Geography Reference
In-Depth Information
mosaicing, two digital maps are integrated to produce a new map (Walter and
Fritsch
1999
; Devogele
2002
; Gal et al.
2003
; Gravano et al.
2003
). Another such
area of use is in data fusion, where raster data received from sensors, is processed,
by means of image-processing techniques, and then integrated (Laurini
1998
;
Lemarie and Raynal
1996
; Sester et al.
1998
). Most of these examples work upon
algorithms for discovering corresponding objects and that can be a possible part of
the solution to these problems. Whereas in geospatial location based problems such
as used in this work, it has been necessary to investigate location-based join of open
source geo-spatial datasets. For the current research work datasets from Open Street
Map (OSM), Geonames and Natural Earth have been used from which an integrated
dataset has been created.
There have been studies (Papakonstantinou et al.
1996
; Park
2001
; Safra and
Doytsher
2006b
; Spaccapietra and Parent
1994
) where two join approaches,
namely, the sequential and the holistic approaches, are presented and compared.
The novelty of those works had been in developing, for each of the two approaches,
effective join algorithms that use only locations of objects. They showed that the
sequential normalized weights method is effective, that is, the result has a high
recall precision combination when the “right” order of joins is being applied. In the
“right” order, the studies joined first the pair of datasets that have the largest overlap
and the fewest errors. For the holistic approach, the studies presented several novel
methods. One version of the holistic normalized-weights method provides high
recall and precision, under all circumstances (Baltsavias
2004
; Spaccapietra
et al.
1992
; ESRI
2004
; Friis-Christensen et al.
2005
). Comparing the two
approaches, it was found by the studies that the time complexity and the space
complexity of the sequential normalized-weights method are lower than those of
the holistic normalized weights method (GSDI
2005
; Hampe et al.
2004
; Lake
2005
). The holistic approach, however, is capable of providing higher precision
than the sequential approach (at the cost of lower recall). Another advantage of the
holistic approach is that each join set is given a confidence value. These studies
marked several problems that remained for future work. One problem they state
was to optimize the run time of the algorithms. This was particularly important if
those algorithms are to be included in real- time applications (Hampe et al.
2004
;
Lake
2005
; Levy
1999
). A second problem was of how to utilize most effectively
locations that are given as polygons or lines, rather than just points (Lu et al.
2007
;
Ma et al.
2000
; Malhotra
2000
). A third research direction is to combine one
approach with other approaches, such as the feature-based approach of (Chen
et al.
2003
; Butenuth et al.
2007
; Beeri et al.
2004
; Budak et al.
2006
), topological
similarity (e.g., Batini et al.
1986
; Beeri et al.
2005
) or ontologies (e.g., Egenhofer
et al.
1989
; Goesseln and Sester
2004
; Rigeaux et al.
2001
; Sattler et al.
2000
;
Walter and Fritsch
1999
; Devogele
2002
; Gal et al.
2003
).
Some studies went forward taking the concept that GIS is a repository for
geographic objects (
Ashton
) and that each object contains information about a
real world entity; a real world entity is described uniquely in the system by one
object. In the system the objects are organized in datasets, where each dataset
contains objects about a certain subject (OGC
2007
; UCGIS
2004
; Batini
et al.
1986
; Beeri et al.
2005
). The input for fusion process is several datasets