The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B4. Beiiins.2008
intersection matrix represents always the true intersections of
the interiors and boundaries of regions. That way the
interpretation of the set of H R relations has to be done by an
algorithm of the application itself. Consequently the algorithm
interprets the respective relation correctly. For example the
intersection-matrix for the situation between operational area 1
and 2 is presented in equation 13.
(13)
The meaning of this matrix is that the intersection of the interior
is empty. However, the intersection of the closure between both
is non-empty. A common point on the closure is existent and
the respective relation is touch. But the algorithm identifies this
set of intersections as disjoint, because the intersections
between the interior and the closure as well as the interiors of
both are empty.
For the application in the disaster domain using the IM and an
algorithm that is able to identify the set of H R relations correctly
has two advantages. The first one is that the IM is widely
implemented in GIS and spatial databases. The second one is
that this GIS component can then also be used for visualization
and user feedback.
2.3 Types of Reasoning
The spatial reasoning process has to provide the needed
information based on the present information state of the
database as well as general and context knowledge. An example
for such reasoning is the query for all operational areas which
are in a damage site containing a fire (cf. Figure 1). For solving
this problem two approaches are possible.
The first one is the geometric approach which is compulsory for
GIS. A spatial query algorithm checks if a point location with
the attribute fire is inside the closure of a region with the
attribute damage site. The next step is to find all operational
areas, which are also inside this damage site.
The second approach is a more elegant way of processing this
question on the level of the ontology. The knowledge base
contains general knowledge about the domain as well as
specific knowledge about a situation. The general domain
knowledge is defined a-priori by modeling the ontology
accordingly (cf. chapter 2.1). For example the part of relation in
the ontology between the two classes damage site and
operational area corresponds to the topological relation inside.
Such relations are universally valid in the whole domain. The
disjoint relation between several operational areas as well as the
contained relation between an operational area and an event are
modeled in a similar manner (cf. Figure 5). In contrast dynamic
knowledge, which is also given by the knowledgebase, is only
valid for a specific situation at a specific moment. For the
example the specific fire event is related to operational area 3
by the inside relation.
All relations which are given by the knowledge base for the
situation of Figure 1, are represented in Figure 5 by solid lines.
That way answering the query for all operational areas which
are in the damage site of the fire is possible. The fire is inside
operational area 3 which is also inside the damage site. Again
all operational areas of this damage site can be provided. The
advantage of this type of reasoning is that the relations are
always present and do not have to be evaluated geometrically.
Figure 5. Ontology based inference net for topological relations
of the spatial scene in Figure 1 (used shortcuts of the H R cf.
Figure 4)
Figure 6. GIS supported inference net for the special scene of
Figure 1 (used shortcuts of the fI R cf. Figure 4)
Another advantage of the reasoning process lies in conditions of
plausible combinations of relations, so called compositions
(dashed lines in Figure 5). A plausible combination is given,
when the relation can be identified unambiguous. For example
when the fire is inside the operational area 3 and operational