RealTime IT News

IBM's 'Veggie Vision' Grows Up

Although IBM is making news today with its Smart Surveillance System (S3) video, which it says can tell a terrorist from a traveling salesman or a scammer from a shopper, researchers began the project with a more modest goal. They wanted to tell a cumquat from a rutabaga.

"Veggie Vision," as it was called a few years back, employed quickly-evolving video interpreting techniques and software to help grocery store cashiers manage the most exotic vegetables and enter the correct price.

Fast-forward three years. Charles Palmer, chief technology officer of security and privacy at IBM Research, is still focused on the check-out line -- only this time he's concentrating on retail theft.

Although the retail industry loses an estimated $50 million to fraud and theft each year, finding the thief often means sifting through hours of video surveillance tapes.

"Humans have a 20-minute limit, then they turn into vegetables," Palmer joked. Another problem: customers have amassed all this video surveillance and they don't know what to do with it, he said.

Enter S3, a system connecting surveillance cameras with software able to watch the video, analyzing the images for what's important, and issuing alerts.

Those alerts can search for the obscure (when an IBM employee walks backwards down a hall in a Korean office) or vital (notify security when a person enters a restricted part of an airport.)

Other interest in the project is related to the U.S. Mexican border. The length, plus rugged landscape, make it a particular challenge, in terms of monitoring for illegal aliens trying to cross the border.

That said, the biggest demand for improved surveillance is coming from the retail sector, Palmer said. IBM has received interest in S3 in order to watch for slip-and-fall scams, phony store returns, even cashiers pilfering from the till.

As demand increases, Palmer emphasized S3 is a surveillance framework able to work with plugins for individual applications. For instance, customers have asked for the ability to look for empty parking spots, animal tracking and other unique uses.

That extensibility and ease of configuration are what will make S3 stand apart from similar products, Frost & Sullivan analyst Karthik Nagarajan said.

3VR Security, which also analyzes surveillance video, is likely the most competitive product, Nagarajan said.

However, 3VR and IBM are focusing on different sectors. While IBM's S3 is gaining fans in the retail area, 3VR's use of biometrics and face recognition technology are targeted at banking, according to the analyst.

IBM insists S3 is not a threat to privacy. Indeed, the product gives users the ability to redact identifying items, such as faces and license plates. Palmer said at one point, the S3 development was halted until a security concern was corrected.

How will IBM's surveillance product expand? Palmer foresees S3's real-time alerts increasing in complexity. For instance, the software might be able to pick a Mercedes out from a row of assorted autos.

IBM is working with major surveillance camera makers, indicating there will likely be R&D collaboration, according to Frost & Sullivan's Nagarajan.

The IBM executive also said the future might hold the ability to automatically monitor "entities," such as terrorists. He quickly added individual persons would not be targets.

"HAL 9000 is not here yet -- give us time," he said about the computer from 2001:A Space Odyssey. (He really was joking.)