 |
Petrosyanc Victor Sergeevich
Faculty: Computer sciences
Speciality: Software engineering
Master's theme:
Research of algorithms, models and methods of recognition of personality on the basis of base.
Leader of work: Zori S.A.
|
Abstract
Introduction
The permanent increase of volumes of the processed information results in tomu, that its greater totally is passed only electronic a way (ò.å. without the receipt of copy on a hard transmitter). The amount of transactions, accomplished by electronic communication means is multiplied, – ot purchases in the Internet - shops and auctions to playing the stock market. There was a necessity of introduction of the so-called «electronic signature». In some countries, laws, confirmative legal force of electronic signature, are accepted. However much the problem of reliable user of resources authentification with the limited access (including komp'yuternykh networks) remains opened. Partly therefore in area of treatment of information the last years there is a boom of development of the systems of authentication of man on biometricheskim signs.
Actuality
The volume of the processed information grows every year, that requires formalization and subsequent algoritmizatsii of processes, before executed by hand. One of key concepts there is “recognition” of objects of certain class in automatic treatment of information. When algorithms execute recognition at the level of expert-man, automation conduces to the acceleration of work of data handling system and increase of their efficiency.
Presently actively biometricheskie technologies develop are «methods and hardwares of receipt and use of biometricheskikh information of man for his authentication». Their purpose is development of the automated systems of authentication of man on the basis of biometricheskikh signs: finger-prints, descriptions of voice and speech, picture of iris of eye, representing a person. It is related to the tem fact, that now by a basic obstacle for further development of informative environments, different sort of virtual services, etc. there is a problem of reliable user identification (verifikatsii).Expected, that application of the similar systems will decrease the amount of crimes, related to íåñàíêöèîíèðî¬âàííûì access substantially, including in komp'yuternykh networks, thus exactly technology of authentication of man on the basis of representing persons is acknowledged most acceptable to mass application, because she does not require a physical contact with a device, unobtrusive, natural and, in potential, can possess high reliability and speed.
Search in the bases of information on the photos of man, automated control of certificates of identity especially actual for law enforcement authorities of most countries in the context of multiplying the amount of assassinations and criminal crimes on a background the increase of general mobility of population.
Overview of existing researches
Among existent works above all things it is needed to select dissertation on the competition of graduate degree of candidate of engineerings sciences Samalya Dmitry Ivanovicha(Minsk, 2002).
In dissertation work of Samal' D. And. conducted research and the followings algorithms are developed:
- algorithm of search of area of person on a photo and localizations of centers of pupils, different ability to find in the real-time mode the centers of pupils on slabokontrastnykh images, got scanning of picture from the certificate of identity.
- algorithm of increase of contrast and normalization of digital photos, allowing to present the analysed photos in a single scale and orientation with near brightness-contrasting descriptions. He differs possibility of aligning descriptions of brightnesses of both slabokontrastnykh (got scanning of certificate of identity) photos and contrasting images of man, actings from a video camera.
- algorithms of search of eight basic anthropometric points of person on digital photos. First experimentally îïðåäå¬ëåíû parameters of filter of Gabora, applied for finding out these points, and parameters of normalization of photos.
The informing signs of representing the face of man are certain on the basis of anthropometric points. The feature of set of signs is robastnost' to the changes of ages the exterior of man, luminosity, to makiyazhu and does not mean l'nym variations of foreshortening of survey, invariance to the scale, brightness and contrast.
The article of research within the framework of master's degree work is an inductive synthesis of implikativnoy base of knowledges and znanieorientirovannoe decision-making within the framework of quantum approach for making identification decisions about personality.
Scientific novelty
Algorithms are developed for authentication of man on digital photos on the basis of geometrical signs . Their basic differences ot existent algorithms it is been: work in the conditions of absence of a priori information about people, stability to the changes of ages of people, represented on the processed photos, possibility straight applications of mathematical receptions and approaches due to clear formalization of knowledges.
Basic part
Chart of functioning of programmatic complex for recognition of personality on a photo
On the stage of forming of database normalization of all photos, presented as blocks «search of area of person», «search of centers of pupils», «improvement of brightness, is executed, contrast». Then, there is extraction of informing features: the function of search of anthropometric points is executed at first, after function of calculation of anthropometric signs. In order to avoid errors possibility of hand correction of co-ordinates of anthropometric points of person is foreseen . The anthropometric signs calculated on their basis are a vector which is kept in a base.
Filling of information (records) about an object is executed an operator. The last name, name, patronymic of object and additional information about him, forming the vector of additional signs, is entered.
On the basis of vector of anthropometric signs and vector of additional signs an eventual vector is formed for adding to the base of information.
Decision about authentication on the looked after object is made on the basis of implikativnoy BkKn by the deductive conclusion of deciding rule.
Conclusion
Every day the systems of recognition of personalities on a photo are perfected, their developers aim to take them to possibility of authentication from a 100%-ñòåïåíüþ authenticity and quality, applying and perfecting all new and new methods, models, methods and algorithms, that their full introduction was justified those results which ot them are expected. Thus, research of methodologies, algorithmic bases and structures of the similar systems is an important scientific and technical problem. In this work authors on the basis of some studied algorithms and new method of storage and selection obtained information offer new approach to the decision set the problem.
This work is not the finished research. At further development the substantial improvement of functioning of this system is possible by modernization of the separate modules, so bringing of new.
Literature
1. Wu C. and Huang J. Human face profile recognition by computer// Pattern Recognition.- 1990. -Vol.23. - No. 3/4. . P. 255-259 p.
2. Daughman J. Face and Gesture Recognition: Overview // IEEE Transactions on Pattern Analysis and Machine Intelligence. – 1997. – Vol. 19. – P. 675-676 ñ.
3. Presentation and use of knowledges: Trudged. with yapon./ Under red. X. Ueno, ì. of Isidzuka. - Ì.: the World, 1989. - 220 p.
4. Osuga Ñ. Îáðàáîòêà knowledges: Trudged. with yapon. - Ì.: the World, 1989. - 293 p.
5. Pospelov D.A. Tipologiya of the intellectual systems.M.: Association of artificial intellkta. 1994.-43 p.
|
|
