A drowsiness warning method and system based on face detection

Abstract

This application discloses a drowsiness alarm method and system based on facial detection, relating to the field of drowsiness alarms. It includes a work data acquisition module, a data detection module, a preliminary analysis module, a tracking and monitoring module, and a comprehensive analysis module. By collecting facial data from a visual camera, it provides a scientific data foundation for drowsiness alarm analysis, helping to constantly remind security personnel to be alert. Through analysis of the target worker's facial data, as well as their blinking and yawning data, the system effectively analyzes the target worker's drowsiness state and provides a reference for subsequent analysis of drowsiness alarms. This effectively improves the accuracy and effectiveness of drowsiness alarms, ensuring that the worker maintains their work status while promptly preventing errors.

Description

Technical Field

[0001] This application relates to the field of drowsiness alarms, and in particular to a drowsiness alarm method and system based on facial detection. Background Technology

[0002] To ensure passenger safety, improve security check service quality, enhance passenger satisfaction, and promote the healthy development of civil aviation, the Security Inspection Department, aiming to improve the quality and efficiency of civil aviation security checks, is accelerating the application of advanced technologies and equipment in the field. However, the anti-drowsiness mechanism for security check operators still has the following shortcomings.

[0003] Existing technologies have historically been used for indirect management of duty status, relying on methods such as employee safety education and training, disciplinary constraints, irregular inspections by security quality control personnel and team leaders, and video surveillance reviews for supervision and inspection. However, the effectiveness of such supervision is often delayed and problems tend to recur, further increasing the security risks at airports. Summary of the Invention

[0004] The purpose of this invention is to provide a drowsiness alarm method and system based on facial detection to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a drowsiness alarm system based on facial detection, comprising:

[0006] Work data acquisition module: used to acquire the work logs of employees and obtain the work data of the target employees;

[0007] Data detection module: used to perform image detection on target workers and obtain detection data of the target workers;

[0008] Preliminary Analysis Module: This module combines the detection data and work data of the target staff to analyze and obtain the preliminary drowsiness alarm coefficient corresponding to the target staff.

[0009] Tracking and monitoring module: used to conduct reaction tests on target staff based on the initial drowsiness alarm coefficient corresponding to the target staff, and obtain the corresponding reaction coefficient of the target staff;

[0010] Comprehensive Analysis Module: This module is used to perform comprehensive analysis based on the initial drowsiness alarm coefficient, the reaction coefficient, and the detection data of the target staff member to obtain the drowsiness alarm result for the target staff member.

[0011] In the preferred embodiment of this solution, the specific execution method of the working data acquisition module is as follows:

[0012] Obtain the work logs of the target staff, and filter the work hours and number of events handled by the target staff based on the work logs and the detection time points. Record the work hours and number of events handled by the target staff as the work data of the target staff.

[0013] In the preferred embodiment of this solution, the data detection module is executed as follows:

[0014] The target worker's face is continuously captured by a pre-set visual camera in front of the target worker, resulting in continuous facial images of the target worker. The facial images of the target worker are then divided into regions to obtain eye sub-images and mouth sub-images of the target worker. Finally, continuous eye sub-images and continuous mouth sub-images of the target worker are obtained by statistical analysis.

[0015] The continuous eye sub-images, continuous eyebrow sub-images, and continuous mouth sub-images corresponding to the target worker are recorded as the detection data corresponding to the target worker.

[0016] In the preferred embodiment of this solution, the specific execution method of the preliminary analysis module is as follows:

[0017] Establish the data extraction relationship between the preliminary analysis module and the database, extract the work fatigue index change curve with working time stored in the database, and obtain the work fatigue index corresponding to the target staff based on the working hours of the target staff.

[0018] Through calculation formula The impact coefficient of the drowsiness alarm on the target staff was calculated. ,in This is represented as the work fatigue index for the target staff. This represents the number of events handled by the target staff member. This represents the working hours of the target staff. This is represented as the preset drowsiness alarm impact coefficient / impact factor;

[0019] Extract the human eye-closing action model and human yawning action model stored in the database;

[0020] Based on the human eye-closing action model stored in the database, the eye-closing recognition is performed on the continuous eye sub-images corresponding to the target staff member to obtain each eye-closing action of the target staff member. The images of each eye-closing action of the target staff member are extracted to obtain the eye sub-images of the target staff member at the beginning of closing the eyes, the eye sub-images of the eyes after closing the eyes, and the eye sub-images of the eyes after opening the eyes for each eye-closing action.

[0021] Image data was acquired from the eye sub-images of the beginning of closing the eyes, the eye sub-images of the completed closing the eyes, and the eye sub-images of the open eyes for each eye closing session, to obtain the eye fissure length and width at the beginning of closing the eyes, the eye fissure length at the completed closing the eyes, and the eye fissure length and width at the open eyes.

[0022] Through calculation formula The eye closure depth coefficient corresponding to the completion of each eye closure was calculated. ;

[0023] The difference in palpebral fissure length and width between the beginning of closing the eyes and the beginning of opening the eyes were obtained by calculation for each instance of closing the eyes.

[0024] Through calculation formula The eye-opening amplitude coefficient corresponding to each instance of closing the eyes was calculated. ,in , These represent the length of the palpebral fissure at the beginning of each eye-closing action and the length of the palpebral fissure at the end of each eye-closing action, respectively. This represents the width of the palpebral fissure at the start of each eye-closing action. , These represent the difference in palpebral fissure length and width between the initial closing and opening positions for each instance of eye closure. This is represented by the number of each time the eyes were closed;

[0025] Through calculation formula The preliminary alarm coefficients corresponding to each instance of closing the eyes are calculated. ;

[0026] Obtain the image acquisition time point of the eye sub-image corresponding to the start of closing the eyes and the image acquisition time point of the eye sub-image corresponding to opening the eyes in each closing process, and calculate the closing time for each closing process.

[0027] The duration of the alarm judgment period corresponding to the alarm system is obtained. Based on the duration of the alarm judgment period corresponding to the alarm system and the image acquisition time point of the eye sub-image corresponding to the start of each eye closing, the number of eye closings corresponding to the alarm judgment period is obtained.

[0028] Obtain the data comparison set corresponding to the alarm coefficient, which includes the standard eye-closing frequency.

[0029] Through calculation formula The preliminary drowsiness alarm coefficient corresponding to the target staff was calculated. ,in This is represented as the standard eye-closing frequency. This represents the number of times the eyes were closed during the alarm detection period. This represents the duration of the alarm detection period. This represents the initial alarm coefficient corresponding to each instance of eye closure within the alarm judgment period. This represents the duration of each eye-closing action within the alarm detection period. This represents the preset standard time spent closing your eyes. This represents the number of each time eyes were closed within the alarm judgment period.

[0030] In the preferred embodiment of this solution, the tracking and monitoring module is executed as follows:

[0031] Establish a data extraction relationship between the tracking and monitoring module and the database, and extract the fatigue inspection dataset corresponding to each preliminary drowsiness alarm coefficient stored in the database. The fatigue inspection dataset includes inspection character length, inspection character brightness, inspection character duration, and inspection character movement model.

[0032] The fatigue inspection dataset for the target staff was obtained by filtering the initial drowsiness alarm coefficients for the target staff.

[0033] Fatigue inspection is performed using the display terminal corresponding to the target worker and the fatigue inspection dataset corresponding to the target worker, and the display coordinate point and display time point of the initial inspection character corresponding to the display terminal are obtained.

[0034] The eye movements of the target staff are detected by a pre-set visual camera in front of the target staff, and the eye movement data of the target staff is obtained. The eye movement data includes the direction and speed of eye movement at each time point. The eye movement data of the target staff after the time point of character display is filtered out and recorded as the eye movement data to be analyzed for the target staff.

[0035] A computer vision model is established based on the eye movement data of the target worker and the work environment of the target worker. The computer vision model is used to obtain the visual target point, the position of the visual target point and the movement speed of the visual target point on the display terminal corresponding to the target worker.

[0036] Extract the theoretical reaction time corresponding to each character length and character brightness combination stored in the database, and obtain the theoretical reaction time corresponding to the target staff by filtering the fatigue inspection dataset corresponding to the target staff.

[0037] Visually match the visual target point of the display terminal corresponding to the target worker with the display coordinate point of the initial inspection character to obtain the time point when the visual target point and the display coordinate point of the initial inspection character coincide. Record this as the visual coincidence point corresponding to the target worker. Record the time interval between the display time point of the initial inspection character and the visual coincidence point corresponding to the target worker as the reaction time corresponding to the target worker.

[0038] When the system recognizes that the visual target point coincides with the display coordinate point of the initial inspection character, the inspection character moves according to the inspection character movement model and movement speed, obtains the time point when the inspection character reaches the end point of the inspection character, obtains the time point when the visual target point reaches the end point of the inspection character, and records the time interval between the time point when the inspection character reaches the end point of the inspection character and the time point when the visual target point reaches the end point of the inspection character as the delay time corresponding to the target worker.

[0039] Based on the position and movement speed of the visual target point, a visual movement model corresponding to the target worker is established. The visual movement model corresponding to the target worker is then compared with the inspection character movement model corresponding to the target worker to obtain the degree of overlap of the visual movement model corresponding to the target worker.

[0040] According to the calculation formula The reaction coefficients corresponding to the target staff were calculated. ,in This represents the theoretical reaction time for the target staff member. This represents the reaction time for the target staff member. This represents the delay duration corresponding to the target staff member. This represents the degree of overlap between the visual movement models corresponding to the target staff member.

[0041] In the preferred embodiment of this solution, the comprehensive analysis module is executed as follows:

[0042] Based on the yawning action model of the person stored in the database, the action is matched with the continuous mouth sub-images corresponding to the target staff to obtain each yawn of the continuous mouth sub-images corresponding to the target staff, and the duration and corner angle of each yawn are obtained.

[0043] Filter alarms to determine the corresponding yawns within a given time period, as well as the duration and angle of the mouth for each yawn.

[0044] Through calculation formula The comprehensive drowsiness alarm coefficient corresponding to the target staff was calculated. ,in This represents the number of yawns within the alarm judgment period. , These represent the duration and angle of the mouth for each yawn within the alarm judgment period, respectively. This represents the number of each yawn within the alarm judgment period;

[0045] The overall drowsiness alarm coefficient corresponding to the target employee is compared and analyzed with the preset standard overall drowsiness alarm coefficient. If the overall drowsiness alarm coefficient corresponding to the target employee is greater than the preset standard overall drowsiness alarm coefficient, the target employee will be alarmed through the alarm device. If the overall drowsiness alarm coefficient corresponding to the target employee is less than or equal to the preset standard overall drowsiness alarm coefficient, the target employee will not be alarmed.

[0046] To achieve the above objectives, the present invention also provides the following technical solution: a drowsiness alarm method based on facial detection, comprising the following steps:

[0047] Obtain the work logs of the staff to get the work data of the target staff;

[0048] Image detection is performed on the target staff to obtain the detection data of the target staff;

[0049] By combining and analyzing the detection data and work data of the target staff, a preliminary drowsiness alarm coefficient corresponding to the target staff is obtained.

[0050] Based on the initial drowsiness alarm coefficient of the target staff, a reaction test is conducted on the target staff to obtain the corresponding reaction coefficient of the target staff.

[0051] Based on a comprehensive analysis of the preliminary drowsiness alarm coefficient, the reaction coefficient, and the detection data of the target staff, the drowsiness alarm result for the target staff is obtained.

[0052] Compared with the prior art, the beneficial effects of the present invention are:

[0053] This invention collects facial data from a visual camera, providing a scientific data basis for drowsiness alarm analysis, further improving the accuracy and effectiveness of drowsiness alarms for target staff, and helping to constantly remind security check operators to be on duty.

[0054] This method analyzes the facial data, blink data, and yawn data of target staff to effectively determine their drowsiness state. It provides a reference for subsequent analysis of drowsiness alarms, effectively improving the accuracy and effectiveness of drowsiness alarms. Through the anti-drowsiness alarm system, it reminds and warns staff to maintain their work status while promptly preventing work errors. Attached Figure Description

[0055] The present invention will be further described with reference to the accompanying drawings, but the embodiments in the drawings do not constitute any limitation on the present invention. For those skilled in the art, other drawings can be obtained based on the following drawings without creative effort.

[0056] Figure 1 This is a schematic diagram of module connections in an embodiment of the present invention.

[0057] Figure 2 This is a schematic diagram illustrating the connection steps in an embodiment of the present invention. Detailed Implementation

[0058] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0059] Please see Figure 1 The present invention provides a drowsiness alarm system based on facial detection, including a work data acquisition module, a data detection module, a preliminary analysis module, a tracking and monitoring module, and a comprehensive analysis module;

[0060] The working data acquisition module is connected to the preliminary analysis module; the data detection module is connected to the preliminary analysis module and the comprehensive analysis module; the preliminary analysis module is connected to the tracking and monitoring module and the comprehensive analysis module; and the tracking and monitoring module is connected to the comprehensive analysis module.

[0061] The work data acquisition module is used to acquire the work logs of staff members and obtain the work data of the target staff members;

[0062] Furthermore, the specific execution method of the work data acquisition module is as follows:

[0063] Obtain the work logs of the target staff, and filter the work hours and number of events handled by the target staff based on the work logs and the detection time points. Record the work hours and number of events handled by the target staff as the work data of the target staff.

[0064] The data detection module is used to perform image detection on target workers and obtain detection data of the target workers;

[0065] Furthermore, the specific execution method of the data detection module is as follows:

[0066] The target worker's face is continuously captured by a pre-set visual camera in front of the target worker, resulting in continuous facial images of the target worker. The facial images of the target worker are then divided into regions to obtain eye sub-images and mouth sub-images of the target worker. Finally, continuous eye sub-images and continuous mouth sub-images of the target worker are obtained by statistical analysis.

[0067] The continuous eye sub-images, continuous eyebrow sub-images, and continuous mouth sub-images corresponding to the target worker are recorded as the detection data corresponding to the target worker.

[0068] The preliminary analysis module is used to combine and analyze the detection data and work data of the target staff to obtain the preliminary drowsiness alarm coefficient corresponding to the target staff.

[0069] Furthermore, the specific execution method of the preliminary analysis module is as follows:

[0070] Establish the data extraction relationship between the preliminary analysis module and the database, extract the work fatigue index change curve with working time stored in the database, and obtain the work fatigue index corresponding to the target staff based on the working hours of the target staff.

[0071] Through calculation formula The impact coefficient of the drowsiness alarm on the target staff was calculated. ,in This is represented as the work fatigue index for the target staff. This represents the number of events handled by the target staff member. This represents the working hours of the target staff. This is represented as the preset drowsiness alarm impact coefficient / impact factor;

[0072] Extract the human eye-closing action model and human yawning action model stored in the database;

[0073] Based on the human eye-closing action model stored in the database, the eye-closing recognition is performed on the continuous eye sub-images corresponding to the target staff member to obtain each eye-closing action of the target staff member. The images of each eye-closing action of the target staff member are extracted to obtain the eye sub-images of the target staff member at the beginning of closing the eyes, the eye sub-images of the eyes after closing the eyes, and the eye sub-images of the eyes after opening the eyes for each eye-closing action.

[0074] Image data was acquired from the eye sub-images of the beginning of closing the eyes, the eye sub-images of the completed closing the eyes, and the eye sub-images of the open eyes for each eye closing session, to obtain the eye fissure length and width at the beginning of closing the eyes, the eye fissure length at the completed closing the eyes, and the eye fissure length and width at the open eyes.

[0075] Through calculation formula The eye closure depth coefficient corresponding to the completion of each eye closure was calculated. ;

[0076] The difference in palpebral fissure length and width between the beginning of closing the eyes and the beginning of opening the eyes were obtained by calculation for each instance of closing the eyes.

[0077] Through calculation formula The eye-opening amplitude coefficient corresponding to each instance of closing the eyes was calculated. ,in , These represent the length of the palpebral fissure at the beginning of each eye-closing action and the length of the palpebral fissure at the end of each eye-closing action, respectively. This represents the width of the palpebral fissure at the start of each eye-closing action. , These represent the difference in palpebral fissure length and width between the initial closing and opening positions for each instance of eye closure. This is represented by the number of each time the eyes were closed;

[0078] Through calculation formula The preliminary alarm coefficients corresponding to each instance of closing the eyes are calculated. ;

[0079] Obtain the image acquisition time point of the eye sub-image corresponding to the start of closing the eyes and the image acquisition time point of the eye sub-image corresponding to opening the eyes in each closing process, and calculate the closing time for each closing process.

[0080] The duration of the alarm judgment period corresponding to the alarm system is obtained. Based on the duration of the alarm judgment period corresponding to the alarm system and the image acquisition time point of the eye sub-image corresponding to the start of each eye closing, the number of eye closings corresponding to the alarm judgment period is obtained.

[0081] Obtain the data comparison set corresponding to the alarm coefficient, which includes the standard eye-closing frequency.

[0082] Through calculation formula The preliminary drowsiness alarm coefficient corresponding to the target staff was calculated. ,in This is represented as the standard eye-closing frequency. This represents the number of times the eyes were closed during the alarm detection period. This represents the duration of the alarm detection period. This represents the initial alarm coefficient corresponding to each instance of eye closure within the alarm judgment period. This represents the duration of each eye-closing action within the alarm detection period. This represents the preset standard time spent closing your eyes. This represents the number of each time eyes were closed within the alarm judgment period.

[0083] The tracking and monitoring module is used to test the reaction of the target staff based on the initial drowsiness alarm coefficient corresponding to the target staff, and to obtain the corresponding reaction coefficient of the target staff.

[0084] Furthermore, the specific execution method of the tracking and monitoring module is as follows:

[0085] Establish a data extraction relationship between the tracking and monitoring module and the database, and extract the fatigue inspection dataset corresponding to each preliminary drowsiness alarm coefficient stored in the database. The fatigue inspection dataset includes inspection character length, inspection character brightness, inspection character duration, and inspection character movement model.

[0086] It should be noted that the view character movement model refers to the regular movement of view characters on the display terminal, such as drawing a circle clockwise and a square counterclockwise, etc.

[0087] The fatigue inspection dataset for the target staff was obtained by filtering the initial drowsiness alarm coefficients for the target staff.

[0088] Fatigue inspection is performed using the display terminal corresponding to the target worker and the fatigue inspection dataset corresponding to the target worker, and the display coordinate point and display time point of the initial inspection character corresponding to the display terminal are obtained.

[0089] The eye movements of the target staff are detected by a pre-set visual camera in front of the target staff, and the eye movement data of the target staff is obtained. The eye movement data includes the direction and speed of eye movement at each time point. The eye movement data of the target staff after the time point of character display is filtered out and recorded as the eye movement data to be analyzed for the target staff.

[0090] A computer vision model is established based on the eye movement data of the target worker and the work environment of the target worker. The computer vision model is used to obtain the visual target point, the position of the visual target point and the movement speed of the visual target point on the display terminal corresponding to the target worker.

[0091] Extract the theoretical reaction time corresponding to each character length and character brightness combination stored in the database, and obtain the theoretical reaction time corresponding to the target staff by filtering the fatigue inspection dataset corresponding to the target staff.

[0092] Visually match the visual target point of the display terminal corresponding to the target worker with the display coordinate point of the initial inspection character to obtain the time point when the visual target point and the display coordinate point of the initial inspection character coincide. Record this as the visual coincidence point corresponding to the target worker. Record the time interval between the display time point of the initial inspection character and the visual coincidence point corresponding to the target worker as the reaction time corresponding to the target worker.

[0093] When the system recognizes that the visual target point coincides with the display coordinate point of the initial inspection character, the inspection character moves according to the inspection character movement model and movement speed, obtains the time point when the inspection character reaches the end point of the inspection character, obtains the time point when the visual target point reaches the end point of the inspection character, and records the time interval between the time point when the inspection character reaches the end point of the inspection character and the time point when the visual target point reaches the end point of the inspection character as the delay time corresponding to the target worker.

[0094] Based on the position and movement speed of the visual target point, a visual movement model corresponding to the target worker is established. The visual movement model corresponding to the target worker is then compared with the inspection character movement model corresponding to the target worker to obtain the degree of overlap of the visual movement model corresponding to the target worker.

[0095] According to the calculation formula The reaction coefficients corresponding to the target staff were calculated. ,in This represents the theoretical reaction time for the target staff member. This represents the reaction time for the target staff member. This represents the delay duration corresponding to the target staff member. This represents the degree of overlap between the visual movement models corresponding to the target staff member.

[0096] Comprehensive Analysis Module: This module is used to perform comprehensive analysis based on the initial drowsiness alarm coefficient, the reaction coefficient, and the detection data of the target staff member to obtain the drowsiness alarm result for the target staff member.

[0097] Furthermore, the specific execution method of the comprehensive analysis module is as follows:

[0098] Based on the yawning action model of the person stored in the database, the action is matched with the continuous mouth sub-images corresponding to the target staff to obtain each yawn of the continuous mouth sub-images corresponding to the target staff, and the duration and corner angle of each yawn are obtained.

[0099] Filter alarms to determine the corresponding yawns within a given time period, as well as the duration and angle of the mouth for each yawn.

[0100] Through calculation formula The comprehensive drowsiness alarm coefficient corresponding to the target staff was calculated. ,in This represents the number of yawns within the alarm judgment period. , These represent the duration and angle of the mouth for each yawn within the alarm judgment period, respectively. This represents the number of each yawn within the alarm judgment period;

[0101] The overall drowsiness alarm coefficient corresponding to the target employee is compared and analyzed with the preset standard overall drowsiness alarm coefficient. If the overall drowsiness alarm coefficient corresponding to the target employee is greater than the preset standard overall drowsiness alarm coefficient, the target employee will be alarmed through the alarm device. If the overall drowsiness alarm coefficient corresponding to the target employee is less than or equal to the preset standard overall drowsiness alarm coefficient, the target employee will not be alarmed.

[0102] Please see Figure 2 To achieve the above objectives, the present invention also provides the following technical solution: a drowsiness alarm method based on facial detection, comprising the following steps:

[0103] Obtain the work logs of the staff to get the work data of the target staff;

[0104] Image detection is performed on the target staff to obtain the detection data of the target staff;

[0105] By combining and analyzing the detection data and work data of the target staff, a preliminary drowsiness alarm coefficient corresponding to the target staff is obtained.

[0106] Based on the initial drowsiness alarm coefficient of the target staff, a reaction test is conducted on the target staff to obtain the corresponding reaction coefficient of the target staff.

[0107] Based on a comprehensive analysis of the preliminary drowsiness alarm coefficient, the reaction coefficient, and the detection data of the target staff, the drowsiness alarm result for the target staff is obtained.

[0108] The above are all preferred embodiments of this application, and are not intended to limit the scope of protection of this application. Therefore, all equivalent changes made in accordance with the structure, shape and principle of this application should be covered within the scope of protection of this application.

Claims

1. A drowsiness alarm system based on facial detection, characterized in that: include: Work data acquisition module: used to acquire the work logs of employees and obtain the work data of the target employees; Data detection module: used to perform image detection on target workers and obtain detection data of the target workers; Preliminary Analysis Module: This module combines the detection data and work data of the target staff to analyze and obtain the preliminary drowsiness alarm coefficient corresponding to the target staff. The specific execution method of the preliminary analysis module is as follows: Establish the data extraction relationship between the preliminary analysis module and the database, extract the work fatigue index change curve with working time stored in the database, and obtain the work fatigue index corresponding to the target staff based on the working hours of the target staff. Through calculation formula The impact coefficient of the drowsiness alarm on the target staff was calculated. ,in This is represented as the work fatigue index for the target staff. This represents the number of events handled by the target staff member. This represents the working hours of the target staff. This is represented as the preset drowsiness alarm impact coefficient / impact factor; Extract the human eye-closing action model and human yawning action model stored in the database; Based on the human eye-closing action model stored in the database, the eye-closing recognition is performed on the continuous eye sub-images corresponding to the target staff member to obtain each eye-closing action of the target staff member. The images of each eye-closing action of the target staff member are extracted to obtain the eye sub-images of the target staff member at the beginning of closing the eyes, the eye sub-images of the eyes after closing the eyes, and the eye sub-images of the eyes after opening the eyes for each eye-closing action. Image data was acquired from the eye sub-images of the beginning of closing the eyes, the eye sub-images of the completed closing the eyes, and the eye sub-images of the open eyes for each eye closing session, to obtain the eye fissure length and width at the beginning of closing the eyes, the eye fissure length at the completed closing the eyes, and the eye fissure length and width at the open eyes. Through calculation formula The eye closure depth coefficient corresponding to the completion of each eye closure was calculated. ; The difference in palpebral fissure length and width between the beginning of closing the eyes and the beginning of opening the eyes were obtained by calculation for each instance of closing the eyes. Through calculation formula The eye-opening amplitude coefficient corresponding to each instance of closing the eyes was calculated. ,in , These represent the length of the palpebral fissure at the beginning of each eye-closing action and the length of the palpebral fissure at the end of each eye-closing action, respectively. This represents the width of the palpebral fissure at the start of each eye-closing action. , These represent the difference in palpebral fissure length and width between the initial closing and opening positions for each instance of eye closure. This is represented by the number of each time the eyes were closed; Through calculation formula The preliminary alarm coefficients corresponding to each instance of closing the eyes are calculated. ; Obtain the image acquisition time point of the eye sub-image corresponding to the start of closing the eyes and the image acquisition time point of the eye sub-image corresponding to opening the eyes in each closing process, and calculate the closing time for each closing process. The duration of the alarm judgment period corresponding to the alarm system is obtained. Based on the duration of the alarm judgment period corresponding to the alarm system and the image acquisition time point of the eye sub-image corresponding to the start of each eye closing, the number of eye closings corresponding to the alarm judgment period is obtained. Obtain the data comparison set corresponding to the alarm coefficient, which includes the standard eye-closing frequency. Through calculation formula The preliminary drowsiness alarm coefficient corresponding to the target staff was calculated. ,in This is represented as the standard eye-closing frequency. This represents the number of times the eyes were closed during the alarm detection period. This represents the duration of the alarm detection period. This represents the initial alarm coefficient corresponding to each instance of eye closure within the alarm judgment period. This represents the duration of each eye-closing action within the alarm detection period. This represents the preset standard time spent closing your eyes. This represents the number of each time eyes were closed within the alarm judgment period; Tracking and monitoring module: used to conduct reaction tests on target staff based on the initial drowsiness alarm coefficient corresponding to the target staff, and obtain the corresponding reaction coefficient of the target staff; The specific execution method of the tracking and monitoring module is as follows: Establish a data extraction relationship between the tracking and monitoring module and the database, and extract the fatigue inspection dataset corresponding to each preliminary drowsiness alarm coefficient stored in the database. The fatigue inspection dataset includes inspection character length, inspection character brightness, inspection character duration, and inspection character movement model. The fatigue inspection dataset for the target staff was obtained by filtering the initial drowsiness alarm coefficients for the target staff. Fatigue inspection is performed using the display terminal corresponding to the target worker and the fatigue inspection dataset corresponding to the target worker, and the display coordinate point and display time point of the initial inspection character corresponding to the display terminal are obtained. The eye movements of the target staff are detected by a pre-set visual camera in front of the target staff, and the eye movement data of the target staff is obtained. The eye movement data includes the direction and speed of eye movement at each time point. The eye movement data of the target staff after the time point of character display is filtered out and recorded as the eye movement data to be analyzed for the target staff. A computer vision model is established based on the eye movement data of the target worker and the work environment of the target worker. The computer vision model is used to obtain the visual target point, the position of the visual target point and the movement speed of the visual target point on the display terminal corresponding to the target worker. Extract the theoretical reaction time corresponding to each character length and character brightness combination stored in the database, and obtain the theoretical reaction time corresponding to the target staff by filtering the fatigue inspection dataset corresponding to the target staff. Visually match the visual target point of the display terminal corresponding to the target worker with the display coordinate point of the initial inspection character to obtain the time point when the visual target point and the display coordinate point of the initial inspection character coincide. Record this as the visual coincidence point corresponding to the target worker. Record the time interval between the display time point of the initial inspection character and the visual coincidence point corresponding to the target worker as the reaction time corresponding to the target worker. When the system recognizes that the visual target point coincides with the display coordinate point of the initial inspection character, the inspection character moves according to the inspection character movement model and movement speed, obtains the time point when the inspection character reaches the end point of the inspection character, obtains the time point when the visual target point reaches the end point of the inspection character, and records the time interval between the time point when the inspection character reaches the end point of the inspection character and the time point when the visual target point reaches the end point of the inspection character as the delay time corresponding to the target worker. Based on the position and movement speed of the visual target point, a visual movement model corresponding to the target worker is established. The visual movement model corresponding to the target worker is then compared with the inspection character movement model corresponding to the target worker to obtain the degree of overlap of the visual movement model corresponding to the target worker. According to the calculation formula The reaction coefficients corresponding to the target staff were calculated. ,in This represents the theoretical reaction time for the target staff member. This represents the reaction time for the target staff member. This represents the delay duration corresponding to the target staff member. This is represented by the degree of overlap between the visual movement models corresponding to the target worker; Comprehensive Analysis Module: This module is used to perform comprehensive analysis based on the initial drowsiness alarm coefficient, the reaction coefficient, and the detection data of the target staff member to obtain the drowsiness alarm result for the target staff member. The specific execution method of the comprehensive analysis module is as follows: Based on the yawning action model of the person stored in the database, the action is matched with the continuous mouth sub-images corresponding to the target staff to obtain each yawn of the continuous mouth sub-images corresponding to the target staff, and the duration and corner angle of each yawn are obtained. Filter alarms to determine the corresponding yawns within a given time period, as well as the duration and angle of the mouth for each yawn. Through calculation formula The comprehensive drowsiness alarm coefficient corresponding to the target staff was calculated. ,in This represents the number of yawns within the alarm judgment period. , These represent the duration and angle of the mouth for each yawn within the alarm judgment period, respectively. This represents the number of each yawn within the alarm judgment period; The overall drowsiness alarm coefficient corresponding to the target employee is compared and analyzed with the preset standard overall drowsiness alarm coefficient. If the overall drowsiness alarm coefficient corresponding to the target employee is greater than the preset standard overall drowsiness alarm coefficient, the target employee will be alarmed through the alarm device. If the overall drowsiness alarm coefficient corresponding to the target employee is less than or equal to the preset standard overall drowsiness alarm coefficient, the target employee will not be alarmed.

2. The drowsiness alarm system based on facial detection according to claim 1, characterized in that: The specific execution method of the working data acquisition module is as follows: Obtain the work logs of the target staff, and filter the work hours and number of events handled by the target staff based on the work logs and the detection time points. Record the work hours and number of events handled by the target staff as the work data of the target staff.

3. A drowsiness alarm system based on facial detection according to claim 2, characterized in that: The specific execution method of the data detection module is as follows: The target worker's face is continuously captured by a pre-set visual camera in front of the target worker, resulting in continuous facial images of the target worker. The facial images of the target worker are then divided into regions to obtain eye sub-images and mouth sub-images of the target worker. Finally, continuous eye sub-images and continuous mouth sub-images of the target worker are obtained by statistical analysis. The continuous eye sub-images and continuous mouth sub-images corresponding to the target worker are recorded as the detection data corresponding to the target worker.

4. A drowsiness alarm method based on facial detection, applied to the drowsiness alarm system based on facial detection as described in any one of claims 1-3, characterized in that: include: Obtain the work logs of the staff to get the work data of the target staff; Image detection is performed on the target staff to obtain the detection data of the target staff; By combining and analyzing the detection data and work data of the target staff, a preliminary drowsiness alarm coefficient corresponding to the target staff is obtained; Based on the initial drowsiness alarm coefficient of the target staff, a reaction test is conducted on the target staff to obtain the corresponding reaction coefficient of the target staff. Based on a comprehensive analysis of the preliminary drowsiness alarm coefficient, the reaction coefficient, and the detection data of the target staff, the drowsiness alarm result for the target staff is obtained.

Images