Big Brother is not watching you
dailymail.co.uk
dailymail.co.uk
The whole "intelligent video" industry is in its infancy. An interesting point, though, is as these analytics improve, so does our ability to limit the scope of video that can be watched. Irrelevant footage can be made off-limits, faces or license plates obscured, etc.
It's not going to happen overnight, but the trend is clear: surveillance video is going to become more structured as time goes on. But this can be used to uphold personal rights as easily as it can be used to bypass them. Laws will just have to adapt in the face of technology.
I'm curious, though--does your company do the kind of thing covered by Steve Rambam in his talk at The Last HOPE? ie, recognition of activities as well as of faces, text, etc.?
http://www.thelasthope.org/media/audio/64kbps/Featured_Speak...(Part_1).mp3 http://www.thelasthope.org/media/audio/64kbps/Featured_Speak...(Part_2).mp3
I know of one company that's hired professional actors to act out different scenarios, in the hopes that they can distill the essence of suspicious activity, and thereby automatically detect it. I'm not optimistic about their chances.
However, it's relatively straightforward to detect people tailgating at security entrances (using one passcard to let two people in), or walking the wrong way into the exit gate at an airport. In both these situations, though, the environment is controlled, and the difference between proper and improper behavior is well-defined. Neither of these things are true on, say, a street corner.
Human interaction and body language is complicated. I expect we'll have a good way of automatically detecting behavior sometime around when we have general artificial intelligences, and not before.
Might be a good opportunity for a real-time image analysis start up, though...
For example, if a crime is committed but the criminals face isn't seen, it's the camera operators that track them until they run into another camera on the system that has been angled to see their face. A system running on autopilot is going to miss a lot of that.
Studies show a trained operator will miss 90% of events looking at a single feed after just 20 minutes. It's even worse over a longer time and over multiple camera feeds.
No one is really watching.
Once automated surveillance gets better, that won't be the case at all. The systems are only currently useful for after-the-fact forensics.
Jesus, that makes it even worse. The only places I've heard that at is high-security corporate and military locations. Certainly not town squares.
If you mean "on the spot" as in "at once, with no delay" then yeah, totally. That's really common in the UK and I support it as it's been shown to have positive effects on crime (often by pushing the crime elsewhere, sure).