General Background
Face recognition is a fairly young technology compared to other biometrics. Research in this field has been going on for decades, but it has been in the last 10 to 15 years that the greatest advances have taken place.
There is new software which can recognize faces within a crowd in the attempt to match them to stored images of known criminals. With the recent terrorist attacks there has been a tremendous push to implement this technology in a variety of public places such as airports, government buildings, border crossings, and other vulnerable areas. Prior to the attacks, however, this technology was still being implemented in high crime areas like the streets of Tampa, as well as other cities. Future uses of this technology may include identification systems for such things as security sensitive areas and ATMs. The eagerness to adopt face recognition can be partially explained by its relatively cheap price. Recognition software can run as cheap as a few thousand dollars, and with the ability to utilize PC cameras the cost is significantly lower than other biometrics hardware.
How It Works
Generally speaking face recognition works by first obtaining an image of a person. This process is usually accomplished by a video camera with at least a 320x240 resolution at 3-5 frames per second. Higher quality cameras will of course produce more accurate results. Second, the computer software analyzes certain features of that image through different techniques, or a combination of techniques. Finally, verifying that person's identity is accomplished by matching those features to other images stored in a database. The analysis process over the years has moved away from using the simple geometry of key facial points to the use of more complex mathematical techniques.
The four main methods that are currently being used are eigenfaces, feature analysis, neural network, and automatic face processing. Eigenfaces is a tool developed by MIT that extracts characteristics through the use of two-dimensional grayscale imagery. Feature analysis, or sometimes referred to as Local Feature Analysis (LFA), is the most widely used technique because of its ability to accommodate for facial changes and aspect. LFA
uses an algorithm to create a face print (84 bytes in size) for comparison. Neural network is a method that extracts features from the face and creates a template of contrasting elements that is then matched to a template in a database. Some people think that the neural network technology is the next step in face recognition. Automatic Face Processing (AFP) is a technique that looks for distances and distance ratios between certain facial features, and is more ideal for poorly lit areas.
The Ongoing Debate
Accompanying the implementation of some of these systems there has been an ongoing debate over two issues of its use: effectiveness and infringement on civil liberties. The implementation of
this type of software in Tampa, Florida became one of the first major battlegrounds over some of these issues.
The underlying fact is that the Tampa police department did stop using the face recognition system after a few months. Face recognition has also been criticized
as being a potential violation of our civil liberties. Privacy becomes a major concern, especially when the general public may be somewhat uninformed of the capabilities of such a system. On the flipside of this debate many people feel that biometric technologies are imperative for our security in the wake of the terrorist attacks. They also argue that it could be a great tool for apprehending other criminals as well. Advocates place emphasis on the fact that if you are obeying the law then you should not worry about the existence of this technology.
Face recognition can be used as a
"confirming" resource to control access to a portal (door, opening,
etc.). By saving images of known authorized individuals, the face recognition
system then becomes a way to identify authorized entry, often in combination
with smart cards, fingerprints, or other methods. Because the system is only
confirming known images, there is no privacy issue at all. The GAITS Bio-Facial
Recognition Portal Control System is one example of this type of system. |
|
 |
|
