RU2541192C1 - Device for determination of characteristics of materials - Google Patents

Abstract

FIELD: measurement equipment.

SUBSTANCE: method of identification of line-of-sight consists in that two or more image detectors are positioned, and their positions are unknown in proper time, in the space area a reference fragment is placed which is an image or object with the known texture, colour and sizes, so that to make it visible for all image detectors simultaneously, pre-calibration using the named reference fragments is performed, or the named calibration is performed directly during measurements, simultaneously the images are received from all image detectors, for each obtained image the object area is localised with selection on it of local features, on each image the local features related to the reference fragment, a three-dimensional model of mutual position of local features with reference to each other is constructed, and also with reference to the image detectors, an actual spatial position of the object segments is determined using the results of calibration, the assessment of the line-of-sight is performed.

EFFECT: possibility of increase of amount of image detectors at absence of requirements to their mutual position, automatic system calibration in real time mode and improvement of measurement accuracy.

9 cl, 4 dwg

Description

The invention relates to a control and measuring technique and can be used to measure the direction of the line of sight in the process of performing cognitive tasks by a person, as well as to implement attention-sensitive interfaces in eye-brain-computer interfaces, in systems that communicate between people with impaired motor functions (print text with an eye using an interactive keyboard).

State of the art

The technique relates to eye movement recording systems. The processing of visual information has been studied in the framework of applied cognitive sciences for many decades [3, 4, 5, 9]. One of the research results was the discovery of the relationship between the dynamics of eye movements and the parameters of perception of visual information [4, 5, 6]. This opened up the possibility of creating fundamentally new methods for measuring visual attention and recognizing user intentions [4, 5, 8, 14]. The practical significance lies in the possibility of creating unique technologies that provide the interface of human interaction with technical systems through gaze. The scope of such systems is quite wide and may include such practical applications as remote control of autonomous robotic systems and technological processes while monitoring the state of the human operator.

Despite the considerable amount of accumulated knowledge in the field of cognitive sciences, the creation of devices that would fully realize the above possibilities presents a difficult technical problem. Many years of experience using traditional methods allowed us to formulate the basic requirements for such systems in general:

1. The possibility of non-contact, remote registration of the movement of the pupil.

2. No restrictions on the user's natural movements.

3. Using passive registration methods (without special highlighting or user training).

The latter requirement is due to the need in some cases to conduct long-term observations of a person, while the presence of backlighting imposes significant time limitations on such measurements (due to the accumulation of radiation dose). Partial achievement of the above requirements became possible only with the advent of relatively cheap image recording devices based on matrix photodetectors (CCD, CMOS). The functionality of such systems is fully provided by image processing algorithms. The composition of such systems includes the image registration device itself (one or several), as well as a computational module (in most commercial options it is a personal computer), which provides processing of the received images in real time. Image processing algorithms at the same time implement complex multi-stage mathematical methods that usually consistently search for the face and eyes of users in the image, search for the pupil, outline it, calculate the spatial position of the user and determine the line of sight. To date, several commercial systems are available that have the architecture described above [7, 8]. Despite significant advantages compared to traditional contact methods, such systems still have a number of significant disadvantages, the main of which are:

- The limited space in which registration is performed, and, as a result, the restriction on the user's natural movements;

- In some cases, the presence of light to calculate the spatial position of the user, which imposes restrictions on the measurement time (due to the accumulation of the radiation dose).

This solution offers a passive method of registering the directivity of the line of sight, and also makes it possible to increase the number of image registration devices with the possibility of their free arrangement. In this case, the system calibration can be performed automatically, in real time.

Overview of analogues

A technical solution is known [14], in which the object is illuminated by several radiation sources that form a light pattern on the surface of the object. As an object, the user's eye is considered. The images of the object containing the reflection of the light pattern are recorded. The object is detected by processing the obtained images containing the light pattern and the image of the object. The position of the pattern determines the spatial position of the object. The main disadvantage of the system is the presence of active illumination, as well as the registration of images in the spectral range of the emitter. In the system, it is necessary to combine the emitter with the field of view of the image recording device. The difference of the proposed solution is the use of passive methods based on registration and image processing, without any irradiation of the object. Instead of detecting reflected light patterns, the proposed method uses the detection of local features of an object belonging to its specific segments. The method is based on image processing without the use of additional sources. Also, the proposed system allows the free arrangement of the components included in its composition (image registration device), while in the above patent a strict position of the registration device and the emitter is necessary.

A solution is known [15] for eye movement registration systems in which an object (user’s face) is illuminated by several low-intensity sources of the near infrared range. Registration of images of an object illuminated by radiation and its segments is carried out. Localization of this object and its segments is performed by threshold image processing. The position of the object and its segments in the image is calculated. The invention does not describe methods for determining the spatial position of an object and its segments, as well as determining a line of sight. The difference between the invention [15] and the present is the method of localizing the object in the image. In the proposed method, localization is carried out by detecting local features inherent in the object and its segments. This method is passive and does not require additional exposure. Determining the spatial position of an object and its segments using [15] can be difficult due to the lack of the required number of local features (to solve this problem, at least five local features are needed).

Also known is the solution [8], in which the object is illuminated by radiation sources, where the radiation is formed into a pattern. The images of the object are registered with the subsequent processing of the image data. The purpose of this processing is to localize the object and its segments in the images (face and eyes), as well as localization in the images of the light pattern reflected from the object. The spatial position of the object is calculated by analyzing the orientation and dimensions of the reflected light pattern from the object and localized in the images. The line of sight is determined. The difference of the proposed method from [8] is the use of passive methods for determining the directivity of the line of sight. The spatial position of the object and its segments is determined by the methods of constructing a three-dimensional model (which includes the mutual spatial position of the local features of the object and its segments) from a series of images obtained from different angles (using epipolar geometry, the methods of which are given in [8]) . In this case, as in previous inventions, the method does not require a clear fixation of the position of image recording devices, whereas in active systems a fixed position of the emitter relative to the detector is required.

Also known is the solution [10], in which the registration of images of an object is carried out, processing of these images with subsequent localization of the object in the images without using active systems, as well as determining the line of sight. The main disadvantage of this solution is the registration of object images from one angle (using a single-camera system). This greatly limits the user's natural movements. The use of several recording devices in order to expand the registration area and remove restrictions on the user's natural movements in this system is not possible. The direction of the line of sight is given in coordinates tied directly to each image recording device. The recording area is determined solely by the field of view of the recording device itself. In the present invention, it is possible to use several devices that take an object from different angles. The determination of the direction of the line of sight is made in the coordinate system attached to all devices simultaneously. It is possible to expand the registration area by increasing the number of recording devices.

Another solution for determining the direction of the line of sight is the method proposed in [1]. In the work, the images of the object and its segments are recorded (the face and eyes of the user), the local features of the object and its segments are searched for on the images, and using the detected local features, the object and its segments are localized on each image. The spatial position of the object and its segments is determined by restoring a three-dimensional model of the object from a series of ordinary two-dimensional images taken from various angles (methods of epipolar geometry) [2].

The three-dimensional model contains the mutual spatial position of the detected local features relative to each other, as well as relative to image recording devices. The last step is determining the direction of the line of sight. The described method has one drawback. The constructed three-dimensional model contains information only on the relative relative position of local features relative to each other and relative to image recording devices. The real scale of the relative position of the features in the work [1] is determined by the previously known position of the recording devices, as well as the known data on the internal parameters of image recording devices (intrinsic parameters). In such a scheme, the system must be pre-configured in terms of camera placement. The displacement of their position during operation will lead to the impossibility of further calculations. In the present invention, the real scale is determined immediately before or in the process of the calculations themselves using the reference fragment, and information on the relative position of the image recording devices, as well as data on their internal parameters, is not required.

The way to determine the direction of the line of sight

Closest to the proposed method for determining the direction of the line of sight is the method described in [13]. In the invention, the line of sight is determined using at least three image recording devices. Similarly to the proposed solution, the object and its segments are registered, and the object segments are localized on each of the images. The spatial position of the segments is carried out by comparing the position of the segments of the object in at least three images. The direction of the line of sight is determined.

The described method has several disadvantages. Since only two points are used to build a three-dimensional model (which are used to select segments of the object), three images are needed to determine their spatial position. In this case, a necessary condition is to fix the position of the image registration devices relative to each other. In the proposed method, the fixed position of the image registration devices is not required, and their relative position is determined during the measurement process. Also, to determine the spatial position of the object and object segments for the proposed method, it is necessary to have only two image registration devices.

Thus, the technical problem to be solved within the framework of this method is to ensure the determination of the line of sight using several images obtained from several angles with an unknown position of the image recording devices themselves. At the same time, the space through which images are recorded can be expanded unlimitedly by increasing the number of image registration devices in the complete absence of any requirements for their mutual position. Calibration of such a system is carried out automatically, without operator intervention, before or directly in the measurement process. All calculations are carried out in real coordinates, which ensures accurate measurements even in case of a change in the relative position of the image recording devices. Thus, this leads to the expansion of the functionality of the method and increase its accuracy.

The technical problem is solved by the proposed set of essential features.

A method for determining the direction of the line of sight, which consists in the fact that at least two image recording devices (SIDs) are set arbitrarily, a SID system is calibrated using a reference fragment (RF), as a result of which the real-scale coefficient of a three-dimensional model is determined, and then an object different angles, find local features (LO) of the segments of the object, based on the received LOs from several images build a three-dimensional model that reflects the spatial position of the LO segments nt object relative to each other, as well as relative to the URI and containing information in relative coordinates, then, taking into account the calibration data, recalculate the positions of the specified LO from the relative coordinates of the three-dimensional model into real coordinates, project the LO of the object segments from the three-dimensional model onto the original images and outline the object segments on these two-dimensional images, they calculate the orientation of these contours in space and determine the direction of the line of sight by drawing a straight line, walking through the center of a circle forming a three-dimensional contour of an object segment and oriented along the normal to the three-dimensional plane in which this contour is located.

Wherein:

- to determine the real scale are used LO belonging to the Russian Federation;

- The position of the Russian Federation is fixed in space;

- the position of the Russian Federation is changed during the measurement process;

- The Russian Federation is located in the field of view of the URZ PZ during the measurement process;

- After calibration, the Russian Federation is placed periodically in the field of view of the URI PP;

- with each measurement taken, the mutual position of the URI is memorized, and the subsequent measurement results are amended to change the position of the URI by comparing the saved old position and the new one;

- as the RF, several different images are used, and calibration is performed according to any of these images;

- as the segments of the object using the iris and / or pupils of the eyes of the object /

When calculating the direction of a person’s gaze, the pupil area and the iris are contoured using methods [11, 12] and using the analysis of the projection of this contour, the direction of the line of sight is calculated, for example, similarly to [13], or using the methodology described in [19].

The method is disclosed in the detailed description below, with reference to the accompanying figures, where the accepted abbreviations:

URI - image registration device;

RF - reference fragment;

PZ - field of view;

LO - local features.

Figure 1 - source image of the object for the case with two URI cameras;

Figure 2 - detection of the face of an object on a given pair of images with the local features of the face LO highlighted on it and a comparison of these features for a given pair of images;

Figure 3 - A) an example image of a reference fragment. B) an example of isolating local features of a reference fragment from a pair of source images.

Figure 4 - The spatial position of local inhomogeneities relative to each other, where the circular markers denote the heterogeneities belonging to the object, which is considered the face of the user, or segments of the object, which are the eyes of the user (the result of processing shown in Figure 2), square markers are shown local features belonging to the reference fragment (the processing result is shown in Figure 3). P1 is the plane in the space in which local features of the face are located, P2 is the plane in the space in which local features of the reference fragment are located, α, β and µ are the angles characterizing the angular orientation of the plane of local features belonging to the user's face.

Figure 4, P1 - plane 1, in which local features that belong to the object (the user's face) are located, or if such local features in space do not lie in one plane, then P1 is constructed as a plane passing through a three-dimensional cloud of points formed from local features of the object and its segments, and passing in such a way that the sum of the distances from each point of the cloud to a given plane is minimal. Such a plane can be found, for example, by least squares methods or by factor analysis methods;

Figure 4, P2 - plane 2, in which there are local features that belong to the reference object.

Method Description

1. At least two image recording devices (SIDs) are installed, and their relative position is not known in advance. The SIDs are set so that each point in space covered by the SID system falls into the field of view (PP) of at least two SIDs at the same time. The intrinsic parameters for each URI are known in advance.

2. The system of separated URIs is calibrated before the start of measurements. The calibration task is to find the coefficient of the real scale of the three-dimensional model. Calibration is carried out by placing a reference fragment (RF) in the URZP, the texture, size and shape of which are known in advance. As such an RF, for example, any three-dimensional object or a flat image can be used. An example of such an image is shown in Figure 3, A. The image of the Russian Federation can be stored in the memory of the computing device. Also, instead of the image of the Russian Federation itself, it is possible to remember LOs belonging to this RF. It also remembers the distance in real coordinates between the LO data. This information can be stored every time before the start, but not during the calculation. Calibration is carried out by registration by all the URIs of the part of the space containing the Russian Federation, after which:

2.1. On each registered image, LOs belonging to the Russian Federation are distinguished. To do this, first retrieve all the LO belonging to this image. This operation can be performed using Harris corner detector methods [17] and / or methods described in [16] or [18] (LO methods SURF and / or SIFT). Detection of LOs belonging to the Russian Federation is carried out by comparing each LO detected in the image with the LO of the RF stored in the memory of the computing device. If the image of the RF itself is stored in memory, the LO is extracted from this image by the methods described above, after which a comparison is made.

2.2. Of all detected LOs belonging to the Russian Federation, those are found that are present on all images simultaneously. As a search method for such correspondences, k-dimensional trees (kd-trees) can be used [20]. An example of RF search in the image is shown in Figure 3, where A is the RF itself, and B is the result of its search for the case of two images.

2.3. Using data on the location of each LO in each image, calculate the relative position of the SID. The calculation of the relative position of the URI is possible with the help of eight, seven, and five correspondences (8, 7, 5 point algorithms) detected in step 2.2 [2].

2.4. To eliminate errors in the search for matches, step 2.3 can be used in conjunction with RANSAC algorithms, for example, its individual modifications from [21]. At this stage, false LOs are filtered out (LOs that were initially recognized as belonging to the Russian Federation, but are not actually such).

2.5. After finding the mutual position of the URI, a three-dimensional model is constructed that displays the spatial position of the LOs belonging to the Russian Federation relative to the URI. Such a model includes only those LOs that were detected on all images simultaneously (the so-called sparse 3D model [2]). The calculation of the spatial position of the LO is possible by the triangulation method described in [2].

The result of steps 2.2.-2.5. is a three-dimensional model containing information on the mutual spatial position of the RF LO relative to the recording chambers and relative to each other, but not reflecting the real coordinates of the LO (the location of the LO is determined to scale). Such a three-dimensional model is presented in Figure 4 - square markers. The model is a three-dimensional cloud of points. Each point in the three-dimensional model corresponds to the LO present on all images simultaneously. Since in the given example the RF is a flat image, all points of the three-dimensional model of the RF lie in the plane P2. For the case when some three-dimensional object is used as the RF, the spatial position of the LO repeats the relief of the RF.

2.6. The calculation of the real scale is carried out by the following method. Take any pair of LOs present on all images (Figure 3, B)) and included in the three-dimensional model. Find the real distance between these LOs using the information about the Russian Federation stored in memory (Figure 3, A)). Moreover, if the list of RF LOs is stored in memory, then at the same time, for each pair of LOs, the distance is also saved. For the case, if the image of the RF is stored in the memory, then the distance between the pair of LOs is calculated based on the known sizes of the RF. Select this pair of LO in a three-dimensional model and calculate the distance between them in the coordinates of this model. The conversion factor from the relative coordinates of the three-dimensional model into real coordinates is given as K = M / N, where K is the conversion factor (real-scale coefficient), M is the distance between the pair of LOs in the coordinates of the image of the Russian Federation in kind, N is the distance between the same a pair of LOs in the coordinates of a three-dimensional model.

3. It is possible to calibrate the system of placed URIs once before starting the measurements. In this case, the calibration data can be remembered, and later the RF itself can be removed from the field of view of the URI.

4. It is possible to calibrate the system of placed URIs directly in the process of measurements themselves after a fixed period of time. For this, it is necessary to keep the RF in the field of view of all cameras during the measurement process.

5. Make the measurements themselves. Register images of some part of the URI space. To do this, for the image obtained from each URI, they search for LOs related to segments of the object (the iris and / or pupils of the human eye). This search for LO can be carried out, for example, by Viola-Jones methods [22]. The method uses wavelet-like functions obtained from the integrated image and describing the LO of the object and its individual segments (face and eyes of the user). An example of the result of processing the original pair of images by this algorithm is shown in Figure 3.

6. Next, find the corresponding LOs that are present in at least two images. The search is carried out similarly to the methods from clause 2.2. Examples of the user's original images obtained from different angles of the URI are shown in Figure 1. The result of the detection of LOs present on all images simultaneously and calculated in accordance with the algorithm described in 2.2 is shown in Figure 2.

7. Next, the mutual position of the URI relative to the LO calculated in the previous step is calculated. The calculation is carried out by methods similar to 2.4-2.5.

8. Based on the results of paragraph 7, triangulation and construction of a three-dimensional model are performed, the elements of which are LOs that belong to the object (face) and its individual segments (eyes). The model is constructed by methods similar to those described in 2.5.

9. As a result of the described processing, a three-dimensional model is obtained, which reflects the mutual spatial position of the LO of the object segments (eyes) relative to each other and relative to the SID. Such a model, similarly to 2.6, is a cloud of points. An example of such a model for a pair of images from Figure 1 is shown in Figure 4 (marked with round markers).

10. In the case of preliminary calibration, using the coefficient of the real scale, recalculate the position of the LO segments of the object from the relative coordinates of the three-dimensional model in real. In this case, use the results of the relations from 2.6.

11. In the event that calibration is performed in real time (or periodically, at regular intervals), then the Russian Federation is in the URI. In the process of processing, the LOs belonging to the segments of the object (Figure 2), as well as those belonging to the Russian Federation (Figure 3) are extracted from the images, and processing 2.3-2.5 is performed for all LOs. This improves the stability of processing algorithms. The three-dimensional model for this case includes two point clouds (corresponding LO of the object, its segments, as well as the Russian Federation). Differentiation of LO points can be performed using data on the RF stored in the memory of the computing device (as images of the RF itself, or as a list of LOs belonging to it). An example is shown in Figure 4 (round and square markers). Similarly to the preliminary calibration, the coefficient of the real scale is calculated with the subsequent calculation of the real coordinates of the LO segments of the object. Based on this, the spatial position of the segments of the object (the iris or / and pupils of the user's eyes) is determined.

12. After calculating the real spatial coordinates of the LO segments of the object, find the projection of the LO data on the original two-dimensional image. Such a projection can be easily carried out using the results from clauses 2.3, 2.4. Outline the boundaries of the segments of the object (iris and pupil) in a two-dimensional image.

13. Using the projection of the segments of the objects, calculate the directivity of the line of sight. This can be done by finding the center of the contour of the boundaries of the segment of the object, calculating its spatial orientation [12, 13]. A similar method can be the method from [19].

14. Then, after constructing the three-dimensional contour of the segment of the object (the iris and / or pupils of the human eye), the directivity of the line of sight is found as a straight line passing through the center of the three-dimensional contour and oriented normal to the plane containing the contour of this segment.

15. Remember the mutual current location of the URI calculated in step 6 (using the methods described in 2.3) and then compare it with the previous one. In the event that these positions are different, a correction is made for the measurement results, adding a shift or rotation of the line of sight.

Information sources

1. Patent US 8339459 Zhang Z., Sankaranarayanan A., Zhang Q., Liu Z., Cai Q., Multi camera head pose tracking, Microsoft Corporation (Redmond, WA).

2. Hartley R.I., Zisserman A. Multiple View Geometry in Computer Vision [Text]. - Cambridge University Press, 2004.

3. Velichkovsky V.M., Cornelissen F., Geusebroek J.-M., Graupner S.-Th., Hari R., Marsman J. B., Shevchik SA, Pannasch S. Measurement-related issues in investigation of active vision [Text] // Measurement with persons: Theory and methods / B. Berglund, GB Rossi, J. Townsend, L. Pendrill (Eds.). - London-New York: Taylor and Francis, 2012 .-- P.281-300.

4. Pannasch S., Velichkovsky V.M. Distractor Effect and Saccade Amplitudes: Further Evidence on Different Modes of Processing in Free Exploration of Visual Images [Text] // Visual Cognition. - 2009 .-- 17 (6/7). - P.1109-1131.

5. Di Stasi LL, Renner R., Staehr P., Helmert JR, Velichkovsky B. M., Canas JJ, Catena A., Pannasch S. Saccadic peak velocity is sensitive to variations in mental workload in complex environments [Text] / / Aviation, Space, and Environmental Medicine. - 2010 .-- 81 (4). - P. 413-417.

6. Majaranta P., Raiha K.-J. Twenty Years of Eye Typing: Systems and Design Issues [Text] // Proceedings of ETRA 2002, Eye Tracking Research and Applications Symposium, 25-27, 2002.

7. SR Research Complete Eye Tracking Solutions [Electronic resource] / SR Research. - Electron, given. - Ontario, Canada, 2012. - Access mode: http://www.sr-research.com/free. - Zagl. from the screen. - Yaz. English

8. SMI Senso Motoric Instruments [Electronic resource] / SMI SensoMotoric Instruments. - Electron, given. - Teltow, Germany, 2013. - Access mode: http://www.smivision.com free. - Title. from the screen. - Yaz. English

9. Velichkovsky B.M., Sprenger A., Pomplun, M., Unema P. Eye movements and encoding manipulations in two visual memory tasks [Text] // Proceedings of the Psychonomic Society 38th Annual Meeting. November 20th-23rd - Philadelphia, USA. 1997 - P.73.

10. Kapur J.N., Sahoo P.K., Wong A.K. A new method for gray-level picture thresholding using the entropy of the histogram [Text] // Computer vision, graphics, and image processing - 1985 - 29 (3). - P.273-285, 1, 2.5, 2.6.

11. Davies E. Machine Vision: Theory, Algorithms and Practicalities [Text]. - Academic Press, 1990 - Chap. 6.

12. Li D., Winfield D., Parkhurst D.J. Starburst: A hybrid algorithm for video-based eye tracking combining feature-based and model-based approaches [Text] // Computer Vision and Pattern Recognition - Workshops, 2005. CVPR Workshops, IEEE Computer Society Conference on, 25-25 June 2005 - P.79

13. Method and device for determining the spatial position of the eyes for calculating the line of sight [Text]: RF patent No. 2444275 / Shevchik SA, Velichkovsky BM, Velichkovsky BB, Vartanov AV - Publ. 09/10/10.

14. Method and installation for detecting and following an eye and the gaze direction therefore [Text]: US patent 7572008.

15. Eye position detecting apparatus [Text]: US patent 5218387.

16. Bay H., Ess A., Tuytelaars T., Van Gool L. SURF: Speeded Up Robust Features // Computer Vision and Image Understanding (CVIU). - 2008 - Vol. 110, No.3 - P.346-359.

17. Harris S., Stephens M. A combined comer and edge detector // Proceedings of the 4th Alvey Vision Conference - 1988 - P.147-151.

18. Lowe G. D. Distinctive Image Features from Scale-Invariant Keypoints // International Journal of Computer Vision. - 2004 - Band 60, Nr. 2 - P.91-110.

19. Safaee-Rad R., Choukanov I., Crless Smith K., Benhabib B. Three-dimensional location estimation of circular features for machine vision // IEEE transactions of robotics and automation - 1992 - Vol.8, No. 5 - P .624-640.

20. de Berg M. et al. Computational Geometry: Algorithms and Applications, 3rd Edition. - Springer, 2008 - P.99-101.

21. Chum O., Matas J., Kittler J. Locally optimized RANSAC // Pattern Recognition Lecture Notes in Computer Science / Eds. B. Michaelis, G. Krell. - 2003 .-- Vol.2781. - P.236-243.

22. Viola P., Jones M.J. Fast multi-view face detection. - Mitsubishi Electric Research Laboratories, TR2003-096 - August 2003.

Claims (9)

1. The method of determining the direction of the line of sight, which consists in the fact that at least two image recording devices (SIDs) are set arbitrarily, the SID system is calibrated using a reference fragment (RF), as a result of which the real-scale coefficient of the three-dimensional model is determined, and then recorded an object from different angles, find the local features (LO) of the segments of the object, based on the received LOs from several images build a three-dimensional model that reflects the spatial position of the LO seg of the object’s coordinates relative to each other, as well as relative to the SID, and containing information in relative coordinates, then, taking into account the calibration data, the positions of the indicated LOs are recalculated from the relative coordinates of the three-dimensional model into real coordinates, the LO segments of the object from the three-dimensional model are projected onto the original images, and the segments are outlined object in these two-dimensional images, calculate the orientation of these contours in space and determine the direction of the line of sight by drawing a straight line, passing through the center of a circle forming a three-dimensional contour of an object segment and oriented along the normal of the three-dimensional plane in which this contour is located. 2. The method according to claim 1, characterized in that for determining the real scale are used LO belonging to the Russian Federation. 3. The method according to claim 1, in which the position of the Russian Federation is fixed in space. 4. The method according to claim 1, in which the position of the Russian Federation is changed during the measurement process. 5. The method according to claim 1, characterized in that the RF is located in the field of view of the SID URI in the measurement process. 6. The method according to claim 1, characterized in that the RF after calibration is placed periodically in the field of view of the URZ PZ. 7. The method according to claim 1, characterized in that, at each measurement, the relative position of the SID is memorized, and the subsequent measurement results are amended to change the position of the SID by comparing the saved old position and the new one. 8. The method according to claim 1, characterized in that several different images are used as the RF, and calibration is performed according to any of these images. 9. The method according to claim 1, characterized in that as the segments of the object using the iris and / or pupils of the eyes of the object.

Images