Better ID through facial recognition
Multiple security applications
Facial recognition is an increasingly powerful technology that is expanding its reach across a broad spectrum of security applications, starting with police forces and customs services, where it is used to check identities and search for known criminals. It is also used by armed forces and homeland security units to identify terrorists. But facial recognition is also rapidly extending its scope to more commercial arenas. For example, casinos use this technology to detect both known cheaters and addicted gamblers who have signed up for a self-exclusion program – not to mention VIP clients who expect special treatment. In stadiums, the system can recognize pre-identified violent fans before they take their seats, while department stores are automatically notified of arriving customers who have already been arrested for shoplifting. The other main applications for facial recognition systems include access control at high-value sites, ID checks for travelers and border control.
Facial recognition : a two-faced conundrum
The effectiveness of facial recognition technology depends on several key factors, starting with image quality, which in turn largely depends on how the facial biometrics are captured. According to Claude Bauzou, "We have to distinguish between cooperative and non-cooperative subjects. Cooperative subjects voluntarily allow us to capture their facial image and they follow instructions, like looking directly at the lens, not smiling, etc. Then there's non-cooperative capture for ID purposes, via surveillance cameras, photos or videos of events taken by witnesses using their smartphones, etc."
Algorithms for identification
The second key performance factor for this biometric technology is the power of the algorithms that are used to determine similarities between facial photos, or matching. For the identification of non-cooperative subjects, matching also involves human input: the system operator has a list of "candidates", sorted by ranked "similarity scores". The operator thus makes a preliminary selection by looking at the person's profile (home address, criminal record, etc.), then makes a decision based on a visual comparison of the face photos.
Facial recognition accuracy also depends on the size and quality of the databases used. To recognize a face, you have to be able to compare it to something! The challenge is to establish matching points between the new image and the source image, in other words, photos of known persons. "The largest image databases in the United States – which only public authorities can access – are those that list holders of driver licenses, passports and other ID documents, not to mention photos of suspects taken during arrests," explains Jim Albers. When police investigators are looking for a suspect, they may also check out photos on Facebook or other social networks, to compare biometric characteristics. There are also private databases, such as those developed by casinos.
A promising outlook
Still a relatively new technology, facial recognition has considerable headroom for improvement. For instance, these systems could add 3D sensors, recognition of moving faces, processing of images captured from above or the side, development of models to integrate aging, and much more. Another area of improvement is the addition of new functions. "In the coming years, systems based on facial recognition will combine official and commercial procedures", says Jim Albers. "For instance, tomorrow's airport checkpoints such as MorphoPASS will not only check the traveler's identity and passport validity, but also their ticket, all in a single passage! Morpho's other proven technologies could be integrated in these checkpoints as well, especially our luggage scanning systems to detect explosives or narcotics."