RU2365995C2 - System and method of recording two-dimensional images - Google Patents

Abstract

FIELD: information technology.

SUBSTANCE: invention relates to biometric systems and methods of identifying people from face features and, particularly to systems and methods of recording biometric data of faces. The technical outcome is achieved due to that, the system for recording two-dimensional face images contains a PIN number reading unit, PIN number decoding unit, memory unit, display, unit for reading face images, processor module, interconnected with the given units and is connected through one of its outputs to an additional lighting unit. The unit for reading face images, made in form of a device for reading three-dimensional images, unit for analysing quality of three-dimensional face images, unit for generating two-dimensional face images and unit for compressing two-dimensional face images are interconnected with the processor module. The system works in accordance with the proposed method.

EFFECT: increased certainty of quantitative evaluation of quality of two-dimensional face images.

14 cl, 12 dwg

Description

The invention relates to biometric systems and methods for recognizing people by facial features and, in particular, to systems and methods for recording biometric face data. It can be used as part of any biometric recognition system for two-dimensional face images.

It is known that in recent years significant progress has been made in the development of technologies for automatic biometric recognition of high-quality two-dimensional facial images. For example, most authoritative tests of face recognition technologies from various manufacturers have demonstrated the high efficiency of using automatic biometric recognition of high-quality two-dimensional face images:

- FRVT2000 and FRVT2002 by governmental organizations of the USA, Great Britain, Canada and Australia (http://www.frvt.org/FRVT2000/default.htm and http://www.frvt.org/FRVT2002/default.htm) [1];

- testing by the German government

(http://www.bsi.bund.de/english/fachthem/BioFace/BioFaceIIReport.pdf) [2];

- testing by the US Department of Defense “DoD Counterdrug Technology Development Program Office's scenario evaluation, Face Recognition at a Chokepoint” (http://www.dodcounterdrug.com/facialrecognition/DLs/ChokePoint Results. Edf) [3];

- testing of CESG 2001 by the National Physical Laboratory commissioned by the UK Government (www.cesg.gov.uk/technology/biometrics) [4];

- testing of Personnel / Visitor Identification at DoD Labs & Technology Centers 2002 by the US Army Research Laboratory

(http://archive.aclu.org/issues/privacy/FINAL_1_Final_Steve_King.pdf) [5];

- interim results of the implementation of the large-budget project Human Identification at a Distance of the Office of Advanced Research of the United States Department of Defense, in the framework of which a multi-biometric system for covert recognition of a person at a large (up to 150 m) distance is created for military purposes (www.darpa.mil/iao/hid.htm) [6];

- testing of the reputable American biometric consulting company International Biometric Group (www.ibgweb.com) [7].

For these reasons, international and national standards have been developed that directly or indirectly introduce a quantitative quality metric for electronic two-dimensional face images:

- ANSI / INCITS 358-2002 - Information technology - BioAPI Specification [8];

- NISTIR 6529-A-2004 - Common Biometric Exchange Format Framework (CBEFF) [9]

- ANSI / NIST-ITL 1-2000 - Standard Data Format for the Interchange of Fingerprint, Facial, & Scar mark & Tattoo (SMT) Information [10];

- ANSI / INCITS 379 - Information Technology - Iris Image Interchange Format [11];

- ANSI / INCITS 385-2004 - Information Technology - Face Recognition Format for Data Interchange [12];

- ISO / IEC 10918 - Information technology - Digital Compression and coding of continuous-tone still images (JPEG) [13];

- ISO / IEC 15444 - Information technology - JPEG 2000 Image Coding System [14];

- NISTIR 6529-A-2004 - Common Biometric Exchange Format Framework (CBEFF) [15];

- ICAO, Facial Image Standard (http://www.icao.int/mrtd/download/documents/Annex D - Facial Image.pdf; http://www.icao.int/mrtd/download/documents/Annex A - Photograph Guidelines.pdf; http://www.icao.int/mrtd/download/documents/Annex B - Facial Image Size Study.no 1.pdf;

http://www.icao.int/mrtd/download/documents/Annex C - Facial Image Size Study. no2. pdf) [16];

- NIST, Image Group, Best Practice Recommendation Version 2 for capturing mugshots and facial images

(http://www.itl.nist.gov/iad/894.03/face/bpr_mug3.htm1) [17];

- U.S. Department of State, Bureau of Consular Affairs, Guidelines for Producing High Quality Photographs for U.S. Travel Documents (http://www.travel.state.gov/passport/pptphotos/index.htm1) [18];

- Project 1603, ANSI / INCITS 395, Information Technology, Signature / Sign Image Based Interchange Format for Data Interchange (currently under development) [19];

- ANSI / INCITS (M1-03-0351), Information Technology, Biometric Performance Testing and Reporting Standard (currently under development) [20];

- Project ISO / IEC CD 19794-5, Biometric Data Interchange Formats - Part 5: Face Image Data (under development) [21];

- Project INCITS 1672-D (M1 / 04-0608), Biometric Sample Quality Standard (under development; http://www.incits.org/tc_home/m1htm/docs/m1040608.PDF) [22];

- Project INCITS 1703-D (M1 / 04-0697), Information technology, Conformance Testing Methodology for ANSI INCITS 358-2002, BioAPI Specification (under development) [23].

In addition, the BioAPI standard (version v1.1, section 2.1.46) [24] defines the structure and gives recommendations for a quantitative subjective assessment of the quality parameters of two-dimensional facial images.

Some other standards (for example, the emerging Biometric Data Quality Standard or Biometric Sample Quality Standard;

http://www.incits.org/tc_home/m1htm/docs/m1040608.PDF) [25] will determine a methodology for calculating objective quantitative assessments of the quality of two-dimensional face images, including expressions and interpretations for use in various applications of biometric face recognition technology:

- compatibility of various face image databases;

- testing and improving face recognition technology;

- real-time quality control when registering a face image to improve operational efficiency and productivity in order to accurately make decisions in biometric recognition: provide access, do not provide access, re-capture biometric data, refuse registration, etc .;

- real-time quality control when administering facial image databases to improve operational efficiency and productivity in order to accurately make decisions in biometric recognition: provide access, do not provide access, re-capture biometric data, refuse registration, etc .;

- use in various face recognition applications;

- use in expert analysis systems;

- use in applications one-to-one matching (1-to-1);

- use in applications one-to-few matching (1-to-few);

- use in applications one-to-many matching (1-to-many);

- use in applications of comparison with the signal list;

- the use of two-dimensional facial images for database administration.

The following is additional information on some of the aforementioned standards and formats for storing biometric data in memory and their exchange between different biometric systems:

- face recognition format with data exchange - M1 (www.incits.org/tc_home/m1.htm) [26] and ISO SC37 (www.jtcl.org) [27]. It guarantees that registered face images will meet the quality standards necessary for both automatic face recognition and expert research. Promotes the use of personal data in applications with a limited amount of memory (for example, passports, visas, driver's licenses, etc.), and provides data exchange between face recognition systems from various manufacturers;

- biometric programming interface -BioAPI: ANSI / INCITS 358-2002 (www.bioapi.org) [28]. Defines the interaction interface of biometric software applications. An open standard agreed upon in the biometric community;

- The format for the exchange of biometric data is CBEFF (www.nist.gov/cbeff) [29]. Defines the data structure for creating files with biometric information, which contributes to the compatibility of various biometric components and systems. Consists of a standard header (biometric specific memory block -BSMB) and an optional digital signature;

- The standard for processing and transmitting biometric data is ANSI X9.84-2002 (www.x9.org) [30]. It defines the basis for data processing and makes demands on the methods of cryptographic protection of biometric data before transferring them by adding additional security elements that are not provided in the BioAPI interface. Compatible with BioAPI and CBEFF;

- Attribute Certificate -X.509 Attribute Certificate (www.nist.gov) [31]. Part of the Public Key Infrastructure (PKI structure). It requires storing a biometric template as an attribute of information within a fixed number of bytes in the Open Key;

- ICAO MRTD standards for biometric machine-readable travel documents (www.icao.int/mrtd/download/technical.cfm) [32].

On the other hand, over the years, government agencies in many countries have compiled various electronic databases of two-dimensional images of persons (for example, registration systems for passports, visas, driver's licenses, etc.) using, at best, the minimum quality requirements for photo images needed for visual analysis by the operator.

Therefore, currently there are many databases containing low-quality two-dimensional images of persons in need of re-registration (re-registration).

In addition, in many countries, new national citizen identification programs are being developed and implemented related to the creation of large-scale databases of high-quality two-dimensional images of persons to improve public safety (identifying terrorists and other criminals, preventing fraud with identity cards and travel tickets, providing authorized access to security zones and systems, etc.):

- US program US-VISIT (www.dhs.gov) [33] biometric monitoring of all citizens entering the country;

- EU program (http://europa.eu.int/eur-lex/en/com/pdf/2003/com2003_0558en01.pdf) [34] on the implementation of biometric data in visas and residence permits;

- The ICAO International Civil Aviation Organization program (www.icao.int) [35] for incorporating biometric data into machine-readable travel documents;

- The UK Government program for the implementation of electronic identity cards with biometric data (www.homeoffice.gov.uk) [36];

- and etc.

Therefore, the market now needs a way of registering two-dimensional face images that would provide an objective quantitative assessment of the quality of two-dimensional face images in an automatic mode close to real time.

Various procedures for recording and recognizing biometric data, for example, are described in the following inventions:

- United States Patent 5457747 October 10, 1995 [37]. System of personal biometric cards;

- United States Patent 6107935 August 22, 2000 [38]. User biometric recognition system for the purpose of access control;

- United States Patent 6246751 June 12, 2001 [39]. A method of providing a service to authorized users using recognition of registered biometric data;

- United States Patent 6421453 July 16, 2002 [40]. A method of access control using the recognition of intentional gestures;

- United States Patent 6735695 May 11, 2004 [41]. A method of organizing access control using registration and recognition of biometric data.

Various procedures for registering and recognizing face data, for example, are described in the following inventions:

- United States Patent 6111517 August 29, 2000 [42]. A way to continuously control access to a protected area using face recognition;

- United States Patent 6724919 April 20, 2004 [43]. A method for verifying the face of a user of a computer system with the ability to continuously scan a face while a user is working at a computer;

- United States Patent 4449189 May 15, 1984 [44]. A method for identifying a person by speech and facial features at the same time;

- United States Patent 5412738 May 2, 1995 [45]. Recognition system (identification and verification) of a person using static and dynamic voice and face data;

- Russian Federation 95115528/09 September 8, 1995 [46]. A method for recognizing a person using a personal PIN number, face recognition algorithms and additional physiological characteristics;

- Russian Federation 96115857/09 August 6, 1996 [47]. A method for recognizing a person from a face image using a mirror to position the head relative to the scanning device;

- United States Patent 6219640 April 17, 2001 [48]. A method of verifying a person by voice using analysis of audio and video data simultaneously;

- United States Patent 6219639 April 17, 2001 [49]. A method for recognizing a person by two signs, one of which is the shape of the lips;

- United States Patent 6504944 January 7, 2003 [50]. Image recognition devices using 3D readers and methods for recognizing the mouth and lips of a talking person;

- United States Patent 6567775 May 20, 2003 [51]. A method for identifying a talking person using audio and video data.

Various methods for controlling the quality of facial data using methods of preliminary analysis, noise reduction, and human location control, for example, are described in the following inventions:

- United States Patent 6005610 December 21, 1999 [52]. Audiovisual way of determining the location and tracking of at least one object;

- United States Patent 6707921 March 16, 2004 [53]. A method of filtering noise in human speech by analyzing the position and movement of the mouth.

Various methods for recovering facial data using an analysis of audiovisual information, for example, are described in the following inventions:

- United States Patent 4757541 July 12, 1988 [54]. Speech recognition device using lip shape;

- United States Patent 4913539 April 3, 1990 [55]. Lip animation device in accordance with the pronounced sounds;

- United States Patent 5608839 March 4, 1997 [56]. A method of synchronizing audio and video data of a person;

- United States Patent 6014625 January 11, 2000 [57]. A method for determining lip movement parameters using audio and video signals;

- United States Patent 6243683 June 5, 2001 [58]. A method for controlling the process of speech recognition using video analysis;

- United States Patent 6449595 September 10, 2002 [59]. A method of restoring a face image in accordance with a speech signal using audio-visual data;

- United States Patent 6594629 July 15, 2003 [60]. A method for detecting speech using speech recognition and analysis of related audio and video data;

- United States Patent 6633844 October 14, 2003 [61]. A method of speech recognition through the analysis of audio and video data simultaneously.

Various methods for registering face data, for example, are described in the following inventions:

- United States Patent 6072891 June 6, 2000 [62]. A method for recording biometric data with preliminary quality control;

- United States Patent 5978494 November 2, 1999 [63]. A method for selecting the best registration image for personal identification.

Various methods for analyzing three-dimensional and two-dimensional face images, for example, are described in the following inventions:

- United States Patent 6026188 February 15, 2000 [64]. A method for recognizing three-dimensional objects by generating a rotated two-dimensional image of an object from a set of two-dimensional registered images;

- United States Patent 6535839 March 18, 2003 [65]. A method for calculating the location and orientation of an object in three-dimensional space;

- United States Patent 6580821 June 17, 2003 [66]. A method for combining computer models of two surfaces in three-dimensional space.

A well-known example of a third-party registration system (source: DoD Counterdrug Technology Development Program Office; testing program: "Face Recognition at a Chokepoint. Scenario Evaluation Results") [67], which uses a block for reading two-dimensional video images and an additional lighting unit made in the form of two primary light sources, two diffusing umbrellas mounted on tripods, and one screen. However, it has significant drawbacks, since it cannot be used to reliably assess the quality and facial features.

Closest to the claimed invention is a method for recognizing three-dimensional objects by generating a two-dimensional image of an object from a transformed three-dimensional model (United States Patent 6,002,782 December 14, 1999) [68], which solves the problem of not assessing quality, but matching the angle by generating a two-dimensional image of the face of the desired angle from a registered three-dimensional image of a face for subsequent comparison with a two-dimensional image of the face of a person in the field of view of the reader of two-dimensional images. This method is selected as a prototype of the claimed invention.

However, unfortunately, the above devices and inventions, including the prototype method, do not contain a description of an acceptable registration method and reliable quantitative assessment of the quality and features of two-dimensional images of a person’s face and, as a result, have significant limitations in use.

Most of them solve the problem of interlinking two images of the same subject registered by different readers, or by the same reader, but at different times. In order to compare two such images, it is necessary to interconnect these images and thus correct relative spatial shifts, differences in video signal amplification, displacements caused by rotations, as well as geometric and brightness distortions of each image.

Typically, patents provide methods and / or devices and / or systems using 3D facial images that solve the problem of adjusting spatial rotations.

In the claimed system, the block for analyzing the quality of three-dimensional face images is a system for understanding images represented by arrays of numbers, which in their simplest form can simply report with some level of reliability (probability) that the image has a given or some unexpected object that characterizes quantify quality and facial features. This problem in the system is solved using a standard system for recognizing two-dimensional images, a block diagram of which is shown in FIG. 10.

However, in the case of registering facial images, the task of choosing the best two-dimensional image and quantifying its quality has so far been solved through expert analysis of two-dimensional images and required the direct participation of an expert operator who subjectively assessed the image quality by comparing it with other images included in the group under consideration.

In this case, methods of verbal description were used, which, as a rule, describe:

- general anatomical facial features of the subject of registration (top to bottom): forehead, eyebrows, ears, eyes, nose, lips, cheeks, wrinkles and chin;

- features characteristic of the subject of registration - that is, special signs: tattoos, scars, lack of parts of the face;

- related characteristics: hairstyle, clothes, toilet articles, jewelry;

- dynamic characteristics: facial expressions.

Typically, all facial features are quantitatively described using a three-level gradation scale (for example, narrow, normal, wide), or five-level (very narrow, narrow, normal, wide, very wide) with known names (for example, a humpback nose, etc. )

A common expert rating scale was the error scale, which measures the degree of distortion of face details from “invisible” to “extremely undesirable” and allows you to get an average confidence score. The disadvantages of the method of scoring include, first of all, the possible non-linearity of the scale for different parts of the face and the need to attract at least twelve experts to obtain a reliable result.

On the other hand, in the case of a large set of face quality parameters, the visual system of the expert operator is simply not able to perceive information about all the details of the face, since it is known that the perception of information by a person is about 50 movements per second, and conventional video systems can generate about 50 million movements every second .

The reason also lies in the fact that the methods available prior to the invention for automatically extracting face details and classifying quality parameters can only evaluate with low accuracy the quality of only two-dimensional face images obtained by fulfilling strict requirements for a specific scale and location of the face with respect to the film set, lighting, foreshortening, facial expressions etc., which in real systems cannot be performed.

Thus, the problem to which the claimed invention is directed is to create a reliable and quick way to highlight the details (features) of a person and quantify their quality with an indication of the level of reliability.

Therefore, such a method for registering biometric face data is necessary, which will provide real-time registration and reliable quantitative assessment of the quality of two-dimensional images of a person’s face in accordance with the requirements of relevant biometric standards. The new method should also ensure a low level of errors caused by the operator’s misunderstanding of the requirements of the registration procedure and, as a consequence, their incorrect execution.

The technical result is achieved through the development of a registration method and reliable quantitative assessment of the quality of two-dimensional (2D) images of a person’s face (SRDOK method), according to which a three-dimensional (3D) model of a face constructed using an appropriate 3D image reader is used.

Next, we will consider in more detail the main advantages of the new registration method of the SRDOK:

- (a) The operator’s participation in the registration process and quantitative assessment of the quality of two-dimensional images of a person’s face is minimal and limited by instructing the subject on the registration procedure and the requirements for positioning the face in front of the reader of three-dimensional images.

Thus, the new registration method can operate in a semi-automatic mode and is characterized by a low level of operator errors in quantifying the quality of 2D face images caused by the "human factor".

In addition, thanks to the new SRDOK method, a low-skilled operator can register, which significantly reduces the cost of operation and opens up wider possibilities for using the system.

The SRDOK method significantly reduces the duration and number of registration iterations and, in general, improves the quality of registered 2D face images;

- (b) accurate automatic quantitative assessment of the quality of 2D images of a person’s face is carried out by a block for analyzing the quality of 3D face images and a block for generating 2D face images from a set of three-dimensional face images and their quantitative quality estimates.

This allows you to automatically provide analysis and reliable quantitative assessment of the quality of 2D facial images and ensure the registration of the most complete information about facial features;

- (c) high quality registration 2D images of a person’s face is achieved due to the possibility of using a reader of 3D images;

- (d) the existing possibility of using an additional lighting unit allows one to achieve the registration of high-quality video images of facial features relatively regardless of the location of the registration system and the influence of most external light sources;

- (f) the registration of biometric 2D face images occurs in strict accordance with the requirements for quality. The quality parameters of 2D facial images can be reliably quantified and, as a result, provide flexible criteria for deciding on the registration or re-registration of 2D facial images in accordance with the requirements of relevant standards. This makes it possible to register biometric face data of a single format, including generally accepted standards for registering and storing biometric two-dimensional images of a person’s face on electronic information carriers. This functionality of the new registration method SRDOK ensures the compatibility of registration data with most third-party human recognition systems and increases its demand in the market;

- (f) a new registration method for SRDOK detects deviations in the quality of recorded 2D images of a person’s face caused by different behavior of the registration subject (variations in the positioning of the subject’s head in front of reading devices, turns of the head during registration, etc.), various conditions of the recording medium (variations in lighting, the presence of foreign objects in the frame, etc.) and the general malfunction of the registration system;

- (g) in cases where the registration subject, due to physical disabilities or age (child, old man), cannot register on his own, the registration of 2D face images is carried out with the help of an operator with wider authority, and a reliable quantitative assessment of the quality of registration data is carried out still in automatic mode (without operator intervention);

- (h) a new method for registering 2D facial images can use video data, thereby increasing the accuracy and reliability of quantitative quality assessments;

- (i) a new method for registering 2D face images allows to reduce the effect of noise due to the use of multidirectional stereo readers of 3D face images (for example, several stereo video cameras);

- (k) a new registration method for SRDOK allows you to restore and / or improve the quality of two-dimensional face images by analyzing the corresponding associated data of three-dimensional face images obtained from independent video channels for reading face data;

- (l) a new registration method for SRDOK allows creating systems for recording three-dimensional and two-dimensional face images simultaneously with low cost;

- (m) a new method for recording SRDOK allows you to register 2D video data of a person with his natural behavior;

- (n) a new method for registering SRDOK provides a preliminary in real time check of the quality and completeness of the read 3D video data;

- (o) a new method for registering SRDOK can be implemented in any standard computer system that contains at least stereo image readers, a display, a control unit and a memory with the possibility of software implementation of a set of instructions;

- (p) a new method for registering SRDOK ensures maximum accuracy of subsequent recognition of a person from a two-dimensional image of a face in various applications and even in relatively unfavorable conditions of the recognition environment: poor lighting, lack of lighting, etc.

Based on the foregoing, it can be argued that the new method for recording and quantifying the quality of two-dimensional images of a person’s face using scanning and analysis of three-dimensional images has extremely wide possibilities for use in most biometric systems.

The inventive method consists of the operations of reading three-dimensional images of the face, the selection and quantitative assessment of the quality of the face and its details with a given level of reliability, the subsequent formation of two-dimensional images of the face and their quantitative assessments of quality from three-dimensional images of the face and their quantitative assessments of quality, respectively, the final compression of two-dimensional images of the face and memory records as reference.

A stepwise method can be represented as follows:

- The subject of registration presents his PIN number to the PIN number reader, which is then read by the PIN number reader and decoded by the PIN number decoding unit, displayed on the display and checked by the processor for presence in the lists of authorized PIN numbers and for the absence of prohibited PIN numbers in the lists that are stored in the memory block.

After a successful check by the processor unit, the processor unit includes an additional lighting unit, three-dimensional face images are read by the three-dimensional face image reader, checked by the three-dimensional face image quality analysis unit for compliance with specified acceptable quantitative estimates of the three-dimensional face image quality parameters, and after successful verification, it is converted by the two-dimensional face image generation unit and their quantitative assessments of quality from three-dimensional images of the face and their number GOVERNMENTAL quality assessments set of two-dimensional images from the respective quantitative estimates of quality parameters dimensional facial images, the face is compressed two-dimensional image compression unit and the processing unit is written into the memory unit as a reference for the two-dimensional face images with quality attributes.

In contrast to existing methods and systems known from the prior art, a new method for recording and quantifying the quality of two-dimensional images of a person’s face provides the following main advantages:

- can work with minimal participation of the operator controlling the registration process;

- reduces the likelihood of re-registration due to more stringent requirements in the exact positioning of the face in front of the reader of three-dimensional images;

- excludes the possibility of registration using the submitted photograph of a person in full size;

- accurately measures the geometric dimensions of face details;

- accurately measures the distance to the face;

- accurately determines the orientation of the head in space;

- accurately measures the shape (relief) of the surface of the face;

- provides high quality registration data through the use of higher-quality two-dimensional face images formed from three-dimensional face images created using a three-dimensional face image reader;

- provides registration of biometric data that meets specified standards;

- provides a reliable quantitative assessment of the quality of two-dimensional face images.

The invention is illustrated by the example of its embodiment in the corresponding system, which is accompanied by explanatory drawings.

The problem is solved in that in the system (FIG. 1) comprising a PIN number reader 1 (for example, a contactless smart card reader), PIN number decoding unit 2 (for example, a data decoder based on PKI technology), a memory unit 4, a display 5 (for example, a computer monitor), a face image reading unit 6, a processor unit 3 (for example, a digital signal processor) interconnected with said units and connected to an additional lighting unit 7 for one of the outputs (for example, two primary light sources, three secondary - two ra screening umbrella and background screen), the face image reading unit 6 is made in the form of a three-dimensional image reader (for example, a stereo video camera), while an additional unit for analyzing the quality and features of three-dimensional face images, a two-dimensional image forming unit 9, and a two-dimensional face image compression unit 10 are additionally introduced interconnected with the processor unit 3.

The system operates as follows (Figure 2):

- upon registration, the subject presents a personal PIN number to the PIN number reading unit;

- the presented PIN number is read using the PIN number reader and decoded using the PIN number decoding unit, displaying the number on the display;

- the decoded PIN number is checked by the processor for presence in the lists of allowed PIN numbers and for the absence in the lists of prohibited PIN numbers that are stored in the memory unit;

- after a successful verification by the processor unit, the PIN numbers include the additional lighting unit using the processor unit;

- three-dimensional images of the face are read using a three-dimensional image reading unit;

- using the block for analyzing the quality of three-dimensional face images, quantitative assessments of the quality and features of the read three-dimensional face image are calculated and checked for compliance with the specified restrictions on quantitative estimates of the quality parameters and features of three-dimensional face images stored in the memory unit;

- after a successful verification of the read three-dimensional image of the face by the block of analysis of the quality of three-dimensional images of the face, it is converted using the block forming two-dimensional images from a three-dimensional image into a set of two-dimensional images. Moreover, using the block for generating two-dimensional face images, quantitative parameters of quality and facial features of each obtained two-dimensional face image are determined using quantitative estimates of the quality parameters and facial features of three-dimensional face images;

- thus, the generated two-dimensional image of the face is compressed further by the compression unit of two-dimensional images of the face and recorded together with quantitative assessments of quality and facial features using the processor unit in the memory unit as registered reference two-dimensional images of the face.

Partial essential features of the invention also contribute to the solution of the problem.

The PIN number reader is capable of automatically reading PIN numbers, including virtual PIN numbers (what the registration subject knows) entered manually from the keyboard and physical personal smart cards (what the registration subject has) presented by the registration subject .

Figure 3 shows a new registration system, which is shown from the side (view A) and rear (view B).

The subject of registration (17 and 28) is positioned in front of the display (12 and 24) in an armchair (18 and 29) with adjustable height.

The display (12 and 24) is made in the form of a computer monitor mounted on a tripod (11 and 19).

In addition to the display on a tripod (11 and 19), a three-dimensional face image reader is installed, presented in the form of a stereo video camera (13 and 23), mounted on an arm (22) with the ability to adjust the height and direction of reading.

The additional lighting unit is made in the form of two light sources (16 and 27), two scattering umbrellas (15 and 26) mounted on separate tripods, and one background screen located behind the subject of registration.

Computer module (20) includes the following blocks of the registration system:

- block 1 (Figure 1) read PIN numbers (virtual PIN number is entered through the keyboard by the operator);

- block 2 (Figure 1) decoding PIN numbers;

- processor unit 3 (Figure 1), implemented on the motherboard;

- block 4 (Figure 1) memory (hard drive);

- block 8 (Figure 1) analysis of the quality of three-dimensional images of the face (software module);

- block 9 (Figure 1) of the formation of two-dimensional images and assessments of quality and facial features (software module);

- block 10 (Figure 1) compression of two-dimensional images of the face (software module).

As a reader of three-dimensional facial images, the system may include a stereo video camera (Figs. 4; 36 and 42) of the STH-MDCS / -C Stereo Head model (manufacturer: Videre Design; website: www.videredesign.com), which is equipped with two independent video channels (30 and 31) for forming a stereo pair and LED indicator (32) for easy operation.

The video camera is located next to the display (33 and 39) and is fixed with a special screw (37 and 43) on a transparent stand (38 and 44) with the ability to adjust the distance to the subject of registration.

The display (33 and 39) displays visual commands (35 and 41), in accordance with which the subject is registered.

To determine the required position of the face relative to the stereo camera (Figure 5), the set of visual commands displayed on the display includes the corresponding visual commands for positioning the head of the registration subject in front of the display, according to which the registration subject can see a stereo pair on the display - two video images of his face.

At the same time, visual commands for positioning the head are displayed on the left video image:

- red frame for centering the head in the frame;

- two blue lines for positioning the eyes.

The position of the subject’s head in front of the display is considered correct when the face is located in the center of the red frame and the two blue lines fall on the centers or between the centers of the eyes.

Figure 6 shows the basic installation diagram of some modules of the registration system:

- a display for displaying visual commands for positioning the head (45);

- stereo video cameras (46);

- stereo microphone (47)

- places of target positioning (chair) of the registration subject (48);

- a background screen with a certain texture and reflective characteristics (49; refers to the additional lighting unit);

- sources of additional lighting with scattering umbrellas (50; refer to the additional lighting unit): left (51) and right (52).

In the new registration system, the block for analyzing the quality of three-dimensional face images is a system for understanding images represented by arrays of numbers, which in their simplest form can simply report with some level of certainty (probability) that the image has a given or some unexpected object, quantifying quality / facial features.

This problem in the system is solved using a standard system for recognizing two-dimensional images, a block diagram of which is shown in Fig.7. A two-dimensional image (54) from the space of two-dimensional images (53) is analyzed by the feature extraction unit (55) using data from the feature space (56). The result of this analysis is the identification of two-dimensional features (57), which can then be assigned to a specific class (60) by the classifier (59) using data from the class space (58).

The use of a stereo camera in the system, which has two independent video channels, ensures high efficiency and reliability of the results of the unit for analyzing the quality of three-dimensional face images, the unit for generating two-dimensional face images and estimates of their quality and the unit for compressing two-dimensional face images.

The system allows to quantitatively evaluate the following quality parameters and facial features of a two-dimensional and / or three-dimensional face image, characterizing:

- anatomical features: type of race; forehead; eyebrow; nose lips the mouth; cheeks; chin; ears, etc .;

- facial features characteristic of the subject of registration: special signs; scars wrinkles; lack of parts of the face; the presence of strabismus, etc.

- specific conditions for registering a person: visibility of the iris and pupil; the presence of points; the degree of transparency of the lenses of glasses; interference due to reflections of light on the lenses of glasses; a frame covering some facial features; interference due to hair covering facial features; angles of deviation of the face from the frontal position; shoulder visibility; uniformity of light scattering on the face; the presence of shadows on the face or in the orbits; the presence of false images due to improper lighting; the background; level of focus on each part of the face; contrast of details of the face; brightness, hue and color saturation; the presence of non-natural colors; the presence of optical distortion of facial features through the detection of abnormal sizes of the face, eyes, eyebrows, nose, lips, mouth and chin; the presence of foreign objects covering some details of the face, etc .;

- short-term physical properties of the face: eye color; type of hairstyle; hair color; the presence and color of the beard and / or mustache; the presence of an open mouth; smiles with a closed or open mouth; the presence of raised eyebrows and the degree of frowning; line of sight; presence of noticeable areas of facial skin (for example, tattoos); completeness of the details of the face in the image: the presence of the forehead, eyes, eyebrows, nose, lips, mouth, cheeks, chin, ears, shoulders, etc., etc.

For shooting in the system, a stereo camera is used, consisting of two video cameras, an additional lighting unit, including two primary light sources, diffusing umbrellas and a background screen of a certain texture.

This arrangement allows you to obtain the following properties of facial images: the background in the image is significantly different from the subject (face and figure of a person); the contrast is significant even with fairly intense lighting.

The presence of two independent video channels, produced two-dimensional video images of faces from different angles, allows you to effectively select and evaluate the above facial features in the image.

In this case, the following features can be used (see W. Prett "Digital image processing", M .: Mir, 1992) [69]:

- brightness;

- bar graphs;

- spatial spectral;

- contour;

- linear and nonlinear contrasting;

- approximation of brightness differences;

- statistical detection of differences;

- color contours;

- spots and lines;

- textural;

- synthesis of textures,

and the following methods of symbolic description of images:

- determination of geometric relationships and connectivity between elements;

- their compression and refinement to reduce to smaller areas;

- description of lines and shapes through approximation and machine analysis of metric, topological and analytical characteristics;

- segmentation by brightness with the help of a threshold restriction, increasing areas;

- contour segmentation: approximation of curves, tracing of contours, connecting differential points;

- texture segmentation;

- form segmentation;

- and etc.

Next, we will consider in detail the allocation of the head in the image using the methods of segmentation by brightness.

The images from the left and right cameras are designated as functions of brightness from coordinates on the plane:

BL (x _L , y _L ) and BR (x _R , y _R ),

where x _L and y _L and x _R and _R are discrete coordinates on the left and right images, respectively.

Images in the system are obtained under normal stereo shooting conditions, i.e. when there is no y-disparity (the difference in the coordinates of the images of the same point along the OY axis is practically zero).

In this case, the disparity of shooting is determined by x-disparity.

In a system, a stereo camera is characterized by the following properties:

- the focal planes of the cameras coincide (i.e., their optical axes are parallel);

- video cameras are equally oriented relative to the optical axes (i.e., can be transferred one into another by plane-parallel transfer);

- the projection of the line connecting the cameras on the images received from the cameras is parallel to the OX axis of these images.

The cameras are spaced apart by distance B, called a stereo base (or simply a basis).

The operation of the system is based on a number of a priori information about the details of the human face as objects of shooting and quantitative parameters of their quality, which make it possible to use distinctive methods for searching for details of the face and assessments of their quality. These include, first of all, physical restrictions on the size obtained by repeated measurements, as well as the fact that a number of details of the face have shapes close to geometric shapes (for example, a full-face image, eye sockets, eyes, pupils, eyeglass lenses, etc. . have shapes close to elliptic). In general, knowledge about the scene of a face image can vary from very simple (for example, the object being searched should be brighter than the background) to very complex scenes (for example, the ordered structure of relations between all parts of the face is known). For example, two large elliptical areas in the eye area can be interpreted as glasses.

In the general case, the selection procedure and quantitative assessment of the quality of face parts include the independent selection of various parts on individual images and subsequent adjustment using a joint analysis of the areas of face details obtained on different images of the stereo pair at the same time. In addition, the obtained quantitative assessments of the quality of the details of the face are accompanied by calculations of the level of reliability of the estimates obtained.

Let us first consider the procedure for highlighting face details using the example of highlighting a face area in a separate image.

It is believed that each area of any part of the face can be associated with a related area, all points of which will have a brightness exceeding a certain threshold value, and which will satisfy the specified criteria of brightness, shape and size.

The main task of identifying and assessing quality is to accurately determine the boundaries of the associated area.

In the case of highlighting the face, it is necessary to find, as a rule, the brightest object in the image (Fig. 8). However, filtering by brightness can lead to errors: the face is part of the associated area (Figa); the face is several connected areas (Fig. 8b), the face is not the largest bright object (Fig. 8c).

Therefore, to effectively determine the location of some parts of the face, their proximity in shape and location to certain geometric shapes and areas is used.

All previously related areas received are sorted by size. Further, when selecting a face, starting with the largest one, the elliptical part is searched until it is successful or there are no significant areas of connected areas (in some cases, at least 10% of the size of the maximum area).

The search algorithm for the elliptical part of the connected area works from top to bottom (Fig.9).

For each value of y, an ellipse is constructed that approximates the scanned part of the probable area of the face.

For each ellipse, the numerical characteristic of the ellipticity of the region is calculated by the following formula:

Q = (S _o / S _e ) {1- (P _c / P _e )},

where S _o is the area of the part of the region enclosed in an ellipse; S _e is the area of the ellipse; P _with - the length of the boundary of the ellipse intersecting with the region; P _e - the perimeter of the ellipse (Figure 10).

If the ellipticity Q of the region under consideration exceeds a certain threshold Q _T , then it will be considered an approximation of the face region.

We note separately that other extremal functionals can also be used as Q.

However, certain parts of the face may be dimly lit or have a low albedo. In this case, they may not fall into the elliptical part of the bound region. For example, in FIG. 8 (c), the eyebrows and eye region did not fall into the associated region. In addition, the image of the face can be divided into several areas (Fig.8, b).

To solve these and other problems, methods of simultaneous "synchronization" of elliptical regions in two stereo image channels are used, i.e. the correlation of regions over two channels is estimated. In this case, preference is given respectively to those areas that were correctly selected in two channels simultaneously.

Next, two stereo image channels are used to reliably estimate the size of face details. This is especially important when you consider that the values of most of the quality parameters of face parts are calculated using their physical dimensions (sizes of glasses, frames, faces, hairstyles, beards, mustaches, etc.).

The physical dimensions of captured objects can be measured using a single camera, the optical parameters of which are known, i.e. determined distance to the subject, focal length of the camera, etc.

However, in real conditions, strong variations are possible in the shooting distance and, therefore, in the frame width in the subject plane.

Using the second camera of the system’s stereo system, it is possible to indirectly measure the distance or frame width from a change in angle due to a change in the observation point.

Using the specified stereo base B (the distance between the optical axes of the cameras), you can set some standard distance L ₀ and by measuring the standard disparity Δx ₀ and the frame width F _{0 on it} , you can calculate the relative image scale S (Fig. 11).

If the standard Δx ₀ and current Δx disparity are measured, then the image scale, defined as S = F / F0 = L0 / L, is calculated by the formula:

The stereo base B is measured in millimeters, and Δx ₀ and Δx are converted from pixels to physical units by the formula:

where Δx _p is the standard disparity in pixels, and F _p is the frame width in pixels.

For objects in the left and right channels, restrictions on the maximum and minimum width and height of the specified objects are checked. If at least one restriction is violated, the brightness threshold is adjusted: increases to cut off excess parts when the minimum size restriction is violated and decreases to capture a larger area when the minimum size restriction is violated. That is, with an increase in the threshold, fewer points fall into the connected areas, and with a decrease, more.

Thus, the threshold is iteratively tuned until the resulting regions satisfy physical constraints.

The above work of the system for extracting face details and quantifying the quality using segmentation methods by brightness and subsequent estimation of the size of face details in the presented stereo images shows excellent results.

Its main advantage is high speed and real-time operation, i.e. the system with sufficient reliability can process video signals, rather than individual images.

Efficiency and correctness of the system operation depends on the accuracy of stereo adjustment.

The reliability of the quantitative assessment of the quality parameters will depend on the duration of the registration session or, in other words, on the number of independent stereo pairs of images in which the area is correctly selected and the physical dimensions have been verified.

We estimate the number of attempts required to obtain reliable data.

If in a series of n attempts the system will receive m occurrences of events of correctly selecting and checking the size of the region in both video channels, then the frequency will be p ^∗ = m / n.

In this case, the probability of detecting a given face detail of a certain person will be p≈p ^∗ .

Let us estimate the smallest number of independent attempts to isolate and verify the size of the region in both video channels, necessary so that with a probability of at least β the frequency p ^∗ deviates from the probability p by no more than α.

In the general case, n is determined from the inequality:

We replace the probability in the left-hand side of the inequality with the approximate Moavre – Laplace integral.

As a result, to determine n, we obtain the inequality:

where Φ {x} is the Laplace integral.

For the worst case, when p ^∗ = 1-p ^∗ = 0.5, for n = 400 and α = 0.05, we have;

This means that with a 95% probability, the difference between the obtained result and the real one does not exceed 5% provided that at least 400 attempts are made to isolate and verify the size of the region in both video channels.

For a stereo video camera of the system, the speed of the processed video stream of each channel is 15 frames per second. That is, to obtain quantitative assessments of quality with the specified reliability, it is necessary to process at least 400 frames or determine the duration of registration at the level of 30 seconds.

The system allows you to increase performance up to 30 frames per second. In this case, for a given reliability, the registration duration will be 15 seconds.

Similar to face extraction, to effectively determine the location of some other details of the face, they also use their proximity in shape to certain geometric shapes and areas, as well as information about the likely location in the image with reference to some reference points (for example, eye centers), necessary to adjust relative spatial shifts, differences in displacement caused by rotation, as well as geometric distortion of the brightness of each area of the image.

As already mentioned above, in addition to highlighting such signs as brightness, the coordinates of contour points, texture parameters, etc. are also highlighted in the image. Contour points are grouped into related areas: line segments or closed curves. In addition, elements with the same texture can also be combined into related areas of image elements.

Using the new system, it is possible not only to isolate and evaluate the reliability of the presence of a face in the image, but also to detect and quantify the following face details: physical dimensions and type of face (for example, race), forehead, eyes (for example, amygdala), eyebrows (for example , arched averages in length and in relation to the eyes), nose (for example, middle, nose tip down), lips (for example, upper lip height middle), mouth (middle), cheeks, chin (height, distance between the lower border of the lip and tip of chin, middle), ears, oso signs, scars, wrinkles, lack of parts of the face, strabismus, visibility of the iris and pupil; type of hairstyle; the presence of points; eyeglass lens transparency; shoulder visibility; the presence of a beard and / or mustache; the presence of an open mouth; smiles with a closed or open mouth; the presence of raised eyebrows and the degree of frowning; line of sight; the presence of strabismus; presence of noticeable areas of facial skin (for example, tattoos); the completeness of the details of the face in the image: the presence of the forehead, eyes, eyebrows, nose, lips, mouth, cheeks, chin, ears, shoulders; Background texture and etc.

In addition, a similar analysis of the appearance of unacceptable connected areas, contour lines, segments and texture parameters allows us to evaluate random abnormal details of the face: the presence of a frame that covers some facial features; the presence of light reflections on the lenses of glasses; interference due to hair covering facial features; uniformity of light scattering on the face; the presence of shadows on the face or in the orbits; the presence of false images due to improper lighting; poor focus on every part of the face; contrast of details of the face; the presence of optical distortion of facial features through the detection of abnormal sizes of the face, eyes, eyebrows, nose, lips, mouth and chin; the presence of foreign objects covering some details of the face; the presence of abnormal areas of facial skin, etc.

The reliability of all estimates is also ensured by the consistency of the selected areas in the images obtained from several independent sources (in our case, these are stereo camera video channels).

Assessment of the brightness, color tone and saturation of the color of the skin, hair of a hairstyle, beard and / or mustache, the presence of non-natural (false) colors on the skin of the face, eyes and hair is carried out using information on the conversion of color coordinates of the original image of the face obtained using color characteristics background screen (color, texture), matched in two video channels of the stereo camera simultaneously.

When evaluating the angles of deviation of the face from the frontal position (spatial orientation of the head), consistent data obtained from two video channels of the stereo camera simultaneously are also used:

- key points, such as the centers of the eyes or pupils, the center of the lips, etc., are located on the face image using the aforementioned mono image processing methods.

- further, for clarification, the spatial position of the points is calculated using all stereo image channels simultaneously;

- the head position is considered to be frontal if the triangle formed in space by the points of the centers of the eyes and lips (supporting triangle) lies in a plane parallel to the image plane, and the axis of symmetry is directed vertically.

A plane in three-dimensional space is given by the equation:

whose coefficients are calculated by the coordinates of the vertices of a triangle lying in this plane;

- the angles of the spatial orientation of the head are determined by the angles of the spatial orientation of the supporting triangle.

The slope (pitch - φ) and rotation (yaw - ψ) of the face are calculated from the equation of the plane of the supporting triangle according to the following formulas:

The lateral tilt (roll - χ) of the face is calculated by the tilt of the axis of the eyes:

or along the slope of the axis of symmetry of the reference triangle.

Other characteristic points of the face whose spatial position can be determined are used to refine these angles.

The task of selecting the best two-dimensional face image is also solved by using consistent data received from two video channels of the stereo camera simultaneously.

Indeed, a certain short-term static two-dimensional result of face visualization, selected by the operator as a reference, may not be the best, because, as noted above, it may contain false details (interference) due to the conditions of face registration varying from measurement to measurement.

All obtained estimates of the quality parameters of three-dimensional images of a face can be used to select reference three-dimensional images that will satisfy the specified restrictions in accordance with certain standards that describe: the format for storing facial data, transferring information about facial features, texture and background color, requirements for taking photographs , other attributes of a digital face image.

The unit for generating two-dimensional face images selects from the set of three-dimensional face images such an image, the quantitative quality parameters of which will be close to the arithmetic average of all similar quality parameters of the corresponding images, coordinated in two video channels of the stereo camera simultaneously.

Also, using the block for generating two-dimensional face images, quantitative parameters of quality and facial features of each obtained two-dimensional face image are determined using quantitative estimates of the quality parameters and facial features of three-dimensional face images.

Stereopairs can be used as two-dimensional face images in the system.

In conclusion, I would like to note that the advantages of the proposed method for registering two-dimensional face images and a reliable quantitative assessment of the quality and facial features over the known ones are manifested in the fact that this technical solution does not require a large number of face image readers, is characterized by registration requirements that are easy and understandable for the subject, minimizes the duration of registration, bringing it almost to the real time mode, eliminates the influence of interference caused by various conditions of stration is not error-prone, has increased resources to ensure adequate registration of two-dimensional images of the face and a quantitative assessment of the quality and features of a person with a given level of confidence is consistent with existing and emerging standards.

Although the above embodiment of the invention has been set forth to illustrate the present invention, it is clear to those skilled in the art that various modifications, additions and substitutions are possible without departing from the scope and meaning of the present invention disclosed in the attached claims.

The proposed method and system is illustrated by drawings 1-11, which also illustrates the possible appearance of the system:

- Figure 1 - composition of the registration system and quantitative assessment of the quality of two-dimensional images of a person's face;

- Figure 2 is a brief block diagram of the operation of the registration system and quantitative assessment of the quality of two-dimensional images of a person's face;

- Figure 3 - view A. Registration system. Side view;

- Fig. 3A is a view B. Registration system. Rear side view;

- Figure 4 - stereo camera installed next to the display;

- Figure 5 - positioning of the face during stereo shooting;

- Fig.6 is a diagram of the installation of the registration system;

- Fig.7 is a block diagram of a standard pattern recognition system;

- Fig. 8 - related areas of the face selection (filtering according to the brightness threshold, black color - maximum associated area);

- Fig.9 - search for the elliptical part through the step-by-step construction of ellipses from the upper boundary;

- Figure 10 - assessment of the ellipticity of the associated area;

- 11 - the geometric parameters of stereo.

Claims (14)

1. A system for recording and quantifying quality parameters and facial features of two-dimensional facial images, which contains:
a PIN number reading unit associated with a PIN number decoding unit;
a memory unit that stores lists of allowed and prohibited PIN numbers, as well as compressed two-dimensional face images with information about image quality;
display;
a face image reading unit in the form of a three-dimensional image reader coupled to a three-dimensional face image quality analysis unit, the three-dimensional face image quality analysis unit quantifying the quality and features of the three-dimensional face image and checking for compliance with predetermined restrictions;
a block for generating two-dimensional face images associated with a block for analyzing the quality of three-dimensional face images, the block for generating two-dimensional face images converts a set of three-dimensional images into a set of two-dimensional images, and also determines quantitative parameters of quality and facial features for each two-dimensional image using quantitative estimates of three-dimensional images persons;
a two-dimensional image compression unit associated with a memory unit;
additional lighting unit;
a processor unit performing PIN number verification, recording in the memory block of compressed two-dimensional face images and the inclusion of an additional lighting unit, interconnected with all of the above blocks. 2. The system according to claim 1, characterized in that the system comprises a stereo camera as a block for reading three-dimensional face images. 3. The system according to claim 1, characterized in that, as a reader of three-dimensional images of the face, the system comprises a stereo video camera with a variable basis. 4. The system according to claim 1, characterized in that, as a PIN number reading unit, the system comprises a contactless radio frequency reader. 5. The system according to claim 1, characterized in that the system comprises a contactless smart card reader as a PIN number reader. 6. The system according to claim 1, characterized in that, as a block for decoding a PIN number, the system comprises a data decoder based on PKI technology. 7. The system according to claim 1, characterized in that as the display system contains a computer monitor. 8. The system according to claim 1, characterized in that as a processor unit, the system comprises a digital signal processor. 9. The system according to claim 1, characterized in that, as an additional lighting unit, the system contains two primary light sources and three secondary, including two diffusing umbrellas and one background screen. 10. A method for registering two-dimensional images of a face, consisting of the following:
a) set the acceptable range of quantitative parameters of quality and facial features in the block of analysis of the quality of three-dimensional facial images;
b) positioning the head of the subject of registration in front of the image reading unit;
c) scan using a face image reader a set A of three-dimensional face images of various quality;
d) evaluate with the help of a block for analyzing the quality of three-dimensional face images, quantitative parameters of quality and facial features of each three-dimensional face image from set A;
e) create from set A a second set B of three-dimensional face images having quantitative estimates of quality parameters and facial features within a given acceptable range;
f) using the block for generating two-dimensional face images, a third set C of two-dimensional face images formed from the second set B of three-dimensional face images is created;
g) using the block for generating two-dimensional face images, quantitative parameters of quality and facial features of each obtained two-dimensional face image from the third set C are determined using quantitative estimates of the quality parameters and facial features of three-dimensional face images from set B;
h) compress each two-dimensional image of the face from the third set C and save it in the memory unit together with certain quantitative estimates of the quality parameters and facial features for a given image as a registered reference two-dimensional image of the face. 11. The method according to claim 10, characterized in that they assess the quality parameters and facial features, while including assessing the presence, position, size and type of face. 12. The method according to claim 10, characterized in that assess the quality parameters and facial features, while including assessing the completeness of the details of the face in the image. 13. The method according to claim 10, characterized in that assess the quality parameters and facial features, while including estimates: type of race; - the presence, position, size and type of forehead; the presence, position, size and type of eyes; the presence, position, size and type of eyebrows; the presence, position, size and type of nose; presence, position, size and type of lips; the presence, position, size and type of mouth; the presence, position, size and type of cheeks; presence, position, size and type of chin; the presence, position, size and type of ears; specific features of the registration subject; characteristic of the subject of registration of scars; wrinkles characteristic of the subject of registration; characteristic of the subject of registration of the absence of certain parts of the face; strabismus and other pathologies characteristic of the subject of registration; visibility of the iris and pupil due to the specific registration conditions of the face; the presence of points; the degree of transparency of the lenses of glasses; interference due to reflections of light on the lenses of glasses; a frame covering some facial features; interference due to hair covering facial features; angles of deviation of the face from the frontal position; shoulder visibility; uniformity of light scattering on the face; the presence of shadows on the face or in the orbits; the presence of false images due to improper lighting; background; level of focus on each part of the face; Contrast details of the face; brightness, hue and color saturation; the presence of unnatural flowers; the presence of optical distortion of facial features; the presence of foreign objects covering some parts of the face; assessment of eye color; type of hairstyle; hair color; the presence of a beard and / or mustache; the colors of the beard and / or mustache; the presence of an open mouth; smiles with a closed mouth and / or open mouth; the presence of raised eyebrows; degree of frowning; line of sight; the presence of noticeable areas of facial skin (for example, tattoos). 14. The method according to claim 10, characterized in that the images are scanned using a stereo camera.

Images