| |
We are developing a full-fledged database
mining system for police department's massive crime database
for identifying crime patterns and criminal behavior. Currently,
we are working on relational data mining in the APD database
of crime reports. Since police departments in Texas have
equipped their cars with video cameras for recording their
proceedings, new video databases will soon start building
up. Criminer is a scalable database mining system using
a novel graph based techniques. The key research will develop
new methods for collecting and understanding large amounts
of video information. The research involves the search for
interesting patterns in video, audio, image, text and other
types of data. We are specifically interested in mining
tasks such as pattern discovery, supervised pattern learning,
clustering, and anomaly detection. The investigation of
the use of graph-based relational data mining (RDM) for
mining video data requires massive computational power such
as the requested cluster. Criminer can extract high and
low level features. The high level features include officer
IDs, dates, types of violations, motorist information, speed
radar readings, global positioning system data, vehicle
telemetries, and light bar (pursuit light) activities. The
low level features include audio and camera and object motions,
colors as well as sizes, trajectories, shapes and speeds
of objects.
|
