Facial recognition system

Computers can learn to see faces.

They do this by looking at a photo. The computer finds your eyes and nose. It measures how far apart they are.

These tools can help at an airport. They can also work on a phone. Some people use them to stay safe. It is a very fast way to check who you are. It is a smart way to use tools.

Computers can learn to recognize human faces. This technology is called facial recognition. It is a type of biometrics. Biometrics are ways to identify people using their body parts.

How does it work? The system follows a set of steps. First, it uses face detection. This step finds a face in a photo. Next, it uses alignment. This makes the face straight in the image. Third, it does facial feature extraction. The computer finds parts like the eyes and nose. It measures them to make a map of the face. Finally, it matches that map against a database of other faces.

People use this in many ways. It helps at airports to check travelers. It also works on smartphones. Some companies use it for security. However, some people worry about it. They fear it might hurt privacy. Some systems can even make mistakes. Because of these worries, some cities have banned it.

Facial recognition is a special kind of technology used to identify people. It is a type of biometrics, which means using body parts to prove who someone is.

These systems can match a human face from a digital photo or video against a large database of other faces. People use this for many jobs, like checking IDs or helping with security. It is also used in robotics and on many smartphones today. While it is not as accurate as scanning an iris or a fingerprint, it is very popular because it is contactless.

To work, the system follows four main steps. First, it uses face detection to find a face in a picture. Next, it uses alignment to make the face straight and account for lighting. The third step is facial feature extraction, where the computer pinpoints and measures parts like the eyes, nose, and mouth.

This creates a mathematical map of the face. Finally, the system matches that map against a database to find a person.

This way, the computer can recognize a person even if the photo is just a flat, two-dimensional image.

This technology has a long history that began in the 1960s. Pioneers like Woody Bledsoe, Helen Chan Wolf, and Charles Bisson worked to teach computers to see faces.

Their early project was called "man-machine" because humans had to help the computer. A person would use a tablet to mark points like the corners of the eyes. The computer would then use those points to calculate 20 different distances. In 1970, Takeo Kanade showed a system that could find facial features without any human help. He later wrote the first detailed book on this technology in 1977.

Many different groups have used these tools over the years. In 1993, a program called FERET was started by DARPA and the Army Research Laboratory. This helped create companies like Vision Corporation and Miros Inc in 1994. Some DMV offices in West Virginia and New Mexico were the first to use it for driver's licenses.

In 1999, Minnesota used a system called FaceIT to help police track criminals. More recently, the Viola-Jones algorithm allowed faces to be detected in real-time video by 2001.

Even though it is useful, facial recognition can cause many worries. Some people say it violates privacy or makes mistakes with race and gender.

Because of these concerns, some cities in the United States have banned it. In 2021, Meta Platforms shut down its Facebook facial recognition system. They deleted the face data of more than one billion users. IBM also stopped offering this technology due to similar worries. These big changes show how important it is to use this powerful tool carefully.

Facial recognition is a specialized type of biometric technology. Biometrics refers to the measurement of a person's unique physiological characteristics to identify them. This system works by matching a human face from a digital image or a video frame against a stored database of faces. It is widely used for ID verification and authentication services. Because the process is contactless, it has seen massive global adoption. However, its accuracy is generally lower than other biometrics like iris recognition, fingerprint scanning, or voice recognition.

To identify a person, the computer must complete a specific four-step process. First, the system performs face detection to segment the face from the rest of the image background. Second, it uses alignment to adjust the image for the face's pose and size. It also accounts for photographic properties like grayscale or lighting levels. Third, the system performs facial feature extraction. During this stage, the computer pinpoints and measures specific parts like the eyes, nose, and mouth.

Finally, the system creates a feature vector, which is a mathematical representation of the face. This vector is then matched against a database of known faces to find a potential match.

The history of this technology began in the 1960s. Pioneers Woody Bledsoe, Helen Chan Wolf, and Charles Bisson worked to teach computers to recognize human faces. Their early project was known as "man-machine" because it required human assistance. A person would use a graphics tablet to mark coordinates for features like the pupil centers or the hairline. The computer then used these coordinates to calculate 20 individual distances, such as the width of the mouth. In 1970, Takeo Kanade demonstrated a system that could locate anatomical features like the chin without human help. Kanade later published the first detailed book on facial recognition in 1977.

In the 1990s, research shifted toward more automated methods. In 1993, DARPA and the Army Research Laboratory established the FERET program. This program evaluated different methods to see if they could work in real-life security and law enforcement settings. The FERET tests were successful enough to spawn three U.S. companies, including Vision Corporation and Miros Inc. During this decade, the technology moved into public service. For example, DMV offices in West Virginia and New Mexico were among the first to use these systems. They used facial recognition to prevent people from getting multiple driver's licenses using different names.

Mathematical models have changed how computers "see" faces over time. In the early 1990s, Matthew Turk and Alex Pentland developed Principal Component Analysis (PCA). This method is also called the Eigenface approach. It uses a linear model to represent a human face as a weighted combination of several "Eigenfaces." This greatly reduced the amount of data a computer had to process. Later, in 1997, researchers improved this using Linear Discriminant Analysis (LDA) to create "Fisherfaces." By the late 1990s, the Bochum system used Gabor filters to link facial features into a grid. This system was robust enough to identify people despite glasses, beards, or different hairstyles.

Modern applications have become even more integrated into daily life. In 2001, the Viola-Jones object detection framework made real-time face detection in video possible. By 2015, this algorithm was small enough to run on handheld devices and embedded systems. Today, the technology is used in many different ways. It is used in video surveillance, passenger screening at airports, and even in robotics.

Some governments even use it in specialized ways. For instance, Ukraine has used Clearview AI software to identify deceased Russian soldiers. This has been used as a form of psychological warfare to raise awareness of Russian activities.

Despite its utility, facial recognition is highly controversial. Many critics claim the technology violates privacy and can make incorrect identifications. There are also serious concerns regarding racial profiling and gender bias. The rise of synthetic media, such as deepfakes, has created new security worries. These issues have led to real-world consequences for major tech companies. In 2021, Meta Platforms shut down its Facebook facial recognition system. This move resulted in the deletion of face-scan data for more than one billion users. IBM also stopped offering its facial recognition technology due to these societal concerns.