Kosakoski_Spatiotemporal Visualization v2

Download Report

Transcript Kosakoski_Spatiotemporal Visualization v2

Spatiotemporal Visualization
Andrew Christie, Adam Kosakoski, Craig Evans, Charles Schmidt, David Altieri, Peter Schooley, Richard Johnson
Advisor: Dr. Terry Mason, Dr. Wan Bae
Universty of Wisconsin - Stout
About The Project
The Spatiotemporal Visualization project is to help analyze movement patterns, behaviors and
trends by:
• Taking cleaned GPS data from the Data Mining Team and displaying it in a 3D or 2D visual
•Allowing selection of areas of interest
•Separating different object types.
• Sorting objects by time.
• Using navigational tools to better visualize the data in the visual environment.
Filtering Data
The researcher is allowed to filter which points should be displayed based upon data values or
the researcher’s selection using a mouse. We have implemented filtering algorithms to filter
data efficiently and accurately. Currently, we are investigating data structures and algorithms
such as R-Trees and hash trees to further increase the efficiency of the filtering while
preserving ease-of-use for the end user.
Handling Missing Data
Since we are getting our data from GPS devices, there’s bound to be places where points
don’t exist, so we had to take this into account when visualizing it. The first attempt we made
to accurately estimate where the object would be based on its position prior to losing
connection and after reestablishing it, was to use B-splines, a Numerical Analysis technique
of estimating a curve, or in this case points at a certain interval on a curve. This works very
well in practice, however it suffers from quite a long run time, as we have to run it on entire
files of data every time we want to open a new file to view the points of.
Pre-filtered 2D and 3D
Filtered 2D and 3D
Distinguishing Data Types
The Applet
Since all of the lines are going to be relatively hard to distinguish differences in them, we are
using a color coding system. Similar to electronic wiring, wires are often striped. Since we are
using color to distinguish between multiple attributes, this seemed the most reasonable. One
object will consist of a large line, colored for object type, and a small line, cycling through multiple
colors to distinguish between objects of same type. Each line will also have a gradient to
represent change in time. We are also displaying different shapes at each data point. For
example, people could be spheres, and cars could be cubes. In a flattened 2D image, the 3D
shapes become 2D as well, e.g. cubes become squares.
The applet promotes online collaboration and provides an easy interface for the researcher to
view and manipulate the spatiotemporal data. The user can open csv data files to analyze on
the applet and save various views of interest. The applet provides information to the user
about what they are viewing and what data is selected.
Here you can see the line striping in random data
(Red-people, Blue-Automobiles)
Displaying different shapes for
different object types