Database Reference
In-Depth Information
describe aggregation of trajectories into a continuous density surface using a spe-
cially designed kernel density estimation method, which involves interpolation
between consecutive trajectory points taking into account the speed and accel-
eration. Density fields built using kernels with different radii can be combined
into one field to expose simultaneously large-scale patterns and fine features.
Andrienko and Andrienko
(
2011
) suggest a method for the tessellation of a ter-
ritory used for discrete spatial aggregation of movement data and generation of
expressive visual summaries in the form of flow maps. The method divides a ter-
ritory into convex polygons of desired size on the basis of the spatial distribution
of characteristic points extracted from trajectories. It uses a special algorithm
for spatial clustering of points that produces clusters of user-specified spatial
extent (radius). Depending on the chosen radius, the data can be aggregated at
different spatial scales for achieving lower or higher degree of generalization
and abstraction.
An example of visualization of flows between locations in the form of an
origin-destination matrix can be found in
Guo
(
2007
). The rows and columns can
be automatically or interactively reordered for uncovering connectivity patterns
such as clusters of strongly connected locations and “hubs,” that is, locations
strongly connected to many others.
To deal with very large amounts of movement data, possibly not fitting in
RAM, discrete spatio-temporal aggregation can be done within a database or
a data warehouse as described by
Raffaet`aetal.
(
2011
). Only aggregated data
are loaded in RAM for visualization and interactive analysis. Using roll-up
and drill-down operators of the warehouse, the analyst may vary the level of
aggregation.
Andrienko and Andrienko
(
2012
) give a comprehensive review and extensive
bibliography of methods, tools, and procedures for visual analysis of movement
data.
