Corresponding Author
1091
MODELING SPATIAL SCENES IN DISASTER DOMAIN ONTOLOGIES
C. Lucas 3 *, S. Werder b
3 Institute of Photogrammetry and Remote Sensing (IPF), Universität Karlsruhe (TH), Kaiserstraße 12, 76128
Karlsruhe, Germany - christian.lucas@ipf.uni-karlsruhe.de
b Institute of Cartography and Geoinformatics (IKG), Leibniz Universität Hannover, Appelstraße 9a, 30167 Hannover,
Germany - stefan.werder@ikg.uni-hannover.de
Commission IV, WG IV/8
KEY WORDS: Disaster Management, Topological Relations, Spatial Modeling, Image Understanding, Geographic
Information Science
ABSTRACT:
Disasters create extremely dynamic situations which have to be handled by emergency operations centers. For fulfilling their task
they rely on up-to-date information from on-site units and passer-bys. Every individual has its own mental image of the situation
which is recorded for situation evaluation as free-form text messages. In order to enable further automatic reasoning processes, this
mental image must be taken into account. This paper focuses on modeling the topological and neighborhood relations of disaster
situations. Also the spatial reasoning process and the important aspects of the used ontology are highlighted. A short introduction to
orientation and distance aspects completes the required components for modeling spatial scenes as a whole.
1. INTRODUCTION
1.1 Background Situation
In the disaster management domain it is fundamental to
visualize detailed up-to-date information of the situation.
Thereby a situation map is an important decision base for an
emergency operations center (EOC) and provides information
sharing between the management staff. The up-to-date
information originates from several on-site units and passer-bys
located at diverse damage sites. Normally it is recorded as free
form text messages. This variety of incoming messages has to
be analyzed with respect to their visualization by one operator
of the management staff. With regard to digitally distributed
situation maps and in order to assist the updates, the aim of this
project is to apply an automated system for simplifying and
speeding up the message analysis.
In order to implement the approach of a human operator with
methods of information technology, several processing steps are
necessary. Firstly, sentence detection and information extraction
play a crucial role for formalizing and analyzing the message
content. Subsequently, a step of semantic augmentation is
necessary to harmonize the content, solve semantic gaps, and
provide content based spatial reasoning. The final step is to
create a graphical representation of the relevant information
with respect to the domain specifics.
Fundamental for processing, particularly for semantic
augmentation, is a knowledge base. An ontology designed with
domain specific considerations is used to provide the necessary
knowledge. This includes background and context information
about objects as well as the relations between them. The
developed ontology, named Disaster Management Data Model
(DM 2 ), was derived from the Command and Control
Information Exchange Data Model (C2IEDM), used for military
interoperability in the NATO (cf. Lucas et al. 2007).
1.2 Spatial Scenes
This paper focuses on modeling spatial aspects of objects in
disaster domain specific ontologies. In this context the common
spatial attributes of objects are their location as well as their
geometric attributes (form, size and feature alignment). These
elementary attributes are traditionally provided in spatial
ontologies as well as geographic information systems (GIS) and
allow describing discrete objects unambiguous by their
dimension and location in space.
Nevertheless, in order to support a spatial reasoning process for
disaster events based on textual descriptions, a more
comprehensive level of spatial information is necessary. The
method of object modeling within the ontology has to be
comparable with the mental model of the reporting person
(Frank, 1998). A mental model represents the level of
geographic perception of an average citizen, and characterizes
thus the perspective view of an observer. This mental image of
the situation is incomplete, represents objects which seem
important to the observer and does not include detailed metrics
(Barkowsky, 2002). This model contains besides discrete
objects also spatial scenes with interactions of two or more
spatial objects in the meaning of neighborhoods or part-of-
relations. Thus a detailed information level is essential for
analyzing e.g. coherences of cities, districts, damage sites, and
operation areas as well as for solving ambiguities. According to
that, the spatial attributes of topology, neighborhood,
orientation, and distance have to be taken into account for
defining spatial scenes.
2. TOPOLOGICAL RELATIONS
2.1 Domain Requirements
The initial information state of a typical situation of the disaster
management domain is presented in Figure 1. The situation is
composed of a fire event and the respective affected regions.