Facial recognition is a biometric access control technology that uses one or more photographic images to recognize a person by measuring points on a face under controlled conditions. Facial recognition systems are not intrusive, require no physical contact with the user, and have a high rate of user acceptance.
Facial recognition is not affected by race or gender-based differences in appearance. It is a robust technology capable of handling a wide range of body types and facial characteristics.
Facial recognition is used throughout the world in industries as diverse as banking, gaming, healthcare, law enforcement, customs, and retail. The technology has been tested successfully in neutral industry comparisons and is currently the fastest growing segment of the biometric access control market.
Facial recognition has a number of desirable aspects.
• The process is intuitive. The user’s interaction with the camera is similar to being photographed for any number of identification applications, from driver’s licenses to passports.
• Facial recognition is non-intrusive. Unlike other biometric access control systems, facial recognition requires minimal interaction with the user.
• Facial recognition is hands-free. The user makes no physical contact with the camera, or detector, and does not contaminate it with body oils or other debris.
• Facial recognition systems retain a photographic record of attempted entries by unauthorized personnel. Although some fingerprint systems can store the fingerprint of an unauthorized user, fingerprint analysis requires special training. Analyzing and recognizing faces is a natural, instinctive human behavior.
• Finally, ambient lighting is sufficient for most facial recognition applications, with no requirement for special white or infrared lighting. However, matching is most effective when the ambient lighting is similar to the lighting used to create the template during enrollment.
Operation
A number of technologies are used in facial recognition access control systems. The two major categories are video imaging and thermal imaging. Video facial recognition analyzes the unique shape, pattern, and positioning of facial features by mapping those features to create a mathematical model . Video can be further subdivided into two-dimensional (2-D) and three-dimensional (3-D) imaging. The number of cameras used is the primary physical difference between the systems.
3-D systems use two cameras and integrate the images to create a 3-D digital template. In 2-D models, a single camera acquires the image. Thermal imaging uses an infrared camera to produce a facial thermograph. The system digitizes the thermal pattern resulting from the heat produced by the blood vessels under the skin.
facial recognition technology enrollment for access control purposes is straightforward, requiring 20 to 30 seconds to take several pictures of the enrollee’s face. This photographic sequence is best done with varying angles and expressions to allow for more accurate matching. The system extracts the relevant information and uses mathematical techniques to create a reference template that is stored in the database.
Verification is similar to enrollment. The user claims an identity through a login name, smart card, or terminal entry and then sits or stands in front of the camera for a few seconds. The system captures an image, creates a template from the extracted information, compares it to the reference template, and then grants or denies access. The point at which the two templates are similar enough to match, known as the threshold value, can be adjusted for different persons, time of day, and other factors.
Applications
Facial recognition is useful for indoor verification applications where the ambient lighting and environment can be controlled. The camera must be situated so that quality facial images can be captured. Facial recognition is not recommended for areas where lighting is not uniform or situations where personnel protective equipment (PPE), such as face masks, is required.
Performance Metrics
The most obvious and useful metric for facial recognition access control is the quality of the capture device. In general, the greater the resolution and contrast of the captured image, the better the system will recognize faces. Processing time is an important factor in user acceptance.
A system should be able to discriminate a live face from attempts to spoof, or deceive, the system using a photograph or a video feed. Finally, any access control system should have a low FMR.
The reliability of some facial recognition systems can be affected by one or more factors involving the user or the environment. Differences in the devices or the light conditions in the enrollment and field environments can increase the false rejection rates. Changes in facial hair, hairstyle, headgear, eyeglasses, body weight, or facial aspect (e.g., angle at which the image is captured) can change the facial shape or outline. Loud or bright clothing can interfere with the system’s ability to distinguish the face in the image. Too much or too little movement on the part of the user can cause some systems to reject falsely or to fail to recognize a live facein facial recognition technology.
Vulnerabilities
Some 2-D facial recognition systems may recognize photographs or video material as a face and allow access to an unauthorized person. Systems that are 3-D are not as vulnerable to spoofing. Recent 2-D systems are capable of recognizing live faces and are not vulnerable to photo or video induced false matches.
References :
- Access Control Technologies Handbook
see also :