Working area safety early warning method

Abstract

The invention discloses a working area safety early warning method, and relates to the technical field of safety early warning. The method comprises the steps of collecting operation tracks and interaction records of workers in a working area, calculating a correlation coefficient through behavior correlation analysis, extracting a risk preference coefficient and a potential cooperation coefficient, and constructing a behavior feature mapping library; real-time interaction of workers is split into simulation and cooperation behaviors, a group simulation and cooperation total value and a concentration coefficient are calculated, cooperation simulation individuals are screened, a group synchronization model is constructed based on a mean field theory, and a simulation orbit and a cooperation orbit violation synchronization rate are split; a dynamic game model is constructed through the compliance hedging coefficient, and violation or compliance game advantages are judged; when compliance is dominant, constructing a screening and reducing early warning model based on the decision tree, and filtering invalid early warning under compliance support; according to the invention, a group violation transmission source can be positioned, the invalid early warning rate is reduced, and the early warning response time is shortened.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of safety warning, in particular to a working area safety warning method. BACKGROUND

[0002] The safety warning of working area (especially the area involving high-risk operation) is the key of safety management. Although the existing technology has realized automatic monitoring through video monitoring identification and sensor collection, and can obtain worker operation trajectory, illegal behavior data and give a warning based on static threshold, there are still significant limitations.

[0003] The existing technology only focuses on individual illegal identification and cannot deal with illegal transmission in group interaction. In the interoperation, workers have imitation and cooperative behavior. The existing technology cannot separate these two types of behavior, cannot identify the core individual leading to illegal transmission, and it is more difficult to quantify the influence of double tracks on group illegal synchronization. The existing technology often waits for the group illegal to form a scale before warning, which is obviously lagging.

[0004] The existing technology does not consider the hedging effect of compliance behavior. The teaching and reminding of compliant workers can inhibit the spread of illegal behavior, but the existing technology only triggers the warning according to illegal data. When compliance can offset the risk, the existing technology still gives a warning, resulting in many invalid warnings, wasting management resources and reducing the response efficiency of real high-risk warning.

[0005] Therefore, the present application provides a working area safety warning method. SUMMARY

[0006] The present application aims to provide a working area safety warning method to solve at least one of the above-mentioned problems of the prior art.

[0007] A working area safety warning method, comprising the following steps: Collecting worker operation trajectory and interaction record data of the working area, mining the implicit features of the operation trajectory and the interaction record through behavior correlation analysis, and constructing a behavior feature mapping library; Based on the behavior feature mapping library, the real-time interaction of the workers is separated into imitation and cooperative behavior, and the cooperative imitation individuals are screened. Based on the cooperative imitation individuals, a group synchronization model is constructed to extract the illegal synchronization rate. The illegal synchronization rate is double-tracked and separated to obtain the illegal synchronization rate of the imitation track and the cooperative track; Based on the illegal synchronization rate of the imitation track and the cooperative track, the compliance behavior hedging analysis is performed to obtain the compliance hedging coefficient of the imitation track and the cooperative track and construct a dynamic game model to determine the game advantage type; If the compliance hedging is dominant, a screening warning model is constructed to identify the illegal synchronization judgment and optimize the illegal synchronization judgment to filter the invalid warning under the support of compliance.

[0008] As a further technical solution of the present application, the behavior correlation analysis is used as follows: The operation trajectory data and interaction record data of workers in high-risk areas are collected, the high-risk stay ratio and violation correction rate in the operation trajectory data are extracted, and the correlation coefficients of the high-risk stay ratio and the violation correction rate are calculated The imitated rate and the synergy dominance in the interaction record data are extracted, the non-imitated rate is obtained by negative conversion of the imitated rate, and the correlation coefficients of the non-imitated rate and the synergy dominance are calculated The correlation coefficients and the correlation coefficients are verified; If the correlation coefficients and the correlation coefficients are verified, the gradient weight hidden feature extraction is performed on the worker operation trajectory data and the interaction record data to obtain the risk preference coefficient and the potential synergy coefficient; Based on the risk preference coefficient and the potential synergy coefficient, the operation trajectory data and the interaction record data are field-split to construct a feature mapping library.

[0009] As a further technical solution of the present application, the double-track splitting method is as follows: Based on the constructed group synchronization model, the violation synchronization rate is extracted and it is judged whether the violation is out of limit, if it is out of limit, the imitation track splitting equation and the synergy track splitting equation are constructed; The imitation field strength and the synergy field strength of the group synchronization model are obtained, and the corresponding imitation track splitting equation and synergy track splitting equation are input combined with the violation synchronization rate to obtain the violation synchronization rate of the imitation track and the violation synchronization rate of the synergy track.

[0010] As a further technical solution of the present application, the group synchronization model is constructed as follows: The individual violation state and the average field variable of all workers corresponding to the synergy imitation individual are obtained; The imitation field strength and the synergy field strength are extracted based on the average field variable; The mean field equation is established to input the imitation field strength and the synergy field strength to obtain the violation synchronization rate of the comprehensive violation.

[0011] As a further technical solution of the present application, the synergy imitation individual is obtained as follows: According to a fixed monitoring period, time slices are divided, and the interaction record of the workers in the time slices is split into imitation behavior and synergy behavior; Based on the imitation behavior and the synergy behavior, and combined with the behavior feature mapping library, the imitation track transmission strength and the synergy track transmission strength of the workers are extracted; ​​Conduct centralized trend analysis based on the imitative track transmission strength and the collaborative track transmission strength to obtain an imitative centralized coefficient and a collaborative centralized coefficient; Screen imitative core individuals from the workers based on the imitative centralized coefficient, and screen collaborative core individuals from the workers based on the collaborative centralized coefficient; Obtain workers that meet the collaborative core individuals and the imitative core individuals at the same time as collaborative imitative individuals.

[0012] As a further technical solution of the present application, the way of extracting the imitative field strength and the collaborative field strength is: Construct an imitative field strength equation, obtain an imitative chain synchronization rate and a group total value parameter, and input the imitative field strength equation combined with an average field variable to obtain the imitative field strength; Construct a collaborative field strength equation, obtain a collaborative circle synchronization rate and a group total value parameter, and input the collaborative field strength equation combined with an average field variable to obtain the collaborative field strength.

[0013] As a further technical solution of the present application, the way of determining the game advantage type is: Conduct compliance behavior hedging analysis based on the violation synchronization rates of the imitative track and the collaborative track to obtain compliance hedging coefficients of the imitative track and the collaborative track; Establish game characteristics of both parties in the game model combined with the compliance hedging coefficients of the imitative track and the collaborative track; Establish a game potential energy equation based on the game characteristics of both parties in the game to output game potential energy, and compare and analyze the game potential energy to determine the game advantage type.

[0014] As a further technical solution of the present application, the way of conducting the compliance behavior hedging analysis is: Establish a compliance behavior hedging equation, and divide the double-track hedging coefficients to obtain the compliance hedging coefficients of the imitative track and the collaborative track.

[0015] As a further technical solution of the present application, the way of constructing the screening early warning model is: Decompose the advantage potential energy difference based on the game potential energy output by the game potential energy equation to locate the weak track; Formulate a weak track reinforcement strategy, and identify whether the compliance scene is over-early warned; If there is over-early warning, construct a screening early warning model to optimize the violation synchronization judgment, and filter invalid early warnings under the compliance support.

[0016] As a further technical solution of the present application, the way of locating the weak track is: Obtain the imitative track compliance potential energy-violation potential energy difference and the collaborative track compliance potential energy-violation potential energy difference, compare and analyze them, and determine the weak track.

[0017] The present application has the following beneficial effects: 1. Collect worker operation trajectory data and interaction record data in the work area. Through behavioral correlation analysis, uncover implicit features between the two types of data. Based on the verified correlations, extract risk preference coefficients and potential collaboration coefficients. Finally, integrate basic fields, related fields, and feature fields to form a behavioral feature mapping library. This provides comprehensive and logically connected data support for subsequent group violation analysis and early warning, reducing subsequent analytical biases caused by incomplete or isolated data.

[0018] 2. Worker real-time interactions are broken down into two core behaviors: imitation and collaboration. Collaborative imitation individuals possessing both core imitation and collaborative attributes are selected, and a group synchronization model is constructed. The overall violation synchronization rate is then broken down into the violation synchronization rates of the imitation track and the collaboration track. This approach helps identify key individuals playing a crucial role in transmitting group violations, while also clarifying the contributions of the imitation and collaboration paths to violation synchronization. This provides direction for subsequent targeted analysis of violation risks along different paths and reduces the overall ambiguity in judging group violations.

[0019] 3. Based on the dual-track violation synchronization rate, the compliance hedging coefficients of the imitation track and the collaborative track are first calculated using the compliance behavior hedging equation. Then, a dynamic game model is constructed with the violating party and the compliant party as the two sides. The game advantage type is determined by calculating and comparing the game potential energy of both parties. Incorporating violation propagation and compliance hedging into the dynamic game framework does not simply trigger warnings based on violation data, but considers the offsetting effect of compliance behavior on violation spread. This approach helps reflect the actual risk status within the work area, identify the trend of violation propagation, and provide a more realistic basis for judging whether a warning is needed and how to handle it, reducing risk misjudgments caused by not considering compliance hedging.

[0020] 4. By using weak link positioning and targeted reinforcement strategies, we can optimize the weak links in the group compliance system in a targeted manner, rather than generalizing to improve overall compliance, which helps to strengthen the group's compliance capabilities more efficiently. The construction of the screening and early warning model can identify and filter excessive warnings, reduce the waste of management resources, and adapt to changes in behavior status through dynamic rule iteration, so that the warning results are more in line with the real-time risk and compliance balance, improving the accuracy and practicality of the warnings. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This is a flowchart of a work area safety early warning method provided by the present invention; Figure 2 This is a flowchart of the determination of whether the limit is exceeded provided in Embodiment 2 of the present invention; Figure 3 This is a module diagram of a work area safety early warning system provided in Embodiment 3 of the present invention. Detailed Implementation

[0023] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. 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 should fall within the scope of protection of the present invention.

[0024] Example 1:

[0025] like Figure 1 As shown in the figure, the work area safety early warning method provided by this embodiment of the invention specifically includes the following steps: S1. Collect worker operation trajectories and interaction records in the work area, and mine the implicit features of operation trajectories and interaction records through behavioral correlation analysis to build a behavioral feature mapping library; The methods for collecting worker operation trajectories and interaction records in the work area are as follows: Preferably, by using historical video surveillance images of the work area and the OpenCV machine learning algorithm framework, the operation trajectory data and interaction record data of workers in pre-divided high-risk areas of the work area during the historical monitoring period are identified; The operation trajectory data includes the ratio of the time a single worker spends in a high-risk area to the total working hours, resulting in the high-risk stay ratio H, and the violation correction rate M for violations. Interaction record data includes: the rate of imitation of behavior (C) and the degree of collaborative dominance (L); It should be noted that the violation correction rate can be obtained by recording the number of violations that were corrected within the monitoring period after the worker's first violation (which was automatically identified by the system) and the behavior was corrected, to the total number of violations by the worker within the monitoring period. The imitation rate needs to be determined by comparing video behavior analysis and operation logs to identify the matching of other workers' operation actions with the target worker, count the total number of times other workers copied the target worker's operation, and compare it with the target worker's own total number of operation times; Collaboration dominance is determined for two-person collaborative operation scenarios by recording action timing (such as the order of equipment operation triggers and the start time of limb movements) to identify the action initiator and to obtain the proportion of the number of actions initiated by the target worker to the total number of actions in two-person collaboration. The way of mining the implicit features of the operation trajectory and the interaction record by the behavior correlation analysis is: collecting worker operation trajectory data and interaction record data in the last N monitoring periods; wherein N is the number of monitoring periods, N>30; by formula one: calculate the correlation coefficient of the high-risk stay ratio H and the violation correction rate M in the N monitoring periods ; wherein, , respectively represent the covariance of the high-risk stay ratio H and the violation correction rate M, the standard deviation of H, and the standard deviation of M; by formula two: calculate the correlation coefficient of the non-imitated rate and the collaborative dominance in the N monitoring periods ; wherein, , respectively represent the covariance of the non-imitated rate and the collaborative dominance L, the standard deviation of , and the standard deviation of L; It should be noted that the non-imitated rate is obtained by negatively transforming the imitated rate (1-C); the correlation coefficient of the non-imitated rate (1−C) and the collaborative dominance is calculated in order to verify whether individuals with strong imitation influence (low non-imitated rate, i.e., high imitated rate C) are more likely to dominate collaboration; based on the correlation coefficient and the correlation coefficient , if the correlation coefficient is negative and the correlation coefficient is positive, and the absolute values of both are higher than 0.6, then it is strongly correlated, i.e., it is verified, then the gradient-weighted implicit feature extraction is performed on the worker operation trajectory data and the interaction record data to obtain the risk preference coefficient and the potential collaboration coefficient; wherein the way of gradient-weighted implicit feature extraction is: by formula one: obtain the risk preference coefficient ; by formula two: obtain the potential collaboration coefficient ; It can be understood that the high-risk stay ratio H reflects the proportion of the length of time that an individual actively exposes to a risk area, and the higher the value, the longer the individual stays in the high-risk area, and the greater the risk of exposure. The rule correction rate M reflects the situation of workers correcting the rule violation, and 1-M quantifies the stubbornness of the rule violation that is not corrected. The larger 1-M is, the more difficult the rule violation is to correct. The positive correlation coefficient verifies the correlation between longer risk exposure and more difficult rule correction. The multiplication of the three integrates the strength of active risk-taking and resistance to correction, and at the same time, the correlation coefficient strengthens the internal relationship between the two, which can effectively locate individuals with high risk exposure and high rule stubbornness. For example, if a worker has a high high-risk stay ratio H, a low rule correction rate M, and a strong correlation between the two, the calculated risk preference coefficient will be larger, indicating that the worker has a higher risk preference and is more inclined to take risks in work; The imitated rate C quantifies the frequency of an individual's operation being imitated by others, reflecting the individual's influence on imitation in the group. The higher C is, the more frequently the individual's operation is imitated, and the greater the influence. The coordination dominance L measures the individual's control over initiating coordination. The larger L is, the more dominant the individual is in the coordination operation. The positive correlation coefficient verifies the correlation between stronger imitation influence and easier dominance of coordination. The multiplication of the three integrates the ability of being widely imitated and dominating coordination, and the correlation coefficient strengthens the coordination relationship between the two, so as to locate individuals with high imitation influence and high coordination control. For example, if a worker has a high imitated rate C and a high coordination dominance L, and a strong correlation between the two, the calculated potential coordination coefficient will be high, which means that the worker can not only attract others to imitate his operation in group coordination, but also play a dominant role in the coordination process, and has a greater influence on group behavior; The risk preference coefficient and the potential coordination coefficient are used as implicit features; Based on the risk preference coefficient and the potential coordination coefficient, the operation trajectory data and the interaction record data are field-split to construct a feature mapping library; The field-splitting method is as follows: Obtain the ID (number) of the worker, and integrate the feature mapping library of the worker's basic field, associated field, and feature field; The basic field includes the high-risk stay ratio H, the rule correction rate M, the imitated rate C, and the coordination dominance L; The associated field includes the positive correlation coefficient and the negative correlation coefficient ; The feature field includes the risk preference coefficient and the potential coordination coefficient.

[0026] S2, based on the behavior feature mapping library, splitting the worker real-time interaction into mimic and collaborative behavior and screening collaborative mimic individuals, constructing a group synchronization model based on the collaborative mimic individuals to extract the violation synchronization rate, splitting the violation synchronization rate into two tracks to obtain the violation synchronization rate of the mimic track and the collaborative track; The way of splitting the worker real-time interaction into mimic and collaborative behavior and screening collaborative mimic individuals is: Preferably, the time slices are divided according to a fixed monitoring period, and the workers in the real-time video monitoring images are obtained; The interaction records of the workers in the time slices are split into mimic behavior and collaborative behavior; It should be noted that the mimic behavior is, for example, worker B mimicking and copying the operation of worker A, and the collaborative behavior is, for example, workers A and B operating collaboratively; Based on the mimic behavior and the collaborative behavior, and combined with the behavior feature mapping library, the mimic track transmission intensity and the collaborative track transmission intensity of the workers are extracted; Exemplarily, the way of extracting the mimic transmission intensity and the collaborative transmission intensity is: The risk preference coefficient of B mimicking A is obtained , ; Among them, The risk preference coefficients of A and B, respectively; For the behavior of B mimicking A, the mimic transmission intensity is obtained by the formula: ; Among them, The number of times B mimics A in the time slice, The total number of times of mimic in the slice time, The violation correction rate change amount of B; The potential collaborative coefficient of A leading B is obtained ; For the behavior of A leading B collaboration, the mimic transmission intensity is obtained by the formula: ; Among them, , The number of times A initiates collaboration in the slice time and the total number of times of collaboration in the slice time, respectively; The collaborative leading degree change amount of B; It can be understood that if a worker does not have mimic track transmission intensity or collaborative track transmission intensity, the value of the mimic track transmission intensity or the collaborative track transmission intensity corresponding to the worker is 0; Based on obtaining and calculating the personal mimic total value and the personal collaborative total value of each worker, the sum of the personal mimic total values of all workers is calculated as the group mimic total value Mz, and the sum of the personal collaborative total values of all workers is calculated as the group collaborative total value Qz; ​​Exemplarily, for worker A, the total imitation transmission intensity I of all the imitations of A as the imitated one is summed up, i.e., the total imitation value of the individual (A) = the sum of the I values of all the B imitating worker A (such as I1 of B1 imitating A + I2 of B2 imitating A +...).

[0027] For example: if worker A is imitated by B and C, I A、B = 1.2, I A、C = 0.8, then the total imitation value of the individual (A) = 1.2 + 0.8 = 2.0; for the worker not imitated, the total imitation value of the individual (A) = 0. For worker A, the total coordination transmission intensity S of all the coordination transmissions of A as the coordination leader is summed up, i.e., the total coordination value of the individual (A) = the sum of the S values of all the A leading B (such as S1 of A leading B + S2 of A leading C +...). For example: if worker A leads B and C to coordinate, S A、B = 1.5, S a、c = 1.0, then the total coordination value of the individual (A) = 1.5 + 1.0 = 2.5; the total coordination value of the individual (A). The proportion of the total imitation value of each worker to the total imitation value of the group is calculated to obtain the imitation concentration coefficient . The proportion of the total coordination value of each worker to the total coordination value of the group is calculated to obtain the coordination concentration coefficient . The imitation core individual in the workers is screened based on the imitation concentration coefficient, and the coordination core individual in the workers is screened based on the coordination concentration coefficient. The worker that meets both the coordination core individual and the imitation core individual is obtained as the coordination imitation individual. It can be understood that the core individual screening threshold , is set, the worker meeting is the imitation core individual, the worker meeting is the coordination core individual, and the worker meeting both is the coordination imitation individual. Among them, the core individual screening threshold 20% comes from historical data statistics, the working area data of the past 12 monitoring periods (7 days per period) is selected, the imitation concentration coefficient and the coordination concentration coefficient of all the workers are calculated, the top 20% quantile is taken as the core individual judgment standard, to ensure that the core individual covers 80% of the imitation or coordination interaction transmission in the group, if the working area personnel scale changes (such as ± 30%), the quantile threshold can be recalculated and adjusted. Based on the coordination imitation individual, the group synchronization model is constructed through the mean field theory algorithm to extract the violation synchronization rate. Among them, the way of constructing the group synchronization model through the mean field theory algorithm is: S201, obtaining individual violation states and average field variables of all workers corresponding to the cooperative mimic individual; Preferably, for all workers in the current time slice, define the individual violation state : For each worker i, identify the severity of its violation operation according to real-time monitoring (0 for complete compliance, 1 for serious violation); For each cooperative mimic individual , identify the severity of its violation operation according to real-time monitoring (0 for complete compliance, 1 for serious violation); Wherein, the value of θ of the cooperative mimic individual needs to be superimposed with its influence weight, that is, the =1+imitation centralization coefficient+cooperative centralization coefficient, used to highlight the driving effect of the cooperative mimic individual on the group; Obtain the average field variable m by the formula: Wherein, The total number of cooperative mimic individuals, the total number of non-cooperative mimic individuals (ordinary workers); It should be noted that based on the θᵢ of all workers, the weight of the cooperative mimic individual is the sum of its imitation centralization coefficient and cooperative centralization coefficient, and the weight of the non-core individual is 1. The average field variable reflects the average violation influence intensity of the group on the individual; S202, extracting the mimic field strength and the cooperative field strength based on the average field variable; Preferably, the mimic field strength is obtained by the mimic field strength equation: ; Wherein, z is the total value of the group imitation, the total value of the group cooperation; The total value of the group imitation and the total value of the group cooperation are collectively referred to as the group total value parameter; The cooperative field strength is obtained by the cooperative field strength equation: ; Wherein, is the imitation chain synchronization rate, that is, the proportion of synchronous violations in the imitators of the cooperative mimic individual; is the cooperative circle synchronization rate, that is, the proportion of synchronous violations in the cooperative objects of the cooperative mimic individual; S203, inputting the mimic field strength and the cooperative field strength into the mean field equation to obtain the violation synchronization rate of the comprehensive violation; Preferably, the violation synchronization rate R of the comprehensive violation is obtained by the mean field equation: Wherein, is the sensitivity coefficient (usually 5-10, controlling the reaction intensity of the group to the change of the field strength),​​​​ for the threshold parameter (when the total field strength is equal to , the synchronization rate is 50%, which can be calibrated by historical data); It can be understood that the mean field equation simulates the phase transition characteristics of the group from asynchronous violation to synchronous violation through a nonlinear transformation, and the final output R value is a quantitative indicator reflecting the overall violation synchronization trend of the group; As Figure 2 shown, the violation synchronization rate is compared with the preset violation synchronization threshold , if the violation synchronization rate is lower than the preset violation synchronization threshold , the change of the violation synchronization rate is continuously monitored; If the violation synchronization rate is higher than or equal to the preset violation synchronization threshold, it is determined that the violation is out of limit, and the violation synchronization rate is double-rail split to obtain the violation synchronization rate of the mimic rail , the violation synchronization rate of the synergy rail ; Through the mimic rail split equation: the violation synchronization rate of the mimic rail is obtained; Through the synergy rail split equation: the violation synchronization rate of the synergy rail is obtained; It can be understood that the extraction of the violation synchronization rate and the double-rail split have the following effects: Effect 1: Locating the key path of group violation transmission, the double-rail split can obtain the violation synchronization rates of the two types of rails through the mimic rail split equation and the synergy rail split equation, respectively, to quantify the contribution of the two paths to the violation synchronization. Reduce the problem of fuzzy group violation attribution, and clearly direct the subsequent intervention; Effect 2: Supporting the dynamic game model to determine the risk, the dynamic game model needs to resist the state data, and the violation synchronization rate after double-rail split is the basis for calculating the compliance hedging coefficients of the mimic rail and the synergy rail. If not split, a single coefficient will mask the path risk difference, and after splitting, the double-rail resistance state can be described to ensure the accuracy of the judgment of the advantageous type of game; Effect 3: The double-rail violation synchronization rate is the core data for locating the weak rail, and the double-rail potential difference can be calculated by combining the transmission strength to identify weak links; at the same time, the comparison with the dynamic early warning threshold can determine whether it is an over-warning, provide a basis for screening the early warning model, filter invalid early warnings, avoid resource waste, and improve the practicality of early warning.

[0028] Embodiment 2: As Figure 1 shown, the working area safety early warning method provided by the embodiment of the application specifically includes the following steps: S3. Conduct compliance behavior hedging analysis based on the violation synchronization rate of the imitation track and the cooperative track, obtain the compliance hedging coefficients of the imitation track and the cooperative track, construct a dynamic game model, and determine the game advantage type. The methods for conducting compliance behavior hedging analysis are as follows: Preferably, establish a compliance behavior hedging equation: Decompose the dual-track hedging coefficient to obtain the mimicry track compliant hedging coefficient. Coordinated track compliance hedging coefficient ; Where P represents the percentage of compliant operations by individuals who collaboratively imitate each other, expressed by the formula: =1 - (Total compliant operation coefficients + Number of violations corrected) / Total operation coefficients, to obtain the proportion of compliant operations P of individuals who collaborate and imitate. Based on the compliance hedging coefficient of the imitation track Coordinated track compliance hedging coefficient Establish a dynamic game model of dual-track hedging and illegal propagation; The method for establishing a dynamic game model is as follows: S301. The game characteristics of the two sides in the game model are established by combining the compliance hedging coefficient of the imitation track and the compliance hedging coefficient of the collaborative track. Preferably, the two parties in the game include the violating party and the compliant party: Among them, the game-theoretic characteristics of the violator include: imitation transmission strength. Imitation transmission strength The overall violation synchronization rate R; The game-theoretic characteristics of compliance parties are: imitation track compliance hedging coefficient. Coordinated track compliance hedging coefficient The proportion of collaborative imitation of individual compliant operations; S302. Based on the game characteristics of both sides, establish a game potential energy equation to output the game potential energy, and conduct comparative analysis on the game potential energy to determine the type of game advantage. Preferably, according to formula one: To gain the game potential of the violating party; Through formula two: To gain the game potential of the violating party; in, The set of interactive objects that collaboratively imitate individual A includes imitation objects and collaborative objects; The total number of individuals that cooperate in imitation; like If the game advantage type is illegal propagation, a critical warning signal is triggered. like Then the type of game advantage is compliance hedging.

[0029] S4, if compliance hedging is dominant, build a screening early warning model, identify and optimize the synchronous judgment of violation, filter invalid early warning under the support of compliance; If compliance hedging is dominant, the way to build a screening early warning model is: S401, based on the game potential energy output by the game potential energy equation, the advantage potential energy difference is decomposed, and the weak track is positioned; S4, if compliance hedging is dominant, build a screening early warning model, identify and optimize the synchronous judgment of violation, filter invalid early warning under the support of compliance; If compliance hedging is dominant, the way to build a screening early warning model is: Preferably, by formula one: Get the mimic compliance ability of A And the potential difference of A to B's mimic violation transmission ; ; By formula two: Get the collaborative compliance ability of A And the potential difference of A to B's collaborative violation transmission ; ; By summation equation one: Get the mimic track compliance potential-energy difference ; By summation equation two: Get the collaborative track compliance potential-energy difference ; Establish the weak track potential energy criterion: Then the collaborative track is the weak track; And Then the mimic track is the weak track; And Then both the mimic track and the collaborative track are weak, triggering the highest level of early warning; S402, develop a weak track reinforcement strategy, and identify whether the compliance scene is over-alarmed; Preferably, based on the determined weak track, get the compliance proportion P of the collaborative mimic individual, the influence coefficient , And the interaction transmission intensity , , screen the core individuals with high influence K≥0.3 and high compliance P≥0.7, and direct them to carry out high-frequency collaboration and mimic tasks (such as compliance teaching) with the interactive objects with low compliance P<0.5 in the weak track, and verify the reinforcement effect of the weak track compliance potential energy through the updated transmission intensity and violation correction rate; By formula: constructing a dynamic early warning threshold model to obtain an updated violation synchronization threshold ; If the absolute deviation of the violation synchronization rate R and the dynamic violation synchronization threshold is less than 20%, and the time , it is determined that there is an over-warning under compliance support; S403, if there is an over-warning, a screening early warning model is constructed to optimize the violation synchronization judgment, and filter invalid early warnings under compliance support; Preferably, the screening early warning model is constructed based on a decision tree and a dynamic rule iteration algorithm to filter invalid early warnings under compliance support; Those skilled in the art can understand that the way to construct the screening early warning model based on the decision tree and the dynamic rule iteration algorithm is: Calculate the deviation of the violation synchronization rate and the dynamic violation synchronization threshold , to obtain the synchronization deviation; Select samples in the historical period , if the violation synchronization rate R does not break through the dynamic threshold in the next 2 periods, mark it as invalid early warning (1), otherwise mark it as valid early warning (0); Take , , the synchronization deviation as the core feature, and integrate the compliance proportion P of the collaborative imitating individual and the influence coefficient , of the core individual as parameters to calculate the features and establish the model data set, train the CART decision tree, and the decision tree automatically learns the nonlinear rules of the double-track potential energy state-synchronization rate deviation-early warning effectiveness-, for example, when the synchronization deviation is less than 20%, it is determined that the early warning is invalid; Every 1 monitoring period (matching S2 time slice) or when the core individual P changes by ≥15%, use the newly added label data to incrementally update the decision tree, and automatically adjust the rule threshold (such as relaxing the screening boundary when the double-track potential energy is enhanced); Real-time extract the parameter calculation features in the model data set, input the decision tree, and output whether to screen the early warning, to realize intelligent filtering of invalid early warnings under compliance support.

[0030] Embodiment 3:

[0031] As shown in Figure 3 , a work area safety early warning system includes the following modules: Feature mapping module: used for collecting worker operation trajectories and interaction records in the work area, mining implicit features of the operation trajectories and the interaction records through behavior association analysis, and constructing a behavior feature mapping library; ​The rule violation analysis module is used for splitting the real-time interaction of workers into mimic and collaborative behaviors based on the behavior feature mapping library, screening collaborative mimic individuals, constructing a group synchronization model based on the collaborative mimic individuals, extracting a rule violation synchronization rate, performing double-track splitting on the rule violation synchronization rate, and obtaining the rule violation synchronization rate of the mimic track and the collaborative track. The game analysis module is used for performing compliance behavior hedging analysis based on the rule violation synchronization rate of the mimic track and the collaborative track, obtaining a compliance hedging coefficient of the mimic track and the collaborative track, constructing a dynamic game model, and determining a game advantage type. The early warning optimization module is used for constructing a screening early warning model if the compliance hedging is dominant, identifying a rule violation synchronization judgment, and optimizing the rule violation synchronization judgment to filter invalid early warnings under the support of compliance.

[0032] The above describes one embodiment of the present application in detail, but the content described is only a preferred embodiment of the present application and cannot be considered as limiting the implementation scope of the present application. Any equivalent changes and improvements made according to the scope of the present application should still belong to the patent coverage of the present application.

Claims

1. A work area safety warning method characterized by, The method comprises the following steps: Collecting worker operation trajectory and interaction record data of a working area, mining implicit features of the operation trajectory and the interaction record through behavior correlation analysis, and constructing a behavior feature mapping library; Based on the behavior feature mapping library, the real-time interaction of the worker is split into mimic and collaborative behaviors, and collaborative mimic individuals are screened, a group synchronization model is constructed based on the collaborative mimic individuals to extract a violation synchronization rate, the violation synchronization rate is split into two tracks, and a violation synchronization rate of a mimic track and a collaborative track is obtained; Based on the violation synchronization rate of the mimic track and the collaborative track, compliance behavior hedging analysis is performed to obtain a compliance hedging coefficient of the mimic track and the collaborative track, and a dynamic game model is constructed to determine a game advantage type; If the compliance hedging is dominant, a screening early warning model is constructed, a violation synchronization judgment is identified, and the violation synchronization judgment is optimized to filter invalid early warnings under the compliance support.

2. The working area safety warning method according to claim 1, characterized in that, The behavior correlation analysis is performed in the following manner: Collect operation trajectory data and interaction record data of workers in high-risk areas, extract high-risk stay ratio and violation correction rate in operation trajectory data, and calculate correlation coefficient of high-risk stay ratio and violation correction rate ; Extract the imitated rate and the synergistic dominance degree in the interaction record data, and convert the imitated rate to the non-imitated rate to obtain the non-imitated rate, and calculate the correlation coefficient of the non-imitated rate and the synergistic dominance degree ; Based on the correlation coefficient and the correlation coefficient Performing verification processing; If the correlation coefficient and the correlation coefficient are verified, the gradient weight of the worker operation track data and the interaction record data is extracted, and the risk preference coefficient and the potential synergy coefficient are obtained. Based on the risk preference coefficient and the potential collaboration coefficient, the operation trajectory data and the interaction record data are split by field, and a feature mapping library is constructed.

3. The method of claim 1, wherein, The split into two tracks is performed in the following manner: Based on the constructed group synchronization model, the violation synchronization rate is extracted, and it is determined whether the violation is out of limit, if so, a mimic track split equation and a collaborative track split equation are constructed; The mimic field strength and the collaborative field strength of the group synchronization model are obtained, and the mimic track split equation and the collaborative track split equation are inputted in combination with the violation synchronization rate to obtain the violation synchronization rate of the mimic track and the violation synchronization rate of the collaborative track.

4. The method of claim 1, wherein, The group synchronization model is constructed in the following manner: The individual violation state and the average field variable of all workers corresponding to the collaborative mimic individual are obtained; Based on the average field variable, the mimic field strength and the collaborative field strength are extracted; The mean field equation is established to input the mimic field strength and the collaborative field strength to obtain the violation synchronization rate of the comprehensive violation.

5. The method of claim 4, wherein the step of determining the safety of the work area comprises: The collaborative mimic individual is obtained in the following manner: According to a fixed monitoring period, time slices are divided, and the interaction record of the workers in the time slices is split into mimic behavior and collaborative behavior; Based on the mimic behavior and the collaborative behavior, and in combination with the behavior feature mapping library, the mimic track transmission strength and the collaborative track transmission strength of the workers are extracted; Based on the mimic track transmission strength and the collaborative track transmission strength, concentration trend analysis is performed to obtain a mimic concentration coefficient and a collaborative concentration coefficient; Based on the mimic concentration coefficient, mimic core individuals are screened from the workers, and based on the collaborative concentration coefficient, collaborative core individuals are screened from the workers; Workers that meet both the collaborative core individual and the mimic core individual are obtained as the collaborative mimic individual.

6. The method of claim 4, wherein, The mimic field strength and the collaborative field strength are extracted in the following manner: A mimic field strength equation is constructed, the mimic chain synchronization rate and the group total value parameter are obtained, and the mimic field strength equation is inputted in combination with the average field variable to obtain the mimic field strength; A collaborative field strength equation is constructed, the collaborative circle synchronization rate and the group total value parameter are obtained, and the collaborative field strength equation is inputted in combination with the average field variable to obtain the collaborative field strength.

7. The method of claim 1, wherein, The game advantage type is determined in the following manner: Based on the violation synchronization rate of the mimic track and the collaborative track, compliance behavior hedging analysis is performed to obtain a compliance hedging coefficient of the mimic track and the collaborative track; The game characteristics of the two parties in the game model are established in combination with the mimic track compliance hedging coefficient and the collaborative track compliance hedging coefficient. The game potential energy equation is established based on the game characteristics of both parties in the game to output game potential energy, and the game advantage type is determined by comparing and analyzing the game potential energy.

8. The method of claim 7, wherein the method further comprises: The manner for performing the compliance hedging analysis is: A compliance hedging equation is established, and a double-track hedging coefficient is obtained to acquire a mimic track compliance hedging coefficient and a collaborative track compliance hedging coefficient.

9. The method of claim 1, wherein, The manner for constructing the screening early warning model is: Based on the game potential energy output by the game potential energy equation, the advantage potential energy difference is disassembled, and the weak track is positioned. A weak track reinforcement strategy is developed, and whether the compliance scene is over-early warned is identified. If there is over-early warning, a screening early warning model is constructed to optimize the synchronous judgment of the violation, and filter the invalid early warning under the support of the compliance.

10. The method of claim 9, wherein the method further comprises: The manner for performing the weak track positioning is: The mimic track compliance potential energy-violation potential energy difference and the collaborative track compliance potential energy-violation potential energy difference are acquired, compared and analyzed, and the weak track is determined.

Images