s, Vol XXXV, Part B4. Istanbul 2004 Inter
International Archives of the Photogrammetry, Remote Sensing and Spatial In ormation Science
8 ; 8 l |
Traditionally, their work is focusing on the technical part of — 3.2.1 The MultirepresentationalRelation: Basically, links
interoperability, i.e. on the specification of data and component could only be represented as a list containing matching pairs
interfaces. With the abstract specification concerning semantics (see figure 3). However, the links between coinciding
and information communities (OGC 1999), the OGC published representations can be further described. In our case, this is 3.2.2
first ideas concerning the realization of semantic necessary because we want to have information about the integ
interoperability. In this document, a system has been proposed degree of similarity of corresponding instances. For this reason, usefi
to reduce the loss of information while transferring data from we introduced a so-called MultirepresentationalRelation object featı
one user community to the other one. However, up to now no to connect multiple representations. Within such an object, all avail
steps have been taken to implement the concept. information on how representations are related can be stored. Its of a
attributes contain the general, geometric, topologic and upde
semantic/meta properties of the representations taking part. appl
3. LINKING AND RELATING MULTIPLE Some of them are listed below: thres
REPRESENTATION DATABASES degr
General attributes: van
In this section it is outlined on which levels spatial databases e The IDs of the corresponding objects. sche
containing multiple representations can be linked. Then, the e The cardinality of the relation.
links or relations, respectively, that can exist on the instance G à 4 ; A lo
level are investigated and some tasks and problems that have to eometric attı ; es: ; can
be dealt with in this context are identified. e Geometric types of corresponding features (see figure 4). t
i : LineString MultiLineString of a
3.1 Levels of integration etc.
mtr) <> 0—9— ©
, ; hides : : Pol A mea:
The integration of different spatial databases.can be performed MultiPolygon MultiPoint simi
on different levels, as it can be depicted from figure 2. E] e e a we
e v
: |
| Schema | «ti, | Schema | : B : : :
Figure 4. Features of different geometric types can take part in a
IN relation. As
: (2) : stati:
Object € 7 Object e Geometric resolution and position accuracy of mea:
geometry | semantics geometry | semantics corresponding objects (e.g. one geometry has been
captured in a large scale of 1:1000, another one in a 3.2.0
(2a) 4 A = 4 (2b) smaller scale of 1:25 000). corr
e Geometric comparisons of corresponding features, amb
(3) depending on the geometric shape (e.g. angles, distance can
measures between objects, comparisons of their area, their feat
length, etc.; see figure 6). attril
Figure 2. Integration of spatial databases on different levels. we
Line Geometric shape relations relat
On the one hand, the different object classes (and attributes) of
the source schemas can be linked (1). On the other hand, the
object instances themselves can be matched by looking at their 1 « Hausdorff distance
a nene. d e length difference
} e average line distance
geometric (including topologic) (2a) and/or semantic properties 3
(2b). Finally, links could only be set up between the geometries
(3) of two data sets. Figure 3 shows that there can be differences Polygon e area difference
between the results of (2) and (3). e centroid distance
= 1. e Hausdorff distance
Geometric links , Object links e
bann QUORUM der Object M Figure 6. Various measures to compare the similarity of
: "— KO ; C1, C2 {e1s Ca} geometric shapes are available. P
ore > >
: 1 d. do. di Objects N,P Topologic attributes: Fi
ned sed adonde ( di, do; ds, da} e Number of adjacent or incident features of corresponding
Object N Object P objects. [mpi
e Their graph-based topology indicators like e.g. the Cons
Figure 3. Geometric linking and object linking can lead to reachability or eccentricity of nodes that constitute an amb
different results. edge, etc. oe
class
s : Semantic/meta attributes: than
3.2 Links between instances e Affiliation of corresponding features to an object class in impt
their source data sets
In our approach, we want to use the results of the integration on
the object level in order to derive an integration on the schema
level. Therefore, we first have to figure out how instances can
be linked or related. Some of the tasks and problems that have
to be faced here are outlined in the following sections.
e Number of corresponding attributes or attribute values of
corresponding features.
e metadata like information about the spatial reference
systems or data quality parameters, e.g. the means of
154
