Intelligent monitoring method and system for fire control room

Abstract

The invention discloses an intelligent monitoring method and system for a fire-fighting control room, and belongs to the technical field of fire-fighting safety monitoring, and the method comprises the steps: obtaining a monitoring video in a fire-fighting monitoring room, and decoding the monitoring video into continuous image frames; acquiring head posture data, body posture data and face key point positions in the image frame; generating a comprehensive fatigue index according to the working duration, the head posture data, the body posture data and the face key point positions; according to the task response time, generating a worker task response time influence coefficient; generating a worker working state comprehensive risk index according to the worker task response time influence coefficient and the comprehensive body posture fatigue index; according to the comprehensive risk index of the working state of the worker, reminding the worker in the fire control room; according to the invention, the state of a worker can be monitored in real time, the risk index is generated, the reliability of a fire control room to fire emergency response is improved, and thus the risk that the optimal fire handling opportunity is missed is reduced.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of fire safety monitoring, and particularly relates to an intelligent monitoring method and system for a fire control room. BACKGROUND

[0002] The fire control room is the core hub of the building fire safety system, and undertakes key functions such as fire monitoring, alarm receiving, equipment linkage and emergency command. The control room staff need to maintain a high level of vigilance and quick response ability, and any state decline caused by fatigue, distraction and the like may delay the processing and cause serious consequences. Therefore, real-time and intelligent monitoring and evaluation of the working state of the fire control room staff, timely detection of fatigue or abnormal signs and intervention, have important significance for guaranteeing fire safety and improving the reliability of emergency response.

[0003] At present, the monitoring of the working state mainly depends on traditional management methods such as regular patrol and personnel self-checking. Regular patrol has the shortcomings of long interval period and inability to realize continuous monitoring, and it is difficult to capture the short-term fatigue or inattention of the staff during the patrol interval. Personnel self-checking is highly subjective and relies on personal consciousness, and its reliability and effectiveness are easily discounted in a long-term monotonous working environment. These traditional methods are difficult to realize objective, continuous and quantitative evaluation of the state of the staff, thereby causing delay in the response of the staff in the fire control room to the initial fire alarm signal and missing the best disposal opportunity of the fire alarm. SUMMARY

[0004] In view of the deficiencies of the prior art, the application provides an intelligent monitoring method and system for a fire control room, which solves the above problems.

[0005] To achieve the above purpose, the application realizes the following technical scheme: an intelligent monitoring method for a fire control room, specifically comprising the following steps:

[0006] Obtaining monitoring video of the staff in the fire control room when working at the workstation, and decoding the monitoring video into continuous image frames;

[0007] Obtaining head posture data, body posture data and facial key point positions of the staff in the image frames;

[0008] Obtaining the working time of the staff, establishing an abnormal analysis model of the staff according to the working time of the staff, the head posture data, the body posture data and the facial key point positions, and generating a comprehensive fatigue index;

[0009] Obtaining the task response time of the staff, and generating a task response time influence coefficient of the staff according to the task response time of the staff;

[0010] According to the staff task response time influence coefficient and the comprehensive body posture fatigue index, a staff work state comprehensive risk index is generated;

[0011] According to the staff work state comprehensive risk index, the staff of the fire control room is reminded.

[0012] On the basis of the above technical solutions, the application further provides the following optional technical solutions:

[0013] Further technical solutions: the generation mode of the comprehensive fatigue index specifically includes:

[0014] According to the staff work time, a basic fatigue index is generated;

[0015] According to the head posture data, the body posture data and the facial key point position of the staff, a body posture comprehensive analysis model is established, and a body posture fatigue index is generated;

[0016] According to the basic fatigue index and the body posture fatigue index, a staff abnormality analysis model is established, and a comprehensive fatigue index is generated.

[0017] Further technical solutions: the generation mode of the basic fatigue index specifically includes:

[0018] Through the formula:

[0019] ;

[0020] Generate the basic fatigue index B f ;

[0021] In the formula, T continuous represents the staff work time, T max represents the preset maximum work time of the staff, represents the fatigue accumulation rate coefficient.

[0022] Further technical solutions: the generation mode of the body posture fatigue index specifically includes:

[0023] According to the head posture data of the staff, a head movement fatigue factor is generated; wherein the head posture data includes the head pitch angle change amount, the head yaw angle change amount and the head roll angle change amount of the staff between adjacent image frames;

[0024] According to the body posture data of the staff, a body posture stability factor is generated; wherein the body posture data includes the shoulder key point position and the neck key point position;

[0025] According to the facial key point position of the staff, a facial action factor is generated;

[0026] According to the head movement fatigue factor, the body posture stability factor and the facial action factor, a body posture comprehensive analysis model is established, and a body posture fatigue index is generated.

[0027] Further technical solutions: the expression of the staff abnormality analysis model is specifically:

[0028] ;

[0029] In the formula, F represents a comprehensive fatigue index, B f represents a basic fatigue index, I f represents a body posture fatigue index, and w is a weight coefficient of the basic fatigue index.

[0030] Further technical solutions: the generation mode of the head movement fatigue factor specifically includes:

[0031] Through the formula:

[0032] ;

[0033] Generate a head movement entropy H head ;

[0034] In the formula, represents a head pitch angle change between the i-th image frame and the i-1-th image frame, represents a head yaw angle change between the i-th image frame and the i-1-th image frame, represents a head roll angle change between the i-th image frame and the i-1-th image frame, and N represents the total number of image frames.

[0035] The head movement entropy H head is normalized to obtain a normalized value of the head movement entropy; wherein the normalization method is maximum-minimum standardization.

[0036] Through the formula:

[0037] ;

[0038] Generate a head movement fatigue factor F h ;

[0039] In the formula, represents the normalized value of the head movement entropy, H0 represents a preset standard normalized value of the head movement entropy, and n is a constant term.

[0040] The generation mode of the body posture stability factor specifically includes:

[0041] Through the formula:

[0042] ;

[0043] generating a body posture stability factor S posture ;

[0044] In the formula, represents the variance of the shoulder key point position in all image frames, represents the variance of the neck key point position in all image frames, represents a stability threshold, and m is a constant term;

[0045] The face action factor is generated in the following manner:

[0046] According to the formula:

[0047] ;

[0048] generating a face action factor F micro ;

[0049] In the formula, F i represents the displacement amount of the face key point in the i-th image frame, represents the maximum standard displacement amount of the face key point, M max represents a preset number threshold of image frames with abnormal face key point positions, N is the total number of image frames, and I(·) is a conditional flag function.

[0050] Further technical solutions: The expression of the body posture comprehensive analysis model is specifically:

[0051] ;

[0052] In the formula, F h represents a head movement fatigue factor, S posture represents a body posture stability factor, F micro represents a face action factor, a1, a2, and a3 are weight coefficients, and a1+a2+a3=1.

[0053] Further technical solutions: The generating manner of the staff task response time influence coefficient specifically includes:

[0054] According to the formula:

[0055] ;

[0056] generating a staff task response time influence coefficient S t ;

[0057] In the formula, T responseT represents the actual response time of the staff standard T represents the standard response time threshold.

[0058] Further technical solutions: the generation mode of the staff working state comprehensive risk index specifically includes:

[0059] Through the formula:

[0060] ;

[0061] Generate the staff working state comprehensive risk index R;

[0062] In the formula, F represents the comprehensive fatigue index, S t T represents the staff task response time influence coefficient, 、 are weight coefficients, and .

[0063] An intelligent monitoring system of a fire control room, which is used to execute the intelligent monitoring method of the fire control room as described above, specifically includes:

[0064] An image acquisition unit is configured to acquire monitoring video in the fire monitoring room and decode the monitoring video into continuous image frames;

[0065] An image data acquisition unit is configured to acquire head posture data, body posture data and facial key point positions of the staff in the image frames;

[0066] A fatigue analysis unit is configured to acquire the working time of the staff, establish an abnormal analysis model of the staff according to the working time of the staff, the head posture data, the body posture data and the facial key point positions, and generate a comprehensive fatigue index;

[0067] A staff reaction analysis unit is configured to acquire the task response time of the staff, and generate a staff task response time influence coefficient according to the task response time of the staff;

[0068] A comprehensive analysis unit is configured to generate a staff working state comprehensive risk index according to the staff task response time influence coefficient and the comprehensive body posture fatigue index;

[0069] A reminding unit is configured to remind the staff in the fire control room according to the staff working state comprehensive risk index.

[0070] The present application provides an intelligent monitoring method and system of a fire control room, which has the following beneficial effects compared with the prior art:

[0071] The application acquires multi-dimensional posture data of the staff, combines work duration and task response time, establishes a comprehensive evaluation model, monitors the staff state in real time and generates a risk index, optimizes the response of the staff in the fire control room to the initial fire alarm signal, improves the reliability of the fire control room in fire emergency response, and reduces the risk of missing the best disposal opportunity of the fire alarm. BRIEF DESCRIPTION OF DRAWINGS

[0072] Figure 1 A flowchart of an intelligent monitoring method of a fire control room provided by the application is provided.

[0073] Figure 2 A flowchart of step S600 provided by the application is provided.

[0074] Figure 3 A flowchart of step S300 provided by the application is provided.

[0075] Figure 4 A flowchart of step S320 provided by the application is provided.

[0076] Figure 5 A structural schematic diagram of an intelligent monitoring system of a fire control room provided by the application is provided.

[0077] Figure 6 A module block diagram of a fatigue analysis unit provided by the application is provided.

[0078] Figure 7 A submodule block diagram of a body posture analysis module provided by the application is provided. DETAILED DESCRIPTION

[0079] In order to make the purpose, technical scheme and advantages of the application clearer, the application is further described in detail below in combination with the drawings and examples. It should be understood that the specific examples described herein are only used to explain the application and do not limit the application.

[0080] The specific implementation of the application is described in detail below in combination with specific examples.

[0081] Please refer to Figure 1 and Figure 2 An intelligent monitoring method of a fire control room provided by an embodiment of the application specifically includes the following steps:

[0082] Step S100: acquiring monitoring video of staff working at a workstation in a fire monitoring room, and decoding the monitoring video into continuous image frames;

[0083] Step S200: acquiring head posture data, body posture data and facial key point positions of the staff in the image frames;

[0084] Step S300: Obtain the worked hours of the staff, and establish a staff anomaly analysis model according to the worked hours of the staff, the head posture data, the body posture data, and the facial key point positions, and generate a comprehensive fatigue index;

[0085] Step S400: Obtain the task response time of the staff, and generate a staff task response time influence coefficient according to the task response time of the staff;

[0086] Step S500: Generate a staff work state comprehensive risk index according to the staff task response time influence coefficient and the comprehensive body posture fatigue index;

[0087] Step S600: Remind the staff in the fire control room according to the staff work state comprehensive risk index;

[0088] Among them, obtaining the monitoring video and decoding it into image frames means capturing the picture in the control room in real time through the video acquisition device, and converting it into a sequence of digital images that can be processed, providing a data basis for subsequent posture analysis;

[0089] The head posture data includes parameters such as pitch angle and yaw angle, which are used to quantify the amplitude of head movement and can be obtained by calculating the key point coordinates extracted by computer vision algorithms;

[0090] The body posture data relates to the spatial position changes of the shoulder and neck, such as tracking the displacement trajectory of the joint nodes using pose estimation algorithms;

[0091] The facial key point positions can be extracted by facial recognition technology to extract the displacement of facial key points; for example, the relative position of the upper lip (the relative position of the upper lip in the face area will change when the staff yawns, for example, the relative position of the upper lip and the cheekbone will change when yawning);

[0092] The worked hours refer to the length of time the staff is on duty continuously, which is used to assess the cumulative effect of basic fatigue;

[0093] The task response time refers to the time interval from receiving an instruction to actually performing an operation, reflecting the emergency response speed;

[0094] The comprehensive fatigue index is generated by integrating basic fatigue and real-time physiological behavior data, and is used to represent the degree of fatigue;

[0095] The task response time influence coefficient is used to quantify the influence weight of response delay on the overall risk;

[0096] The work state comprehensive risk index is generated by combining the fatigue index and the response coefficient, and is used as the basis for triggering the reminder.

[0097] Specifically, the method obtains head, body and face dynamic data through continuous image frame analysis, and establishes a basic fatigue model combined with working time. For example, head posture data can detect abnormal movements such as frequent nodding or head stillness, body posture data can identify states such as stiff shoulders and neck or frequent shaking, and face data can capture features such as yawning or long-time eye closure. The working time as a basic parameter of fatigue accumulation is input into the analysis model together with the real-time physiological behavior data to generate a comprehensive fatigue index. The task response time is calculated by comparing the actual value with the preset threshold value, and its contribution to the overall risk is calculated. Finally, the fatigue index and the response coefficient are weighted and fused to form a comprehensive risk index. When the index exceeds the preset threshold, a reminder instruction for the worker is triggered, so that the worker is reminded; the reminding mode includes but is not limited to sound and light prompts or management platform notifications.

[0098] Compared with the prior art, the traditional method relies on manual patrol and subjective self-checking, and has defects such as discrete data collection and non-uniform evaluation standard. The present application realizes continuous data collection through video analysis, eliminates monitoring blind spots; a quantitative model is established based on multi-dimensional physiological behavior characteristics to avoid misjudgment of a single indicator; the task response time is introduced as a dynamic correction factor to improve the real-time and accuracy of state evaluation. For example, the prior art cannot associate working time with real-time posture changes, while the present application considers long-term fatigue accumulation and instantaneous state fluctuation through model fusion, making the reminder mechanism more forward-looking.

[0099] Through the above technical solutions, the present application can capture the subtle changes in the head, body and face state of the worker in real time, and dynamically evaluate the working state risk level in combination with the working time and task response speed. For example, when the worker has a low head, long-time eye closure and delayed task response due to long-time duty, the system can timely identify that the comprehensive risk index is out of standard and trigger a reminder to avoid emergency response delay caused by fatigue. The method solves the problems that the traditional method cannot continuously monitor and the evaluation is highly subjective, and provides objective and quantitative decision support for state management of fire control room personnel.

[0100] Preferably, please refer to Figure 3 , the present application further proposes that the generation mode of the comprehensive fatigue index specifically includes:

[0101] Step S310: generating a basic fatigue index according to the working time of the worker;

[0102] Step S320: establishing a body posture comprehensive analysis model according to the head posture data, body posture data and face key point position of the worker, and generating a body posture fatigue index;

[0103] Step S330: establishing a worker abnormality analysis model according to the basic fatigue index and the body posture fatigue index, and generating a comprehensive fatigue index;

[0104] The basic fatigue index refers to a fatigue accumulation quantitative index established by the working time parameter, which can be specifically implemented by using an exponential function in combination with a preset maximum working time and a fatigue accumulation rate coefficient, and is used to reflect the physiological function decline law caused by continuous work.

[0105] The body posture fatigue index refers to a real-time state evaluation index established based on multi-dimensional dynamic sign data, which can be specifically implemented by head movement trajectory analysis, body key point stability detection and facial micro-expression (i.e., change of facial key point position) capture technology, and is used to capture action abnormalities caused by sudden fatigue.

[0106] The comprehensive fatigue index refers to a weighted calculation result of the basic fatigue and real-time sign data, which can be specifically implemented by using a linear weighting model in combination with a dynamically adjusted weight coefficient, and is used to balance the comprehensive evaluation of long-term load influence and short-term state fluctuation.

[0107] Specifically, the basic fatigue index quantifies the progressive influence of working time on fatigue degree through the nonlinear relationship between the preset maximum working time and the real-time working time. The body posture fatigue index calculates the motion entropy value through the change amount of the head pitch angle, yaw angle and roll angle, analyzes the body stability in combination with the shoulder and neck key point position variance, and forms a multi-modal sign fusion evaluation by counting the abnormal frequency of facial key point displacement. The comprehensive fatigue index superimposes the basic fatigue and real-time sign data through a weight distribution model, so that the long-term cumulative fatigue and sudden posture abnormalities can jointly act on the final evaluation result.

[0108] Compared with the prior art, the traditional method only relies on working time or single sign data for isolated evaluation, while the present application processes the basic fatigue and real-time sign data in stages through hierarchical modeling, which not only retains the continuous influence law of working time on fatigue degree, but also captures sudden abnormalities such as attention distraction and slow action by using multi-dimensional sign data, thereby realizing dynamic monitoring of fatigue state in all time periods.

[0109] Through the above technical solutions, the present application solves the technical problem that the comprehensive fatigue state cannot be quantitatively evaluated in the traditional monitoring, and can accurately evaluate the fatigue phenomenon caused by working time accumulation and real-time posture abnormalities, thereby providing an objective evaluation basis for the state judgment of fire control room staff through multi-dimensional fusion, and avoiding the misjudgment or omission of single data source.

[0110] Preferably, the present application further proposes that the generation mode of the basic fatigue index specifically comprises:

[0111] Through the formula:

[0112] ​

[0113] Generate the basic fatigue index B f ;

[0114] In the formula, T continuous represents the worked time of the staff, T max represents the preset maximum working time of the staff, represents the fatigue accumulation rate coefficient;

[0115] wherein the worked time T continuous refers to the length of time during which the staff continuously performs tasks during on-duty, which can be realized by recording the time difference from the beginning of work to the current time by using a timer module, and this parameter is used to quantify the cumulative effect of working time on fatigue degree;

[0116] The preset maximum working time T max refers to the upper limit of continuous working time of the post set according to ergonomics, which can be realized by using the maximum allowable value specified in the industry standard or post specification, and this parameter is used to establish a benchmark reference value for fatigue assessment;

[0117] The fatigue accumulation rate coefficient refers to the adjustment factor reflecting the growth speed of fatigue under different working intensities, which can determine the numerical range by experimental data fitting or post characteristic analysis, for example, the coefficient of high-intensity post can be set to 0.8-1.2, and the coefficient of medium-intensity post can be set to 0.5-0.8, and this parameter is used to dynamically adjust the growth curve of fatigue index with working time.

[0118] Specifically, the generation process of the basic fatigue index establishes a nonlinear correlation between working time and fatigue degree through an exponential function model. When the worked time approaches the preset maximum value, the exponential term tends to 0, and the basic fatigue index B f approaches the limit value 1, which conforms to the physiological law that human fatigue accelerates with working time. The normalized ratio eliminates the influence of the absolute value of individual working time, making the evaluation results of different posts comparable. By adjusting the fatigue accumulation rate coefficient, differentiated fatigue growth curves can be generated for different task intensity posts in the fire control room, such as the monitoring post during the high-incidence period of fire and the equipment inspection post during the regular period, so as to more accurately reflect the influence of actual work load on physiological state.

[0119] Compared with the prior art, the conventional method usually calculates the fatigue value by using a linear cumulative model, and cannot reflect the non-linear characteristics of the accelerated deterioration of human fatigue in the later period of long-time work. The present application can accurately depict the curve change trend of the fatigue degree with the increase of the working time by modeling with an exponential function, especially significantly improves the fatigue index when approaching the maximum working time, and is more in line with the actual physiological response law. In addition, the fatigue evaluation parameters in the prior art are mostly fixed values, which cannot be adapted to different working environments, while the present application introduces an adjustable fatigue accumulation rate coefficient, which can dynamically optimize the model parameters according to the actual working intensity of the fire control room, and improve the scene adaptability of the evaluation results.

[0120] Through the above technical solutions, the present application realizes dynamic quantitative evaluation of the basic fatigue degree of the fire control room staff, and solves the evaluation deviation problem caused by linear modeling in the conventional method. The fatigue accumulation effect is accurately reflected by using a non-linear function, and the false alarm or missed alarm caused by fixed threshold judgment is avoided, for example, timely triggering a reminder when the staff continuously work close to the maximum time. The parameter dynamic adjustment mechanism enables the model to adapt to different working intensity scenes, for example, automatically increasing the value of the fatigue accumulation rate coefficient during the fire emergency response, so as to more sensitively capture the state change under high-intensity work.

[0121] Preferably, referring to Figure 4 The present application further proposes that the generation mode of the body posture fatigue index specifically comprises:

[0122] Step S321: generating a head movement fatigue factor according to the head posture data of the staff; wherein the head posture data comprises the head pitch angle change amount, the head yaw angle change amount and the head roll angle change amount of the staff between adjacent image frames;

[0123] Step S322: generating a body posture stability factor according to the body posture data of the staff; wherein the body posture data comprises the shoulder key point position and the neck key point position;

[0124] Step S323: generating a facial action factor according to the facial key point position of the staff;

[0125] Step S324: establishing a body posture comprehensive analysis model according to the head movement fatigue factor, the body posture stability factor and the facial action factor, and generating a body posture fatigue index;

[0126] The head posture data refers to the detected head pitch angle, yaw angle and roll angle change amount between adjacent image frames, and specifically, a three-dimensional posture estimation algorithm can be used to extract Euler angle parameters for quantifying the amplitude of the slight tremor or abnormal deflection of the head;

[0127] The body posture data refers to a position coordinate sequence of key points of a shoulder and neck part, and can be specifically obtained by a human body skeleton key point detection algorithm, and is used for analyzing a stability degree of a body posture deviating from a reference state.

[0128] The facial key point position refers to a displacement amount of a facial key point, and can be specifically captured by using a facial feature point tracking technology to track an activity feature of the facial key point, and is used for identifying facial expression changes caused by fatigue, and specifically includes, but is not limited to, facial expression changes caused by yawning.

[0129] Specifically, the head motion entropy is calculated by using a head three-dimensional angle change amount of continuous image frames, a feature of frequent nodding or static of the head caused by fatigue is captured, the posture stability is calculated based on a shoulder and neck key point position variance, body tilt or swing abnormalities are identified, and the micro-expression activity is analyzed by using a facial key point (for example, a position of an upper lip) displacement statistic. The three types of features are used to construct quantitative indexes from the dimensions of kinematics, posturology and expression, single-dimension data noise interference is eliminated by using weighted fusion, and a mathematical model that comprehensively reflects a body posture fatigue degree is established.

[0130] Compared with the prior art, the traditional monitoring method usually only uses a single posture parameter or facial expression recognition, for example, only monitors a head tilt angle or a blink frequency, and is easy to make a misjudgment due to environmental light change or temporary action interference. The prior art lacks a cooperative analysis mechanism for multiple source biomechanical features, and cannot distinguish between normal operation actions and abnormal postures caused by fatigue.

[0131] By using the technical solution, the problem of insufficient body posture related fatigue evaluation precision in the traditional monitoring method is effectively solved, dynamic quantitative analysis of head motion of a worker, body stability and facial expression is realized, the accuracy and real-time performance of fatigue state recognition are significantly improved, and reliable technical support is provided for fire control room personnel state monitoring.

[0132] Preferably, the expression of the worker abnormality analysis model is specifically as follows:

[0133] ;

[0134] In the formula, F represents a comprehensive fatigue index, B f represents a basic fatigue index, I f represents a body posture fatigue index, and w is a weight coefficient of the basic fatigue index.

[0135] The basic fatigue index refers to a fatigue accumulation degree calculated based on a working time length, and can be specifically realized by using an exponential decay function combined with a preset maximum working time length, and is used for reflecting physiological load caused by continuous work.

[0136] The body posture fatigue index refers to a real-time state index generated by head posture, body posture and facial state data, which can be specifically realized by multi-dimensional motion entropy analysis combined with key point displacement statistics, and is used for capturing sudden posture abnormalities.

[0137] The weight coefficient refers to a parameter for adjusting the proportion of the basic fatigue index and the body posture fatigue index in the comprehensive evaluation, which can be specifically realized by a dynamic adjustment mechanism, for example, automatically optimizing the weight distribution according to the work stage or the environmental risk level.

[0138] Specifically, the calculation of the comprehensive fatigue index fuses the basic fatigue index and the body posture fatigue index by a linear weighting method. The basic fatigue index reflects the objective cumulative trend of fatigue through the ratio of the working time length to the preset threshold, while the body posture fatigue index reflects the real-time state fluctuation through the head motion entropy, body stability variance and facial micro-motion frequency. The weight coefficient is set as an adjustable parameter, for example, set as 0.6 during the regular monitoring period to focus on the basic fatigue evaluation, and adjusted to 0.3 after the fire alarm is triggered to enhance the sensitivity of the real-time posture data. This fusion mechanism makes the model pay more attention to the influence of the working time length when the worker is in a long-time work but stable posture, and automatically increases the evaluation weight of the real-time data when there are frequent head shaking, body tilting or facial expression abnormalities.

[0139] Compared with the prior art, the traditional method usually only uses a single index for fatigue evaluation, for example, only monitors the working time length or only analyzes the head posture. While the present application establishes a dynamic weighting model of the basic fatigue index and the body posture fatigue index, covering both the long-term cumulative effect and the short-term abnormal state of fatigue. The defects of fixed weight or single-dimensional analysis in the prior art are eliminated, for example, the head movement caused by normal operation cannot be distinguished from the posture instability caused by fatigue, while the present application effectively reduces the misjudgment rate through multi-dimensional data fusion and dynamic weight adjustment.

[0140] Through the above technical solutions, the present application solves the limitations of single-index evaluation and realizes the fine evaluation of the fatigue state of the fire control room workers. By fusing the working time length data and real-time posture features, it can accurately identify the decline in attention caused by long-term work and sudden posture abnormalities. The dynamic weight mechanism further improves the adaptability of the model, ensuring that posture abnormalities are captured preferentially during high fire periods, and fatigue accumulation is emphasized during the regular monitoring stage.

[0141] Preferably, the present application further proposes that the generation method of the head motion fatigue factor specifically includes:

[0142] Through the formula:

[0143] ;

[0144] Generate head motion entropy H head ;

[0145] In the formula, represents the change of head pitch angle between the i-th image frame and the i-1-th image frame, represents the change of head yaw angle between the i-th image frame and the i-1-th image frame, represents the change of head roll angle between the i-th image frame and the i-1-th image frame, and N represents the total number of image frames.

[0146] The head motion entropy H head is normalized to obtain a normalized value of the head motion entropy; wherein the normalization method is maximum-minimum standardization.

[0147] Through the formula:

[0148] ;

[0149] Generate head motion fatigue factor F h ;

[0150] In the formula, represents the normalized value of the head motion entropy, H0 represents the preset standard normalized value of the head motion entropy, and n is a constant term.

[0151] The generation method of the body posture stability factor specifically includes:

[0152] Through the formula:

[0153] ;

[0154] Generate body posture stability factor S posture ;

[0155] In the formula, represents the variance of the position of the shoulder key point in all image frames, represents the variance of the position of the neck key point in all image frames, represents a stability threshold, and m is a constant term.

[0156] The generation method of the face action factor specifically includes:

[0157] Through the formula:

[0158] ;

[0159] Generate face action factor F micro ;

[0160] In the formula, Fi represents the displacement amount of the face key point in the i-th image frame, represents the maximum value of the standard displacement amount of the face key point, M max represents the preset number threshold of the image frame with abnormal face key point position, N is the total number of image frames, and I(·) is a conditional flag function;

[0161] wherein the head motion entropy refers to the cumulative amount of head posture change between consecutive image frames to quantify the abnormality degree of head motion, and can be specifically calculated by the root mean square of three-dimensional angle change, and is used to reflect the unnatural shaking or static state of the head of the staff due to fatigue;

[0162] wherein the normalization processing refers to adjusting the numerical range of the head motion entropy to a preset interval, and can be specifically implemented by linear scaling or standard deviation normalization method, and is used to eliminate the influence of the motion amplitude difference of different individuals on fatigue evaluation;

[0163] The body posture stability factor refers to evaluating the stability state of the body posture by the dispersion degree of the positions of the shoulders and the neck key points in the time sequence, and can be specifically implemented by variance calculation combined with stability threshold comparison, and is used to detect the frequent shaking or tilting of the body caused by fatigue;

[0164] The face motion factor refers to quantifying the abnormal frequency of the micro-motions of the face by counting the number of abnormal image frames with face key point displacement exceeding the preset threshold, and can be specifically implemented by key point tracking technology and threshold comparison method, and is used to capture fatigue features such as eye abnormality (key point position abnormality caused by long-time closed eyes) or facial expression abnormality (key point position abnormality caused by yawning);

[0165] Specifically, by continuously collecting image frame sequences of the monitoring video, the head posture angle change amount of each frame image is calculated. For example, the root mean square operation is performed on the three-dimensional angle change amount of the pitch angle, yaw angle and roll angle to generate an entropy value reflecting the complexity of head movement. After normalization processing, the entropy value is compared with a preset standard motion entropy value. When the staff appears frequent head shaking or long-term static due to fatigue, the motion entropy will deviate significantly from the standard value, thereby generating a corresponding head movement fatigue factor. At the same time, the coordinate data of the shoulder and neck key points are extracted by the human posture recognition technology, and the position variance of the key points in the continuous image frames is calculated. For example, when the staff causes frequent forward leaning or left and right shaking due to fatigue, the position variance of the shoulder and neck key points will exceed the stability threshold, generating a body posture stability factor representing unstable posture. In addition, the displacement of the facial key points is tracked frame by frame, and the number of abnormal frames with displacement exceeding the standard displacement threshold is counted. For example, when the staff causes eye blinking interval to be extended or facial muscles to be relaxed due to drowsiness, the proportion of abnormal frames will significantly increase, generating a corresponding facial action factor. Through the cooperative calculation of the above three types of factors, multi-dimensional quantitative evaluation of the fatigue state is formed.

[0166] Compared with the prior art, the traditional method relies on manual observation or single-dimensional action detection, and cannot distinguish between normal operation and abnormal posture changes caused by fatigue. For example, the prior art may only monitor the head static state, but cannot identify the abnormal mode of sudden head shaking recovery caused by drowsiness. However, the present application can effectively distinguish between normal operation and non-autonomous movement caused by fatigue by capturing the time sequence characteristics of head movement through motion entropy calculation, combined with body posture variance analysis and facial micro-motion statistics, solving the problem of high false positive rate and insufficient sensitivity of the traditional method.

[0167] Through the above technical solutions, the present application realizes precise quantitative analysis of the head abnormal movement, body posture instability and facial micro-motion change of the fire control room staff, solving the technical defect that the traditional monitoring method cannot objectively identify fatigue characteristics.

[0168] Preferably, the present application further proposes an expression of the body posture comprehensive analysis model, which is specifically:

[0169] ;

[0170] In the formula, F h represents the head movement fatigue factor, S posture represents the body posture stability factor, F micro represents the facial action factor, a1, a2 and a3 are weight coefficients, and a1+a2+a3=1.

[0171] The head movement fatigue factor refers to a quantitative index generated by calculating the change amount of the head pitch angle, yaw angle and roll angle between adjacent image frames, and can be specifically realized by adopting the Euclidean distance to accumulate and average the three-dimensional angle change amount and then normalizing the same, and is used to reflect the head shaking or abnormal posture caused by fatigue;

[0172] The body posture stability factor refers to a stability index calculated based on the variance of the positions of the key points of the shoulders and the neck, and can be specifically realized by the ratio of the variance of the positions of the key points in the image frame sequence to a preset stability threshold, and is used to represent the stability attenuation degree of the body posture;

[0173] The facial action factor refers to an index generated by counting the displacement amount of the facial key points, and can be specifically realized by adopting the facial key point tracking algorithm in combination with the displacement amount threshold, and is used to identify the facial micro-motion abnormality caused by the distraction of attention.

[0174] Specifically, the technical scheme integrates the data of the head, body and face in three dimensions by the linear weighting method to generate the body posture fatigue index. The head movement fatigue factor captures the frequent adjustment or abnormal shaking of the head caused by fatigue by calculating the change amount of the three-dimensional angle of the head in the continuous image frames. The body posture stability factor quantifies the stability decline trend of the body posture by analyzing the variance of the positions of the key points of the shoulders and the neck. The facial action factor reflects the facial action abnormality caused by the distraction of attention or fatigue by counting the number of displacement amount abnormalities of the facial key points. The constraint of the weight coefficient enables the model to dynamically adjust the contribution degree of each index according to different scenes.

[0175] Compared with the prior art, the traditional method usually only relies on a single index such as the head posture or the body inclination angle for fatigue evaluation, for example, only monitors the frequency of the head drooping or the body inclination angle threshold. Such a method is easy to ignore the relevance of multi-dimensional fatigue features. The present application constructs a comprehensive analysis model by integrating the data of the head movement, body stability and facial action in three dimensions, which can more comprehensively capture the performance differences of the fatigue state in different body parts and avoid the misjudgment or omission caused by a single index.

[0176] Through the above technical scheme, the present application solves the problem of inaccurate fatigue evaluation caused by relying on a single body posture index in the traditional method, and realizes the fine evaluation of the fatigue state of the worker. Through the dynamic weighted integration of multi-dimensional data, the model can adapt to the needs of different working scenes, for example, in the night shift scene, the sensitivity of the facial action factor is enhanced, thereby improving the accuracy and scene adaptability of the fatigue judgment.

[0177] Preferably, the present application further proposes a generation method of the worker task response time influence coefficient, which specifically comprises:

[0178] By formula:

[0179] ;

[0180] Generate staff task response time influence coefficient S t ;

[0181] In the formula, T response Actual response time of staff, T standard Standard response time threshold value;

[0182] Wherein, the actual response time T response The time interval from receiving task instruction to starting operation of staff, specifically, the time difference between task triggering time and operation starting time can be recorded by timer, which is used to quantify the reaction speed of staff to sudden task;

[0183] The standard response time threshold T standard It refers to the maximum response time allowed, which can be determined according to the response time requirement specified in the fire control room operation procedure, which is used as the reference value to judge whether the response efficiency meets the standard.

[0184] Specifically, when the actual response time does not exceed the standard response time threshold, the influence coefficient remains 1, indicating that the response efficiency meets the safety operation requirement; when the actual response time exceeds the standard threshold, the influence coefficient is calculated by using exponential function, so that the ratio of the time difference exceeding the standard to the standard threshold is used as the input variable of the exponential function. Through the nonlinear attenuation characteristics of the exponential function, the more the response time exceeds the standard, the greater the influence coefficient decreases, so as to accurately reflect the risk accumulation effect caused by response delay in comprehensive risk assessment.

[0185] Compared with the prior art, the traditional method usually uses fixed threshold to judge whether the response is overdue, only makes binary qualified determination, and cannot quantify the influence difference of different overtime on overall risk. While the present application constructs a continuous dynamic evaluation model through exponential function, which can accurately reflect the nonlinear relationship between response time and risk level, and solves the problem of insufficient sensitivity of traditional method to slight overtime.

[0186] Through the above technical scheme, the present application realizes dynamic quantitative evaluation of response efficiency of fire control room staff, so that the deviation degree of response time can be accurately mapped into comprehensive risk index. This evaluation method avoids misjudgment caused by inaccurate identification of response delay, ensures the timeliness and accuracy of work state risk reminder, and effectively prevents safety hazards caused by response delay.

[0187] Preferably, the application further proposes that the generation mode of the staff working state comprehensive risk index specifically comprises:

[0188] The staff working state comprehensive risk index R is generated by the formula:

[0189] ;

[0190] The staff working state comprehensive risk index R is generated by the formula:

[0191] In the formula, F represents a comprehensive fatigue index, S t represents a staff task response time influence coefficient, , are weight coefficients, and ;

[0192] The comprehensive fatigue index R is a quantitative index established by head posture, body posture, facial state and working time data, and can be specifically implemented by a multi-source physiological data fusion model, and is used to reflect the physiological fatigue degree of the staff due to long-time work.

[0193] The task response time influence coefficient is a quantitative parameter generated based on the comparison between the actual response time and the preset standard threshold, and can be specifically implemented by a piecewise function or an exponential decay model, and is used to measure the reaction efficiency of the staff to the sudden task; the greater the actual response time is greater than the standard response time, the smaller the staff task response time influence coefficient is, and the greater the actual response time is greater than the standard response time, the greater the staff working state comprehensive risk index R is through the term (1-S t ).

[0194] The weight coefficient is a proportional factor used to adjust the contribution degree of fatigue state and response efficiency to the comprehensive risk, and can be specifically implemented by an empirical value or a dynamic adjustment algorithm, and is used to adapt to the risk assessment requirements in different working scenarios.

[0195] Specifically, the technical scheme linearly weights the comprehensive fatigue index and the task response time influence coefficient to construct a multi-dimensional risk assessment model. The comprehensive fatigue index F integrates physiological posture changes and working time data, is calculated by a preset abnormal analysis model, and can capture the decline of body function caused by fatigue. The task response time influence coefficient S t is generated according to the degree of deviation of the actual response time from the standard threshold, and when the response time exceeds the standard threshold, the value thereof exponentially decays with the delay time, so that the term (1-S t ) converts the response delay into a risk increment. The comprehensive risk index R generated thereby can reflect the dual risks of physiological fatigue accumulation and operation efficiency decline, and provides a quantitative basis for real-time reminding.

[0196] In some embodiments, the calculation of the comprehensive fatigue index can combine head movement entropy, body posture stability factors and other sub-indicators, such as by weighted summation to fuse each sub-indicator. The task response time influence coefficient can be handled by a segmented function, for example, the coefficient is 1 when the response time is less than or equal to the standard threshold, and is attenuated according to an exponential function when it exceeds. The weight coefficient can be dynamically optimized based on historical data, for example, using a machine learning model to automatically adjust the weight coefficient according to the environmental complexity and .

[0197] Compared with the prior art, the traditional method relies more on single-dimensional subjective evaluation or fixed threshold alarm, such as only monitoring the working time or separately analyzing the response time. While the present application builds a multi-dimensional dynamic evaluation model by fusing physiological fatigue data and task response efficiency data, solving the problem of one-sidedness of single indicator evaluation. The response time evaluation in the prior art usually adopts binary judgment, while the present application realizes continuous quantification of response delay degree through (1-S t ) term, which can more finely reflect the risk change trend.

[0198] Through the above technical solutions, the present application realizes objective and continuous quantitative evaluation of the working state of the fire control room staff, overcoming the monitoring blind area problem caused by the dependence of the traditional method on manual patrol. By dynamically adjusting the weight proportion of fatigue and response efficiency, the risk evaluation model can adapt to different work intensity scenarios, such as increasing the fatigue weight coefficient during night shift. The final generated comprehensive risk index provides a quantifiable decision basis for the prompting system.

[0199] Preferably, please refer to Figure 5 , the present application further proposes an intelligent monitoring system for fire control room, which is used to execute the intelligent monitoring method for fire control room as described above, specifically comprising:

[0200] An image acquisition unit 100 is used to acquire monitoring video in the fire monitoring room, and decode the monitoring video into continuous image frames;

[0201] An image data acquisition unit 200 is used to acquire head posture data, body posture data and facial key point position of the staff in the image frames;

[0202] A fatigue analysis unit 300 is used to acquire the working time of the staff, and establish an abnormal analysis model of the staff according to the working time of the staff, the head posture data, the body posture data and the facial key point position, to generate a comprehensive fatigue index;

[0203] A staff reaction analysis unit 400 is used to acquire the task response time of the staff, and generate a staff task response time influence coefficient according to the task response time of the staff;

[0204] The comprehensive analysis unit 500 is configured to generate a staff working state comprehensive risk index according to the staff task response time influence coefficient and the comprehensive body posture fatigue index;

[0205] The reminding unit 600 is configured to remind the staff in the fire control room according to the staff working state comprehensive risk index.

[0206] Preferably, referring to Figure 6 The fatigue analysis unit 300 further includes

[0207] The working time analysis module 310 is configured to generate a basic fatigue index according to the staff working time;

[0208] The body posture analysis module 320 is configured to establish a body posture comprehensive analysis model according to the head posture data, the body posture data and the facial key point position of the staff, and generate a body posture fatigue index;

[0209] The comprehensive fatigue index generation module 330 is configured to establish a staff abnormality analysis model according to the basic fatigue index and the body posture fatigue index, and generate a comprehensive fatigue index.

[0210] Preferably, referring to Figure 7 The body posture analysis module 320 further includes

[0211] The head analysis submodule 321 is configured to generate a head movement fatigue factor according to the head posture data of the staff; wherein the head posture data includes a head pitch angle change amount, a head yaw angle change amount and a head roll angle change amount of the staff between adjacent image frames;

[0212] The posture stability analysis submodule 322 is configured to generate a body posture stability factor according to the body posture data of the staff;

[0213] The face analysis submodule 323 is configured to generate a facial action factor according to the facial key point position of the staff;

[0214] The body posture fatigue index generation submodule 324 is configured to establish a body posture comprehensive analysis model according to the head movement fatigue factor, the body posture stability factor and the facial action factor, and generate a body posture fatigue index.

[0215] Although the embodiments of the present application have been shown and described, it is to be understood that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. A method for intelligent monitoring of a fire control room, characterized in that, Specifically comprising the following steps: Obtaining monitoring video of a worker in a fire control room when working at a workstation, and decoding the monitoring video into continuous image frames; Obtaining head posture data, body posture data and facial key point positions of the worker in the image frames; Obtaining a worked time length of the worker, and establishing a worker abnormality analysis model according to the worked time length, the head posture data, the body posture data and the facial key point positions of the worker, and generating a comprehensive fatigue index; Obtaining a task response time of the worker, and generating a worker task response time influence coefficient according to the task response time of the worker; Generating a worker working state comprehensive risk index according to the worker task response time influence coefficient and the comprehensive body posture fatigue index; According to the worker working state comprehensive risk index, reminding the worker in the fire control room.

2. The intelligent monitoring method of a fire control room according to claim 1, characterized in that, The generation mode of the comprehensive fatigue index specifically comprises: Generating a basic fatigue index according to the worked time length of the worker; Establishing a body posture comprehensive analysis model according to the head posture data, the body posture data and the facial key point positions of the worker, and generating a body posture fatigue index; Establishing a worker abnormality analysis model according to the basic fatigue index and the body posture fatigue index, and generating a comprehensive fatigue index.

3. The intelligent monitoring method of the fire control room according to claim 2, characterized in that, The generation mode of the basic fatigue index specifically comprises: Through the formula: ； Generating a base fatigue index B f ; In the formula, T continuous represents the worked hours of the worker, T max represents the preset maximum working hours of the worker, represents the fatigue accumulation rate coefficient.

4. The intelligent monitoring method of the fire control room according to claim 2, characterized in that, The generation mode of the body posture fatigue index specifically comprises: Generating a head movement fatigue factor according to the head posture data of the worker; wherein the head posture data comprises a head pitch angle change amount, a head yaw angle change amount and a head roll angle change amount of the worker between adjacent image frames; Generating a body posture stability factor according to the body posture data of the worker; wherein the body posture data comprises a shoulder key point position and a neck key point position; Generating a facial action factor according to the facial key point positions of the worker; Establishing a body posture comprehensive analysis model according to the head movement fatigue factor, the body posture stability factor and the facial action factor, and generating a body posture fatigue index.

5. The intelligent monitoring method of a fire control room according to claim 2, wherein, The expression of the worker abnormality analysis model is specifically: ； In the formula, F represents a comprehensive fatigue index, B f represents a basic fatigue index, I f represents a body posture fatigue index, and w is a weight coefficient of the basic fatigue index.

6. The intelligent monitoring method of a fire control room according to claim 4, characterized in that, The generation mode of the head movement fatigue factor specifically comprises: Through the formula: ； generating head motion entropy H head ; In the formula, represents the head pitch angle change between the i-th image frame and the i-1-th image frame, represents the head yaw angle change between the i-th image frame and the i-1-th image frame, represents the head roll angle change between the i-th image frame and the i-1-th image frame, and N represents the total number of image frames. Head movement entropy H head The normalized value of the head movement entropy is obtained by normalization processing; wherein the normalization manner is maximum-minimum standardization. Through the formula: ； generating a head motion fatigue factor F h ; In the formula, H0 represents a preset head motion entropy standard normalized value, and n is a constant term. Wherein, the generation mode of the body posture stability factor specifically comprises: Through the formula: ； Generating a body pose stability factor S posture ; In the formula, represents the variance of the position of the shoulder key point in all image frames, represents the variance of the position of the neck key point in all image frames, represents a stability threshold, and m is a constant term. Wherein, the generation mode of the facial action factor specifically comprises: Through the formula: ； Generating a facial action factor F micro ; In the formula, F i represents the displacement amount of the face key point in the i-th image frame, represents the maximum value of the standard displacement amount of the face key point, M max represents the preset number threshold of the image frame with abnormal face key point position, N is the total number of image frames, and I(·) is a conditional flag function.

7. The intelligent monitoring method of a fire control room according to claim 4, characterized in that, The expression of the body posture comprehensive analysis model is specifically: ； In the formula, F h represents a head movement fatigue factor, S posture represents a body posture stability factor, F micro represents a facial movement factor, a1, a2, and a3 are weight coefficients, and a1+a2+a3=1.

8. The intelligent monitoring method of a fire control room according to claim 1, characterized in that, The generation mode of the worker task response time influence coefficient specifically comprises: Through the formula: ； A generation worker task response time influence coefficient S t ; In the equation, T response represents the actual response time of the worker, T standard represents the standard response time threshold.

9. The intelligent monitoring method of a fire control room according to claim 1, characterized in that, The generation mode of the worker working state comprehensive risk index specifically comprises: Through the formula: ； Generating a worker working state comprehensive risk index R; In the formula, F represents the comprehensive fatigue index, S t represents the worker task response time influence coefficient, , are weight coefficients, and .

10. An intelligent monitoring system for a fire control room, characterized in that, The system is used for executing the intelligent monitoring method of the fire control room according to any one of claims 1-9, and specifically comprises: An image acquisition unit, configured to obtain monitoring video in a fire control room, and decode the monitoring video into continuous image frames; An image data acquisition unit, configured to obtain head posture data, body posture data and facial key point positions of a worker in the image frames; The fatigue analysis unit is configured to acquire a worked time length of the staff, establish a staff anomaly analysis model according to the worked time length of the staff, the head posture data, the body posture data and the facial key point positions, and generate a comprehensive fatigue index; The staff reaction analysis unit is configured to acquire a task response time of the staff, and generate a staff task response time influence coefficient according to the task response time of the staff; The comprehensive analysis unit is configured to generate a staff working state comprehensive risk index according to the staff task response time influence coefficient and the comprehensive body posture fatigue index; The reminding unit is configured to remind the staff in the fire control room according to the staff working state comprehensive risk index.

Images