‘Computers get an insight into people’s lives’
Dr Hannah Dee is a lecturer in the Department of Computer Science at Aberystwyth University
The aim of my research is to build computer programs that understand what’s going on in an image or a video.
I want systems to be able to take a digital image or video and say: “That’s a person”, or: “There’s something moving around there”, or even: “That car’s going the wrong way down a one-way street” and “That person is behaving erratically”.
In order to do this, we try to find patterns in the pixels of an image or video and group these into things that represent objects we understand, like cars or faces.
Researchers in this area tend to work on small parts of the problem.
Our aim is to build a machine that can see like we do, but the whole problem is too big.
My particular corner is aimed at building systems that can understand the way things move through space.
If you point a camera at a station, can the computer program learn the paths that people take through it? Can a program work out how traffic flows around a roundabout? Can it find parking spaces, benches, and other geographical features?
The obvious applications of this kind of research are in surveillance. The UK is the most watched nation in the world, but at the moment, surveillance control rooms are reactive.
This means the vast majority of these camera images are never seen by anyone, unless there’s some kind of event – Cardiff council runs 200 CCTV cameras but it doesn’t don’t employ 200 people to watch them.
If we’re successful, CCTV systems will become more reliable as a computer will be able to filter the hundreds of feeds so the human observers up in the control room only have to look at stuff that’s interesting.
But it’s not all Big Brother – anything that can use video can take advantage of these systems.
It’ll be useful for architects and city planners, as they’ll be able to see where people walk in their designs.
This kind of technology can be used to help the elderly and disabled, using cameras to work out when they’ve had a fall or if they’re not using areas of the home they need to.
It’s useful in computer games because if we can automatically learn from video how real people move around, we can make better virtual agents and virtual worlds.
Biologists are interested in knowing whether the same techniques can be used to work out how plants grow from speeded-up video.
Anything we use our eyes to do, a computer vision system could be able to do too – without getting tired or distracted.
To contact Hannah please email hmd1@aber.ac.uk.
This article first appeared in the Western Mail‘s Health Wales supplement on the 19th December 2011, as part of the Welsh Crucible series of research profiles.