mm E E UE
Objectives
The main field of interest of the WG VII/4 is the development
of data fusion and combination techniques for multisensoral
data analysis and the development of algorithms to analyse
spatial structure diversity. With the availability of low to very
high spatial resolution satellite data with different spectral
characteristics and the disposition of several radar satellites,
the use of information from different data types for a certain
application becomes more and more attractive. In order to
extract the needed information, data fusion and combina-
tion techniques will be of increasing importance. The issue
will undoubtedly continue to challenge researchers in years
to come. Along with the disposition of many very high spa-
tial resolution satellites within the next decade, the structure
information in satellite data will become most important for
many applications. Even though there are some texture
algorithms already available, the development in this direc-
tion is still a challenge, especially for the very high resolution
datasets, which will have a different statistical behaviour
from data with 20m and 30m spatial resolution. The subject
provides a wide range of activities to be undertaken in dif-
ferent application fields or interrelated techniques, like digi-
tal photogrammetry.
The main objective of the working group is to co-ordinate
the efforts of researchers and developers in the aforemen-
tioned fields and support the interaction between algorithm
developer and the application side. This should be sup-
ported by exchanging ideas in the form of working group
meetings and circular letters. It is also planned to co-ordi-
nate meetings between working groups of different organi-
sations, like EARSeL and IUFRO.
The discussion during the WG VII/3 Workshop ‘Sensor
fusion and advanced classification algorithms’ was focused
mainly on the sensor fusion topic. After the DLR presenta-
tion entitled “Overview of DLR-forest projects and future
perspectives" delivered by Wolfgang Steinborn, the follow-
ing topics were discussed:
Sensor Fusion:
"Operational issues of multisensor data fusion for visual
image exploitation" Werner Schneider, University of
Vienna, Austria, *Image information fusion in remote sens-
ing: towards a framework and a consistent terminology"
Roland Fritz, FeLis, University of Freiburg, Germany,
"Practical Application of Multisensor Data Fusion for For-
est Inventory Mapping" Mathias Schardt, Joanneum
Research, Austria, "Combining Satellite data and auxiliary
GIS data".
Advanced Classification Algorithms and Procedures "Eval-
uation of the KNN method for combining NFI sample plot
data and satellite data” Matthias Dees, Felis, University of
Freiburg, Germany, "Integrating satellite and GIS data into
a large scale sample-based forest inventory - the classical
sample-based approach” Klaus Steinnocher, Department
of Environmental Planning, Austria, “Feature-based image
fusion” Alois J. Sieber, JRC, Italy, “Needs for data fusion in
the area of landmine survey and detection” Silvana
Dellepiane, University of Genova, Italy and Gianni Ver-
nazza, University of Cagliari, Italy and “Model regularisa-
tion in remote-sensing image analysis” Tobias W. Kellen-
berger, RSL, Switzerland.
The discussion showed that all participants agreed that sen-
sor/data fusion will be one of the important topics within the
next years. According to the amount of earth observation
208
satellites already in orbit and to the future satellite pro-
gramme the topic of fusing the information from different
satellites will gain increasing relevance. It was agreed that
until now the remote sensing society is missing a standard-
isation of definitions in the field of data fusion. For example:
- what kind of data are included in the topic data fusion,
- outline of the benefits of sensor and data fusion tech-
niques
- outline of examples for educational purposes
It was pointed out that there is still a gap between devel-
oper of fusion algorithms and users. In order to improve
the contact
- there should be more joint meetings between devel-
oper of fusion algorithms and users.
- new algorithms of the developer community should be
implemented in standard software and be available as
public domain software.
Only if the algorithms are provided to the large group of
users they will be used and verified. The algorithms must
be transparent to the users to estimate the reliability of the
results. The developer is the algorithm developer, the sci-
entific users are defined as the scientists working in the
field of remote sensing applications and the end-user is
from the application side, which wants to use the informa-
tion from remote sensing without being involved in remote
sensing.
The final proposal of the working group meeting was:
- There is a need for the working group to address the
topic data fusion also in the future
The approach of the working group to the topic should be
a scientific user stand-point, complementary to the algo-
rithm developer groups
The ISPRS WG VII/4 participated in the ECO BP'98 Sym-
posium from 1st — 4th of September in Budapest. WG VII/4
had two sessions, one under the topic of advanced classi-
fication techniques and one under the topic of automatic
image analysis. Altogether, nine proposals were submitted.
There was some emphasis on presentations dealing with
neural network classifiers and spectral unmixing. During
the meeting, the first call of the planned Joint ISPRS and
EARSeL workshop which took place from the 3rd to 4th of
June 98 in Valladolid, in Spain was presented.
The working group was active in setting up the above
mentioned joint workshop on “Fusion of sensor data,
knowledge sources and algorithms for extraction and
classification of topographic objects". The programme
and content was circulated around by e-mail and mail.
The workshop was in co-operation with WG Ill/5, WG
IV/III.2 and the EARSeL SIG Group on 'Data Fusion' which
is a continuation of the workshop on 'New classification
algorithms and data fusion' which took place in Freiburg
in 1997. This was a joint meeting between WG VII/4 and
the EARSeL SIG group ‘Forestry and Land use Planning’.
The Proceedings are available at the Dept. of Remote
Sensing and Landscape Information Systems, University
Freiburg.
The EARSeL/ISPRS workshop on "Fusion Sensor Data,
Knowledge Resources and Algorithms for Extraction and
Classification of Topographic Objects” took place from 3rd-
4th of June in Valladolid, Spain. The workshop was organ-
ised from ISPRS WG Ill/5 “Remote Sensing and Vision
Theories for Automatic Scene Interpretation", WG 1V/3
“Integration of Image Analysis and GIS” and WG VII/4
“Computer Assisted Image Interpretation and Analysis”
International Archives of Photogrammerty and Remote Sensing. Vol. XXXIII, Part A. Amsterdam 2000.