Tnm; m a ll NE A €
re | Tr ES N
SW P) WES -— Rn gh.
Figure 7. Unsupervised classification based on different
Based on the knowledge obtained during the previous exercises
in the courses SATK1 and SATKA2, the studens produce an
enhanced False Colour Composite suitable for general visual
interpretation. A colour-hardcopy, which is later used during the
field check, is made by the Tektronix ink-jet plotter.
The main contents of this exercise is unsupervised classi-
fication or "clustering". As for the other exercises in the SATK-
courses, this one is also carried out in groups of two and two
students. Each group shall carry out two unsupervised
classifications. The algorithms ISODATA and CLUSTR are
used.The number of classes is however not the same for the
different groups. By comparing the results, it is then possible
to get an impression of how the choice of algorithm and
number of classes influences the result. An example is shown
in Figure 7.
The result of the unsupervised classification is also produced as
a colour hardcopy, which is used during the field-work. With
support from the already existing vegetation map and
observations in the field, the aim is to combine the "spectral-
classes" from the unsupervised classification into so-called
"information classes". Eventually the final result is also
produced as a hardcopy on the Tektronix colour ink-jet plotter.
clustering algorithms.
Research and development activities related to ERDAS.
ERDAS is also considered as a useful tool in connection with
the research and development activities related to satellite-
mapping which are carried out at the Department of Surveying.
A comparative study of the effect of different rectification
procedures applied to SPOT data is now carried out utilizing
ERDAS to a large extent. In connection with the evaluation of
the geometric quality in this project, the "chip extraction"
capability in GCP (new in version 7.4) has shown most useful.
As part of the same study, a computer program for automatic
measurement of geometric deviations between images rectified
with different procedures, based on image correlation, has been
developed. The program itself runs outside ERDAS, but it uses
387
ERDAS -.LAN and -.GCP files as input. The result is written
to a -.GCP file. It is then possible to use the GCPERR program
for further analysis of the results. (For some reason GCPERR
is not included in ERDAS versions later than 7.3. This is
considered as unfortunate, as the lack of this program reduces
the total functionality of the system).
In connection with teaching and training given outside the
university, often only standard VGA graphic capability is
available for presentation of satellite images and thematic maps
in raster format. Therefore software which makes it possible to
use standard VGA graphics for the presentation of -.GIS/-.TRL
files has been developed. VGA presentation of -.LAN files can
be obtained by conversion via -.GIS/-TRL files using
RGBCLUS and similar methods. In order to utilize equipment
for slide production in the "DTP-world" outside the university,
software for conversion of -.LAN files to 24 bit TARGA
format has also been developed.
Concluding remarks.
In general we consider PC-ERDAS as a well developed and
"mature" system. The way of thinking which has been the basis
for the design of the system's functionality, has evidently been
strongly influenced by wide experience within the field of map-
and satellite-image application. PC-ERDAS, even without
LAPLINK, offers good possibilities for integration of image
data and GIS.
Since geodesy and photogrammetry are central diciplines at the
Department of Surveying, problems related to geometry play an
important role also in the teaching of satellite mapping. As
described above, ERDAS is a useful tool in this connection.
There seems however to be some inconsistency between the
different ERDAS programs related to geometric rectification.
The constants in transformation equations are for example
denoted differently in the various programs and supplementary
documentation. This leads to some confusion for the students,
and should be improved in future revisions. Dealing with
geometric rectification, we would also like to point out that
what we miss in PC-ERDAS, is a real "split screen" function-
ality, which can be useful, particularly in connection with
image to image registration.
We have got the understanding that version 7.5 is the last stage
in the development of the ERDAS software under MS-DOS.
Transition to "Imagine" under UNIX for PC users implies
considerble investments. For many university institutions like
ours, funding is mostly rather limited. In an educational
environment, continued use of PC-ERDAS may in addition to
the economical also have some other advantages. The studens
are familiar with MS-DOS and the menu-driven user interface
is well suited for teaching. From this point of view it should be
possible to teach digital satellite image processing in a
meaningful way with PC-ERDAS well into the next century.
On the other hand, the lack of a graphic user interface, as for
example the one in "Imagine", will give the students the
impression that they are working with obsolete equipment,
which may have a bad influence on their motivation and their
interest in the subject. It is therefore hoped that ERDAS will
reconsider to develop new versions running under MS-DOS,
tentatively with a graphic user interface.