Advanced software programs now assist human monitors of security cameras at airports or other high profile public venues. Automated interpretation of raw data can help the human monitor by finding the suspicious amid a haystack of data. But there are practical limits to what a computer can be expected to accomplish.
Tom Meservy of the University of Arizona is working on video analytics software that is designed to decode the motions of a person's body during an interview. This type of analysis can allow video cameras at security checkpoints to determine suspicious individuals. Telltale behaviors often mount while people are waiting in line, Meservy said.
Systems like these are at the leading edge of the technology, but less complex software is finding its way into mainstream security applications. The most basic of these techniques is motion detection. Instead of having a human sit in front of a bank of monitors 24 hours a day, software can now send an alert based on a programmed condition. Studies have shown human operators can only stay sufficiently alert for about 20 minutes watching mostly eventless images.
For example, a person climbing over the fence could trigger such an alert, and the software can detect that whatever is climbing over the fence is of sufficient mass, so that cats and birds don't cause false alarms. Any surveillance system must have adequate camera resolution for human operators or analytical software to be successful. No matter how good the software: garbage in = garbage out.
The system can send detailed alarms over the Internet, SMS messages to cell phones, turn on or increase the sample rate of a video recorder, or trigger an existing general alarm system, depending on the urgency of the event.
No comments:
Post a Comment