each of the above steps, blunders may be identified and the
system will automatically take the user to the desired images
and to the desired points. The system then presents the user
with the appropriate options. Thus in only seconds a blunder
can be remedied on multiple images or a point can be replaced.
There is no burden of reloading images or redoing interior
orientation. Again, the images are "rectified" to aid in the
remeasurement process. An interesting point that is being
brought out in the literature is that digital triangulation may be
more accurate in practice than conventional triangulation. This
may be primarily attributed to the ease with which the user can
correct very tiny blunders. In analytical triangulation on
analytical plotters or comparators, the time to reset stereo
models and fix tiny blunders is considerable. Thus small
blunders would be allowed to stay in the bundle adjustment as
long as the overall accuracy of a project was being met. With
digital photogrammetry and the judicious use of disk systems,
small blunders can be readily fixed with a few "button clicks"
within very large image blocks.
3.4 Further Automation is Currently in Testing
An even more automated method for block triangulation is
starting to be used which calls for the automatic elimination of
blunders and no remeasurement is generally necessary. This is
because the automatic measurement process has measured
many more points in the block than are necessary. This high
redundancy permits more reliable blunder elimination using
automatic methods. Thus, blunders can be eliminated and
sufficient points are still remaining so as to yield a high quality
triangulation result. This method can improve the timeline even
further since the semi-automatic remeasurement is not
necessary or is substantially reduced. More will be written on
this technique in the near future. It has already been tried out
successfully by customers. Today, substantial automation has
been obtained in digital triangulation and much more can and
will be done.
3.5 Usage of orientation parameters
Although it is an intrinsic part of digital photograrametry, one
often forgets the importance of the property that, once
triangulation has been completed, all models are available for
immediate use on any workstation in the network. No further
interior, relative or absolute orientation work is required. We
have also added a Model Manager, in which pairs of images are
predefined as models; thereafter the operator can switch
between models with a delay of only about one second - much
faster that on analytical plotters with big stage plates like the
BCIS and DSR15-18, and with more than two models too.
Software has also been written for exchange of orientation data
between SOCET SET and various other workstations, for
example many Leica analytical stereoplotters and the DVP.
4.0 GENERATION OF DIGITAL TERRAIN MODELS
4.1 Automatic Elevation Extraction
Automatic elevation extraction using area and/or feature based
methods have been around for some time and continue to
improve. Much has been published about these methods, for
example by Helava (1988) and Miller and DeVenecia (1992),
252
and many users are using them daily to produce elevation
products such as contours and digital terrain models (DTM).
These methods have brought huge improvements in productivity
for specific photogrammetric projects. In particular, small scale
and/or open terrain conditions are very well suited to automatic
elevation measurement. In many of these cases, automation is
two to ten times faster than conventional contouring or profiling
(Miller, Walker and Walsh, 1995).
Confidence in these automated approaches has grown to the
extent that they have been applied in change detection studies,
Customers have generated DTMs from photography acquired at
different epochs in order to pinpoint urban change in Japan or
forest change in Oregon and New Zealand. The objective of the
latter is to quantify both cleared areas and growth; with
properly chosen post spacing and computational strategies,
change in the forest canopy can be estimated accurately enough
to compute timber volume (Carson, Miller and Walker, 1996).
4.2 Adaptive Automatic Elevation Extraction
Leica-Helava continue to invest in this automatic process and
many small practical improvements have been made. These
include semi-automatic editing tools such as the automated
removal of trees and buildings from a DTM when the bare
ground is desired. Batch editing facilities have been added
whereby previously acquired planimetric data such as ditches
or stone walls in rural areas can be used to eliminate errors in
the DTM caused by these features. Currently, we are testing an
expert system based approach to automatic elevation
production. This approach, which we are currently calling
Adaptive Automatic Terrain Extraction (AATE), brings several
practical benefits to the production flow. These include:
1. One step processing with multiple images: A single job or
map sheet can be generated in one batch process. Multiple
images can be selected covering the job in a more or less
random configuration. An expert system is used to decide
which images should be used and how to proceed to
produce the best data. Image reshaping and image pyramid
creation are performed as necessary.
2. Expert system driven adaptability: An expert system uses
only a few inputs from the user and many automatically
derived inputs to measure the surface. This makes the
process much more adaptive to changing image and terrain
characteristics. Thus, the output is more accurate and more
successful.
3. Ease of use: Due to the adaptability of the expert system,
the user no longer needs to prespecify large numbers of
inputs such as multiple regions and strategies. Large areas
can be covered in one step without pre-rectification and
without post processing to merge the resulting files.
These improvements allow the computer to work harder while
the user can concentrate on those areas where human intuition
is required. The improvements will also minimizes disk usage.
The result is a more streamlined production flow and higher
quality data for less cost.
5.0 ORTHOPHOTOS AND MOSAICKING
A significant use of current digital photogrammetric systems is
for the production of orthogonally corrected images or image
maps. Although this process is theoretically straightforward,
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B2. Vienna 1996
provi
time
amot
often
b-— "FX Wax am 4
510
To ir
explo
more
the pr
proce:
and c
as be
DTM
ortho]
a sing
5.2 R:
As m
image
acute
Intera
à mos
user t
obtain
seldor
One r
proble
standa
each i
the al;
to reel
border
part o
radion
The r:
portior
is peri
mosaic
differe
reduce
of very
3.3 Se:
Depen
conten
à Varie
Some c
buildin
Teason:
