Slashdot Mirror
3D Face Cameras
newsblaze writes "Now there is a fast, inexpensive device for simultaneous acquisition of accurate three-dimensional and two-dimensional human faces. It will allow law enforcement and security agencies to capture both types of mug shots in seconds as a single snapshot and provides incredible accuracy in correctly mapping the individual being booked in 3D."
The real problem with this type of work is when software is used to match mugshots to actual pictures and video. Even leading researchers are still having a very difficult time doing this. I think the government should be focusing more on advancing the work of these researchers than on gathering more data.
Voice your opinion!
It's nice to see that their crack team of token racially-diverse employees
http://www.bbninternational.com/company.html can make sure that everyone who's between the ages of 32 and 36 will be so catalogued.
The article says about as much as the summary. It would be nice to know how the 3D aquisition is done!
I used to work at the computer vision research lab at Notre Dame, and we had a pretty cool device that was used to capture 3-D frontal images of anything (we used it for faces). IIRC, it scanned a horizontal laser line down across the subject and measured reflected light using two sensors, triangulating to compute 3D information. I wonder if this camera uses the same concept? Some drawbacks of the afore-mentioned technique are that you can only capture a terrain-style 3D map of one side of an object, and subjects that open their eyes cause holes in the terrain where the reflected light was scattered by the wet surface. To get a 360 degree capture requires multiple scans and software to merge the resulting models.
Very cool technology. Casinos will most likely implement something like this first. In Nevada it is state law that if someone has registered as a compulsive gambler the burden is on the casino to keep them out.
My sigs offend the max # of people all over the world, regardless of race, religion, color, sex or creed. It's a gift.
There's at least one company with a product that makes a good 3D image from a 2D image. There are applications in security for matching two images taken from different angles - you want to match in 3D, not 2D.
Sure, it's making up information, but the human head isn't an unknown amorphous blob, there is a certain regularity to it.
If you have the conditions to use special hardware and photography techniques you might get better 3D, but if this is being aimed at security, that's not a luxury you usually have.
My God, it's Full of Source!
OUTSIDE_IP=$(dig +short my.ip @outsideip.net)
Your comment is more insightful than you may realize. 3D face technologies were originally intended to overcome some of the issues of lighting, pose, and shadow that ruined traditional 2D recognition methods. But *everything* has sensitivities and 3D was impaired by facial expression in ways that were arguably even worse. Some folks also investigated thermal methods, but ever see a face scan of someone who just drank hot coffee?
We want face recognition (and all biometrics in general) to be invariant to environmental factors and subject variations. But in practice, we just trade one issue (light, pose, motion, cost, speed, etc) for another.
Oh, and how quickly can you take a 2D picture? Fast. 3D isn't so cheap and timely. Watch for the Face Recognition Vendor Test 2005 (later this year or early next) to see how 3D really stacks up.