Canada's front-line border officers back the idea of a perimeter security arrangement with the United States, with a few caveats.Facial recognition often gets knocked around a bit because casual observers think it is supposed to function well as an unattended system. Maybe one day it will, but not yet.
The Customs and Immigration Union wants more intensive screening of travelers, including a biometric face-recognition tool to pinpoint security threats and wanted criminals.
That, however, doesn't make it useless. It is a valuable tool in the hands of trained personnel.
Computers don't look at the world the way we do. Whether that's a good thing or not depends on what you're trying to accomplish. For facial recognition in a law enforcement context, it's a good thing to have a radically different point of view applied to a challenge.
First, faces are probably the most meaningful objects in human existence. It's not too much of an exaggeration to say that for millennia human survival has depended upon our abilities at one type of facial recognition: recognizing people you know. Sorting through hundreds of thousands of pictures of people we don't know in order to match the two that are of the same person, however is not something we're inherently good at.
Computers can do that in less than a second, then give the two pictures to a human which is very good at making the single comparison.
The computer does this by treating the face as a mathematical formula. It finds the eyes, then the nose, then other points, measures the distances and angles between the points and turns that all into a long number. Then it does that for everyone in the database. Then it just compares numbers to see which ones come close to matching. It doesn't care about age, race, or gender. It'll match a white man and a black woman. It will match a picture of an old person taken one year ago with a picture of a young person taken today (i.e. a match that's impossible because people don't get younger).
People don't do that. When you ask a person to describe someone they're likely to tell you they were Hispanic, middle aged, round faced, good looking, brown eyes, etc. The computer says: Their eyes were 57.55 mm apart; tip of nose 25 mm below that line 2mm off center to the left, etc.
People and Face Rec make a good team because it's hard to fool them both at the same time. People can't memorize the faces of 1000's of people they don't know & you can fool them by cross-dressing or disguising yourself as a much older person. Computers don't know anything about gender, ethnicity, age or attractiveness, but they can do the math on 1000's of faces in an instant.