Database Reference
In-Depth Information
are overlayed. Theme overlays allow users to view and explore geographic data
in different contexts. Being able to visualize data in context is an important
functionality in integrating diverse types of geospatial data.
A theme can be represented by either vector data or field-based data. A
road network with roads being individual features, for example, would be rep-
resented as vector data, whereas a vegetation index would be represented as
field-based data (more specifically a raster image). If the data for the lay-
ers come from different sources, two problems can occur. First, the system
used to integrate the data has to be
interoperable
with the other systems the
geospatial data are retrieved from. Here,
interoperability
means that systems
can exchange information and data using standard protocols and formats.
As we will discuss below, a high degree of interoperability can be achieved
when distributed and heterogeneous geospatial data sources can be uniformly
accessed using geospatial Web services.
Second, the geospatial data may come in different formats with conflicting
structures and semantics. For example, if two sources provide vector data for
the same theme and region, the data might conflict in terms of their spa-
tial components as well as their descriptive components (see Section 10.2.1).
Such a situation can even occur if both datasets are based on the same pro-
jection (spatial reference system), have been georeferenced/aligned, and have
the same scale (spatial resolution). Re-projection, georeferencing, and scaling
are tasks that are frequently used in the context of remotely sensed imagery
and are typically performed on the datasets prior to their overlay or integra-
tion. Another typical example often occurring in practice is when some raster
imagery is overlayed with vector data. Phenomena in the image might not
align or match up with the features modeled by the vector data. Approaches
to resolving these types of conflicts are known as
conflation
, meaning to “re-
place two or more versions of the same information with a single version that
reflects the pooling, weighted averaging, of the sources.”
21
The key in dealing with conflicting spatial components of two or more
datasets to be integrated is to make use of the location information asso-
ciated with geospatial objects (and cells/pixels in a raster image), something
unique to spatial datasets. Several approaches have been proposed that deal
with the integration of vector data and road maps in particular and the com-
bination of imagery with vector data.
48
-
50
More fine-grained approaches have
been developed for finding corresponding objects in datasets to be integrated.
Corresponding objects (features) represent the same real-world entity but are
possibly misaligned across different data sources. Approaches for point-based
data have been presented by Beeri et al.
51
,
52
referred to as
location-based join
.
Related approaches are so-called entity resolution techniques, which try to de-
termine the true location of a real-world entity in case geospatial data about
the entity comes from a collection of data sources.
53
For resolving spatial conflicts, that is, if the same real-world entity has
conflicting feature location information in the different sources, nonspatial
attributes associated with the features can help in resolving such conflicts. For
