Biometric identification systems.
Biometric identification systems
As noted in the Photonics Spectra magazine, the control systems of any modern society are based on the use of computers and information. Such systems require high speed, ease of use when performing various transactions and reliable protection against unauthorized actions, i.e. security. Fulfillment of all these requirements is associated with a certain risk, and when assessing the degree of this risk, it is necessary to maintain a certain balance between the ease of use of such systems and their necessary security:
The main problem of ensuring the security of systems and objects is the identification of persons attempting to gain access to the system or object. In the process of solving this problem, various methods and technical means based on the use of computers, image formation systems and complex mathematical algorithms have been proposed and developed.
A special place among identification systems is occupied by biometric systems, based on the use of certain individual and unique features or characteristics for each person, also called biometric keys. Such features include the pattern of the iris, fingerprints, hand shape, face or voice.
Image processors compare images of these features with previously recorded images or samples obtained by mathematical modeling methods to verify the identity of the person possessing these features. Identification can also be carried out using control images stored in a central database.
However, despite some success in the application of these promising systems, their wider distribution is held back by a number of problems (cost, speed, reliability, public acceptance). However, new advances in photoelectronics, microelectronics and video technology give reason to believe that these problems will be successfully resolved.
Improvement in the production technology of charge-coupled devices (CCD) and increasing customer requirements for these devices have led to the creation of a new generation of digital and color video cameras. Such cameras have become an important element of security systems — from government agencies to casinos, due to their compactness and high sensitivity.
A digital video camera with a CCD matrix has an analog-to-digital converter on KMOS semiconductor devices located directly behind the matrix. The digital image signals produced by such a camera can be easily input into the memory of a computer or a digital video recorder, which can store five times more data than a conventional VCR. Accordingly, the price of color video cameras has fallen, which currently does not exceed $ 200. Compact image processing software provides a number of advantages, including instant recall of a specific image, image quality improvement, and real-time comparison of the presented image with a stored reference image. A biometric identification or recognition system uses these capabilities to solve a variety of problems, including restricting access to certain physical areas, computer networks, and specific resources.
The US National Security Agency (NSA) management has shown interest in developing biometric facial recognition systems. This interest was supported by the fact that existing inexpensive digital color video cameras with accurate 24-bit color reproduction, high resolution (640×480 image elements) at 30 frames per second and a wide dynamic range will ensure the success of such systems.
A number of US companies (Miros, Lau Technologies, Identification Technologies International) have already developed systems for recognizing a person by face, which operate similarly to a police officer checking a driver's license and comparing his face with the photograph in the presented document.
Lau Technologies' Face-in-the-Crowd system uses a Hitachi (Japan) autofocus video camera to capture an image of the face of the individual being scanned. The camera's frame capture board converts analog image signals to digital signals, and Lau Technologies' proprietary software creates a control image sample, represented by 128 bytes, based on the digital signals. These samples are included in a biometric signature file. Such a signature does not represent the entire image, but only a certain part of it or a certain biometric feature obtained using a unique mathematical algorithm. The formation of such samples partially eliminates the problem of the individual being scanned directly in the comparison process. His control sample can be included in the identification card he is issued, from which it is read, instead of requesting it from a database.
According to the company, its system can identify terrorists at airports or other places where there are large numbers of people. During testing of the covertly installed system, its camera could .select up to four wanted faces in a crowd. To do this, it is necessary to point the camera to where the person looks natural, similar to how he looks when passing through the inspection arch of the metal detector.
A similar system from Identification Technologies International eliminates the problems associated with ambient light during filming by using a near-infrared LED with a CCD camera. Bandpass filters in front of the camera's lens block most of the light beyond 880 nm, and the company's proprietary software creates a topographic map of the face in the frame for comparison with reference images in a file.
Both companies believe their systems will find application in the increasingly common use of ATMs.
The Miros True Face system, developed for this application, uses two Pacific Electro Optics monochrome cameras (USA) to create a stereoscopic image of a person's face. By comparing this image with a reference image, the system can determine the differences between the two, which are reflected in the perspective shifts of individual parts of the face. Uniform shifts indicate that the stereoscopic image is a true image of the face.
Miros operates seven True Face ATM kiosks with an electronic payment company in Fort Worth, Texas, and is quite satisfied with their performance.
Despite these successes by Miros, developers of biometric facial recognition systems still have a number of challenges to overcome in order to create iris scanning systems to dominate the ATM market.
The first iris scanning system developed by Sensar (USA) has attracted the interest of a number of banks in the US and other countries, who have expressed a desire to finance further development of such systems.
«The heart» of the Sensar system was the iris-alignment algorithm developed by Iriscan (USA). The system used CCD chips from the Japanese company Chughiboyeki, which were designed to perceive near-infrared radiation reflected from different points of the iris pattern. After converting the output signals of the chips into digital form, a special program of the system selects one of 30 frames reproduced in 1 second, and the system performs 192 radial measurements (angle and distance from the center) and forms an image sample represented by 256 bytes. The Sensar system can obtain images of the human retina from a distance of up to 1 m (compared to 30 cm for other systems). This distance depends on the diameter of the input aperture of the system's lens.
A similar system, Irisident, from Iriscan uses two fixed-focus video cameras to detect a person approaching. An algorithm controls the cameras to determine the position of the person, their head, eyes, and iris in an X,Y coordinate system. The resulting image signals are passed through a pyramidal imaging microchip (developed by Sarnoff Research Laboratory for the U.S. Air Force), which turns the video cameras into variable-resolution image sensors. When the best resolution is achieved, the microchip forms a three-dimensional image in an X, Y, Z coordinate system.
Based on the signals from the microchip, a tracking mechanism focuses a third camera on the iris with a depth of field of up to 25 mm.
Sensar's developers hope to replace the gold-plated retinal scanning mirrors and magnifying optics with a micromirror array, thereby making the system more compact.
There are different opinions on the use of biometric identification systems in banks. Thus, the American Banking Association's public relations representative, while acknowledging the importance of the problem of combating theft in banks, believes that the cost of switching to ATMs and vending machines with biometric identification systems will create new problems. Other bank representatives fear that the introduction of biometric systems may cause a negative attitude towards them from clients.
Biometric identification, in particular fingerprint identification, is widely used in systems using credit cards.
According to Master Card (USA), which developed the optical biometric fingerprint identification system, 6,700 visitors have been checked in the company's offices since the system was installed in 1996. The company believes that this system is the most convenient for credit card holders.
The San Bruno (USA) identification system uses a near-infrared LED to illuminate the fingers from the side and produce a raised fingerprint pattern. A plastic microprism, manufactured using a proprietary method, deflects the light coming into it and directs it to the 113C chip. The company offers three models of its system: two of them have internal analog-to-digital converters, and the third, the most compact system, the size of two cubes of sugar, can be combined with an intelligent mouse-type manipulator or a computer keyboard.
There are currently 700 ATMs with biometric identification in Spain. If the image quality deteriorates, for example due to scratches, the identification system can be easily replaced. Image contamination is often the reason for a critical attitude towards optical readers.
Fingermatrix (USA) has developed printers for one and ten fingers, in which the optical system is located under a bath of alcohol and water. The liquid layer protects the surface on which the image is reproduced from contamination and increases light transmission.
The ten-finger printer is the only biometric device that meets US government standards. The US National Institute of Standards and Technology, together with the Federal Bureau of Investigation (FBI), developed requirements for images produced by scanners. They must have a resolution of up to 200 dots per centimeter and 8-bit grayscale.
To meet these requirements, the te-printer system, designed specifically for forensic investigations, uses an additional linear CCD array.
Because federal agencies require full fingerprints, the operator must rotate the defendant's fingers to take the prints.
Surface or planar scanners like those used in access control systems cannot produce clear images of moving fingers, so a linear CCD array tracks the fingers and moves with them as the prints are taken.
During 1997, the Boston Police Department (USA) participated in a program to develop a pilot system for electronic transmission of optical scanner images to the FBI's central database. This system would reduce the requirements for scanning systems, since low-quality images could be rejected at the database entry.
Searching for a required fingerprint in a database is associated with significant difficulties and takes a lot of time. New methods have been developed using spatial light modulators and Fourier transforms, which reduce the search time by 100 times.
The optical correlator developed by the institute uses a helium-neon laser and a collimating lens. The laser beam passes through a spatial light modulator on a liquid ferroelectric crystal to form an input image. The lens of the Fourier converter focuses the beam on a second filter registered on a hologram carrier located in the Fourier plane. The coincidence of the input image on the hologram carrier is accompanied by bursts of light intensity on the CCD matrix of the video camera located in the output plane.
Another American company, Quatalmage, has developed a more advanced correlator that uses a high-speed spatial light modulator (response time less than 1 μs) with a resolution of 200 lines/mm, developed by the company. The computer-generated image is sent to two ferroelectric spatial light modulators, irradiated with light from a laser diode with a wavelength of 830 nm. The laser beam passes through the lens of a Fourier transform. The high-speed spatial light modulator amplifies the Fourier-transformed image. A second laser beam with a wavelength of 850 nm reads the amplified image and transfers the results back through the lens of the Fourier transform to an intelligent sensing element capable of detecting correlation peaks when comparing up to 4,000 fingerprints in 1 second.
Quantalmage hopes to increase the system's throughput from 1,000 to 4,000 fingerprints per second by replacing the two input ferroelectric spatial light modulators with one high-speed spatial light modulator and one ferroelectric light modulator.
Experts say current electronic technology can match fingerprints at rates of up to 3,000 per second, but only after the number of potential matches is limited by pre-processing suspicious information such as gender and age.
Biometric identification systems will be used more widely in state border security services. This will be facilitated by the IBM (USA) Fastgate system, developed on the model of the system used by the US Immigration and Naturalisation Service. The Fastgate system, currently undergoing testing, uses hand geometry scanning technology from Recognition Systems (USA).
The system uses a light-emitting diode (LED) emitting near-infrared light at 690 nm to illuminate four fingers of a hand. The hand rests on a prismatic substrate of the type used in road sign construction. The surface of the substrate reflects incident light back to the source and a nearby camera. Broadband filters block as much ambient light as possible. The system creates the smallest commercially available biometric sample (represented by only 9 bytes) by taking 91 measurements of fingers, knuckles, hands, etc. The biometric samples, combined with other information (credit card and other card numbers and phone numbers) in a central database, can be used by other services.
After testing, the system will be prepared for serial production. IBM is already in talks with many governments and airports in a number of countries.
While Fastgate will be the first mass-produced biometric system, it is only the latest advance in the growing field of biometric security systems. Further progress will accelerate as these systems become more familiar to customers.
At one of the conferences on biometric systems, it was noted that in 1996 there were about 10,000 such systems in operation, with a total cost of over $17 million. According to Personal Identification magazine, sales of biometric systems were expected to grow by 45% in 1997, reaching $25 million, and by 1999 it would increase to $500 million. Based on its analysis, the Freedonia Group (USA) predicts an increase in sales of hardware and software for information security to $4 billion by 2001. This growth will occur in the face of fierce competition from companies producing photoelectronics, microelectronics, and video technology, and creating various systems based on these products.
я