EA038293B1 - Method and system for detecting troubling events during interaction with a self-service device - Google Patents

Abstract

The invention relates in general to the field of video image analysis. A method for detecting troubling events during interaction with a self-service device comprises the steps of obtaining data from a video surveillance camera characterizing an image of one figure of a person situated in the vicinity of a self-service device; conducting an analysis of the obtained data using a machine learning model, during the course of which the obtained images are processed in order to detect keypoints of the figure of a person; analyzing the keypoints detected in the preceding step in order to determine the pose assumed by the figure of a person; determining the pose of the figure of a person in a sitting, crouching or squatting position; analyzing the interaction of the figure of a person in the detected position with the self-service device; determining the duration of the detected interaction. The technical result is an increase in the accuracy of detection of troubling events, which is achieved by using machine learning models to analyse the poses of a person during interaction with a self-service device within a given time interval.

Description

Technology area

The presented technical solution refers in general to the field of video image analysis for detecting alarming events when interacting with self-service devices (hereinafter CS), and in particular to methods and systems for detecting alarming events when interacting with ATMs.

State of the art

Currently, there is a problem of prompt and high-quality processing of video data for the analysis of illegal actions in relation to bank property or for the purpose of committing fraudulent actions, in particular when interacting with ATMs or payment terminals. The principle of using external computing devices installed in the compartment for accepting cards of ATMs for reading payment card data was widespread at one time. There are also methods of physical impact on ATMs in order to access the contents of cash cassettes. One way or another, all these events are caused by the corresponding characteristic actions performed by intruders who fall into the object of surveillance cameras, however, when monitoring with the help of security personnel, it is sometimes not enough to quickly and accurately identify the fact of committing one or another alarming event, which with a large more likely to lead to damage or theft of valuable property.

There are various methods for analyzing video data, for example, inside bank branches, which allow us to identify non-standard and / or dangerous behavior of visitors, which may indicate the likelihood of alarming or critical events. Such approaches, as a rule, are based on the recognition of the postures of visitors, as well as their movement in order to classify the factor of the onset of an event, alarming or critical, based on existing patterns of behavior.

From the prior art, a solution is known for analyzing the behavior of visitors based on the use of machine learning models that determine deviations from the norm of behavior, for example, the detection of non-standard or illegal actions based on the processing of images received from CCTV cameras (US 20160005050, publication date: 07.01.2016 ).

From another solution, a system for monitoring the position of the hands of a user interacting with an ATM is known for detecting various kinds of illegal actions (WO 2012178202, publication date: 27.12.2012). The analysis of the position and movement of the hands is carried out on the basis of the construction of a digital skeletal model of the hands, which makes it possible to construct and track the movement of fingers for the analysis of a particular event.

A significant drawback of the known solutions is the lack of accuracy in determining a certain type of position of a human figure, in particular when a person is in a bent, seated or semi-seated position in a certain time range near the US, which allows clearly classifying an event as alarming.

Disclosure of invention

The technical problem or technical problem to be solved using the proposed approach is the creation of a new effective method for identifying alarming events when users interact with the RS, for example, in a bank.

The technical result is to improve the accuracy of detecting alarming events, which is achieved through the use of machine learning models for analyzing human postures when interacting with the EOS in a given time period.

An additional technical result is an increase in the efficiency of the system's response to the fact of an alarming event.

In a preferred embodiment, a method is claimed for detecting alarming events when interacting with a control system, performed using at least one computing device, comprising the steps of obtaining data from at least one CCTV camera characterizing the image of at least one figure of a person in the vicinity US;

analyzing the data obtained using a machine learning model, during which the obtained images are processed to identify the key points of the human figure;

analysis of key points is performed to determine the posture taken by the human figure;

determine the posture of a human figure in a sitting, bent or crouching position;

analyzing the presence of the interaction of the human figure in the identified position with the US;

determine the time interval of the identified interaction; and if the time interval is greater than the set threshold value, then the presence of an alarm event is identified when operating with the DC.

In one of the particular embodiments of the method, the image obtained from the US camera is additionally captured.

In another particular embodiment of the method, data is additionally recorded, receiving

- 1 038293 from the US sensors.

In another particular embodiment of the method, the resulting image from the CCTV cameras is analyzed for identifying the identity of a person who is near the CD.

In another particular embodiment of the method, the resulting image from the CCTV cameras is analyzed for the presence of elements that cover the face of the person interacting with the CD.

In another particular embodiment of the method, the analysis of a person's posture is carried out by constructing a graphical skeleton of a human figure using a machine learning model.

In another particular embodiment of the method, posture is determined by analyzing the angles between key points of the digital skeleton of a human figure.

In another particular embodiment of the method, vectors are constructed between key points to analyze the angles between them.

In another particular embodiment of the method, upon detection of an alarm event, it transmits it via a communication channel to the rapid response console and / or activates the alarm.

In another particular embodiment of the method, access to the room in which the CD is installed is additionally blocked.

In another particular embodiment of the method, upon identification of an alarm event, a notification is generated containing at least an image of a human figure near the CD, the time and address of the event fixation.

In another particular embodiment of the method, the notification is transmitted over data transmission channels to at least one remote device.

The specified technical result is also achieved by implementing a system for detecting alarms when interacting with a control system, which contains at least one processor and a data storage medium containing machine-readable instructions, which, when executed by at least one processor, provide data from at least one video surveillance camera, characterizing the image of at least one figure of a person located near the US;

analyzing the data obtained using a machine learning model, during which the received images are processed to identify the key points of the human figure;

performing analysis of key points to determine the posture taken by the human figure;

the implementation of the determination of the posture of a human figure in a sitting, bent or crouched position;

analysis of the presence of the interaction of the human figure in the identified position with the US;

determination of the time interval of the identified interaction; and if the time interval is greater than the set threshold value, then identifying the presence of an alarm event when working with the DC.

In one of the particular embodiments of the system, upon identification of an alarm event, a notification is generated containing at least an image of a human figure near the US, the time and address of the event fixation.

In another particular embodiment of the system, the notification is transmitted via data transmission channels to at least one remote device.

Description of drawings

Features and advantages of the present technical solution will become apparent from the following detailed description of the invention and the accompanying drawings, in which: FIG. 1 illustrates the general scheme of interaction of the elements of the claimed solution;

fig. 2 illustrates the process of performing the claimed method;

fig. 3 illustrates a digital skeleton model;

fig. 4 illustrates the construction of a digital skeleton based on a real image;

fig. 5 illustrates a schematic example of the analysis of the time spent by a person in the identified posture when working with the control system;

fig. 6 illustrates an example of the operation of data exchange between system elements.

Implementation of the invention

Below will be described the concepts and terms necessary to understand this technical solution.

In this technical solution, the system means, including a computer system, a computer (electronic computer), CNC (numerical control), PLC (programmable logic controller), computerized control systems and any other devices capable of performing a given, well-defined sequence operations (actions, instructions).

A command processing device means an electronic unit or an integrated circuit (microprocessor) that executes machine instructions (programs).

The command processor reads and executes machine instructions (programs) from one

- 2,038293 or more storage devices. The role of data storage devices can be, but are not limited to, hard disks (HDD), flash memory, ROM (read only memory), solid state drives (SSD), optical drives.

A program is a sequence of instructions intended for execution by a computer control device or a command processing device.

As shown in FIG. 1, in the general case, the automated system (100) includes an installed CD (110), which is located in the area of operation of one or more video cameras (120). The information, in particular the video stream received from the camera (120), is transmitted to the remote server (130) using the selected data transmission protocol, for example, TCP / IP, WLAN, or other suitable type of communication, but not limited to. Various types of cameras can be used as video surveillance cameras (sometimes video cameras), in particular PTZ cameras, IP cameras, stationary anti-vandal cameras, a circular view camera, etc., but not limited to.

Information can also be transmitted to the server (130) from systems for monitoring the state of the US, for example, from vibration, temperature, gyroscope sensors built into the US, as well as data from the biometric US system (built-in camera, retina scanner, microphone, etc.) ... This information can be used for additional analysis of the situation arising near the EOS (110).

Also, the system (100) includes an alarm system (140), which generally includes a door lock sensor (141) to the room where the CD (110) and / or sound notification (142) is installed. Additionally, the alarm system (140) may include other devices that will be used in the event of an alarm event. The door lock sensor (141) is connected by the corresponding data transmission channel directly with the door (143) installed in the area of the US (110).

The operation of the system (100) consists in obtaining images of the figures of people in the area (111) of interaction with the EOS (110), which is subsequently processed using an analytical module built on the basis of one or more machine learning models to ensure the process of recognizing the posture taken a person (10) interacting with the CS (110) to make a judgment about the likelihood of committing a particular illegal action.

A method for detecting an alarming event (200), implemented in the course of implementing the presented solution, consists in performing the following sequential computational steps. At the first stage (201), a video stream is obtained from one or more CCTV cameras (120), which characterizes the presence of one or more figures of people (10) located in the zone (111) of the location of the US (110).

The received video stream from the camera (120) is transmitted to the server (130) for further processing. On the server (130), using the analytical module, a frame-by-frame parsing of the received video stream is performed for the purposes of analyzing the poses of people (10) interacting with the US (110). The video stream is analyzed using a machine learning model such as the OpenPose model. Using the OpenPose model, at step (202), key points are determined on the obtained images (frames) extracted from the video stream. The OpenPose machine learning model processes the received frames and, at the output, generates a JSON file with x / y coordinates in pixels, as well as a probabilistic characteristic for each key point. As a probabilistic characteristic, we understand the probability of correctly finding key points relative to their belonging to the digital skeleton. An example of such an algorithm is disclosed in the article by Cao et al. OpenPose: Realtime Multi-Person 2D Pose Estimation using Part Affinity Fields published: 12-18-2018.

Step (202) is performed by building a digital human skeleton for each input image. An example of a general view of a digital skeleton is shown in FIG. 3. Based on the OpenPose model algorithm, pixel coordinates are constructed for each human figure (10) on frames received from surveillance cameras (120). Additionally, OpenPose uses the Caffe deep learning framework, which is able to switch between GPU and CPU during computations, resulting in a performance boost. OpenPose is based on training models using a Convolutional Neural Network (CNN).

The model accepts a color image of size w-h as input, and outputs the location of the key points of each object of observation at the output. Determination of key points occurs in three stages. The first 10 layers of the training sample (VGGNet) are used to create a map of the feature points for the original image. The SNN uses two branches: the first predicts a map (S) of a set of points on the human body (shoulder, arm, etc.), the second a set of vectors (L) that contain the degree of difference between body parts. (S) -map and (b) -vectors are analyzed to create key points for all people (10) in the image.

Next, at step (203), according to the coordinates of the human figure (10), on the obtained frames using the OpenPose model, key points of the body are determined to recognize the pose taken by the person (10). The coordinates of the key points are used to construct a vector, on the basis of which the angle between the vectors thigh-lower leg and thigh-spine is calculated, and if the mentioned angle is less than the specified one and

- 3 038293 while the person (10) in this position is near the US (110), then a notification about the presence of an alarm event is generated, the fact of the presence of which is also recorded in the data store, for example, a database.

The input data for analyzing the presence of an alarm event based on the analyzed data is a frame in .PNG format and a JSON file containing two-dimensional coordinates of all key points of human figures (10) detected by OpenPose on the basis of a digital skeleton, as well as a probabilistic characteristic for each key coordinate in the range from 0 to 1. Example of JSON file structure {version: 1.1, people: [{

pose_keypoints_2d: [169.449,147.875,0.734807,127.885,254.116,0.078005, ...]}], part_candidates: [{

0: [169.449,147.875,0.734807,127.885,254.116,0.078005], l: [270.113,210.873,0.476116],

2: [302.885,103.86,0.266788],

3: [127.885,254.116,0.078005],

4: [283.015,332.986,0.665039],

5: [346.876,213.112,0.321377,172.104,221.875,0.554128],

6: [],

7: [],

eight:[],

nine:[],

ten:[],

H: [],

12:[],

13:[],

14: [182.112,131.880,0.563214],

15: [203.867,130,0.686417],

16:[],

17: [258.116,124.853,0.77654]}

]}

New OpenPose files process the incoming data in an iterative manner. If a new * json file appears, the recognition algorithm reads the results of OpenPose work from it - if no human figures were found, the file and the corresponding frame are deleted, otherwise processing begins, which is in several stages.

For the involved key points of the detected figures (10), it is checked that the probability value exceeds the set threshold value, for example, 0.4, which makes it possible to cut off the overwhelming number of false positives. In this case, the threshold value can vary depending on the illumination of the viewing area of the cameras (120), the location of the cameras (120), and also the threshold value can be set separately for a specific processed coordinate. Since there may be windows in the areas captured by cameras (120), the algorithm checks that each figure (10) that gets into the frame is not outside the territory of the US installation site (110), for example, outside the territory of a bank branch. FIG. 4 shows an example of building a digital skeleton based on the obtained image.

Next, the angles are calculated between the key points of the digital skeleton (LHigh, LKeleno, LKeleno), (LHigh, PKKeleno, PKeleno), (Neck, Middle Part of the Thigh, PKKeleno) and (Neck, Middle Part of the Thigh, LKeleno). If all four angles are in the range from 0 to 120 °, then the algorithm determines,

- 4 038293 that the figure is in a sitting, squatting or bent position.

Additionally, a different position of the figures can be fixed, which can be non-standard and can also be fixed using the stated solution. To detect recumbent figures, the frame areas corresponding to the floor of the location (111) of the US (110) are set, and vectors that define the floor plane are determined (it is assumed that the lying person will try to position himself along the walls, and not at an arbitrary angle - the latter case is difficult to determine in two-dimensional coordinates). The figure is considered to be in the supine position if both points of the MidHip and Neck are within the area (111), and the angle between the indicated points and one of the vectors (2) does not exceed 10 °. Also, the figure is considered to be lying if the points LAnkle and PAnkle are located above the Neck point. After determining the pose at step (203), the analysis of the time interval (204) is carried out, during which the person (10) interacts with the US (110) in the identified pose. FIG. 5 shows a schematic example of the analysis of frames with the identified posture of a person (10) during a certain time interval T. then at step (205), the algorithm makes a judgment about the presence of an alarm event.

Upon the fact of fixing an alarm event at step (205), a corresponding record is formed on the server (130), which may include the event identifier (ID), the number of the institution of the US installation, the number of the video camera that recorded the event, the time / date of the incident, the name of the alarm event, the image recorded by a surveillance video camera.

Next, the hardware and software component of the claimed solution, presented in FIG. 6. As mentioned above, in addition to the information received from the camera or surveillance cameras (120), information received directly from the devices installed in the CD (110) can also be transmitted to the server. As such devices can be various types of sensors and sensors that respond to the emerging external influence. The biometric information of persons interacting with the US (110) can also be analyzed, in particular, their image received from a video camera built into the US (110). The data received from the US (110) and the video camera (120) are transmitted via the data transmission channel to the server (130). The information received from the devices (110, 120) is transmitted for further analysis of the graphic information (video stream) to the analytical module for determining the pose (131). The analytical module (131), built on the basis of machine learning algorithms, contains a module for determining key points (1311) and a module for determining the pose (1312) of a human figure in frames from a surveillance camera.

Module (1311) can do its work using the OpenPose model algorithm. The data from the surveillance camera (120) can be transmitted using well-known video streaming protocols, for example, RTSP, RTMP. The module (1312) classifies the posture of the human figure (10), and also analyzes the time spent by the person (10) in this pose when working with the US (110). Also, the module (1312) takes into account the marking of the zones of the location (111) of the US (110) with reference to the US number (110), and also the number of the department in which it is installed can be additionally taken into account. Mirror areas and transparent objects outside the controlled area, such as windows, glass doors, are separately highlighted to exclude these areas from analysis to reduce false positive incidents. The zones are marked in pixels along the two axes X and Y. The reference point with coordinates (0,0) is the upper left corner of the frames broadcast from cameras (120). The near-bank zones (111) are taken into account when calculating the time T of a person's stay in them (10) and are correlated with events from other sources in the correlation module (151) of the event collection and analysis system (150) (ССАС).

From the analytical module (131), the processed data on the identified alarm events, together with their identifying information, are transferred to the database (132) (Time / date, branch number, ATM number, type of alarm event, image), which performs long-term storage of identified incidents and serves data source for the console (133).

The console (133) is a software solution for dealing with identified incidents. The logic for aggregating incidents in the console (133) implies combining several alarm events into one incident (for display only), which will include data from the very first alarm event when combined, and photographs of all combined events. Different time intervals are provided for aggregation. When a new alarm event appears, it will be checked for the possibility of combining with already existing elements for the selected period. Module (131) also sends the mentioned data to SSAC (150) (SIEM - Security information and event management). Sending is carried out using the syslog protocol. CCAC (150) can run on a separate server associated with server (130).

Additionally, using image analysis, an additional type of information that falls into the field of view of cameras can also be analyzed (120). For example, the presence of foreign objects and / or tools in the frame can be analyzed, with the help of which illegal actions can be committed, for example, cutting equipment, cylinders, wires, etc. There may also be a fixation of extraneous actions directly from the US (110) or in its vicinity, for example, the presence of smoke, sparks, open flames, etc.

- 5 038293

System (150) is a typical solution and consists, as a rule, of three modules - connectors (collecting data), manager (receiving and correlating events), and database (performing long-term storage). In the presented solution, the correlation module (151) CCAC (150) receives events from the following information sources: US sensors (110), models for analyzing the posture of a human figure (130) and from the biometric module (134).

The biometric module (134) is a platform for the analysis of biometric features obtained by means of the US (110), for example, a camera or a microphone. The module (134) includes a face recognition module (1341) and a face hiding recognition module (1342). Modules (1341, 1342) can be executed based on one or more machine learning models, for example, they can be an artificial neural network, a random forest algorithm, etc.

With the help of the module (1341), the identity of a person (10) located in the fixation zone of the US camera (110) is performed. This approach allows you to obtain additional information to predict the possibility of an alarming event. For example, when identifying a person who is on the wanted list, or a person who has been caught / convicted of illegal actions in relation to the property of the bank. Analysis and response to this type of information can be carried out by comparing the received data with data stored in an external or internal database, for example, by promptly sending biometric data through the CCAS system (150) to the console (160) to respond to the fact of recognition this or that person's personality (10) about US (110). The module (1342) is designed to analyze the presence of people near the US (110) in the elements that hide the face, for example, scarves, masks, a helmet, etc. Fixing this information can be an additional parameter for making an operational decision when analyzing the possibility of an alarming event.

After processing the events, the correlation module (151) transmits information about events to the event database (152) for long-term storage and reporting. When an alarm event occurs, the CCAS (150) sends an alert to responsible persons, for example, to the emergency response panel (160) of internal services or to the alarm system (140), which is associated with the control of access control to the room, in particular, the door lock sensor (141).

For example, an event is alarming when the correlation module (151) receives an event from a vibration sensor installed in the DC (110), and an event from the posture analysis model (131), which recorded that a person sits near the DC (110) for a long time. In this case, the identifiers of such an event as the time, the ATM number and the number of the ATM installation department must match, which is necessary for the purpose of unambiguous correlation of the incoming data to record the reality of the occurrence of an alarm event.

Correlation rules for identifying alarming events can be set for each DC (110) separately. For example, you can set a list of predefined alarm event identifiers, which, when they occur and match a predefined pattern, will clearly identify the fact of its occurrence. Such identifiers can be: the type of a person's posture (10) near the US (110) or the zone number near the US (111), signals from specific US sensors (110) or the presence of foreign objects in the frame received from surveillance cameras (120), the time interval fixing a person (10) in a given position, etc.

FIG. 7 shows an example of a general view of a computing device (300) that provides an implementation of the presented solution. For example, the server (130), the CCAC (150) and the hardware and software component of the CD (110) can be formed on the basis of the computing device (300).

In general, the device (300) contains one or more processors (301) united by a common bus of information exchange, memory means such as RAM (302) and ROM (303), input / output interfaces (304), input / output devices (305 ), and a device for networking (306).

The processor (301) (or multiple processors, multi-core processor, etc.) can be selected from a range of devices currently widely used, for example, from manufacturers such as Intel ™, AMD ™, Apple ™, Samsung Exynos ™, MediaTEK ™ , Qualcomm Snapdragon ™, etc. Under the processor or one of the processors used in the system (300), it is also necessary to take into account a graphics processor, for example, NVIDIA GPU or Graphcore, the type of which is also suitable for full or partial execution of the method (100), and can also be used for training and applying machine learning models in various information systems.

RAM (302) is a random access memory and is intended for storing computer-readable instructions executed by the processor (301) for performing the necessary operations for logical data processing. RAM (302), as a rule, contains executable instructions of the operating system and corresponding software components (applications, software modules, etc.). In this case, the available memory of the graphics card or graphics processor can act as RAM (302).

ROM (303) is one or more persistent storage devices such as a hard disk drive (HDD), solid state data storage device (SSD), flash memory (EEPROM, NAND, etc.), optical storage media (CD-R / RW, DVD-R / RW, BlueRay Disc, MD), etc.

- 6 038293

To organize the operation of system components (300) and to organize the operation of external connected devices, various types of I / O interfaces (304) are used. The choice of appropriate interfaces depends on the specific design of the computing device, which can be, but are not limited to, PCI, AGP, PS / 2, IrDa, FireWire, LPT, COM, SATA, IDE, Lightning, USB (2.0, 3.0, 3.1, micro, mini, type C), TRS / Audio jack (2.5, 3.5, 6.35), HDMI, DVI, VGA, Display Port, RJ45, RS232, etc.

To ensure user interaction with the computing system (300), various means (305) I / O information are used, for example, a keyboard, display (monitor), touch display, touch pad, joystick, mouse manipulator, light pen, stylus, touch panel, trackball , speakers, microphone, augmented reality, optical sensors, tablet, light indicators, projector, camera, biometric identification (retina scanner, fingerprint scanner, voice recognition module), etc.

The means of networking (306) provides data transmission via an internal or external computer network, for example, an intranet, the Internet, a LAN, etc. One or more means (306) may be used, but not limited to: Ethernet card, GSM modem, GPRS modem, LTE modem, 5G modem, satellite communication module, NFC module, Bluetooth and / or BLE module, Wi-Fi module and etc. Additionally, satellite navigation means can also be used as part of the system (300), for example, GPS, GLONASS, BeiDou, Galileo.

The specific choice of elements of the device (300) for the implementation of various software and hardware architectural solutions can vary while maintaining the required functionality provided.

The submitted application materials disclose preferred examples of the implementation of the technical solution and should not be construed as limiting other, particular examples of its implementation, not going beyond the scope of the claimed legal protection, which are obvious to specialists in the relevant field of technology.

Claims (15)

CLAIM 1. A method for detecting alarming events when interacting with a self-service device (US), performed with the help of at least one computing device, comprising the steps of obtaining data from at least one CCTV camera, characterizing the image of at least one figure of a person located near the US; analyzing the data obtained using a machine learning model, during which the obtained images are processed to identify the key points of the human figure; analyze the key points identified in the previous step to determine the pose taken by the human figure; determine the posture of a human figure in a sitting, bent or crouching position; analyzing the presence of the interaction of the human figure in the identified position with the US; determine the time interval of the detected interaction, and if the time interval is greater than a predetermined threshold value, then the presence of an alarm event is identified when working with the control system. 2. The method according to claim 1, characterized in that the image obtained from the US camera is additionally recorded. 3. The method according to claim 2, characterized in that the data received from the US sensors are additionally recorded. 4. The method according to claim 1, characterized in that the resulting image from the CCTV cameras is analyzed for identifying the identity of a person who is near the US. 5. The method according to claim 1, characterized in that the resulting image from the surveillance cameras is analyzed for the presence of elements that cover the face of the person interacting with the US. 6. The method according to claim 1, characterized in that the analysis of a person's posture is carried out by constructing a graphical skeleton of a human figure using a machine learning model. 7. The method according to claim 6, characterized in that the determination of the pose is carried out by analyzing the angles between the key points of the digital skeleton of the human figure. 8. The method according to claim 7, characterized in that vectors are constructed between key points to analyze the angles between them. 9. The method according to claim 1, characterized in that when an alarm event is detected, it is transmitted via a communication channel to a rapid response console and / or alarm is activated. 10. The method according to claim 9, characterized in that the access to the room in which the CD is installed is additionally blocked. 11. The method according to claim 1, characterized in that when identifying an alarm event, a notification is generated containing at least an image of a human figure near the US, time and - 7 038293 rescription of the event. 12. The method according to claim 11, characterized in that the notification is transmitted over data transmission channels to at least one remote device. 13. A system for detecting alarms when interacting with a self-service device (SD), containing at least one processor and data storage means containing machine-readable commands, which, when executed by at least one processor, provide data from at least one CCTV camera , characterizing the image of at least one figure of a person who is near the US; analyzing the data obtained using a machine learning model, during which the received images are processed to identify the key points of the human figure; performing analysis of key points to determine the posture taken by the human figure; the implementation of the determination of the posture of a human figure in a sitting, bent or crouched position; analysis of the presence of the interaction of the human figure in the identified position with the US; determination of the time interval of the identified interaction; and if the time interval is greater than the set threshold value, then identifying the presence of an alarm event when working with the DC. 14. The system according to claim 13, characterized in that upon identification of an alarm event, a notification is generated containing at least an image of a human figure near the CD, the time and address of fixing the event. 15. The system of claim 14, wherein the notification is transmitted over data transmission channels to at least one remote device.