ACCESS CONTROL SYSTEM WITH FACE AND PSYCHO-EMOTIONAL STATE RECOGNITION ALGORITHMS
ACCESS CONTROL SYSTEM WITH FACE AND PSYCHO-EMOTIONAL STATE RECOGNITION ALGORITHMS
Vildan Bayazitov
student of Ufa State Petroleum Technological University,
Russia, Ufa
ABSTRACT
The article considers the relevance of computer vision, a widely used technology, on the example of face recognition as part of an access control system. The basic methods employed in the implementation of conventional access control systems are considered. The plan of use and the specifics of the application of the access control system with face recognition technology are described. This technology enables increasing of the enterprise information security level and, ultimately, reduces the possible financial damage from the attacks on their asset, either from illegitimate penetration into the protected area through the access control system by means of legal user passes.
Video analytics - the basic principle of an intelligent video surveillance system, is a technology based on methods and algorithms for pattern recognition, image processing, and automated data collection upon video stream analysis. Without human involvement, such equipment is capable of detecting and tracking the real-time specified objects (a car, a group of people), potentially dangerous conditions (smoke, fire, unauthorized intervention in the operation of video cameras), programmed events, and prompt issuing of an alarm signal. Screening of irrelevant video data significantly reduces the load on communication channels and the historical database.
Face recognition is one of the most promising methods of biometric non-contact facial identification.
The first face recognition systems were implemented in a form of computer programs. Nowadays, face recognition technology is most often used in video surveillance systems, access control, as well as on various mobile and cloud platforms.
By 2019, facial recognition algorithms have reached unprecedented perfection: now the technology outperforms the man, and algorithm developers compete in terms of accuracy already in the 5th or 6th decimal place [2].
The main drivers of this segment are the growth of the video surveillance market, requests from government agencies, as well as the increase in the use of biometric applications in various industries. Apart from traditional means of the object security provision, a modern security system is nowadays substantially dependent on biometrics. Its capabilities are used not only to control access to premises or protect for information protection, but also to carry out financial transactions, identify people and objects, and ensure industrial security. The use of an access control system (ACS) based on face recognition ensures object access to authorized persons only, and provides flexible differentiation of rights.
Access control systems (ACS) introduced in enterprises allows for automatically controlled process of the admission of employees to the facilities, and analysis of time spent there. Automated control helps both – to set access rules individually for each employee, and monitor the activities of each employee in a particular room. The database stores information about the employee coming and going, name of his position and how much time he spent at the facility. Such systems enable to reduce the cost of manual control, as well as the time of passing through the entrance turnstile or door due to no need to record the employee data and in / out time in a paper log with alternative registration in an electronic log. In addition, the security specialist always obtains up-to-date information about the presence of each employee at the enterprise. If necessary, the pass may be locked immediately.
In addition, the security specialist always has up-to-date information about the presence of each employee at the enterprise. If necessary, an immediate blocking of the pass is possible.
A typical ACS, as a rule, is based on a network of controllers connected to a computer and includes:
− controlled blocking devices (turnstiles, doors equipped with controlled locks, gates, barriers);
− devices for identification data input (readers);
− electronic microprocessor modules that implement object authentication, and authorization logic for access to certain premises;
− software that allows centralized management of controllers from a personal computer, generation of reports, various additional functions;
− converters (interfaces) of the environment for connecting ACS hardware modules to each other and to a personal computer;
− auxiliary equipment (cameras, power supplies, routers, etc.) [3].
Efficient system operation envisages several types of IP cameras with different performance characteristics. The object within the controlled area is detected by all-around cameras with a resolution of 1 megapixel and a focal length of 1 mm, and brought into focus of scanning devices. These are advanced cameras (from 2 megapixels, from 2 mm) performing simple-method recognition. Object identification employs cameras with good quality, sufficient for applying complex algorithms (from 5 megapixels, 8-12 mm).
The recognition technology is based on neural networks, each of which is a mathematical apparatus consisting of several layers and performing parallel calculations.
The first neural network is an “aligner” that analyzes the input image by detecting, cutting out and aligning faces.
The second is the "recognizer" neural network. It takes a prepared image and, after analyzing it, produces vectors, which in most cases are characterized by a power of the number 2. When comparing two images, the distance between the vectors is calculated (the greater the distance, the less similarity).
The practical implementation of an access control system based on the facial recognition requires describing the main components that can be included in the system.
The first component is the analysis platform, which can be implemented as a separate server to which the client sends a HTTP-request. In response to the request, the server forwards a unique user identifier.
The second component is the analyzer. In order to reduce time costs and improve the quality of research, pre-trained neural networks should be used to create a model as a deep learning analyzer, which, in turn, simplifies the development and implementation of ACS.
The third component is the decision module. The input images pass through two "equal" networks used for two different inputs. After that, the outputs are compared and a decision is made about the similarity of the images. In case of successful recognition, the system issues a unique user ID.
The main recognition technologies of static biometric identification methods allows for the conclusion that an access control system based on approaches to face shape recognition through the use of neural networks for face shape detection and feature recognition may most probably be developed [1].
The specificity of the facial recognition technology implementation differs in different criticality to errors depending on the scope of application:
1. Access control systems;
2. Face recognition in transport;
3. Time tracking;
4. Face recognition in the crowd;
5. Age determination;
6. Sex determination.
Installing smart video cameras saves millions of rubles by preventing fraud in the lending and payments segments. In 2022, the essential infrastructure network will be created, and all operations at ATMs will become possible only after client’s biometric facial identification.
In the next decade, high technology will allow opening a network of unattended markets, like, the buyer walks through the storefronts, selects the product he needs and leaves. The face and image recognition system will determine the identity of the buyer and his purchase contents, and write off the necessary amount from his account. Creating of systems for recognizing the psycho-emotional state is in progress. The analysis of human emotions will be in demand in multimedia areas, namely in animation, cinematography, the industry of creating computer games [4].
Errors are anticipated in process of comparing the information received from the biometric reader with the template. Such errors in biometrics are divided into two types: errors of the first kind and errors of the second kind. They inevitably arise, which is the main reason for the technology as a whole to face criticism.
False Rejection Rate (FRR), the errors of the first kind occur when the scanner is unable to recognize a registered user. This is not critical for security, but creates inconvenience, since it is necessary to carry out a secondary verification of the biometric parameter. Errors of the second kind (FAR - False Acceptance Rate) mean that the system takes an unregistered user as registered. This type of error is security-critical because a trespasser gains access to the facility.
FRR and FAR errors are interrelated; decrease in the number of one kind brings to the increase in the number of the other kind. Therefore, in each specific case, systems and devices are configured in accordance with the tasks for biometrics to solve at particular facility [6].
Thus, biometrics in ACS is a fairly reliable, but not 100% accurate, identification method. However, every year technologies become more and more advanced, which means that biometric identification methods will become even more popular and available in the security market.
References:
- Averenkova, M. A. Development of an access control and management system using biometric identification methods / M. A. Averenkova, A. V. Tezin. - Text: direct // Young scientist. - 2021. - No. 51 (393). - p. 8-12. — URL: https://moluch.ru/archive/393/86863/ (date of access: 13.04.2022). [Averenkova, M. A. Razrabotka sistemy kontrolya i upravleniya dostupom s primeneniyem biometricheskikh metodov identifikatsii / M. A. Averenkova, A. V. Tezin. — Tekst : neposredstvennyy // Molodoy uchenyy. — 2021. — № 51 (393). — S. 8-12. — URL: https://moluch.ru/archive/393/86863/ (data obrashcheniya: 13.04.2022).]
- Biometric readers/terminals for ACS and VDA based on face identification technologies Effective Companion of Professionals — URL: https://www.secuteck.ru/company (date of access: 13.04.2022). [Biometricheskiye schityvateli/terminaly SKUD i URV na osnove tekhnologiy identifikatsii po litsu Effektivnoye kompniti professionalov — URL: https://www.secuteck.ru/company (data obrashcheniya: 13.04.2022).]
- Development of an Access Control System Based on Face Recognition Antipova — URL: https://cyberleninka.ru/article/n/razrabotka-sistemy-kontrolya-dostupa-na-osnove-raspoznavaniya-lits (date of access: 13.04.2022). [Razrabotka sistemy kontrolya dostupa na osnove raspoznavaniya lits Tekst nauchnoy stati po spetsialnosti «Kompyuternyye i informatsionnyye nauki» S.A. Antipova — URL: https://cyberleninka.ru/article/n/razrabotka-sistemy-kontrolya-dostupa-na-osnove-raspoznavaniya-lits (data obrashcheniya: 13.04.2022).]
- Facial recognition access control systems: choosing the right one - URL: https://www.drdoors-msc.ru/stati/sistemy-kontrolya-dostupa-s-raspoznavaniem-lic-kak-pravilno-vybrat.html 04/13/2022). [Sistemy kontrolya dostupa s raspoznavaniyem lits: kak pravilno vybrat — URL: https://www.drdoors-msc.ru/stati/sistemy-kontrolya-dostupa-s-raspoznavaniem-lic-kak-pravilno-vybrat.html (data obrashcheniya: 13.04.2022).]
- Terminals and integrated access control systems with biometric face recognition. Journal "Protection Technologies" - URL: http://www.techportal.ru/review/sistemy-kontrolya-dostupa-s-raspoznavaniem-lits/#raspoznavanie-lits-v-sigur-shirokie-vozmozhnosti-vnedreniya- (date of access : 13.04.2022). [Terminaly i integrirovannyye sistemy kontrolya dostupa s biometricheskim raspoznavaniyem lits. Zhurnal "Tekhnologii Zashchity" — URL: http://www.techportal.ru/review/sistemy-kontrolya-dostupa-s-raspoznavaniem-lits/#raspoznavanie-lits-v-sigur-shirokie-vozmozhnosti-vnedreniya- (data obrashcheniya: 13.04.2022).]
- Face Recognition Technology "A" to Z" - URL: https://securityrussia.com/blog/face-recognition.html#141 (accessed 13.04.2022). [Tekhnologiya raspoznavaniya lits «A» do YA» — URL: https://securityrussia.com/blog/face-recognition.html#141 (data obrashcheniya: 13.04.2022).]