|
Major Research Subjects of Ito, K. (2002-) Assistant Professor
- Analysis and Visualization of Spatio-Temporal Data
In the study of spatial information science, spatio-temporal data
are also important. But without any appropriate processing method,
we cannot make full use of the spatio-temporal data. Therefore
it is necessary to develop a method to model the data source which
explains and describes the data. We have proposed those methods
such as; modeling the spatial distribution and birth-death process
of objects and integrating both of them into a description length
[bit]; visualizing spatio-temporal state on a drawing in color
(a hue axes and saturation axes) to increase the dimensionality
of expression. We are also trying animation technique.
- Formulation of Optimum Area-Segmentation
In order to find out the underlying structure out of huge amount
of data, the observation in suitable complexity of resolution
is indispensable. Spatial unit of analysis should be an area with
data of essential similarity, and should avoid picking up an accidental
fluctuation. We tried area-segmentation by using MDL (Minimum
Description Length) criterion as a basis of model selection. Firstly
a class of models is defined, and then, the optimum model in the
class is selected. We described a class of model by a binary tree
segmentation and obtained an algorithm to search for the global
optimum efficiently. We applied this area-segmentation method
to the analysis of spatial distribution and lifetime of tenants
in central Tokyo and also to the analysis of composition of building
sizes in Damascus Old City.
- Object Identification
Spatio-temporal data can be created from a time sequence of spatial
data. Since spatial information has various random error and ambiguity,
good identification of objects in a series of spatial data is
necessary. We proposed a method to match the objects allowing
for the geometric and attributive ambiguity. First we cluster
buildings (tenants) in all different years by a dissimilarity
measure, which is defined by their spatial position and attributes
(names, years, & etc.), and then, the buildings (tenants) in a
same cluster are supposed to be the identical building (tenant).
We applied the method to data of five-years residential map of
central Tokyo. The total of 1.65 million building-years and 4.85
million tenant-years are identified, and the spatio-temporal data
of buildings and tenants are made.
- Space and Events
By combining various information, spatial information science
can describe and express not only matters which have marked on
the past printed map but also various events. Our world is essentially
consisted of many events. The spatial information science provides
techniques to overview the events in the world, which will help
advance both conceptual and quantitative approaches in our study
of "space and events".
|
