A. Meid (1991) reports on the integration of knowledge based
techniques for the measurement of image coordinates. With his
approach error sources can be kept minimum or remaining
errors will be eliminated.
The digital orthophoto is meanwhile a standard tool of
photogrammetry (H.P. Baehr/T. Voegtle, 1991) - a cost-
benefit analysis was given by H.P. Baehr/J. Wiesel (in H.
Ebner et al, 1991) It is no isolated product of
photogrammetry; in GIS it serves as orientation map for vector
data in which it is handled as an additional layer. Using these
orthoprojections of image data in postprocessing bottom-up
procedures nodes and segments of edges can be extracted,
which are the basis in further vectorisation processes (D.
Fritsch in H. Ebner, 1991).
Image matching was also deepened in the period of 1988 -
1992. While R. Li (1990) investigates area and edge matching
techniques the research work carried out by C. Heipke (1991),
B. Straub (1991) and C.T. Schneider (1991) give first
experiences when using object space image matching. The
same approach is under investigation at the Institute of
Photogrammetry and Cartography, Technical University of
Darmstadt (B.P. Wrobel et al. in W. Foerstner/S. Ruwiedel,
1992). One main problem in this matching definition is the
computer effort for the solution of large linear equation
systems which might be overcomed by adequate solution
strategies such as multigrid and by regularisation techniques;
furthermore the progress in computer architectures (e.g.
Advanced RISC) gives its full support. J. Piechel (1991) points
out an increase in accuracy for automatic digital terrain model
generation if matching is solved area-based.
Image matching is not only interesting from a theoretical point
of view: F. Ackermann/M. Hahn (in H. Ebner et al., 1991) and
R. Krzystek/D. Wild (in W. Foerstner/S. Ruwiedel, 1992)
present an operational program package for image matching
which uses a chain of several algorithms in a multiresolution
mode. The chain consists of the setup of the image pyramide,
image normalization, feature detection and location, and the
matching process itself. In this sense, large scale aerial
photographs can be processed leading to very accurate DTM's.
Problems still unsolved are the location and elimination of
man-made constructions and other natural phenemena for
example trees, bushes etc.. But it should be noted that today
image matching techniques represent also standard tools of
digital photogrammetry.
Using the reseach and experience in image matching one
further step forward is image understanding which extracts
geometry and semantics from digital images. Also here some
research activities were initiated by german scientists - a good
review of existing techniques can be found in W. Foerstner
(1991). Other contributions to this subject are given by the
proceedings of the workshop on Robust Computer Vision
which was organized by WG III/2 (W. Foerstner/S. Ruwiedel,
1992). Image understanding will become more and more
important; it is expected that for future periods
photogrammetric research is deepened here.
3. Development of Remote Sensing
Reserach carried out in remote sensing belongs to the work
coordinated by Commission VII ISPRS. But also for
Commission III some references will be cited in particular
those dealing mostly with algorithmic aspects. One important
topic is the radiometric adjustment of different image sceneries
also called mosaicing which is a problem to be overcomed not
38
only in processing of remote sensing data but also in digital
orthoprojections. The contribution given by M. Kaehler (1989)
shows radiometric corrections for satellite imagery -the
textbook of J. Albertz (1991) points out furthermore this
problem.
Within data analysis of remote sensing the classification
remains still a problem which is not yet sufficiently solved.
Classification algorithms structure pixel oriented remote
sensing images according to objects introduced before. This
object oriented approach is the basis for an integration of
remote sensing data in GIS. J. Albertz (1991) and W. Goepfert
(1991) present classical algorithms for supervised and
unsupervised classification; a new model for supervised
classification can be found in H. Schumacher (1991). The
application of aerial image matching techniques has been
transferred to SPOT data by C. Heipke/W. Kornus (in H.
Ebner et al., 1991); it shows that matching problems in aerial
photogrammetry and remote sensing might be solved using the
same software.
4. Development of GIS and Digital Mapping
Progress in GIS is concentrated on several topics: first it is
well-known that photogrammetry delivers three-dimensional
coordinates nowadays to be handled by two-dimensional GIS-
products. The task is the extension of GIS theory by at least
the third dimension in geometry. A review and solution
strategies of this problem are given by D. Fritsch (1991) - it is
pointed out that interfacing a DTM with GIS is one of three
solutions. In this context further DTM research can be seen: H.
Kuhn (1989) deals with visualization aspects, and W.
Reinhardt (1991) proposes strategies for an interactive DTM
setup.
The integration of image data in GIS presupposes a DTM as
reference surface. Thus, not only image data can be rectified
but also image understanding will have some profits
integrating the third dimension. D. Fritsch (in M. Schilcher/D.
Fritsch, 1989) presents a three level approach for image
integration: the lowest level is given by rectified image data,
the medium level manages object oriented raster data, and the
top level extracts vectorial geometry which is semantically
structured. The dissertation of H. Yang (1992) fits exactly into
the medium level, because homogeneous raster objects can be
coded very efficiently by quadtree structures.
But only with the highest (top) level of image data integration
spatial databases of GIS can be filled and updated. Therefore
we have an analogy to the image understanding task which
should not be seen as inverse problem but as information
fusion using GIS data.
5. Education and Presentation of Research
Within the period 1988 - 1992 several new textbooks in
photogrammetry, remote sensing and GIS have been published
written by german scientists. These books could fill
deficiencies in scientific literature and help to improve the
education considerably. The following textbooks should be
named: H.P. Baehr/T. Voegtle (1991) and W. Goepfert (1991)
in digital image processing, J. Albertz (1991) in interpretation
of aerial photographs and satellite images, and R. Bill/D.
Fritsch (1991) with introductory readings in GIS. As further
progress in education some tutorials were organized. Not only
the intercommission WG III/VI did education work here; also
the WG' s itself introduced some workshops with tutorials. The
situation in education within 1988 - 1992 was much better than
in the quadrennium before: it started with a tutorial on
EO ode 7M mod D ML]
x —RL
e
jme mh Pe V 9 PPM
ne am — uh | J- AS pj uud