Geography Reference
In-Depth Information
sliders, buttons and menu items to filter the data and instantly view the result in the display.
Furthermore, other pre-processing may be achieved through knowledge discovery (KD)
and data mining (DM) methods to find patterns in large spatial datasets (MacEachren et al. ,
1999). In addition, the plain quantity of data often exacerbates the problem. Not only does
it slow the interactive process, but saving and loading the data takes time and effort. Hence
caching techniques would also speed up the interaction.
Storing and caching the processed (shoebox) data are important not only for interactivity,
but they can also be utilized by the investigator as a temporary copy or by other investigators
to confirm a hypothesis. Thus, there is a need for both curation of geographical data (espe-
cially shoebox information) and storage of information on the provenance of the data and
the session. For instance Thomson et al . (2005) discuss issues of lineage and provenance
in geovisualization and Koua and Kraak (2004a-c) discuss some ideas on tools to support
'knowledge construction throughout the exploration process', but much work is still to be
achieved.
While filtering methods pre-process or dynamically select a smaller demonstration set,
other processing methods aim to group, aggregate and assign structure to the data. Patterns
that are assigned through KD and DM techniques can be visualized. Many of the techniques
are non-trivial as it is often not possible to provide brute-force algorithms to find similar-
ities within the data (Miller and Han, 2001). Parameters of these algorithms can be tightly
coupled to the interface, such that the results are consequently updated (Maceachren et al. ,
1999; Koua and Kraak, 2004a-c). For example, by adapting values to (say) the k -means
algorithm, the user can explore different partitioning configurations (Torun and Duzgun,
2006). Thus, it is possible to combine knowledge discovery, data mining and visualization
techniques. In fact, various researchers have called for the closer integration of data min-
ing and visual exploration techniques (Kreuseler and Schumann, 2002; Kocherlakota and
Healey, 2005). Classification hierarchies can be used to explore the data (Kemp and Tan,
2005); these can be displayed as graphs and other structured organizations (Rodgers, 2004).
Classifications are important, and if, in particular, they are visualized, they allow the user to
understand the underlying structure of the data and furthermore can be used for filtering
and processing. Diverse categorizations provide different ontologies (Arpinar et al. , 2006).
These classifications can then be visualized separately and thus provide multiple views on
the same information; see Section 3.3. Not only are there challenges with processing the
quantity of data, but there are also challenges in displaying it all at once. Venkataraman
et al. (2006) write: 'LIDAR data can easily exceed 100 Mpixels making it impossible to have
detail as well as context on a single desktop display'. Some researchers, particularly from the
database community, have looked at pixel-based visualizations (Keim, 2000) and some work
has been done in spatial visualization using pixel displays (Keim, et al. , 2004), but there is
still a limited number of pixels on a screen and hence a limit to the amount of information
that can be displayed at once.
The crucial factor is not just the size but the quantity of pixels it can display. High-quality
large-format cinema-style video projectors are still very expensive and have huge running
costs, typically because of the short life-span of the bulbs. Hence researchers have developed
tiled screens. Such multiple display devices have been successfully used for geospatial explo-
ration (Venkataraman et al. , 2006). However, these displays have their problems. They need
special device drivers to control all the multiple displays, the screens need to be housed in a
customized rig, there may be problems with the viewing angles, and there will always be a
gap between each screen due to the screen's housing.
Search WWH ::




Custom Search