The International Archives of the Photogrammetrv. Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B4. Beiiine 2008
1092
The regions are both static a-priori defined regions with
political background like cities and districts (solid lines) and
dynamic regions of the emergency administration, like damage
sites and operational areas (OA) (dashed lines).
The binary topological relations of the diverse regions are easy
to identify for human operators. The fire is inside the
operational area 3 which is disjoint to the operational area 2 and
contained by the damage site. The damage site intersects district
A and B and all regions are inside the city and so on.
Figure 1. A fire event (AM) and the resulting interactions of
involved administrative regions (OA - operational area)
But first of all a definition of the degree of the needed
topological relations as well as the region characteristic is
necessary. Based on the situation in the figure above, equation 1
defines a set fl R of essential topological relations in R 2 , which
are adequate for representing domain specific spatial scenes.
ù R ={disjoint, intersect, contain, inside, equal} (1)
Further possible relations like touch* (OA 2 touches OA3),
cover (damage site covers OA3) and covered by (inverse
relation to cover) are excluded deliberately because no benefit
of gaining further information was found. As a basis restriction
regions have to be regular closed and without holes.
For processing, the information content of the scene in Figure 1
is also represented in a knowledge base, here the domain
ontology DM 2 . It has to be emphasized that modeling relations
in such ontologies has a semantic background which is quite
different from the topological one. Semantic relations are
focused on entities and objects in general. So, a semantic
relation is for example the part-of relationship of a fire engine
and a fire brigade (a composite of engines) which does not
contain topological information about the spatial situation of
both. It is not possible to know at a specific point in time, if the
fire engine is contained by, or disjoint to the convoy of the fire
brigade. The needed information content for a correct
* The touch relation is not necessary, because neighbourhood
relations are defined separately.
topological representation is derivable by using background and
context knowledge which is also inherent in the ontology. For
example it is possible to deduce the location of a fire engine
based on information of its activity. That means that an engine
is located, where it extinguishes a fire. Its location is therefore
independent of the location of the fire brigade.
The scene information of Figure 1 is modeled in the DM 2
according to the class-scheme in Figure 2. The five object
classes of fire, operational area, damage site, district and city
are obviously required. For modeling the complete semantic of
this situation, the classes context, building and address are also
necessary. Semantic modeling means here that more
information is represented in the knowledge base than is
inferable by the visualization solely. That way an event in
general relates to an object which is affected, here a building.
The relation between building and fire in the DM 2 is affected by
and the inverse relation is affects. Building itself has a
geographical location as well as a “semantic” location, the
address. The relation between them is an instance of
relationship in the ontology. That way the content given by the
ontology holds much more potentially useful information for a
respective reasoning process.
Figure 2. Part of the class scheme of the DM 2 , in IDEF1X-
notation
This kind of object modeling and reasoning is a bit more
complex than a spatial query function in a GIS for relating a fire
and the respective city, but it is done with respect to the systems
semantic and the mental model of the staff members. This is
important for adopting the reasoning process.
2.2 Topological Reasoning
Diverse methods are available for different ways of topological
reasoning. They can be categorized in the three different groups
of position based methods, dimension extended methods and
calculus based methods. Position based methods derive the
respective topological relation between regions from the
position to each other. Such methods like the weighted
walkthroughs model (WWM) (Cicerone and Clementini, 2003)
as well as the model of ternary projective relations (TPR)
(Billen and Clementini, 2004) are suitable for topological
reasoning about dynamic and moving regions. More popular are
dimension extended methods like the intersection model (IM)
(Egenhofer and Herring, 1991) and calculus based methods like
the region connection calculus (RCC) (Renz, 2002). The latter
two concepts of representing a set of topological relations are
quite different.
The formal categorization of topological relations by the IM is
based on a comparison between the interiors Sx and the
boundaries x° of objects. That way, a 4-intersection-matrix