Geography Reference
In-Depth Information
Figure 3. Drawing and visualizing vertical Bézier curves in 3D-GIS environment
(ArcScene).
2.2. Space-Time Kernel Density Estimation
As discussed above, the temporal information of movements in geogra-
phic space is important to detect the spatio-temporal trends of underlying
human mobility. But with the increasing number of aggregated human/vehicle
trajectories in urban space, the space-time path representation model will be
hard to interpret because of the overlapping and cluttering issues. To solve this
problem, an extension of kernel density estimation (KDE) (Silverman, 1986)
was suggested. The KDE has been widely used in spatial analysis to charac-
terize a smooth density surface that shows the geographic clustering of point
or line features in 2D space. In order to incorporate the time information, the
space-time kernel density estimation (STKDE) can be taken as a generali-
zation approach of the 2D-space KDE into the 3D space-time cube which can
support the exploration of spatio-temporal patterns, clusters and changes. Such
STKDE techniques have been used in several studies, such as crime clustering
analysis (Brunsdon et al., 2007; Nakaya and Yano, 2010), trajectory data
mining (Demšar and Virrantaus, 2010), publication citation analysis (Gao et
