Over the last several decades, the increasing popularity of law enforcement television shows has made the general public become familiar with the concept of facial recognition software. A common plot device for screenwriters looking to introduce a character, facial recognition systems are any device able to identify an individual from a digital image or a frame from a video file. These systems work in several ways and are used in a surprisingly large variety of ways throughout the technological world.
For the most part, facial recognition software works on simple, basic programming functions. Algorithms are written and mathematical models are made in order to identify several main facial features in the image being analyzed, which are then compared to images stored in a database of some kind. These programs are becoming more and more sophisticated as time passes, taking note on everything from skin blemishes to bone structure in order to ensure that the wrong images are not matched together.
Originally, facial recognition capabilities began to be developed in the early 1960s, by researchers employed by a still unidentified intelligence agency. The original method, as developed by scientists Blesdoe, Chan, and Bisson, was to measure the distance between important features on the face such as the eyes and nose. These distances were then mapped against those stored in the database, allowing a crude if fairly reliable method of facial recognition.
Over the next five decades, the ability of facial recognition software to match images of faces together has improved drastically. By the 1990s the facial recognition systems of Peter Hart were able to reliably outperform volunteers attempting to match the images manually. By 2010, these programs could successfully distinguish between images of identical twins, far out matching human beings’ ability to do the same.
There are four main ways that facial recognition software is used to map out the layout of a human face. The traditional method, involving the distribution of facial landmarks, is the first of these. The second method involves the use of three dimensional scanning technology, and is the most reliable method currently available. The third method is the use of skin texture analysis, storing information on the slightest skin discoloration or blemish. And the fourth and final method is the use of thermal cameras, in order to render disguises useless.
The traditional method of scanning an image for the facial recognition program is the identification of certain facial landmarks, which are then broken down individually and used to search for images which contain the same features. There are hundreds of algorithms and methods used over the decades in order to improve the accuracy and efficiency of this process. Three dimensional face recognition relies on the use of delicate sensors to map the contour of faces, so that the angle of the image is less important. Skin texture analysis is usually used in combination with the other methods, improving their reliability by up to twenty percent. And the use of thermal imaging technology allows the software to focus on shape of the face alone, ignoring facial hair and scars.