What is facial recognition technology and how does it work?
Facial recognition, a technology that has experienced a huge growth in usage over the last several years, is a way of identifying someone based on a previous photograph of their face. Once a biometric is assessed (a measure of a person’s facial features: the width of their nose, the height of their forehead, the fullness of the lips, etc) and stored, this information is used to identify a person in other digital images. While the human eye can recognize people it knows with ease, it is less successful at identifying strangers. The human also tires of looking at hundreds, sometimes thousands of images, but an electronic “eye” experiences no such fatigue. This technology has been successfully employed by police, in surveillance, in forensics, and even in social media.
The technology was first developed in the late 1980s but was not used commercially until the 1990s. It was put to the test on January 28, 2001, during Super Bowl XXXV in Tampa Bay, Florida. Without their knowledge, everyone who entered the Raymond James Stadium that day was digitally photographed, over 100,000 mages were then compared to a police database. According to ABC News, nineteen “petty criminals” were arrested as a result.
Although its first trial’s success was arguable, the technology continued to be used and improved. Today, facial recognition is widely used by law enforcement, the government, and the military. Its use in helping to find missing children has been one of the most laudable uses of facial recognition. Nearly half of all states use it for this purpose. Private use has also embraced facial recognition. Casinos, in particular, universally employ facial recognition technology to catch card counters and other dishonest patrons. It is also routinely employed by private companies, used in airports, and at ATMs, to name a few more common places you might encounter a facial recognition camera. Social media sites, like Facebook, also use facial recognition when you upload photos. If you have previously “tagged” a person, the next time you upload a picture of that friend or family member, the software will tag that person for you (you do have the option to decline the tag).
Facial recognition software first compiles images into a database, converting images into numbers to represent various facial feature. These number sets represent a single face and are compared to other images and the numbers representing other faces.
Not all facial recognition technology is the same. Some are much more nuanced and reliable than others, depending on the quality of the image captured as well as the quality of the mathematical algorithms employed in the software. Things that can affect image quality include how much light is present (too much or too little affects quality), the background (solid backgrounds are preferable to cluttered ones), and the angle of the head (straight on produces better image results than heads that are titled or turned). Images can be captured either when the subject is moving as well when the subject is still. Stills are often captured in mug shots by law enforcement. Those in motion are often images gathered by security cameras. These are often harder to correctly identify, but useful enough to continue the gathering of data.
There are two techniques used for facial recognition: geometrical and/or photometric. Geometrical data relies on shapes on the face, called “nodal points”(the distance between the eyes, nose width, jawline, cheekbones, depth of eye sockets and chin length) and the position of the head. Photometric data creates a template of the features of the face and uses that template for identifications.
The thing that most separates one type of facial recognitions software from another is the quality of the algorithm employed. Algorithms are propriety and typically kept very secret. Despite improvements, there is still no system available that can report with one hundred percent accuracy. Despite the qualitative differences, all facial recognition software follows the same process steps:
1) acquisition of the image
3) normalizing the image, if needed (scaling and rotating)
4) feature extraction
5) creation of a numerical map of the face
6) storage of the image in a database.
The stored images then may be compared either one-one-one or one-to-many. Facial recognition technology has been used to identify criminals from the petty thief to the terrorist, from missing children to finding old friends online. Still in its relative infancy, expect facial recognition technology to become ever more accurate and ubiquitous.
Source: World of Forensic Science, ©2006 Gale Cengage. All Rights Reserved.
Facial recognition technology was undoubtedly one of those "futuristic" things which was once featured in some old book or science fiction movie; however, it has become a reality. In the shortest terms, this technology is a computer application which is able to identify a person based on either a digital photograph or video image. It does require a database of faces with which to compare them.
One of the two ways facial recognition works is by identifying the particular identifying features of a face (in other words, all of the things which make your face different from everyone else's). This is called the geometric approach, and it measures all of the facial features, also known as landmarks, of any face.
Each human face has approximately 80 nodal points. Some of these measured by the software are
These nodal points are measured creating a numerical code, called a faceprint, representing the face in the database.
- Distance between the eyes
- Width of the nose
- Depth of the eye sockets
- The shape of the cheekbones
- The length of the jaw line
Unfortunately, these measurements are not generally able to be compared to a perfect facial specimen when seen in 2D. By that I mean that a face captured, let's say, in a passport control line, may not be directly facing the camera, may be partially in a shadow, or have a particular expression which is unreadable by the technology. Lack of resolution or the addition of sunglasses or facial hair can also have an impact. All of these uncontrollable factors have an impact on how the image is read by the computer application.
3D facial recognition technology does do a better job of identifying unchangeable features such as the curves of eye sockets, noses, and chins. It is even able to recognize a face by profile only. These 3D images are converted to 2D images which can then be compared to other 2D images, such as photographs.
The second type of facial recognition technology works by breaking the image down to small pixels, giving them each a value, and then eliminating the variances.
The forerunners to this technology are the more familiar biometric processes of fingerprint or iris recognition systems.
This facial recognition technology is used all over the world in such diverse places as airports, casinos (to identify card counters or blacklisted players), customs, criminal justice systems, law enforcement, and even the Super Bowl XXXV. One of the largest users of this technology is the United States Department of State, which
operates one of the largest face recognition systems in the world with over 75 million photographs that is actively used for visa processing.
Despite its ubiquitous presence, most applications of this program have been disappointingly ineffective. Work on facial recognition technology began in the 1960s and it has certainly been improved; however, there is still much work to be done to make it an effective technology for accurate, consistent usage.
Facial recognition technology is a type of technology that has the ability to captures ones face and recognize it through a series of capturing unique facial features likes one's jaw line, nose bridge, forehead, etc. Facial recognition has been integrated into a lot of today's technology such as phones, computers, camera systems. This technology enables criminals to be caught and security to become extra secure.