Event Detectives
Every visitor to the Calit2 Building at UC Irvine really counts.
Devices called “people-counters” have been installed near every door by Project ResCUE researchers, who use the small optical sensors to track the number of people entering and exiting the building – information that could prove crucial during an emergency evacuation.
Now, the sensors are aiding another Calit2 research project. By learning the building’s normal traffic patterns, researchers have developed a way to determine the probability that an unexpected event could be occurring.
Padhraic Smyth, professor of computer science, postdoctoral scholar Alexander Ihler and graduate student Jon Hutchins tracked the people-counter data every 30 minutes over a period of three months.
This allowed them to learn normal activity in the building’s “living lab” environment. Then, for the next nine weeks, they compared realtime data from the people counters to the normal activity patterns they had learned previously.
The majority of the events detected during that timeframe were later correlated to scheduled Calit2 Building events.
Determining Probability
The research application isn’t confined to predicting the occurrence of an unusual event. Rather, it determines the probability of an atypical event occurring.
The software system isn’t limited to data from cameras or people-counters, either. It can analyze any type of sensor that measures human activity, including highway sensors, ambient noise sensors and Web usage tracking.
The results have multiple real-world applications. The atypical event might indicate a large meeting in a building or it could signify something more ominous: a malicious attack on a Web server, a traffic accident on the freeway or a hostage situation.
Crisis Aid
Now that the software program is operational, the team is collaborating with Project ResCUE to develop an interface for the program based on Google Earth.
The interface connects with the sensor database, enabling anomalous events to be identified on a map of the designated area. The end result will be an “information dashboard” that can be used by emergency operations centers for crisis assessment and response.
As part of this effort, researchers are working with data from the California Department of Transportation to analyze traffic patterns from 20,000 highway embedded sensors in Los Angeles and Orange County.
The research team is also working with the fire department in Ontario, Calif. To implement the software as part of its emergency preparedness plan.
“It’s just another information tool,” says Hutchins. “Once you know what ‘normal’ is in a given situation, it’s much easier to determine when something unusual is occurring.”
- Anna Lynn Spitzer
Reprinted with permission from Interface, Calit2's magazine
