Biometric identification in integrated security systems. Face recognition systems

Today, several types of such systems are presented on the market and they are fulfilled differently in terms of the complexity of the problem: from remote recognition in the crowd to accounting of working time in the office. Personal recognition solutions are available to customers on different platforms - this is a server architecture, mobile and embedded solutions and cloud services.

Modern systems operate on neural network algorithms of deep learning, therefore the accuracy of recognition is maximum even for low-quality images, they are resistant to head turns and have other advantages.

Example 1. Public Security

Ensuring security is a kind of starting point with which the implementation of biometric identification systems began. Systems of remote recognition of persons are used to ensure the safety of objects of mass finding people.

The most difficult task is the identification of a person in the crowd.

The so-called non-opoperative recognition, when a person does not interact with the system, does not look into the camera lens, turns away or tries to hide the face. For example, in transport and transfers, metro, major international events.

Cases

One of the most significant projects 2017 for our company has become the largest international exhibition EXPO-2017, which took place in Kazakhstan this summer. Specialized chambers were used in the system of remote biometric recognition of persons.

Selecting persons in the frame occurs in the chamber itself and only the image of the face is transmitted to the server, it unloads the channel and significantly reduces the cost of network infrastructure. Cameras controlled four input groups, in different parts of the complex. The system architecture was designed in such a way that the input groups worked separately or all together, while the correct operation of the system was provided with only 4 servers and 48 cameras.

With the help of video analytics online at large territorially distributed objects, they are looking for suspects, missing people, investigating incidents and incidents, carry out the analysis of passenger traffic.

At some airports, by the end of 2017, biometrics will begin to apply and to register passengers to fly. According to the TADVISER portal, the Smart Gate systems are also planning to introduce 12 European countries (Spain, France, Netherlands, Germany, Finland, Sweden, Estonia, Hungary, Greece, Italy, Romania).

And the next step should be the introduction of individual recognition systems for the passage of border and migration control. With state support, the introduction of identification on the face can be as commonplace as the framework of metal detectors in the future of the next three to five years.

Example 2. To know your buyer in the face

Business also makes a bet on biometric identification by face. First of all, this is retail.

Systems recognize the floor and age of buyers, the frequency and time of visiting outlets, accumulate statistics on each individual network store.

After that, for automatic mode, detailed reports are displayed in automatic mode both in the network and disaggregated by outlets. Based on these reports, it is convenient to make a "client portrait", plan effective marketing campaigns.

Unfortunately, we cannot disclose customers. Among them, the largest retailers and DIY (Do IT Youself) network, in the assortment of which there is an expensive tool and components.

How it works

Many are afraid of leaks of confidential information, but we emphasize that no personal data of the recognized people is stored in the archives. Moreover, not even an image is stored, but its biometric pattern, by which the image does not restore.

When repeated visits, the biometric pattern of the face "pulls up", so the system knows exactly who and how many times in the store. For the safety of personal data you can be calm.

For small shops, car dealerships, pharmacies Marketing analytics collection mechanism are implemented in the cloud recognition service. For small and medium-sized businesses, this option is preferable because it does not require costs for server equipment, hiring additional personnel, software update, and so on, first, a convenient tool for assessing the efficiency of outlets, and secondly, an excellent assistant To identify thieves. That is, one system performs several functions at once.

Example 3. Access Control and Access Control Systems

In addition to the above functions, the face recognition system is conveniently used as an alternative to proximity-cards in access control and access control systems.

They have a number of advantages: Provide high reliability of recognition, it is impossible to deceive, copy or steal the identifier, they are easily integrated with existing security equipment. You can even use already available surveillance cameras. Biometric identification systems are remotely and very quickly with fixing events in the archive.

Based on biometric scois, it is convenient to keep accounting working hours of employees, especially in large office centers.

Case

We have implemented such a system at a large Indian enterprise, which specializes in the field of logistics last year. The number of permanent employees is more than 600 people. At the same time, the company operates around the clock and practices a "floating" labor schedule. With the help of our remote biometric identification system, the Customer received a complete and reliable accounting of staff working hours, a tool of preventive security of an object and akud.

Example 4. Passing the fan to the stadium

At the time of buying a ticket at the box office, the face of each buyer is automatically photographed and loaded into the system. This is formed the base of the visitors of the match. If the purchase was via the Internet or mobile application, then authorization is possible remotely using Selfie. In the future, when a person comes to the stadium, the system recognizes it without any passports.

Identification of visitors to sports competitions has become mandatory according to Federal Law No. 284-FZ "On Amendments to Article 20 of the Federal Law" On Physical Culture and Sports in the Russian Federation "and Article 32.14 of the Code of Administrative Offenses of the Code of Administrative Offenses.

The stadium will be held exactly the one who bought a ticket, transfer the ticket to another person or go through a fake ticket is impossible. The remote recognition of persons in the stadium works on the same principle as on large territorial distributed transport objects: if a person is entered into lists of persons who are prohibited from the stadium, the system will not miss it.

Case

In March 2016, in the framework of the joint project of the vocoda and the Khanty-Mansiysk branch of PJSC Rostelecom, the remote recognition system was used to ensure the security of the World Cup in Biathlon, held in Khanty-Mansiysk. Since 2015, the same system has been successfully operating in the Multifunctional Sports Complex "Arena Omsk". It enters the six largest sports facilities in Russia, is the largest sports and entertainment object of Siberia and the Basis of the Hockey Club "Avangard".

Example 5. Internet banking and ATMs

Another niche in which the recognition of persons is settled is the banking sphere. Here the introduction of new technologies is intensively, since the financial sector is more interested in the reliability and preservation of personalized information.

Today, biometrics gradually begins, if you do not force out the usual and established "paper" documents, then to go to them. At the same time, the degree of protection during payments is significantly increasing: it is enough to see the camera of his smartphone to confirm the transaction. At the same time, the biometric data itself is not transmitted, respectively, it is impossible to intercept them.

The introduction of biometric identification technologies is directly related to the massive use of electronic services and devices, the development of Internet trading and the spread of plastic cards instead of cash.

With the advent of high-performance graphics processors (GPU) and ultra-compact hardware platforms on their base - such as NVIDIA Jetson - the recognition of persons began to be introduced into ATMs. Now only the owner of the card can be removed or to carry out the account operations, for example, through the Tinkoff Bank ATMs. And the PIN soon can retire.

Biometria is a combination of methods and devices for identifying a person, which are based on its unique physiological or behavioral characteristics.

This type of identification can be used to prevent prohibited access to the building, computers, ATMs, mobile phones, and so on.

Biometric properties is:

• fingerprints;
• geometry of the face;
• rainbow eye sheath;
• retina pattern;
• vote;
• handwriting;
• printing on the keyboard;
• veins pattern in hand, etc.

Science 2.0 identification of personality

Advantages of biometric identification

Biometric protection gives a greater effect compared, for example, using passwords, smart cards, PIN codes, tokens or public key infrastructure technology. This is due to the possibility of biometrics to identify non-device, but a person.

Conventional protection methods are fraught with loss or theft of information that becomes open to illegal users. An exceptional biometric identifier, for example, fingerprints, is a key not subject to loss.

Classification of biometry methods

By type of information used, biometric identification is divided into:

• Static methods based on unique properties, data man from birth and integral from him. Physiological indicators (palm geometry or papillar finger pattern) are unchanged for man '
• Dynamic methods based on behavioral (i.e. dynamic) characteristic of the personality. These features are characteristic of subconscious movements when playing any actions (speeches, signatures, keyboard dynamics). Such behavioral characteristics are influenced by managed and not very controlled mental factors. Because of their changeability, biometric samples should be updated when used.

Personality identification methods for biometric parameters

This recognition method is the most common. It uses the unique of the papillary patterns of the fingers for each person. A special scanner get an image of a fingerprint. It is transformed into digital code and is mapped to the template entered earlier.

The process of identifying more than a few seconds. A certain disadvantage of the deterrent development of this method is to prejudice some people who do not want to leave data on their fingerprints. The counterproof of the device developers is that the information about the papillary pattern is not stored, and only a short identification code is stored, built on the fingerprint and does not allow to recreate the pattern for comparison. The advantage of the method is ease of use, reliability and convenience.

Hand shaped identification

This static method is based on measuring hand brush shape. It is also a unique biometric human parameter. A special device allows you to get a three-dimensional view of the brush. As a result, measurements are obtained to create a unique digital code identifying a person.

This method on its technology and accuracy is comparable to the method of identifying a fingerprint, although the device itself has a lot of space for the implementation of the method. Extremely small probability of presence of two identical hands with the same geometry, although hands change with age.

Today, the identification of the geometry of the hand is used in the legislative bodies, hospitals, international airports, etc.

Rainbow shell authentication

The basis of this method is the exclusivity of the pattern on the iris. To perform it, you need a camera to get an eye image with sufficient resolution, and special software to highlight the image from the resulting image on the rainbow shell. It creates a digital code that serves to identify a person.

The advantage of scanners is that the person does not require focus on the target, since the sample of the spots of the iris is concentrated on the surface of the eye. Scanning is possible at a distance of less than 1 m. It is convenient for use, for example, in ATMs.

Retid Identification

The retina is scanned using low-intensity infrared light, which is directed to the blood vessels of the rear wall of the eye through the pupil. The retina scanners are widespread in access systems to secret objects, since they almost never have incorrect access permission. Errors can be explained by the deviation of the head from the reference position and incorrect focusing of the view on the light source.

Even twins differ the capillary pattern of the retina. That is why this method can be successfully used to identify the personality.

The disadvantage of such systems can include a psychological factor: not every person can look into a dark hole in which something shines in the eye. In addition, these systems are sensitive to incorrect orientation of the retina, so you need to closely monitor the position of the eye towards the hole.

The form Persons as an object for identification

This static identification method is to create a two-or three-dimensional image of a person's face. The camera and specialized software in the image of the face emphasize the contours of the eyes, lips, eyebrows, nose, etc. Then they calculate the distances between these elements and other parameters. According to this information, an image is created that for comparison is transformed into a digital form.

This method refers to the most dynamically developing areas in the biometry industry. His attractiveness is based on the fact that special expensive equipment is required. Enough personal computer and camcorder. In addition, there is no physical contact with devices. No need to touch anything, or stop, specifically waiting for the system's triggering.

Manuscript Recognition

The basis of the identification of the handwriting is the uniqueness and stability of this factor for each person. Characteristics are measured, they are translated into digital form and exposed to computer processing. That is, for comparison, not a letter is selected as a product, but the process itself.

Two data processing methods are common: normal comparison with sample and dynamic verification. The first is unreliable, because the signature is not always the same. This method leads to a large percentage of errors. Dynamic verification consists in more complex calculations. This method in real time registers the parameters of the signature process: the speed of the hand in different parts, the pressure force and the duration of different signatures. This eliminates the fake, as it is impossible to copy the signature hand movements in accuracy.

Recognition on the keyboard handwriting

This method is generally similar to the above, but the signature in it is replaced with a certain code word, and only a normal keyboard is needed from the equipment. The main identification characteristic is the dynamics of the keyboard set of the code word.

According to modern research, the keyboard handwriting has a certain stability, due to which one can uniquely identify the personality. The source data is the time between clicking the keys and their retention. And the time between the press shows the pace of work, and hold is the style of work, that is, a smooth press or a sharp blow.

At first, at the filter stage, data on "service" keys - functional, cursor control, etc. are deleted.

The following user characteristics are allocated:

• the number of errors during the set;
• time between keystrokes;
• set speed.
• time to hold keys;
• arrhydrichy when set .

Voting recognition

The biometric method of identifying the voice is convenient for use. The reasons for its implementation are the wide distribution of telephone networks and embedding microphones in computers. The disadvantages can be considered the factors affecting the recognition: interference in microphones surrounding noises, errors in the process of pronouncing, the different emotional state of a person when identifying, etc.

The main thing in building authentication devices vote is the choice of parameters, best describing the individuality of the voice. These signal parameters are called individuality features. Such signs, except for the peculiarities of the voice, should have other properties. For example, they should be easily measured, and do little depend on noise and interference. In addition, they must have stability in time and resist imitation.

Systems with the use of a combined voice analysis method with facial scan are developed. It turns out that the Mimica of the speaker distinguishes only it and will be different in the same person's spent the same words.

Thermographic observation of facial arteries and veins

The identification of a person in person is very simplified, if you go to the infrared range of light waves. The thermography of the identifiable person identifies the uniqueness of the location on the face of the arteries supplying the skin with blood. The backlight question for these biometric devices does not exist, since they perceive only the temperature differences of the face and they do not need the light. The recognition efficiency does not depend on overheating or supercooling the person, natural aging of the personality, plastic operations, as they do not change the inner position of the vessels.

The method of facial thermography can be distinguished by twins, the facial blood vessels of which are greatly different.

This identification method uses a specialized infrared range camcorder.

Identification on hand veins

The biometric market contains devices that are built on the analysis of the individual veins of the veins in the hands. Intect the vein drawer, located on the back of the brush compressed in the hand fist. Observation of the drawing of veins carries out a television camera with infrared illumination. When entering the image, its binarization is produced, emitting veins. Such equipment produces the only English company Vinchek.

Perspectives of biometry

The detection of fingerprints remains the dominant way of identifying the personality. For this, there are two main reasons:

• in many countries, the transition to passports with biometric data began;
• development of updated models of fingerprint scanners for use in small devices (cell phones, pocket PCs, laptops).

A significant extension can be expected in the Signature Identification Sector in connection with the wide implementation of the digital electronic signature. Voice recognition can also gain momentum due to the implementation of large projects in the construction of intelligent buildings.

The main forecasts are reduced to the fact that the introduction of biometric safety devices in the near future will acquire avalanche. Fighting global terrorism will require the practical use of any achievements in this area. Thanks to the intensive development of multimedia and digital technologies and their further redundancy will allow to develop and implement fundamentally new identification systems.

Certain biometric technologies are currently undergoing the development stage and some of them are recognized as promising:

1. face thermogram in the infrared range;
2. dNA characteristics;
3. spectroscopy of the skin of the fingers;
4. palm prints;
5. the shape of the oh sink;
6. manifold parameters;
7. individual smells of man;
8. skin salinity level.

These methods of biometric identification today can be considered formed. Perhaps soon they will move from scientific research to commercial technologies.

Recently, there are many articles on Huglovsky identification systems on persons. To be honest, then from many of them and brings journalism and to put it mildly incompetence. And I wanted to write a good article on biometrics, it's not in the first! There are a couple of good articles on biometrics on Habré - but they are short enough and incomplete. Here I will try briefly outline the general principles of biometric identification and modern achievements of humanity in this matter. Including in identification by persons.

The article has a continuation that, in essence, is her bull.

As the basis for the article will be used joint with a colleague publication in the journal (BDI, 2009), recycled under modern realities. Colleagues while Habré is not, but it supported the publication of the recycled article. At the time of publication, the article was a brief overview of the modern market of biometric technologies, which we conducted for themselves before moving out their product. Estimated judgments about the applicability put forward in the second part of the article are based on the opinions of people who used and implemented products, as well as in the opinions of people engaged in the production of biometric systems in Russia and Europe.

general information

Let's start with azov. In 95% of biometric cases, in its essence, it is mathematical statistics. And Matstat is an accurate science, algorithms from which are used everywhere: in radar and in Bayesian systems. As the two main characteristics of any biometric system, you can take errors of the first and second sort). In the theory of radar, they are usually called "false alarm" or "passing target", and in biometrics, the most established concepts - FAR (False Acceptance Rate) and FRR (False Rejection Rate). The first number characterizes the likelihood of a false coincidence of biometric characteristics of two people. The second is the likelihood of failure to access a person who has tolerance. The system is better than the less FRR value with the same FAR values. Sometimes the comparative characteristic of EER is also used, which determines the point in which FRR and FAR graphics intersect. But it is far from always representative. More details you can see, for example.
You can note the following: if the system characteristics are not given FAR and FRR on open biometric databases - that the manufacturers did not declare its characteristics, this system is most likely incapacitated or strongly weaker competitors.
But not only FAR and FRR define the quality of the biometric system. If it were only so, then the leading technology would be recognition of people on DNA, for which FAR and FRR tend to zero. But it is obvious that this technology is not applicable at the present stage of human development! We have developed several empirical characteristics that allow you to evaluate the quality of the system. "Fake resistance" is an empirical characteristic that generalizes how easily the biometric identifier is easily deceived. "Environmental resistance" is a characteristic, empirically evaluating the stability of the system under various external conditions, such as changing the illumination or room temperature. "Easy use" shows how difficult it is to use the biometric scanner, whether the identification "on the go" is possible. An important characteristic is the "speed of work", and the "cost system". Do not forget that the biometric characteristics of a person can change over time, so if it is unstable, it is a significant minus.
The abundance of biometric methods is striking. The main methods using the static biometric characteristics of the person are identifying the papillary pattern on the fingers, the iris, the geometry of the face, the retina, the drawing of the veins of the arm, the geometry of the hands. There is also a family of methods using dynamic characteristics: voting identification, handwritten dynamics, heart rhythm, gait. Below is the distribution of the biometric market a couple of years ago. In each second source, these data ranges 15-20 percent, so this is just an estimate. Here, under the concept of "hand geometry", two different methods are hidden about which will be described below.

In the article, we will consider only those characteristics that are applicable in access control and access control systems (SCS) or in the tasks close to them. By virtue of its superiority, it is primarily static characteristics. From the dynamic characteristics to today, only recognition by voice has at least some statistical significance (comparable to one hundred static algorithms FAR ~ 0.1%, FRR ~ 6%), but only in ideal conditions.
To feel the probabilities of FAR and FRR, you can evaluate how often false coincidences will arise if you set the identification system on the passing organization with the number of human personnel. The likelihood of a false coincidence of the fingerprint received by the scanner for a database from n fingerprint is equal to FAR ∙ N. And every day, about N man passes through the access control point. Then the probability of error for the working day FAR ∙ (N ∙ N). Of course, depending on the objectives of the identification system, the probability of an error per unit of time can vary greatly, but if you accept one error permissible during the working day, then:
(1)
Then we obtain that the stable operation of the identification system at Far \u003d 0.1% \u003d 0.001 is possible with the number of personnel N≈30.

Biometric scanners

To date, the concept of "biometric algorithm" and "biometric scanner" is not necessarily interrelated. The company can produce these elements by solitary, and may be together. The largest differentiation of manufacturers of scanners and software manufacturers is achieved in the biometric market of the papillary pattern of fingers. The smallest of 3d face scanners market. In fact, the level of differentiation in many ways displays the development and saturation of the market. The more choice - the more themes are worked out and brought to perfection. Various scanners have a different set of abilities. Basically, this set of tests for checking forged the biometry object or not. For finger scanners, this may be a relief check or temperature testing, for the eye scanners, this may be the checker of the accommodation of the pupil, for the face scanners - the movement of the face.
Scanners are very strongly affected by the received Far and FRR statistics. In some cases, these figures may vary in tens of times, especially in real conditions. Typically, the characteristics of the algorithm are given for a certain "ideal" base, or just for a well-appropriate, where the unaware and blurred frames are thrown. Only a few algorithms honestly indicate the base and the full issuance of Far / FRR on it.

And now in more detail about each of the technologies

Fingerprints

Dactyloscopy (fingerprint recognition) - the most developed biometric identification method of identification today. The catalyst for the development of the method was its widespread use in criminalistics of the 20th century.
Each person has a unique papillar fingerprint pattern, thanks to which identification is possible. Usually algorithms use characteristic points on fingerprints: end of the pattern line, line branching, single points. Additionally, information on the morphological structure of the fingerprint is attracted: the relative position of the closed lines of the papillary pattern, "arched" and spiral lines. Features of the papillary pattern are transformed into a unique code that retains the infinction image inflammativity. And it is the "fingerprint codes" it is stored in the database used to search and compare. The image of the fingerprint image in the code and its identification usually does not exceed 1C, depending on the size of the database. The time spent on the dealing hand is not taken into account.

As a source of data on FAR and FRR, Verifinger SDK statistics were used, obtained using the fingerprint scanner DP U.are.u. Over the past 5-10 years, the characteristics of the recognition of the finger did not step much forward, so that the figures showed well show the average value of modern algorithms. The Verifinger algorithm itself won the International Fingerprint Verification Competition international competition, where the finger recognition algorithms compete.

The characteristic FAR value for the fingerprint recognition method is 0.001%.
From formula (1) we obtain that the stable operation of the identification system at Far \u003d 0.001% is possible with the number of personnel N≈300.

The advantages of the method. High reliability - method statistical indicators Better indicators of identification methods for person, voice, painting. Low cost of devices scanning fingerprint image. A simple simple imprint scanning procedure.
Disadvantages: a papillar fingerprint pattern is very easily damaged by small scratches, cuts. People who used scanners at enterprises with the number of personnel about several hundred people declare a high degree of scanning. Many of the scanners are inadequately belong to dry skin and do not miss old people. When communicating at the last exhibition MIPS, the head of the security service of a large chemical enterprise told that their attempt to introduce finger scanners in the enterprise (the scanners of various systems were tried) failed - the minimum impact of chemical reagents on the fingers of the staff caused the scanner security systems failed - scanners declared the fingers of the fake. There is also insufficient protectedness from the falsification of an imprint image, partly caused by the wide distribution of the method. Of course, not all scanners can be fooled by methods from destroyers legends, but still. For some people with "inappropriate" fingers (body temperature, humidity), the probability of failure can be achieved 100%. The number of such people varies from the interest of interest for expensive scanners up to ten percent for inexpensive.
Of course, it is worth noting that a large number of disadvantages are caused by a wide prevalence of the system, but these shortcomings have a place to be and manifest themselves very often.

Situation in the market

At the moment, more than half of the biometric market occupy more than half of the biometric market. Many Russian and foreign companies are engaged in the production of access control systems based on the method of dactyloscopic identification. Due to the fact that this direction is one of the most pressing, it got the most disseminated and is today the most developed. Fingerprint scanners have passed a really long way to improve. Modern systems are equipped with various sensors (temperatures, pressing force, etc.), which increase the degree of protection against fakes. Every day the system is becoming increasingly comfortable and compact. In fact, the developers have already achieved a certain limit in this area, and develop a method further. In addition, most companies produce ready-made systems that are equipped with everything necessary, including software. Integrators in this area simply do not need to collect the system independently, as it is unprofitable and will take more time and forces than buying ready and already inexpensive the system, the more the choice will be really wide.
Among foreign companies involved in fingerprint recognition systems, SecuGen can be noted (USB scanners for PC, scanners that can be installed on businesses or embed in locks, SDK and software communication with a computer); Bayometric Inc. (Fingerprint Scanners, TAA / Access Control Systems, FingerPrint SDKS, Embedded Fingerprint Modules); DigitalPersona, Inc. (USB Scanners, SDK). Russia in this area employs companies: Biolink (Dactyloscopic scanners, biometric access control devices, software); Sonda (Dactyloscopic Scanners, Biometric Access Control Devices, SDK); Smartlock (Dactyloscopic Scanners and Modules), etc.

Iris

The Rainbow Eye Shell is a unique characteristic of a person. The image of the iris is formed in the eighth month of intrauterine development, finally stabilized at the age of about two years and practically does not change during life, except as a result of severe injuries or sharp pathologies. The method is one of the most accurate among biometric methods.
The Identification System for the Rainbow Shell is logically divided into two parts: an image capture device, its primary processing and transmission of a computer and a calculator, which makes a comparison of images with images in the database that transmits the command to the executive device.
The time of primary image processing in modern systems is approximately 300-500ms, the rate of comparison of the resulting image with the base has a level of 500,000-150000 comparisons per second on a regular PC. Such a reference speed does not impose restrictions on the use of the method in large organizations when used in access systems. When using specialized computers and algorithms for search optimization, it becomes even possible to identify a person among residents of the whole country.
Immediately I can answer that I am somewhat biased and positively treat this method, since it was on this Niva that we launched our startup. A small self-based paragraph will be devoted at the end.

Statistical characteristics of the method

Features FAR and FRR for the iris are the best in the class of modern biometric systems (except, perhaps the method of recognition of the retina's retina). The article presents the characteristics of the recognition library of the Rainbow Shell of our algorithm - EYER SDK, which correspond to the Verieye algorithm verified on the same bases. Casia bases obtained by their scanner were used.

The characteristic value of FAR - 0.00001%.
According to formula (1) N≈3000, the number of personnel of the organization, in which the identification of the employee occurs fairly stable.
It is worth noting an important feature that distinguishes the recognition system on a rainbow shell from other systems. In the case of using the permission cameras from 1.3MM, you can capture two eyes on one frame. Since the probabilities FAR and FRR are statistically independent probabilities, then when recognizing in two eyes, the FAR value will be approximately equal to the square of the Far value for one eye. For example, for Far 0.001%, when using two eyes, the probability of false tolerance will be equal to 10-8%, with FRR is only two times higher than the corresponding FRR value for one eye at FAR \u003d 0.001%.

Advantages and disadvantages of the method

The advantages of the method. Statistical reliability of the algorithm. The capture of the image of the iris can be made at a distance from several centimeters to several meters, while the physical contact of a person does not occur with the device. The rainbow shell is protected from damage - which means it will not change in time. Also, it is possible to use a high number of methods that protect against fake.
Disadvantages of the method. The price of a system based on a rainbow shell above the price of a system based on finger recognition or face recognition. Low availability of ready-made solutions. Any integrator that will come to the Russian market today and will say "Give me a ready-made system" - most likely breaks down. For the most purpose, expensive turnkey systems are sold, installed by large companies, such as Iridian or LG.

Situation in the market

At the moment, the proportion of the Identification technologies on the Rainbow Eye of the eye on the global biometric market is at different estimates from 6 to 9 percent (while recognition technologies for fingerprints occupy over half of the market). It should be noted that from the very beginning of the development of this method, its strengthening on the market slowed down the high cost of equipment and components necessary to assemble the identification system. However, with the development of digital technologies, the cost of a separate system has decreased.
The leader in the development of software in this area is the company Iridian Technologies.
The entry into the market to a large amount of manufacturer was limited to the technical complexity of scanners and, as a result, their high cost, as well as a high price of Iridian's monopolized position on the market. These factors allowed to develop in the field of recognition of a rainbow shell only to large companies, most likely to produce some components suitable for identification system (high-resolution optics, miniature chambers with infrared illumination, etc.). Examples of such companies can be LG Electronics, Panasonic, OKI. They concluded an agreement with Iridian Technologies, and as a result of collaboration, the following identification systems appeared: IRIS Access 2200, BM-ET500, OKI IRISPASS. In the future, improved systems of systems arose, due to the technical capabilities of these companies independently evolve in this area. It should be said that the above companies also developed their own software, but as a result, Iridian Technologies software preferred in the finished system.
The products of foreign companies dominate in the Russian market. Although this can be bought with difficulty. For a long time, the company Papilon assured everyone that they have recognition on a rainbow shell. But even representatives of Rosatom are their direct purchase, for whom they did the system tell that it is not true. At some point, another Russian firm was manifested, which made the scanners of the iris. Now I will not remember the name. Algorithm they have purchased someone, perhaps at the same Verieye. The scanner itself was a system of 10-15 years ago, by no means contactless.
In the last year, a pair of new manufacturers was published on the global market due to the expiration of the primary patent for human recognition in the eyes. The greatest confidence of them, in my opinion, deserves AOPTIX. At least their previews and documentation causes suspicion. The second company is SRI International. Even at first glance, a person engaged in the iris recognition systems of their rollers seem very false. Although I will not be surprised if in reality they can do something. And the same system does not show data on FAR and FRR, as well as, apparently, is not protected from fakes.

Face recognition

There are many methods of recognition on the geometry of the face. All of them are based on the fact that the features of the face and the shape of the skull of each person are individual. This area of \u200b\u200bbiometrics seems attractive to many, because we learn each other primarily in the face. This area is divided into two directions: 2-D recognition and 3-D recognition. Each of them has advantages and disadvantages, but much also depends on the application and claims submitted to a specific algorithm.
In brief, I will tell you about 2-D and go to one of the most interesting methods for today - 3-d.

2-D face recognition

2-D face recognition is one of the most statistically inefficient methods of biometrics. It appeared for a long time and was applied, mainly in forensic, which contributed to its development. In consequence, computer interpretations of the method appeared, as a result of which he became more reliable, but certainly inferior and every year more and more inferior to other biometric identification methods. Currently, due to bad statistical indicators, it is used in multimodal or, as it is also called crossbiometrics, or on social networks.

Statistical characteristics of the method

For FAR and FRR used data for Verilook algorithms. Again, for modern algorithms, it has very ordinary characteristics. Sometimes algorithms with FRR 0.1% are planetized with a similar Far, but the bases for which they are obtained are also very dubious (carved background, the same facial expression, the same hairstyle, lighting).

The characteristic value of FAR - 0.1%.
From formula (1) we obtain N≈30 - the number of personnel of the organization, in which the employee identification occurs fairly stable.
As can be seen, the statistical indicators of the method are quite modest: this levels the advantage of the method that you can carry out the hidden shooting of persons in crowded places. It's funny to watch how a regular project is funded for the discovery of criminals through camcorders installed in crowded a couple of times a year. Over the past ten years, the statistical characteristics of the algorithm have not improved, and the number of such projects has grown. Although it is worth noting that to maintain a person in a crowd through many cameras, the algorithm is quite suitable.

Advantages and disadvantages of the method

The advantages of the method. With 2-D recognition, in contrast to most biometric methods, no expensive equipment is required. With appropriate equipment, the possibility of recognition at considerable distances from the camera.
Disadvantages. Low statistical accuracy. Requirements for lighting (for example, it is not possible to register the faces of people from the streets of people on a sunny day). For many algorithms, the unacceptability of any external interference, such as, for example, glasses, beard, some elements hairstyles. Be sure to frontal face image, with very small deviations. Many algorithms do not take into account possible changes in the facial features, that is, the expression should be neutral.

3-D face recognition

The implementation of this method is a rather difficult task. Despite this, there are currently many methods of 3-D face recognition. Methods cannot be compared with each other, as they use different scanners and bases. Not all of them are issued Far and FRR, absolutely different approaches are used.
Transient from 2-D to 3-D method is a method that implements the accumulation of information about the face. This method has the best characteristics than 2D method, but as well as it uses only one chamber. When entering the subject to the database, the subject turns the head and the algorithm connects the image together, creating a 3d template. And when recognized, several frames of the video stream are used. This method is more likely to experimental and implementing for SCS systems I have never seen.
The most classical method is the method of projecting the template. It is that the grid is projected on the object (face). Next, the camera takes pictures with a speed of dozens per second, and the resulting images are processed by a special program. The beam falling to the curved surface is bent - the larger the curvature of the surface, the stronger the bending of the beam. Initially, the source of the visible light supplied through the "blinds" was used. Then the visible light was replaced with infrared, which has a number of benefits. Typically, at the first stage of processing, images are discarded on which the faces are not visible at all or there are extraneous objects that interfere with identification. A 3-D model of a person on which unnecessary interference (hairstyle, beard, mustache and glasses) is distinguished and removed. The model analysis is then performed - anthropometric features are allocated, which are eventually recorded in a unique code that is entered into the database. The capture time and image processing time is 1-2 seconds for the best models.
The 3-D method of recognition on the image obtained from multiple cameras is also gaining popularity. An example of this can be a company Vocord with its 3D scanner. This method gives the accuracy of positioning, according to the assurances of developers, above the method of projecting the template. But, until I see Far and FRR at least at their own database - I will not believe !!! But it has already been developed for 3 years, and there are no progress at exhibitions yet.

Statistical methods of method

Full data on FRR and FAR for the algorithms of this class on the sites of manufacturers are open not shown. But for the best models of Bioscript (3D Enrolcam, 3D FastPass) operating according to the design of the template at FAR \u003d 0.0047% FRR is 0.103%.
It is believed that the statistical reliability of the method is comparable with the reliability of the identification method for fingerprints.

Advantages and disadvantages of the method

The advantages of the method. No need to contact the scanning device. Low sensitivity to external factors, both on the person itself (appearance of glasses, beards, change hairstyles), and in its environment (illumination, head turn). A high level of reliability compared to the identification method for fingerprints.
Disadvantages of the method. Highness equipment. Complex commercials exceeded even the price even the rainbow shell scanners. Changes in facial and interference on the face worsen the statistical reliability of the method. The method is not yet well designed, especially in comparison with the long-applied dactyloscopy, which makes it difficult to wide use.

Situation in the market

The recognition of the geometry of the face is classified as "three large biometrics" along with recognition of fingerprints and a rainbow shell. It must be said that this method is quite common, and it is still preferred before recognizing the eye iris. The proportion of recognition technologies for the geometry of the face in the total volume of the global biometric market can be estimated within 13-18 percent. In Russia, there is also greater interest in this technology than, for example, to identifying the iris. As mentioned earlier, there are many 3-D recognition algorithms. In most, the company prefers to develop ready-made systems, including scanners, servers and software. However, there are those who offer the consumer only SDK. To date, the following companies engaged in the development of this technology can be noted: Geometrix, Inc. (3D Face Scanners, By), Genex Technologies (3D Face Scanners, Software) in the USA, Cognitec Systems GmbH (SDK, Special Computers, 2D Camera) in Germany, Bioscrypt (3D Face Scanners) - a subsidiary of the American company L- 1 Identity Solutions.
In Russia, the ARTEC Group (3D Face Scanners) is operating in this direction - the company, the head office of which is located in California, and the development and production are conducted in Moscow. Also, several Russian companies own the technologies of 2D face recognition - Vocord, ITV, etc.
In the field of recognition of 2D faces the main subject of development is software, because Conventional cameras are perfectly coping with image image capture. The solution to the task of recognition on the image of a person to some extent went into a deadlock - for several years there has been practically no improvement in the statistical indicators of algorithms. In this area there is a planned "work on errors".
3D Face Recognition is now a much more attractive area for developers. It employs many teams and regularly audible about new discoveries. Many works are in a state "That's and let's release." But so far only old offers on the market, in recent years the choice has not changed.
One of the interesting moments over which I sometimes conceive and for which the Habr may answer: And the accuracy of Kinect is enough to create such a system? Projects for pulling out 3D models of a person through it quite themselves.

Hand recognition

This is a new technology in the field of biometrics, its widespread use began only 5-10 years ago. The infrared chamber takes pictures of an external or inside hand. The veins pattern is formed due to the fact that blood hemoglobin absorbs IR radiation. As a result, the degree of reflection decreases, and veins are visible on the chamber in the form of black lines. A special program based on data creates a digital convolution. No contact person with scanning device is required.
The technology is comparable by reliability with recognition on the iris, in something surpassing it, and something yielding.
The FRR and FAR value are given for the Palm Vein scanner. According to the developer's data at FAR 0.0008% FRR is 0.01%. A more accurate chart for several values \u200b\u200bdoes not produce a single firm.

Advantages and disadvantages of the method

The advantages of the method. No need to contact the scanning device. High reliability - method statistical indicators are comparable with the readings of the iris. Host Specifications: Unlike all of the above, this feature is very difficult to get from a person on the street, for example, photographing it.
Disadvantages of the method. Invalid illumination of the scanner of the sunny rays and the rays of halogen lamps. Some age diseases, such as arthritis, deteriorate greatly and FRR. The method is less studied in comparison with other static biometry methods.

Situation in the market

The recognition of the hand drawing of the arm is quite a new technology, and in connection with this, its share in the global market is small and amounts to about 3%. However, a greater interest is manifested for this method. The fact is that, being quite accurate, this method does not require such expensive equipment, such as the recognition methods on the geometry of the face or iris. Now many companies are developing in this field. For example, by order of the British company, TDSI was developed for a biometric reader of the Palmvein Palmvein, represented by Fujitsu. The scanner itself was developed by Fujitsu primarily to combat financial fraud in Japan.
Also in the field of identification, the following VEID PTE companies work. Ltd. (Scanner, Software), Hitachi Veinid (Scanners)
In Russia companies engaged in this technology, I do not know.

Retina

Until recently, it was believed that the most reliable method of biometric identification and identification authentication is a method based on scanning of the retina. It contains the best features of the Identification on the rainbow shell and on the veins of the hand. The scanner reads the pattern of capillaries on the surface of the retina. The retina has a fixed structure, continued in time, except as a result of a disease, for example, cataracts.
The retina scanning occurs using infrared light of low intensity directed through the pupil to the blood vessels on the rear wall of the eye. The retina scanners of the eye were widely distributed in access control systems to highly secret objects, since they have one of the lowest percentages of refreshing the registered users and is practically no erroneous access permission.
Unfortunately, a number of difficulties occur when using this method of biometry. The scanner here is a very complex optical system, and a person should not move a considerable time while the system is presented, which causes unpleasant sensations.
According to Eyedentifying for the ICAM2001 scanner at Far \u003d 0.001%, the FRR value is 0.4%.

Advantages and disadvantages of the method

Benefits. High level of statistical reliability. Due to the low prevalence of systems, the likelihood of the development of the method of their "deception".
Disadvantages. Complex when using a high-time system with high processing time. High cost system. Lack of a wide market offer and as a result, the lack of intensity of the development of the method.

Geometry of the hands

This method is quite common for another 10 years ago and the occurrence of criminals in recent years has declined. It is based on the receipt of the geometric characteristics of the hands: the lengths of the fingers, the width of the palm of the I.T.D. This method, like the retina, is dying, and since he has much lower characteristics, they will not even introduce it to a complete description.
It is sometimes considered that in the recognition systems on veins use geometric methods of recognition. But on sale, we have never seen such a clearly stated declared. And besides, often when recognizing on the veins, a snapshot of only palms is taken, whereas when the geometry is recognized, a picture of the fingers is taken.

A little self-based

At one time we have developed a good recognition algorithm in the eyes. But at that time, such a high-tech thing in this country was not needed, and in Burzhuystan (where we were invited after the first article) - I did not want to go. But suddenly, the year a year and a half there were investors who wanted to build a "biometric portal" - a system that would eat 2 eyes and used the color component of the iris (for which the investor had a global patent). Actually now we are doing this. But this is not an article about self-director, it is a brief lyric retreat. If anyone is interested, there is a bit of information, and someday in the future, when we go to the market (or do not leave) I will write here a few words about the biometric project of the biometric project in Russia.

conclusions

Even in the class of static biometric systems there is a large selection of systems. What kind of choose them? It all depends on security requirements. The most statistically reliable and resistant to fake access systems are the tolerance systems for the Rainbow shell and on the veins of the hands. For the first of them there is a wider market of proposals. But this is not the limit. Biometric identification systems can be combined by reaching astronomical accuracy. The most cheap and simple to use, but possessing good statistics, are tolerances to the tolerances. The admission of 2D to the person is convenient and daughtered, but has a limited area of \u200b\u200bapplications due to bad statistical indicators.
Consider the characteristics that will have each of the systems: resistance to fake, environmental resistance, ease of use, cost, speed, biometric sign stability in time. Select estimates from 1 to 10 in each column. The closer the score to 10, the better the system in this regard. Principles of selection of estimates were described at the very beginning of the article.


Also consider the Far and FRR ratio for these systems. This ratio determines the efficiency of the system and the breadth of its use.


It is worth remembering that for a rainbow shell, you can increase the accuracy of the system almost quadratically, without loss for time, if you complicate the system by making it two eyes. For a dactyloscopic method - by combining several fingers, and recognizing on the veins, by combining two hands, but such an improvement is possible only with an increase in time spent when working with a person.
Generalizing the results for methods, we can say that for medium and large objects, as well as for objects with a maximum safety requirement, you should use a rainbow shell as biometric access and, possibly recognition on the veins of the hands. For objects with the number of personnel, up to several hundred people will be optimal to access fingerprints. Recognition Systems 2D Face Image Specific. They may be required in cases where recognition requires the lack of physical contact, but it is impossible to put the control system on the iris. For example, if you need to identify a person without its participation, a hidden camera, or an outdoor detection chamber, but maybe it is only with a small number of subjects in the database and a small stream of people removed by the camera.

Young technique for a note

For some manufacturers, for example, Neurotechnology, the site has a demo version of the biometry methods that they produce, so it is possible to connect them and play. For those who decide to dig in the problem of more serious, I can advise the only book that I saw in Russian - "Biometry Guide" R.M. Boll, j.x. Connel, Sh. Pankanti. There are many algorithms and their mathematical models. Not everything is fully and not all corresponds to modernity, but the base is not bad and comprehensive.

P.S.

In this opus, I did not go into the problem of authentication, but only affected the identification. In principle, all conclusions on authentication are asked by the Far / FRR feature and the possibilities of fake.

In recent years, biometrics have increasingly penetrate our lives. The leading countries of the world have already been introduced into circulation or in the near future plan to introduce electronic passports containing information on the biometric characteristics of their owner; Many office centers have introduced biometric sensors to corporate access control systems; Laptops have long been equipped with funds of biometric user authentication; In armament of security services, modern means of identifying any wanted criminal in the crowd of people appear

Andrei Khrulev
Head of biometric department
and integrated security systems
Group of Companies "Technoserv", Ph.D.

Examples of using biometric systems are becoming more and more. Biometric success is easy to explain. Traditional identification means of identification based on the principles "I am what I have" (identification cards, tokens, certifying documents) and "I know what I know" (passwords, PIN-codes) are not perfect. The card is easy to lose, the password can be forgotten, besides, any attacker can take advantage of them, and no system can distinguish you from a passing person.

In addition, traditional identification tools are absolutely useless, if we are talking about the tasks of the hidden identification of the personality, and more such tasks are becoming more and more:

• recognize a criminal in the crowd;
• check if the passport really places his owner;
• find out whether the person is wanted;
• to find out whether the person was not previously involved in financial fraud with loans;
• reveal potentially dangerous fans at the entrance to the stadium, etc.

All these tasks can only be solved using the means of biometric identification of the personality based on the principle "I am what I am." This principle allows the information system to identify directly by a person, and not the items that it places, or the information reporting.

The uniqueness of the biometry of the face

Among the variety of biometric characteristics of the person used to identify the personality, it is especially worth noting the image of the face. Face biometrics are unique in that it does not require the creation of specialized sensors to obtain an image - an image of a person can be obtained from a conventional video surveillance system. Moreover, the person's photograph is present almost on any identity document, which means that the introduction of this technology is not involved in practice with a variety of regulatory problems and the difficulties of social perception of technology.

It is also worth noting that the image of a person can be obtained implicitly for the person himself, which means that the biometry of the person is optimally suitable for building monitoring and hidden identification systems.

Any face recognition system is a typical pattern recognition system, the task of which is reduced to the formation of a certain set of features, the so-called biometric pattern, according to the mathematical model system. It is this model that is the key know-how of any biometric system, and the effectiveness of the recognition of the face directly depends on factors such as the stability of the biometric pattern to various kinds of interference, distortions in the source photo or video image.

The effectiveness of the recognition of the face directly depends on factors such as the stability of the biometric template to various kinds of interference, distortions in the original photo or video image

Despite the huge variety of face recognition systems, presented in the Russian market and in the world, in many of them are used by the same biometric engines - the actual software implementation of methods for building and comparing mathematical models of the face. In Russia, biometric engines such as Cognitec (Development of Cognitec Systems GmbH, Germany), "Cascade Stream" (Development of Technoserv, Russia), FRS SDK (Development of ASIA Software, Kazakhstan), Faceit (Development of the Company Company L1 Identity Solutions, USA).

As a rule, the face recognition in any biometric engine is performed in several stages: face detection, quality assessment, constructing template, comparison and decision making.

Step 1: Face Detection

At this stage, the system automatically highlights (detects) in the flow of video frames or in the photographs of the face of people, and the range of angles and the scale of individuals can vary significantly, which is extremely important for building security systems. It is absolutely optional that all selected faces will be recognized (as a rule, it is impossible), but to detect the maximum number of persons in the stream and, if necessary, to place them in the archive is extremely useful (Fig. 1).

Face detection is one of the key recognition steps, since the passage of the person by the detector automatically means the impossibility of further identification. The quality of the detector work is made to characterize the probability of detection of the face P0. For modern biometric systems operating under the flow of people, the probability of a person detection is from 95 to 99% and depends on the conditions of video recording (illumination, camera resolution, etc.).

One of the most promising trends in the development of the biometric market is the appearance of intelligent digital video cameras that implement the face detection function based on the built-in logic (Fig. 2). Intelligent video cameras make it possible to receive not only high-quality video streams, but also related metadata containing information about those found.

This approach can significantly reduce the load on the hardware power of the recognition system, which, in turn, reduces the final value of biometric complexes, making them more accessible to the end user. In addition, the requirements for data transmission channels are reduced, since with this approach, we do not require gigabit links for transmitting high-quality video, and it is enough to have standard networks for transmitting a compressed video and a minor flow of detected persons.

Stage 2: Quality Assessment

This is a very important stage of recognition, on which the biometric engine selects from the entire array of detected persons only those images that satisfy the specified quality criteria.

Often the developers of biometric systems are custody, arguing that their system provides a high level of recognition in accordance with the image images in the video stream of quality requirements defined in GOST R ISO / IEC 19794-5. However, this GOST imposes very rigid (practically ideal) conditions on the quality of the photographs of the face (the frontal perspective of the person with a deviation of no more than 5 degrees; uniform illumination; neutral facial exposure, etc.) that cannot be performed in real conditions of systems Video surveillance. Such GOST requirements are fully justified by the fact that, in fact, this standard is intended to unify the storage format of the electronic photography in the new generation of passport and visa documents - the so-called biometric passports. In practice, biometric identification system is forced to deal with much less favorable working conditions:

• deviation of the face from the frontal position on the corners exceeding 20 degrees;
• strong illumination;
• overlap part of the person;
• the presence of shadows on the face;
• small image size, etc.

It is the stability of a biometric engine in such difficult conditions and determines its quality. In modern biometric engines at the quality assessment stage, as a rule, are estimated:

• persons persecution (should not exceed 20-30 degrees.);
• the size of the person (estimated at the distance between the pupils of the eyes and must be more than 50-80 PKS);
• partial closure of the face (the closure of the face should not be more than 10-25% of the total area of \u200b\u200bthe face).

There is a common misconception that if the face of the face is closed (blinking or glasses), then the system cannot recognize the person. Really early face recognition algorithms used eye pupil centers as a base for further image processing, in particular for standard face scaling. However, at the moment, many modern biometric engines (for example, cognitec or "cascade flow") use more complex schemes of face coding and are not tied to the position of the centers of pupils.

Stage 3: Building Template

This is one of the most complex and unique stages of the face recognition, which constitutes the key know-how of the biometric engine technology. The essence of this stage consists in a non-trivial mathematical conversion of a face in a set of features combined into a biometric pattern.

Each person corresponds to his unique biometric pattern. The principles of constructing biometric patterns are extremely diverse: the template can be based on the texture properties of the face, on the geometric features, at the characteristic points, on combinations of various heterogeneous signs.

The most important characteristic of the biometric template is its size. The larger the size of the template, the more informative signs it includes, but the lower the speed and efficiency of the search for this template. The typical value of the facial template size in biometric systems is from 1 to 20 KB.

Stage 4: Comparison and decision making

This combined stage of the recognition system, which produces a biometric template of the face built on a detected person, with an array of templates stored in the database. In the simplest case, the comparison is carried out by simply searching for all templates and the assessment of the measure of their similarity. Based on the estimates and their comparison with the specified thresholds, a decision is made on the presence or absence of an identical personality in the database.

In modern systems, the comparison is implemented according to complex optimal comparison schemes, providing the rate of comparable from 10,000 to 200,000 comparisons per second and more. Moreover, it is necessary to understand that the mapping process can be fastened, which allows you to work with identification systems in almost real-time, even by large image arrays, for example, 100,000 people.

The quality of the work of individual recognition systems is made characterized by probabilities of identification. Obviously, with biometric identification, two types of errors are possible.

1. The first error is related to the ability to miss and not recognize a person, in fact, which is in the database, is often called the first kind error. And often indicate not the very value of the error of the first kind, and the unit minus the probability of the error of the first kind. Such a value is called the probability of correct recognition of PPR.
2. The second mistake reflects the cases when the system recognizes a person, in fact, not in the database or confuses it with another person, it is customary to be called a second-sort error. For modern face recognition systems, the typical value of the likelihood of the correct recognition is usually in the range of 80 to 97%, with a second kind error not exceeding 1%.

Terms of successful identification

It is worth understanding that face recognition is not absolute technology. Often you can hear criticism in the address of biometric systems, which in real objects it is not possible to achieve as high indicators as in the "laboratory" conditions. This statement is true only in part. Indeed, it is possible to effectively recognize a person only under certain conditions, which is why it is extremely important when the biometry is introduced to understand, under what conditions the system will be operated. However, for most modern recognition systems, these conditions are quite achievable in real objects. Thus, to increase the efficiency of face recognition in identification areas, you should organize a directional flow of people (doorways, frameworks of metal detectors, turnstiles, etc.) to ensure the possibility of a short-term (no more than 1-2 second) fixation of the face of each visitor. At the same time, video phixation cameras should be set with such a condition that the angle of deviations of fixed persons from the frontal position did not exceed 20-30 degrees. (for example, the installation of chambers on the distance from the passage zone of 8-10 m with a suspension height of 2-3 m).

Compliance with these conditions in the implementation of recognition systems allows to effectively solve the task of identifying the identity and search for people who are of particular interest, with probabilities, as close as possible to the values \u200b\u200bof successful identification indicators declared developers.

All are well known scenes from fantastic films: the hero is suitable for the door and the door opens, having learned it. This is one of the visual demonstrations of the convenience and reliability of the use of biometric technologies to control access. However, in practice, not so simple. Today, some firms are ready to offer consumers to access access using biometric technologies.

Traditional identification methods are based on various identification cards, keys or unique data, such as, for example, the password are not reliable to the extent that is required today. A natural step in improving the reliability of identifiers was attempts to use biometric technologies for security systems.

The range of problems whose solution can be found using new technologies, extremely wide:

1. prevent the penetration of intruders on protected areas and in the premises by fake, theft of documents, cards, passwords;
2. limit access to information and ensure personal responsibility for its safety;
3. provide admission to the responsible objects of only certified specialists;
4. avoid overhead costs related to the operation of access control systems (cards, keys);
5. eliminate inconvenience associated with loss, damage or elementary forgetting keys, cards, passwords;
6. organize accounting of access and attendance of employees.

The development of technologies for the recognition of images on various biometric characteristics began to be engaged at sufficiently long ago, the beginning was laid in the 60s. Significant success in developing the theoretical foundations of these technologies, our compatriots have achieved. However, practical results are obtained mainly in the West and only "yesterday." The power of modern computers and advanced algorithms allowed to create products that in their characteristics and the ratio have become available and interesting to a wide range of users.

The idea to use individual characteristics of a person to identify it is not Nova. To date, a number of technologies have been known to be involved in safety systems for identifying personality by:

1. fingerprints (both separate and hands as a whole);
2. face features (based on optical and infrared images);
3. rainbow Eye Shell;
4. vote;
5. other characteristics.

In all biometric technologies, there are general approaches to solving the identification problem, although all methods are characterized by ease of use, the accuracy of the results.

Any biometric technology is applied in stages:

1. scanning object;
2. extract individual information;
3. formation of a template;
4. comparison of the current template with the database.

The biometric recognition system establishes the compliance of the specific physiological or behavioral characteristics of the user to some specified template. Typically, the biometric system consists of two modules: the registration module and the identification module.

Registration module "Trains" system identify a particular person. At the registration stage, the video camera or other sensors scan the person in order to create a digital representation of his appearance. Face scanning lasts about 20 - 30 seconds, as a result of which several images are formed. In the ideal case, these images will have a slightly different perspective and expressions of the person, which will obtain more accurate data. A special program module processes this presentation and determines the characteristic features of the individual, then creates a template. There are some parts of the person who practically do not change over time, this is, for example, the upper outlines of the eyes, the area surrounding cheekbones, and the edges of the mouth. Most algorithms developed for biometric technologies allow you to take into account possible changes in the hairstyle of a person, as they are not used to analyze the field of the face above the hair growth boundary. The image template of each user is stored in the biometric system database.

Identification module Gets from the camcorder, the image of a person and converts it to the same digital format in which the template is stored. The data obtained is compared with stored in the database template in order to determine if these images meet each other. The degree of similarity required for verification is a certain threshold that can be adjusted for a different type of personnel, PC power, time of day and a number of other factors.

Identification can be performed in the form of verification, authentication or recognition. When verification, the identity of the data and the template stored in the database is confirmed. Authentication - confirms the correspondence of the image obtained from the camcorder to one of the templates stored in the database. When recognizing, if the characteristics obtained and one of the stored patterns are the same, the system identifies a person with a corresponding template.

When using biometric systems, especially the recognition system in the face, even when the correct biometric characteristics is not always introduced, the authentication decision is correct. This is due to a number of features and, first of all, with the fact that many biometric characteristics may vary. There is a certain degree of system error. Moreover, when using various technologies, an error can differ significantly. For access control systems, when using biometric technologies, it is necessary to determine what is more important not to miss the "alien" or skip all the "their".

An important factor for users of biometric technologies in security systems is the ease of use. A person whose characteristics are scanned should not have any inconvenience at the same time. In this regard, the most interesting method is, of course, the recognition technology for the face. True, in this case other problems arise primarily, with the accuracy of the system.

Despite the obvious advantages, there are a number of negative prejudices against biometrics, which often cause questions about whether biometric data will be used for surveillance of people and violating their right to privacy. Due to sensational statements and unreasonable noise, the perception of biometric technologies is sharply different from the real state of affairs.

Nevertheless, the use of biometric identification methods has gained special relevance in recent years. Especially acute this problem manifested itself after the events of September 11 in the United States. The world community realized the degree of increase in the threat of terrorism around the world and the complexity of organizing reliable protection by traditional methods. It is these tragic events that served as the starting point to enhance attention to modern integrated security systems. It is well known that if control in airports was stricter, then misfortunes could be avoided. Yes, and today, the search for those perpetrators in a number of other incidents could be significantly facilitated by using modern video surveillance systems in integration with individual recognition systems.

Currently, there are four main face recognition methods:

1. "Eigenfaces";
2. analysis of "distinctive features";
3. analysis based on "neural networks";
4. method "Automatic Image Image Processing".

All these methods differ in the complexity of the implementation and purpose of use.

"Eigenface" can be translated as "own face". This technology uses two-dimensional images in grayscale grades, which represent the distinctive features of the face image. The "Eigenface" method is often used as a basis for other face recognition methods.

Combining characteristics 100 - 120 "Eigenface" You can restore a large number of persons. At the time of registration, the "Eigenface" of each particular person is presented in the form of a number of coefficients. For authentication mode, in which the image is used to check the identity, the "live" pattern is compared with the already registered template, in order to determine the difference ratio. The degree of difference between templates and determines the fact of identification. The technology "EiGenface" is optimal when used in well-lit rooms, when there is the possibility of scanning a face in the FAS.

Methods for analyzing "distinctive features" - the most widely used identification technology. This technology is similar to the "Eigenface" technique, but is more adapted to the change in the appearance or facial expressions of a person (smiling or frowning person). In the technology of "distinctive features" are used dozens of characteristic features of various field areas, and with regard to their relative location. An individual combination of these parameters determines the features of each particular person. Man's face is unique, but quite dynamically, because A person can smile, let go of his beard and mustache, wear glasses - all this increases the complexity of the identification procedure. Thus, for example, with a smile, some displacement of the parts of the person located near the mouth is observed, which in turn will cause such a movement of adjacent parts. Given such displacements, you can definitely identify a person and with various facial changes. Since this analysis considers local sections of the person, permissible deviations can be in the range of up to 25 ° in the horizontal plane, and approximately 15 ° in the vertical plane and requires a sufficiently powerful and expensive instrument, which corresponduously reduces the degree of propagation of this method accordingly.

In a method based on the neural network, the characteristic features of both individuals - registered and checked compared to the coincidence. The "neural networks" use an algorithm that establishes the compliance of the unique parameters of the person being checked and the parameters of the template in the database, while the maximum possible number of parameters is applied. As the compares, the inconsistencies between the person being checked and the template from the database are determined, then the mechanism is started, which, using the appropriate weight coefficients, determines the degree of compliance of the face being checked by a template from the database. This method increases the quality of the identification of the person in difficult conditions.

The "automatic image processing method" method is the most simple technology that uses the distance and the ratio of distances between the easily defined points of the face, such as the eyes, the end of the nose, the corner of the mouth. Although this method is not so powerful as "Eigenfaces" or "neural network", it can be quite efficiently used in low light conditions.

Face recognition systems present in the market

To date, a number of commercial products designed to recognize persons have been developed. Algorithms used in these products are different and is still difficult to evaluate which technologies has advantages. The leaders are currently the following systems: Visionic, Viisage and Miros.

• The visionic FaceIt application is based on the algorithm for analyzing local signs developed at the University of Rockefeller. One commercial company in the UK integrated Faceit to a television anticriminal system called Mandrake. This system is looking for criminals for video data, which come from 144 cameras combined into a closed network. When identity is established, the system reports this security officer. In Russia, the representative of Visionic is the company "Dank".
• Another leader in this area, Viisage, uses the algorithm developed at the Massachusetts Institute of Technology. Commercial companies and government structures in many US states and in a number of other countries use the Viisage system along with identification certificates, such as driver's license.
• Zn Vision Technologies AG (Germany) offers a number of products in the market, which uses face recognition technology. These systems are presented in the Russian market by Soling.
• The Miros TrueFace Recognition System is used by neural network technology, and the system itself is used in the Mr.Payroll Cash Corporation Complex and is installed in the Casino and other fishering institutions of many US states.

In the US, independent experts carried out comparative testing of various face recognition technologies. Test results are presented below.

Fig. 1. Comparative analysis of the effectiveness of recognition of persons in different systems

In practice, using the recognition systems of persons as part of standard electronic security systems, it is assumed that a person who should be identified, looks straight into the chamber. Thus, the system works with a relatively simple two-dimensional image, which significantly simplifies the algorithms and reduces the intensity of calculations. But even in this case, the task of recognition is still not trivial, since the algorithms should take into account the possibility of changing the level of lighting, a change in the expression, the presence or absence of makeup or glasses.

The reliability of the face recognition system is very dependent on several factors:

• Image quality. The probability of unmistakable operation of the system is noticeably reduced, if the person we is trying to identify, does not look directly into the chamber or removed with poor lighting.
• The relevance of the photo recorded in the database.
• The value of the database.

Face recognition technologies work well with standard video cameras that transmit data and manage a personal computer, and require permission of 320x240 pixels per inch at a video stream speed, at least 3 - 5 frames per second. For comparison - acceptable quality for video conference requires the speed of the video stream from 15 frames per second. The higher speed of the video stream at a higher resolution leads to an improvement in the quality of identification. When recognizing persons from a large distance, there is a strong dependence between the quality of the camcorder and the result of identification.

The amount of databases when using standard personal computers does not exceed 10,000 images.

Conclusion

The methods of recognition methods offered today are interesting and close to widespread introduction, but it is not yet possible as in the cinema to trust the door opening only to face recognition. It is good as an assistant for a security guard or another access control system.

It is this method that is used in many situations when it is required to make sure that the document is really belonging to his person. This happens, for example, at the international airport, when the border guard twists the photo on the passport with the person of the passport owner and makes a decision, it is a passport or not. A computer access system is also valid for a similar algorithm. The difference consists only that the photo is compared with the template database already stored in the database.

Already appeared technologies that are based on the recognition of persons in infrared light. The new technology is based on the fact that the thermal picture created by heat radiation by blood vessels or, differently, the thermogram of a person's face is unique for everyone and, therefore, can be used as a biometric characteristic for access control systems. This thermogram is a more stable identifier than the geometry of the person, since it almost does not depend on the change of human appearance.