Face Recognition Technology

Facial recognition records the spatial geometry of distinguishing features of the face. Different vendors use different methods of facial recognition, however, all focus on measures of key features of the face. Because a person’s face can be captured by a camera from some distance away, facial recognition has a clandestine or covert capability (i.e. the subject does not necessarily know he has been observed). For this reason, facial recognition has been used in projects to identify card counters or other undesirables in casinos, shoplifters in stores, criminals and terrorists in urban areas.

Although the concept of recognizing someone from facial features is intuitive, facial recognition, as a biometric, makes human recognition a more automated, computerized process. What sets apart facial recognition from other biometrics is that it can be used for surveillance purposes. For example, public safety authorities want to locate certain individuals such as wanted criminals, suspected terrorists, and missing children. Facial recognition may have the potential to help the authorities with this mission.

Facial recognition offers several advantages. The system captures faces of people in public areas, which minimizes legal concerns for reasons explained below. Moreover, since faces can be captured from some distance away, facial recognition can be done without any physical contact. This feature also gives facial recognition a clandestine or covert capability.

For any biometric system to operate, it must have records in its database against which it can search for matches. Facial recognition is able to leverage existing databases in many cases. For example, there are high quality mugshots of criminals readily available to law enforcement. Similarly, facial recognition is often able to leverage existing surveillance systems such as surveillance cameras or closed circuit television (CCTV).

As a biometric, facial recognition is a form of computer vision that uses faces to attempt to identify a person or verify a person’s claimed identity. Regardless of specific method used, facial recognition is accomplished in a five step process.

• First, an image of the face is acquired. This acquisition can be accomplished by digitally scanning an existing photograph or by using an electro-optical camera to acquire a live picture of a subject. As video is a rapid sequence of individual still images, it can also be used as a source of facial images.
• Second, software is employed to detect the location of any faces in the acquired image. This task is difficult, and often generalized patterns of what a face “looks like” (two eyes and a mouth set in an oval shape) are employed to pick out the faces.
• Once the facial detection software has targeted a face, it can be analyzed. As noted in slide three, facial recognition analyzes the spatial geometry of distinguishing features of the face. Different vendors use different methods to extract the identifying features of a face. Thus, specific details on the methods are proprietary. The most popular method is called Principle Components Analysis (PCA), which is commonly referred to as the eigenface method. PCA has also been combined with neural networks and local feature analysis in efforts to enhance its performance. Template generation is the result of the feature extraction process. A template is a reduced set of data that represents the unique features of an enrollee’s face. It is important to note that because the systems use spatial geometry of distinguishing facial features, they do not use hairstyle, facial hair, or other similar factors.
• The fourth step is to compare the template generated in step three with those in a database of known faces. In an identification application, this process yields scores that indicate how closely the generated template matches each of those in the database. In a verification application, the generated template is only compared with one template in the database – that of the claimed identity.
• The final step is determining whether any scores produced in step four are high enough to declare a match. The rules governing the declaration of a match are often configurable by the end user, so that he or she can determine how the facial recognition system should behave based on security and operational considerations.

People are generally very good at recognizing faces that they know. However, people experience difficulties when they perform facial recognition in a surveillance or watch post scenario. Several factors account for these difficulties: most notably, humans have a hard time recognizing unfamiliar faces. Combined with relatively short attention spans, it is difficult for humans to pick out unfamiliar faces. Considerable evidence supports this claim. For example, in a British study, trained supermarket cashiers were tested on their ability to screen shoppers using credit cards that included a photograph of the card owner. Each shopper was issued four cards: one with a recent picture of the shopper, one that included minor modifications to the shopper’s hairstyle, facial hair or accessories (e.g., glasses, hat), another card with a photograph of a person similar in appearance to the shopper, and the last card with a photograph of a person who was only of the same sex and race as the shopper. When the various cards were presented to the checkout clerks, more than half of the fraudulent cards were accepted. The breakdown was as follows: 34 percent of the cards that did not look like the shopper were accepted, 14 percent of the cards where the appearance had been altered were accepted, and 7 percent of the unchanged cards were rejected by the clerks. In addition to unfamiliar face recognition problems, the ability of human beings to detect critical signals drops rapidly from the start of a task, and stabilizes at a significantly lower level within 25 to 35 minutes. Thus the ability of people to focus their attention drops significantly after only half an hour.

Machines also experience difficulties when they perform facial recognition in a surveillance or watch post scenario. Dr. James L. Wayman, a leading biometrics expert, has explained that performing facial recognition processes with relatively high fidelity and at long distances remains technically challenging for automated systems. At the most basic level, detecting whether a face is present in a given electronic photograph is a difficult technical problem. Dr. Wayman has noted that subjects should ideally be photographed under tightly controlled conditions. For example, each subject should look directly into the camera and fill the area of the photo for an automated system to reliably identify the individual or even detect his face in the photograph. Thus, while the technology for facial recognition systems shows promise, it is not yet considered fully mature.

The “Facial Recognition Vendor Test 2000” study makes clear that the technology is not yet perfected. This comprehensive study of current facial recognition technologies, sponsored by the Department of Defense (DoD) Counterdrug Technology Development Program Office, the Defense Advanced Research Projects Agency (DARPA), and the National Institute of Justice, showed that environmental factors such as differences in camera angle, direction of lighting, facial expression, and other parameters can have significant effects on the ability of the systems to recognize individuals.

By controlling a person’s facial expression, as well as his distance from the camera, the camera angle, and the scene’s lighting, a posed image minimizes the number of variables in a photograph. This control allows the facial recognition software to operate under near ideal conditions – greatly enhancing its accuracy. Similarly, using a human operator to verify the system’s results enhances performance because the operator can detect machine-generated false alarms.