If you’re a fan of spy films, you’ll probably have come across facial recognition technology. These systems aren’t just for unlocking the door to a secret lair, though. From Face ID on the iPhone X to airport border control, facial recognition technology is finding its way into our lives more and more every day.
While it sounds like the stuff of sci-fi, this technology isn’t as complicated as you might think. Get a grasp on these ten buzzwords and you’ll be an expert.
An ePassport carries biometric data, with at least the holder’s portrait securely stored in a contactless chip. Your Emirates ID card does this too.
This information is stored in a standard format defined by the International Civil Aviation Organization. It can be retrieved and verified by an agent using a passport reader or at a self-service Automatic Border Control (ABC) gate anywhere in the world.
Automated Border Control (ABC) Gate
At border control, your identity is verified either by an officer who compares your face to the picture on your passport, or digitally by a camera-enabled system in an ABC gate.
Instead of scanning your passport photo from the page, an ABC Gate reads the data stored on the contactless chip. This makes sure it is not confused by any physical blemishes to the passport page and ensures that the document has not been altered or fraudulently issued. In Dubai International Airport, this technology is used to process passengers in less than 15 seconds!
Live Facial Recognition Systems
Facial recognition systems are already being used for security in public places, as well as airports. The software identifies people as they walk past a camera, highlighting people of interest to security teams.
It’s this technology that Dubai Police plan to use to prevent crime in tourist destinations across the city in the run-up to Expo 2020. High-tech surveillance cameras will use facial recognition software to identify people involved in criminal activity.
The system for a security camera is more complex than an ABC Gate as it will have to search for faces and isolate them from the crowd. Then, it will likely have to tilt, stretch, distort and re-orient them before being able to locate the facial features and align them for accurate measurement.
From the captured photo, the facial recognition system generates a faceprint, unique to the single individual. This is a data-set generated from precise measurements of up to 85 precise points, such as the distance between the eyes and the length of the jaw line. A biometric system will then use a faceprint for either verification or identification.
For verification, the faceprint is compared to one biometric profile. The Smart Gates at Dubai airport verify by comparing a faceprint to the information on a passport.
Identification occurs when a system has to recognise an unknown person. The person’s biometrics have to be checked against all the others in a reference database, such as a watch list.
Facial Recognition Rate
The automated recognition of individuals based on their biological characteristics involves the transformation of analogue information (face) to digital information (faceprint). This is where technical challenges arise, reducing accuracy.
The accuracy of a facial recognition system is shown by its Recognition Rate (RR). If nine faces out of 10 are correctly verified by a system, its RR would be 90%. The larger the data set a system’s algorithm is trained with, the more accurate it can become.
True Positives, True Negatives, False Positives and False Negatives
Accuracy in verification is measured with these outcomes:
- True positive: recognition of a legitimate user
- True negative: recognition of an illegitimate user
- False positive: recognition of an illegitimate user as legitimate.
- False negative: recognition of a legitimate user as illegitimate.
False Acceptance and Rejection Rates
Identification occurs when a system assigns a faceprint to the person with the most similar biometric template. To prevent false positives and negatives, the similarity has to lie between certain levels. This threshold affects the system’s False Acceptance Rate (FAR) and False Rejection Rate (FRR) – the instance of false positives and negatives.
With similarity threshold that is too low, FRR will be minimised but FAR will increase, and if the threshold is too high the opposite will occur.
Because of this, the thresholds of facial recognition systems have to be adjusted depending on the application.
In a system used for passenger clearance, a 1:100 FAR would be considered low security as one unauthorised person out of 100 could pass through the immigration process. For sensitive uses such as this, the minimum recommended FAR is 1:1000.
AI & Machine learning
Central to facial recognition, AI is the technology that allows machines to act like humans. With AI, machines use algorithms to solve complex problems and learn from a stream of inputs, identifying and categorising data than taking action.
AI and deep learning enable machines to improve their performance, with better speed and accuracy, thereby tackling the challenge of human error.
The possibilities are endless
The possibilities for the development and application of facial recognition technology are endless. From eliminating airport waiting times to helping apprehend criminals, this technology will soon be everywhere in a range of different forms.
Dubai International Airport has developed a facial recognition system that makes security a novelty. Outbound passengers will soon be able to clear security by walking through a virtual aquarium lined with cameras that identify them as they pass by!
Get ready for a time when showing an ID means you’ll just have to flash a smile – the future is now! Facial recognition is about to change security as we know it.