Regine April 24, 2006 security, software

CCTV to spot individuals carrying concealed firearms

Surveillance cameras are not only said to make people feel better, they might one day be able to spot individuals carrying concealed firearms.

For Loughborough University‘s multi-environment deployable universal software application (Medusa, see PDF) project, CCTV footage of people carrying concealed firearms will be analyzed to identify characteristics associated with the behaviour of criminals (body stance, gait, movement and eye contact with cameras) before they commit a gun-related crime. This information will be used to develop a machine-learning system for behavioural interpretation. Armed with this data, the CCTV cameras will scan footage and match behavioural characteristics that indicate if an individual might be carrying a gun.

The system might be developed to study knives as well.

Via The Engineer. Image from Lonely Radio photostream.

3Comments

Add yours

1

Justin on April 24, 2006 at 7:59 pm

Great. So next time you’re walking with a limp or bump into someone, the police will descend on you and try to take away their imaginary gun. Or knife.
it is kind of neat and all, but how much surveillance do we need? Freedom depends on some level of anonymity.
2

Sascha on April 25, 2006 at 12:16 am

Hm, I guess that it’s close to impossible to recognize a gun that you’re hiding (depends on the gun, though ;) but what’s scaring me is the eye contact-part. That’s definitely possible (to register, not to read) and I have looked so many times at cameras with the same slight rush that you get when passing through electronic anti-theft gates even though you’re sure that you haven’t done anything wrong.
There’s a hilarious project by Sebastian Bodirsky which provides a back-channel for CCTV: http://www.digital.udk-berlin.de/~sebastianbodirsky/
3

David Mery on April 25, 2006 at 3:04 am

Reintroduce death penalty, add a long range precision rifle and it’s The End of Violence.
Hopefully profiling techniques will have improved.
Also some British company had a video on the web a few years ago showing how they were predicting whether people entering the field of vision of the cameras would break in cars in a parking lot. Individuals were initially represented as green dots, then turned orange when suspected and red when the system determined they would commit a crime. Can’t remember the name of the company and don’t know what happened to this system.
br -d