evaluating the
Helava DCCS, and in [3] evaluating triangulation of SPOT
data), but there are clear advantages in speed and cost
in the production of Digital Elevation Models (DEMs) and
image maps, although the reliability of DEMs derived
automatically is not yet equivalent to that obtained by
manual means.
Many new sensors are now planned with a range of new
applications: along track stereo systems such as JERS-1
OPS and MOMS-02; JERS-1 SAR, ERS-1 SAR and
Radarsat are producing or will produce microwave data
which can be obtained independent of cloud cover and
have been shown to produce DEMs from interferometric
methods; and further in the future high resolution optical
systems.
A third new data source is the scanner. Expensive high
cost scanners have been available for a long time but
scanners have only been developed for photogrammetry
relatively recently. Zeiss and Intergraph announced their
scanner in 1989, but since then the Vexcel VX3000
scanner and the Wehrli Photoscanner PQ-2000 have
become available. It is also possible to use low cost
scanners developed for publishing for photogrammetric
work. Sarjakoski has calibrated the Sharp JX-600 [18]
and shown that it can be calibrated and the corrected
data give accuracies of 0.2 pixels. He also suggests that
the scanner could be improved with little difficulty. Now
scanners with 1200 dpi (dots per inch) are on the market.
Huurneman has described a low cost small format
scanner producing 5000 dpi [10].
The use of scanners produces problems. In order to ob
tain images with resolution to compare with photographs,
pixels of 5 - 10 nm are required: a single 230 mm x 230
mm photograph digitised at 10 ^m produces 0.5 Gbytes
of data, and thus produces problems in storage and
access. Such a small pixel size is not needed for many
automatic processes and may not be needed for all
display purposes, therefore some interactive operation is
required to give the high accuracy only where needed.
368
NGT GEODESIA 93 - 8