'Intelligent video' is useful, but can't do everything.

VIDEO CAMERAS ARE A FAMILIAR piece of security operations worldwide, but the information captured in these images is limited to what the human eye can see. What if video cameras could detect things humans can't see?

That is the question being addressed by firms that are working on a technology called video analytics. A marriage between advanced computer science and artificial intelligence, video analytics uses algorithms to turn camera images into data.

Companies in the business of video analytics are touting their products as "intelligent video," but some experts think this label is misleading.

"Video is still vision-based; it's not yet intelligent," said Ed Troha, director of marketing and communications for Object Video, an intelligent video software firm in the Washington D.C. area.

Video analytics can do face and number recognition, for example, but is still limited by a number of factors, Troha said at the GovSec homeland defense conference.

It can't see through fog or darkness and "there is no video analytics system that can spot a terrorist" in a large crowd, Troha asserted.

What video analytics can do is pick out human forms and vehicles. It can also detect a person before the eye can, he said.

Video analytics detects certain objects or activity by following rules outlined by the user. For example, if security personnel want to know when someone walks across a certain road at a specific time, they can write this rule into the software by drawing a line on the video screen shot. The video program will then alert the user when someone crosses that line.

Such software can be used for physical security, retail, banking, casinos, traffic management, crowd monitoring, left object detection and building management, Troha said.

Email your comments to [e-mail]