For us law-abiding citizens, the biggest botheration with security cameras is feeling spied upon – or the way the things instantly add 10 pounds on the monitor.
The biggest problem for companies using the cameras is something altogether different.
They don’t care if the cameras make us feel violated or look fat, but they do care it takes a lot of manpower, and thus cash, to sit there and watch them.
It also takes a savvy viewer not to make a bunch of mistakes that result in running after someone who the viewer only “thought” looked or did something suspicious – and an even savvier viewer not to fall asleep.
The University of Arizona is working on fixing that, with security cameras that not only record people’s actions but go so far as to interpret these actions, according to a report from UA News.
If the camera pans pan out, security cameras will be able to note when someone caught on camera is behaving suspiciously. This way human security guards or other viewers don’t have to waste their own time doing so.
Researchers are surely hoping it pans out, as the project just received a $2 million grant from the Defense Advanced Research Projects Agency, with the promise of another $3 million if the first two years are successful.
So how does it work?
“Drawing on ways the human brain processes visual information, the research team plans to build a visual detection and tracking system, models of human behaviors and simulators to generate possible future scenes,” the report tells us.
“One of the project’s major components is the incorporation of visual imagination, a process by which the brain decodes what the eyes see while also guessing what should occur next.
“‘The same parts of the brain are involved in vision and visual imagination. To some extent the brain is telling the eyes what they should be seeing. So we want to model this sort of heavy involvement of the brain,’” the report quotes team leader and UA computer science department head Paul Cohen.
So far such brainy cameras have worked in small pilot programs, but more activities need to be added to cover the massive range of behavior that could be caught on video and deemed suspicious.
To kick off the program, cameras will be programmed to interpret 48 actions based on verbs like “run,” “escape,” “climb,” and “carry.”
Researchers also want to add another level of brainpower, where the cameras will understand suspicious behavior when someone is interacting with another person or has an object, perhaps such as a tire iron or bomb.
Whew.
There is a reason many of us majored in English.
Let’s throw out a hypothetical example to see how an equally hypothetical version of the newfangled security camera might work.
Say a camera is focused on a man who gets out of his car. The camera interprets: “Stand.” He then opens the trunk. The camera goes: “Dead body.”
Then the man picks up the object and walks towards a garbage can. “Destroying evidence.”
A woman happens to walk by with her dog and she nods at the man. “Accomplice. Accomplice with deadly weapon.”
At this point a law enforcement officer – or eventually a law enforcement robot – would be on the scene, perhaps arresting them both and throwing the dog in the pound.
Possible suspicious behavior thwarted. Problem solved.
Now if researchers could just create a camera that didn’t automatically add 10 pounds, we’d really be on the way to a perfect future.
[tnipoll]
What do you think?
Does a security camera with brains sound feasible?
Would you trust the interpretations?
Have you ever been wrongly busted for something caught on camera?
What about rightly busted?
