The UK Panopticon
Ars Technicha has a nice article about the closed-circuit television cameras (CCTV) that pervade the United Kingdom. In short, they haven't done much of anything to either reduce crime, or increase the effectiveness of crime solving and prosecution. But, folks keep demanding it, so they keep building it. At the moment, this is just security theater.
As the article says, "millions are spent on facial software recognition systems that have yet to be shown to have any merit." These are the sorts of systems that are supposed to identify faces, associate them with a specific person, and mark their location. In theory, if the whole system were working the way it is intended, when someone, "is captured on camera roughly every six seconds," it can provide a very detailed way to track where people go. Unfortunately for the security apparatus, the software systems we have now aren't anywhere close to capable of doing this reliably. You need a high quality, full to mostly-full shot of the face to do it. A simple hooded sweat-shirt, with the hood drawn as far over your head as possible can foil a lot of the camera angles required to get a good 'hit'.
The other problem they're facing is a simple data overload one. So much data is being gathered, that effective sifting technologies don't really exist yet. Especially when asking for footage from privately owned CCTV systems.
These are both problems where sufficient application of emerging technology can help. So long as the security apparatus can resist ripping out all this infrastructure they've built, future storage systems can capably handle the information load generated by these cameras. Couple that with advances in software, and you have the beginnings of an actually effective CCTV system. Computer-based tracking of license-plate numbers should come well before then, optical character recognition in noisy environments being a fundamentally easier problem than facial recognition, which will make tracking vehicle movements much easier as well.
Right now, the system is not very useful. On the other hand, this is exactly the sort of ground-work needed for a system that could actually be useful. Once the "back office" details are sufficient, the data flood will become manageable. Once that happens, the utility of all that data becomes much higher.
As the article says, "millions are spent on facial software recognition systems that have yet to be shown to have any merit." These are the sorts of systems that are supposed to identify faces, associate them with a specific person, and mark their location. In theory, if the whole system were working the way it is intended, when someone, "is captured on camera roughly every six seconds," it can provide a very detailed way to track where people go. Unfortunately for the security apparatus, the software systems we have now aren't anywhere close to capable of doing this reliably. You need a high quality, full to mostly-full shot of the face to do it. A simple hooded sweat-shirt, with the hood drawn as far over your head as possible can foil a lot of the camera angles required to get a good 'hit'.
The other problem they're facing is a simple data overload one. So much data is being gathered, that effective sifting technologies don't really exist yet. Especially when asking for footage from privately owned CCTV systems.
These are both problems where sufficient application of emerging technology can help. So long as the security apparatus can resist ripping out all this infrastructure they've built, future storage systems can capably handle the information load generated by these cameras. Couple that with advances in software, and you have the beginnings of an actually effective CCTV system. Computer-based tracking of license-plate numbers should come well before then, optical character recognition in noisy environments being a fundamentally easier problem than facial recognition, which will make tracking vehicle movements much easier as well.
Right now, the system is not very useful. On the other hand, this is exactly the sort of ground-work needed for a system that could actually be useful. Once the "back office" details are sufficient, the data flood will become manageable. Once that happens, the utility of all that data becomes much higher.
posted by Corwin at 12:18 PM
<< Home