MODEL-BASED ASSISTANCE
FOR ANALYZING REMOTE SENSOR DATA
Wolf-Fritz Riekert (Siemens, Munich and FAW, Ulm, West Germany)
Thomas Ruwwe, Günter Hess (FAW, Ulm, West Germany)
Bitnet: RIEKERT@DULFAW1A, RUWWE@DULFAW1A, HESS@DULFAW1A
ABSTRACT
The global objective of the RESEDA (REmote SEnsor Data Analysis) project, being conducted at the FAW Research
Institute for Applied Knowledge Processing in Ulm (West Germany), is to make remote sensing technology available
to a larger group of users in environmental management.* For this purpose, a knowledge-based advisory system is being
developed, called the RESEDA Assistant. Processing models represented in a knowledge base are available to the
RESEDA Assistant, which is thus able to facilitate the use of software tools for image processing or handling of spatial
data.
KEY WORDS: remote sensor data analysis, advisory system, knowledge-based, processing model
1. REMOTE SENSING AS A KNOWLEDGE-
BASED ANALYSIS TASK
Remote sensing produces a large amount of data that is
relevant to the state of the environment. Analyzing this
data requires both a great deal of computational power
and a high degree of knowledge. Although the first
requirement may be met by using conventional hardware
and software, the second is very demanding on the
experts working in this field. Because the number of
qualified experts is small, there is a need for automated
techniques that make remote sensing technology availa
ble to a wider community of users.
The goal of remote sensing in environmental protection is
always to derive a certain piece of geographic informa
tion. Moreover, the remote sensing data to be analyzed
may itself be considered as a kind of geographic informa
tion. That is, the task of processing remote sensing data is
a typical task of transforming and analyzing geographic
information. Remote sensing data as well as ancillary
spatial and factual data describing geographic entities are
input into the analysis process. The output of the analysis
consists of environmental data related to the geographic
entities to be analyzed. The analysis is controlled by the
expert’s knowledge about concepts and methods of remo
te sensing, image processing, and the geo-sciences. In
RESEDA, we are trying to represent these concepts and
methods in the knowledge base of an expert system
(figure 1) (Riekert, 1990).
Human experts in remote sensing are able to find a
computational pathway from the source data to the target
data. They know how to apply an appropriate sequence of
image processing methods and manipulations of spatial
data. This ability is based on two kinds of structural
knowledge:
Knowledge about remote sensing targets and their
features, which are to be analyzed and computed in
the course of the analysis.
Knowledge about dependencies between these fea
tures, implicating algorithms suited to compute cer
tain features from one another.
In RESEDA, an object-oriented formalism is used to
describe this structural knowledge:
- The various remote sensing targets and their features
are represented by abstract target classes and target
attributes. The concrete manifestations of these two
concepts appear in the form of geographic data-,
these are called classifications and attributions.
The dependencies between features and the algo
rithms to compute them are represented in an object-
oriented form by abstract processing models and
concrete computations of geographic data.
In addition to the object-oriented representation techni
que, rules are used to describe the conditions under which
a processing model is adequate to compute certain geo
graphic data and what the constraints are between the
input and the output data of such a computation.
* The RESEDA project is supported by the Federal State of Baden-
Württemberg and by Siemens, Munich
146