sur-
ige of
2s Can
areas
REKLIP
)).
i from
Mitte-
laa oos H..
atolo-
sensor
at-TM.
p. 173
h, H.-
e Ana-
edings
P-414,
1996.
ERS-1
he ag-
Par-
mental
tions,
otter-
A EUROPEAN PROPOSAL FOR TERMS OF REFERENCE IN DATA FUSION
L. Wald
Groupe Télédétection & Modélisation, Ecole des Mines de Paris
BP 207, 06904 Sophia Antipolis cedex, France
Commission VII
KEY WORDS: data fusion, concepts, definitions, remote sensing, EARSeL
ABSTRACT
Data fusion is a subject becoming increasingly relevant as scientists try to extract more and more information from remotely sensed
data. Archives are growing, as well as the number of space missions devoted to Earth observation. It is generally correct to assume
that improvements in terms of classification error probability, rejection rate, and interpretation robustness, can only be achieved at
the expenses of additional independent data delivered by more separate sensors. Sensor data fusion allows to formalise the
combination of these measurements, as well as to monitor the quality of information in the course of the fusion process.
A Special Group of Interest 'data fusion' has been established jointly within the European Association of the Remote Sensing
Laboratories (EARSeL) and the French Society for Electricity and Electronics (SEE). This Group has defined several major tasks to
be handled in order to increase our understanding and use of data fusion. One of these tasks is the establishment of terms of reference
that are accepted by both the scientific and the industrial communities at least in Europe.
A definition of the data fusion is proposed, which allows to set up a conceptual approach to the fusion of Earth observation data by
putting an emphasis on the framework and on the fundamentals in remote sensing underlying data fusion. Several other definitions
'are given which are useful to describe any problem of data fusion.
RESUMÉ-
La fusion de données revét une importance grandissante au fur et à mesure que les scientifiques tentent d'extraire de plus en plus
d'information des mesures et images d'observation de la Terre. La taille des archives croit rapidement, de méme que le nombre de
missions spatiales dédiées au domaine. Il est généralement observé que les améliorations en termes de qualité, taux de classification,
taux de rejet, robustesse d'analyse ou d'interprétation, ..., sont atteintes par l'augmentation des observations independantes effectuées
par des instruments indépendants. La fusion de données permet de formaliser la maniére dont les observations sont combinées, et
également de mieux suivre l'évolution de la qualité au cours des différents processus de traitement.
Un groupe de travail (SIG) 'fusion de données' a été créé au sein de l'association européenne des laboratoires de télédétection
(EARSeL) et de la société francaise d'électricité et d'électronique (SEE). Ce groupe a défini plusieurs táches principales à traiter afin
d'augmenter notre compréhension de la fusion de données et son usage. L'une d'entre elles est l'établissement de termes de
référence, qui soient acceptés par les communautés scientifiques et industrielles, tout au moins en Europe. Cette communication
présente ces termes. La définition proposée pour la fusion de données, met l'accent sur la formalisation de la fusion de données et sur
les problémes fondamentaux sous-jacents en télédétection, et par conséquent, permet de définir une approche conceptuelle de ce
domaine. On définit également d'autres termes qui sont utiles dans la description des problémes de fusion de données.
1 INTRODUCTION
Data fusion is a very recent word. It means an approach to
information extraction spontaneously adopted in several
domains. An illustration is given by the human system which
calls upon its different senses, its memory and its reasoning
capabilities to perform deductions from the information it
perceives. However the operation by itself is not new in remote
sensing: classification procedures are performed for more than
twenty years, and are obviously relevant to data fusion (see e.g.,
Mangolini 1994, Pohl 1996).
The quantity of information available to describe our
environment increases rapidly. Archives are growing, as well as
the number of space missions devoted to Earth observation.
Accordingly, data fusion is a subject becoming increasingly
relevant as scientists try to extract more and more information
from remotely sensed data using the concept of synergy. Sensor
data fusion allows to formalise the combination of these
measurements, as well as to monitor the quality of information
in the course of the fusion process.
The European Association of Remote Sensing Laboratories
(EARSeL) and the French society for Electricity and
Electronics (SEE) are fully aware of the importance of data
fusion in remote sensing. They jointly organised a conference in
1996, called ‘Fusion of Earth Data - merging point
measurements, raster maps and remotely sensed images', which
was the first of a series with a venue every 2 years. During the
round table of that meeting, the need for a working group,
called special interest group (SIG) in the EARSeL jargon, was
expressed, and it was created later this year.
2. THE CREATION OF A SIG WITHIN EARSeL - SEE
The SIG 'data fusion' will contribute to a better understanding
and use of data fusion by tackling the fundamentals in remote
sensing underlying data fusion. Further the SIG will certainly
be helpful in designing and engineering tools and methods for
assessing a priori or a posteriori the quality of a fusion product,
that is answering the following questions: is it worth
performing a fusion process ? was it worth doing it ?
During the 1996 conference, it has been proposed to restrict the
SIG to the so-called radiometric aspects. The so-called
geometrical aspects are already tackled by several working
groups at national and international levels. The exact meaning
of these words (radiometric, geometrical) is still unknown to me
International Archives of Photogrammetry and Remote Sensing. Vol. XXXII, Part 7, Budapest, 1998 651
