Vol. 77, No. 2New technology

Putting a name to a face

Calgary police gets new facial recognition technology

This side-by-side photo comparison of Boston bomber Dzhokhar Tsarnaev illustrates how facial recognition technology uses an image to find a match. Credit: NEC Corporation

By

A teenager was communicating online with an unknown man. She only had a first name, and the Calgary Police Service (CPC) detective wasn't even sure if that name was real or fake.

In fact, Det. David Palmer had no leads and had nowhere left to turn in the child abuse investigation — until he got a profile picture that the offender had used online.

"I remembered we had access to new facial recognition software and thought, 'Why don't we submit it and see?' " says Palmer.

Quick match

CPS is the first police agency in Canada to get the technology to help solve crimes.

"We've really been keeping our eye on facial recognition technology for some time," says Jan Gregory, a supervisor with the CPS Criminal Identification Unit. "We were waiting for the technology to mature enough to a point where it began to be useful in investigations."

The process has always been manual at CPS. Photo lineup clerks would go through a mug-shot database of images of those charged with criminal offences in Calgary and manually try to identify suspects or victims that way.

Afzal Baig, a project manager in CPS's Information, Communication and Technology Section, says the new technology is far more efficient.

"Sometimes you have an incident where you can see a face, but you can't put a name to the face," says Baig. "In the past, someone would look for a male with brown hair of a certain age and height and then they would take one picture at a time and go through it. Now, the system takes less than 30 seconds to do a probe and you're done."

The technology is similar to a fingerprint matching process. It takes an image of a face, applies a matching algorithm — plotting landmark features like the eyes, nose and mouth — and establishes a series of measurements that describe that face in mathematical terms.

It then analyzes incoming images and compares those facial features to eliminate non-matches and bring back the best possible matches.

Palmer thought it was a shot in the dark because the software only searches the CPS mug-shot database of approximately 300,000 images. If the offender wasn't previously arrested in Calgary, it wouldn't turn up a match.

"But I had no other options at that point, so I sent it in and then we got this hit back. He was charged exactly a year earlier for doing exactly the same thing," says Palmer. "It was perfect and the match is unbelievable."

Added value

The software isn't a positive means of identification, says Gregory. CPS is using it to further investigations.

"We provide that information to an investigator and then the investigator has to use that and take further investigative steps to determine whether or not that's a solid suspect for their case," says Gregory.

Several investigators have used the new technology, and several matches have been made, but they can't say what crimes have been solved at this point because most are still in the follow-up stages.

But Gregory says, as with any biometric or fingerprint system, there really is no way to quantify the value of any one particular match, but it will increase their efficiency going forward.

"I've been a part of the fingerprint world for 25 years and my most exciting moments were getting latent fingerprints from the crime scenes and identifying someone," says Gregory. "To be able to do that using another piece of technology is always exciting. You feel like you're contributing to a new world out there."

And while Palmer was originally skeptical that the technology would turn up a match, he now knows its value.

"I'm thinking this technology is something that's going to be more accepted in the future, but this was certainly successful," says Palmer. "It was the break in the case that I needed and everything fell into place."

Date modified: