CCTV software that scans behaviour could ‘spot pickpockets before they strike’

MARK BLUNDEN

The inventors of London’s Oyster card are developing technology with the potential to analyse CCTV and predict if a passenger is about to have a heart attack, dodge a fare — or even go into labour.

It is claimed Cubic Corporation’s “automated scenario recognition” system could help to anticipate platform fights or spot pickpockets before they strike.

Experts say that in future the technology could also play a role in fighting crime or identifying early indicators of potential terrorist activity.

The neural network trains itself to predict unusual situations by learning patterns from thousands of hours of high-definition station CCTV footage, according to papers lodged with the European Patent Office.

Cubic suggests the system would be more efficient than human CCTV controllers as the robots can watch multiple cameras and never forget. In an emergency, alerts would be sent straight to police, the fire brigade or paramedics.

The patent says: “In contrast to the human, the neural network can be alert 24/7, simultaneously watch multiple cameras to look for similar scenarios it has been taught to look out for, respond instantaneously, and constantly learn new scenarios without forgetting anything. The judgment could replace the human element or may be used as a filtration process.”

The firm suggests scenario recognition would be able to scan the “demeanour” of pickpockets and fare-dodgers trying to use busy crowds to “disguise the behaviour” from humans.

The patent says: “If the user has practised fare evasion in the past, the neural computing system may alert the nearest fare inspector … to carefully watch the particular user as he navigates the validation area of the transit station.” Security expert Philip Ingram, a former Army intelligence officer, said: “They are trying to get machines to analyse the huge amount of CCTV footage out there now and look for unusual patterns, so you can identify potential terrorist activity or someone potentially carrying contraband and flag that up to security operators.

“Machines can analyse an awful lot more activities much more quickly than humans can do, but you’re always going to need a human in the loop to make sense of what they find.”

Cubic declined to comment on the patent.