Dynamic early warning system for safety management and control of subway field section

Abstract

The invention relates to the technical field of subway safety management and control, and discloses a dynamic early warning system for subway field section safety management and control, which comprises a monitoring module, an identification module, a boundary evolution module, a violation prediction module and an early warning module, wherein the monitoring module is used for collecting field data of each management and control area of a subway field section; the identification module is used for identifying a hard violation behavior and a soft violation behavior of each management and control area; the boundary evolution module is used for learning a discrimination boundary of the soft violation behavior; the violation prediction module is used for constructing a violation behavior chain of each management and control area and predicting a safety management and control risk of each management and control area based on the violation behavior chain; the early warning module carries out supplementary identification on the safety management and control risk based on the illegal behavior chain and carries out risk early warning based on the safety management and control risk; according to the invention, intelligent identification and accurate early warning of hidden violation of the subway field section are realized, and the safety management and control level of the subway field section is improved.

Description

Technical Field

[0001] The present application relates to the technical field of subway safety management and control, and specifically to a dynamic early warning system for subway yard safety management and control. Background Art

[0002] Existing technologies primarily rely on traditional manual inspections, video surveillance, access control systems, and other basic methods for subway yard security management. These traditional inspection methods rely on subjective judgment, provide incomplete coverage of potential hazards, and lack the granular, dynamic identification of personnel behavior and potential violations within the yard.

[0003] Existing safety monitoring systems mostly rely on hard-coded violation criteria, such as not wearing a hard hat, not wearing required work clothes, and illegally crossing a boundary. For softer violations that are difficult to define and have ambiguous boundaries, such as people frequently approaching restricted areas or repeatedly entering and exiting key areas, existing systems lack effective dynamic identification mechanisms. This results in hidden violations remaining in regulatory blind spots for a long time, making it difficult to detect safety hazards in a timely manner.

[0004] Existing violation identification methods generally use static threshold settings, which cannot be flexibly adjusted to the actual conditions of different areas and different types of work tasks. The on-site risk levels and personnel behavioral characteristics corresponding to different work tasks vary significantly. The use of fixed violation identification standards can easily lead to excessively high misjudgment rates or high missed judgment rates. In addition, the existing subway site safety management system has an imperfect early warning mechanism for violation risks. It usually only issues simple alarm prompts after a violation has occurred or constituted a fact. It lacks comprehensive analysis and prediction methods for violation trends and violation intentions, and cannot achieve early perception and proactive warning, which affects the overall level of safety management and control.

[0005] For example, Chinese patent application CN119294834B discloses an intelligent subway construction safety risk assessment system, which includes a field data acquisition module, an equipment data import module, a construction data analysis module, a construction risk assessment module, a risk collation and display module, a database, and an early warning execution and feedback terminal. This technical solution addresses the current shortcomings in data collection scope and depth by setting up key areas, collecting real-time construction site data, and importing cumulative construction hours and real-time equipment monitoring data from these key areas. It also conducts risk assessments across three key areas: personnel, equipment, and structure, forming an independent and complete risk indicator analysis system. Furthermore, by constructing risk indicator maps within each key area, it facilitates comprehensive risk monitoring and management for construction managers, improving the effectiveness of risk control and feedback.

[0006] For example, patent application publication number CN116882938A discloses a method for fine-grained control of engineering activities in subway protection zones, including the following steps: obtaining subway protection zone data and determining control parameters and safety evaluation indicators; constructing and training a neural network model based on the control parameters and safety evaluation indicators; drawing a four-dimensional cloud map of the relationship between the control parameters and safety evaluation indicators based on the neural network model; and, based on the four-dimensional cloud map, obtaining the determined items in the control parameters of the engineering activities in the protection zone to be controlled, and determining the control standards or safety of the engineering activities in the protection zone. Compared with existing technologies, this technical solution finely determines the boundaries of subway protection zones for different engineering activities and, through visualization, enables technicians to intuitively, quickly, and quantitatively determine the impact of engineering activities on subway structural safety.

[0007] The above technical solutions all have the problem raised by this background technology: lack of comprehensive analysis and prediction methods for violation trends and violation intentions.

[0008] The information disclosed in this background technology section is only intended to enhance the understanding of the overall background of the application and should not be regarded as an admission or any form of suggestion that the information constitutes the prior art already known to ordinary technicians in this field. Summary of the Invention

[0009] The technical problem to be solved by this application is to overcome the defects of the existing technology and provide a dynamic early warning system for subway yard safety management and control, so as to realize the intelligent identification and accurate early warning of hidden violations in subway yards and sections, and improve the safety management and control level of subway yards and sections.

[0010] To solve the above technical problems, this application provides the following technical solutions:

[0011] This application provides a dynamic early warning system for subway station security management, including a monitoring module, an identification module, a boundary evolution module, a violation prediction module, and an early warning module; wherein:

[0012] The monitoring module is used to collect on-site data of each control area in the subway depot;

[0013] An identification module identifies hard violations and soft violations in each control area based on the field data;

[0014] The boundary evolution module learns the discrimination boundary of the soft violation behavior based on historical soft violation records;

[0015] The violation prediction module constructs a violation chain for each control area based on the hard violation, the soft violation, and the discrimination boundary, and predicts the security control risk of each control area based on the violation chain;

[0016] The early warning module performs additional identification of security control risks based on the illegal behavior chain, and issues risk early warnings based on the security control risks.

[0017] As a preferred solution of the dynamic early warning system for subway section safety management described in this application, wherein: the field data includes restricted area boundary range, personnel location data, and monitoring screen data;

[0018] The identification module includes a first identification unit and a second identification unit;

[0019] The first recognition unit is used to identify hard violations in each control area; the hard violations include at least not wearing a safety helmet or not wearing prescribed work clothes;

[0020] The first recognition unit is configured with an image recognition algorithm; the first recognition unit processes the monitoring screen data based on the image recognition algorithm to identify the hard violation behavior;

[0021] The second identification unit identifies soft violations in each controlled area based on the restricted area boundary range and personnel positioning data, and records the violation indicators of each soft violation; the soft violation behaviors include at least restricted area approaching behavior and frequent entry and exit behavior; wherein, the violation indicator corresponding to the restricted area approaching behavior is the duration of the restricted area approaching behavior; the violation indicator corresponding to the frequent entry and exit behavior is the number of times frequent entry and exit of the designated area is detected per unit time.

[0022] As a preferred embodiment of the dynamic early warning system for subway section safety management described in this application, the boundary evolution module includes a database unit; the database unit is configured with historical soft violation records for each control area; each historical soft violation record includes a violation indicator and a soft violation label for a soft violation behavior; the soft violation label includes explicit soft violations and implicit soft violations;

[0023] The explicit soft violation indicates that the violation index of the soft violation behavior is greater than or equal to the corresponding judgment boundary; the implicit soft violation indicates that the violation index of the soft violation behavior is less than the corresponding judgment boundary;

[0024] Any piece of historical soft violation data also includes the job task type corresponding to the soft violation behavior.

[0025] As a preferred solution of the dynamic early warning system for subway section safety management described in this application, the boundary evolution module further includes a dynamic boundary unit; the dynamic boundary unit is configured with a boundary evolution strategy for learning the discrimination boundary of the soft violation behavior; the boundary evolution strategy specifically includes:

[0026] Obtain the current job task type of each control area; select historical soft violation records with the same job task type as the current one for each control area from the historical soft violation records, and use them as reference soft violation records for the corresponding control area;

[0027] For any control area, classify the reference soft violation records based on the type of soft violation to obtain a reference violation record for each soft violation;

[0028] For any soft violation, further classify the reference violation records based on the soft violation labels; count all violation indicators corresponding to explicit soft violations in the reference violation records of the soft violation, and form a first boundary range of the corresponding soft violation indicators; count all violation indicators corresponding to implicit soft violations in the reference violation records of the soft violation, and form a second boundary range of the corresponding soft violation indicators;

[0029] A determination boundary for soft violation behavior is determined based on the first boundary range and the second boundary range.

[0030] As a preferred solution of the dynamic early warning system for subway section safety management described in this application, the boundary evolution strategy further includes determining the judgment boundary of any soft violation behavior based on the first boundary range and the second boundary range, as follows:

[0031] Setting a false positive weight value and a missed positive weight value; the false positive weight value and the missed positive weight value are both greater than 0 and less than 1, and the sum is 1;

[0032] When there is no intersection between the first boundary range and the second boundary range, the judgment boundary is determined based on the false positive weight value and the missed judgment weight value; specifically, if the false positive weight value is greater than the missed judgment weight value, the judgment boundary is the maximum value of the violation indicator in the second boundary range; otherwise, the judgment boundary is the minimum value of the violation indicator in the first boundary range.

[0033] As a preferred solution of the dynamic early warning system for subway section safety management described in this application, the boundary evolution strategy further includes: if the first boundary range and the second boundary range have an intersection, obtaining the values ​​of all violation indicators in the intersection as alternative judgment boundaries; any alternative judgment boundary is greater than or equal to the minimum value of the violation indicator in the first boundary range and less than or equal to the maximum value of the violation indicator in the second boundary range;

[0034] Calculate the false positive rate and missed positive rate of any candidate decision boundary as the decision boundary, and perform weighted summation of the false positive rate and missed positive rate to obtain the confidence level of the corresponding candidate decision boundary, wherein the weight of the false positive rate is the false positive weight value, and the weight of the missed positive rate is the missed positive weight value;

[0035] The candidate decision boundary with the highest confidence is taken as the decision boundary.

[0036] As a preferred solution of the dynamic early warning system for subway yard safety management described in this application, wherein: the violation prediction module includes a behavior modeling unit and a violation matching unit;

[0037] The behavior modeling unit is configured with a violation modeling strategy for constructing a violation behavior chain in any control area; the violation modeling strategy specifically includes:

[0038] Specifying the order of each hard violation and soft violation, and initializing a violation chain; each element in the violation chain corresponds to a hard violation or a soft violation;

[0039] Based on the identification results of hard violations, the corresponding elements in the violation chain are assigned values. If any hard violation exists in the control area, the corresponding element is assigned a value of 1; otherwise, the corresponding element is assigned a value of 0.

[0040] Based on the identification results of soft violations, violation indicators, and corresponding judgment boundaries, the corresponding elements in the violation chain are assigned values, as follows: if the control area does not identify any hard violations, the corresponding element is assigned a value of 0; if the control area identifies any hard violations, the soft violation label is determined based on the violation indicators and the corresponding judgment boundary. If the soft violation label is an explicit soft violation, the corresponding element is assigned a value of 1; otherwise, the corresponding element is assigned a value of r, where r is a positive number less than 1.

[0041] As a preferred solution of the dynamic early warning system for subway section safety management described in this application, wherein: the violation matching unit is configured with a reference violation chain for each control area; any reference violation chain is constructed based on the violation modeling strategy;

[0042] The violation matching unit is also configured with a risk prediction strategy for predicting the security control risk of any control area; the risk prediction strategy specifically includes:

[0043] Calculate the similarity between the violation chain and each reference violation chain in turn;

[0044] Continuously collecting the field data and updating the violation chain in real time; when the value of at least one element in the violation chain changes, triggering a recalculation of the similarity between the violation chain and each reference violation chain;

[0045] The security control risk is predicted based on the similarity between the violation chain and the reference violation chain.

[0046] As a preferred solution of the dynamic early warning system for subway yard safety management described in this application, the safety management risk is predicted based on the similarity between the violation chain and the reference violation chain, specifically including:

[0047] The violation matching unit is further configured with a first similarity threshold and a second similarity threshold;

[0048] If it is detected that the similarity between the violation chain and any reference violation chain is greater than the first similarity threshold, there is a security control risk, and a first risk alert is sent to the early warning module;

[0049] If it is detected that the similarity between the violation chain and any reference violation chain is greater than the second similarity threshold, and the similarity between the violation chain and the corresponding reference violation chain continues to increase in at least m consecutive similarity calculations, there is a security control risk, and a first risk alert is sent to the early warning module; m is a positive integer.

[0050] As a preferred solution of the dynamic early warning system for subway yard safety management described in this application, wherein: the early warning module includes a risk identification unit and a risk early warning unit;

[0051] The risk identification unit is configured with a risk identification strategy for supplementary identification of security control risks; the risk identification strategy specifically includes: performing a weighted summation on each element in the violation chain to obtain a risk value for the corresponding control area;

[0052] The risk identification unit is further configured with a risk value threshold. If the risk value of any control area is greater than the risk value threshold, there is a security control risk, and a second risk alert is sent to the risk warning unit.

[0053] The risk warning unit performs a risk warning based on the security management and control risk, specifically including: if at least one of the first risk warning and the second risk warning is received, performing a risk warning.

[0054] Compared with the prior art, the beneficial effects achieved by this application are as follows:

[0055] This application improves the scope and accuracy of subway station identification of hidden violations by extracting and identifying characteristic parameters of these behaviors. By collecting historical violation data and non-violation data, it automatically defines the boundaries for identifying hidden violations for each area and behavior type, ensuring that the system's identification criteria are objective and dynamically optimizable.

[0056] This application uses historical data-driven dynamic demarcation of judgment boundaries, operational task linkage, and chain analysis of violations to achieve intelligent identification of hidden violations in subway yards, prediction of development trends, and precise early warning, thereby comprehensively improving the level of safety risk management. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For those skilled in the art, other drawings can be derived from these drawings without inventive effort. Among them:

[0058] Figure 1 A schematic diagram of the structure of a dynamic early warning system for subway station safety management provided in this application;

[0059] Figure 2 A functional diagram of a dynamic early warning system for subway yard safety management provided in this application. DETAILED DESCRIPTION

[0060] The technical solution of the present application is described in detail below through the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present application and the specific features in the embodiments are detailed descriptions of the technical solution of the present application, rather than limitations on the technical solution of the present application. Unless there is a conflict, the embodiments of the present application and the technical features in the embodiments can be combined with each other.

[0061] This embodiment introduces a dynamic early warning system for subway station security management and control. Figure 1 The system includes monitoring module, identification module, boundary evolution module, violation prediction module and early warning module; the functions of the above modules are as follows: Figure 2 As shown, where:

[0062] The monitoring module is used to collect on-site data of each control area in the subway depot;

[0063] The monitoring module includes a camera unit, a positioning unit, and a map unit;

[0064] The on-site data includes the restricted area boundary range, personnel positioning data, and monitoring screen data; the camera unit is used to collect the monitoring screen data; the positioning unit is used to collect the personnel positioning data; the map unit is used to obtain an electronic map of the subway section and mark the restricted area boundary range of each controlled area.

[0065] An identification module identifies hard violations and soft violations in each control area based on the field data;

[0066] The identification module includes a first identification unit and a second identification unit;

[0067] The first recognition unit is used to identify hard violations in each control area; the hard violations include at least not wearing a safety helmet or not wearing prescribed work clothes;

[0068] The first recognition unit is configured with an image recognition algorithm; the first recognition unit processes the monitoring screen data based on the image recognition algorithm to identify the hard violation behavior;

[0069] The second identification unit identifies soft violations in each controlled area based on the restricted area boundary range and personnel positioning data, and records the violation indicators of each soft violation; the soft violation behaviors include at least restricted area approaching behavior and frequent entry and exit behavior; wherein, the violation indicator corresponding to the restricted area approaching behavior is the duration of the restricted area approaching behavior; the violation indicator corresponding to the frequent entry and exit behavior is the number of times frequent entry and exit of the designated area is detected per unit time.

[0070] The preferred method for identifying soft violations in this embodiment is as follows: identifying behavior approaching the restricted area based on the restricted area boundary range and personnel positioning data; for example, when the distance between the detected personnel positioning and the restricted area boundary is less than a preset detection threshold, then the behavior approaching the restricted area is identified; the boundary range of a designated area, such as a construction area, is determined based on an electronic map, and the number of times the personnel enters and exits the designated area within a unit time, such as 15 minutes, is identified in combination with the personnel positioning; if the number is greater than a preset entry and exit threshold, then frequent entry and exit behavior is identified.

[0071] The boundary evolution module learns the discrimination boundary of the soft violation behavior based on historical soft violation records;

[0072] The boundary evolution module includes a database unit and a dynamic boundary unit;

[0073] The database unit is configured with historical soft violation records for each control area; each historical soft violation record includes a violation indicator and a soft violation label of a soft violation behavior; the soft violation label includes explicit soft violations and implicit soft violations;

[0074] An explicit soft violation indicates that the violation indicator of the soft violation is greater than or equal to the corresponding discrimination boundary; a hidden soft violation indicates that the violation indicator of the soft violation is less than the corresponding discrimination boundary. In this embodiment, soft violations reflect potential violation intent. An explicit soft violation is one in which the violation intent is obvious and the probability of actual violation is high; conversely, a hidden soft violation is one in which the violation intent is not obvious and the probability of actual violation is low.

[0075] Each piece of historical soft violation data also includes the work task type corresponding to the soft violation. In this embodiment, the work task types preferably include equipment maintenance, temporary fault repair, and line construction. Each work task type corresponds to different violation identification criteria. For example, for temporary fault repair, the violation criteria for equipment-related restricted area approach are relaxed, while for line construction, the violation criteria for frequent entry and exit are tightened.

[0076] The dynamic boundary unit is configured with a boundary evolution strategy for learning the discrimination boundary of the soft violation behavior; the boundary evolution strategy specifically includes:

[0077] Obtain the current job task type of each control area; select historical soft violation records with the same job task type as the current one for each control area from the historical soft violation records, and use them as reference soft violation records for the corresponding control area;

[0078] For any control area, classify the reference soft violation records based on the type of soft violation to obtain a reference violation record for each soft violation;

[0079] For any soft violation, further classify the reference violation records based on the soft violation labels; count all violation indicators corresponding to explicit soft violations in the reference violation records of the soft violation, and form a first boundary range of the corresponding soft violation indicators; count all violation indicators corresponding to implicit soft violations in the reference violation records of the soft violation, and form a second boundary range of the corresponding soft violation indicators;

[0080] A determination boundary for soft violation behavior is determined based on the first boundary range and the second boundary range.

[0081] The boundary evolution strategy further includes determining a judgment boundary for any soft violation based on the first boundary range and the second boundary range, as follows:

[0082] Setting a false positive weight value and a missed positive weight value; the false positive weight value and the missed positive weight value are both greater than 0 and less than 1, and the sum is 1;

[0083] In this embodiment, the preferred method for setting the weights for misjudgments and missed judgments is as follows: a weight comparison table is established; the weight comparison table is used to specify the weights for misjudgments and missed judgments for different task types and different soft violations. For example, if the task type is temporary fault repair and the soft violation is approaching a restricted area, the weight for misjudgments is 0.3 and the weight for missed judgments is 0.7. This means that a smaller weight for misjudgments is used, indicating that in this case, misjudgments are tolerated appropriately to avoid excessive misjudgments due to complex on-site operations.

[0084] When there is no intersection between the first boundary range and the second boundary range, the judgment boundary is determined based on the false positive weight value and the missed judgment weight value; specifically, if the false positive weight value is greater than the missed judgment weight value, the judgment boundary is the maximum value of the violation indicator in the second boundary range; otherwise, the judgment boundary is the minimum value of the violation indicator in the first boundary range.

[0085] The boundary evolution strategy further includes: if the first boundary range and the second boundary range have an intersection, obtaining the values ​​of all violation indicators in the intersection as candidate judgment boundaries; any candidate judgment boundary is greater than or equal to the minimum value of the violation indicator in the first boundary range and less than or equal to the maximum value of the violation indicator in the second boundary range;

[0086] Calculate the false positive rate and missed positive rate of any candidate decision boundary as the decision boundary, and perform weighted summation of the false positive rate and missed positive rate to obtain the confidence level of the corresponding candidate decision boundary, wherein the weight of the false positive rate is the false positive weight value, and the weight of the missed positive rate is the missed positive weight value;

[0087] The candidate decision boundary with the highest confidence is taken as the decision boundary.

[0088] In this embodiment, the method for calculating the false positive rate and missed positive rate of any candidate decision boundary as the decision boundary is as follows:

[0089] The first boundary range and the second boundary range are sampled n times, and a violation indicator value is extracted as a target violation indicator in each sampling; n is a positive integer;

[0090] If the target violation indicator is smaller than the candidate judgment boundary, the determination label of the target violation indicator is implicit soft violation; otherwise, the determination label of the target violation indicator is explicit soft violation;

[0091] If the soft violation label of the target violation indicator is an explicit soft violation and is inconsistent with the judgment label, the target violation indicator is a missed sample; if the soft violation label of the target violation indicator is a hidden soft violation and is inconsistent with the judgment label, the target violation indicator is a misjudged sample;

[0092] The number of misjudgment samples is counted and divided by n to obtain the misjudgment rate; the number of missed judgment samples is counted and divided by n to obtain the missed judgment rate.

[0093] Current subway violation detection relies primarily on rigid rules, such as whether a driver is wearing a helmet or entering a restricted area. However, a large number of soft violations occur in practice, such as frequent approaching restricted areas without entering, or wandering around the edges of restricted areas. This application dynamically learns the boundaries for soft violations, enabling adaptive detection based on factors such as the operational scenario and the location of the controlled area.

[0094] The violation prediction module constructs a violation chain for each control area based on the hard violation, the soft violation, and the discrimination boundary, and predicts the security control risk of each control area based on the violation chain;

[0095] The violation prediction module includes a behavior modeling unit and a violation matching unit;

[0096] The behavior modeling unit is configured with a violation modeling strategy for constructing a violation behavior chain in any control area; the violation modeling strategy specifically includes:

[0097] Specifying the order of each hard violation and soft violation, and initializing a violation chain; each element in the violation chain corresponds to a hard violation or a soft violation;

[0098] Based on the identification results of hard violations, the corresponding elements in the violation chain are assigned values. If any hard violation exists in the control area, the corresponding element is assigned a value of 1; otherwise, the corresponding element is assigned a value of 0.

[0099] Based on the soft violation identification results, violation indicators, and corresponding discrimination boundaries, corresponding elements in the violation chain are assigned values ​​as follows: If the control area does not identify any hard violation, the corresponding element is assigned a value of 0; if the control area identifies any hard violation, a soft violation label is determined based on the violation indicator and the corresponding discrimination boundary. If the soft violation label is an explicit soft violation, the corresponding element is assigned a value of 1; otherwise, the corresponding element is assigned a value of r, where r is a positive number less than 1. In this embodiment, the specific value of r is set based on the difference between the discrimination boundary and the violation indicator, and the larger the difference, the smaller r.

[0100] The violation matching unit is configured with a reference violation chain for each control area; any reference violation chain is constructed based on the violation modeling strategy; in this embodiment, any reference violation chain corresponds to a security control risk.

[0101] The violation matching unit is also configured with a risk prediction strategy for predicting the security control risk of any control area; the risk prediction strategy specifically includes:

[0102] Calculate the similarity between the violation chain and each reference violation chain in sequence; in this embodiment, the reciprocal of the Euclidean distance is preferably used as the similarity;

[0103] The violation matching unit is further configured with an update cycle. At the beginning of each update cycle, a data collection instruction is sent to the monitoring module; the monitoring module responds to the data collection instruction and re-collects the field data;

[0104] The violation chain is updated in real time based on the re-collected field data; when the value of at least one element in the violation chain changes, a recalculation of the similarity between the violation chain and each reference violation chain is triggered;

[0105] Predicting security control risks based on the similarity between the violation chain and the reference violation chain, specifically including:

[0106] The violation matching unit is further configured with a first similarity threshold and a second similarity threshold; wherein the first similarity threshold is greater than the second similarity threshold;

[0107] If it is detected that the similarity between the violation chain and any reference violation chain is greater than the first similarity threshold, there is a security control risk, and a first risk alert is sent to the early warning module;

[0108] If it is detected that the similarity between the violation chain and any reference violation chain is greater than the second similarity threshold, and the similarity between the violation chain and the corresponding reference violation chain continues to increase in at least m consecutive similarity calculations, there is a security control risk, and a first risk alert is sent to the early warning module; m is a positive integer.

[0109] The early warning module performs additional identification of security control risks based on the illegal behavior chain, and issues risk early warnings based on the security control risks.

[0110] The early warning module includes a risk identification unit and a risk early warning unit;

[0111] The risk identification unit is configured with a risk identification strategy for supplementary identification of security control risks; the risk identification strategy specifically includes: performing a weighted summation on each element in the violation chain to obtain a risk value for the corresponding control area;

[0112] In this embodiment, when calculating the risk value, the weight value of each element in the violation chain can be set by those skilled in the art based on actual needs.

[0113] The risk identification unit is further configured with a risk value threshold. If the risk value of any control area is greater than the risk value threshold, there is a security control risk, and a second risk alert is sent to the risk warning unit.

[0114] The risk warning unit performs a risk warning based on the security control risk, specifically including: if at least one of the first risk warning and the second risk warning is received, then a risk warning is issued. In this embodiment, the risk warning includes at least one of an on-site sound and light alarm and a pop-up warning on the monitoring interface.

[0115] Existing subway yard safety management and control systems primarily rely on single-behavior analysis to generate immediate alerts, such as failure to wear a helmet, inconsistent workwear, and crossing boundaries. This approach lacks a holistic understanding of the chain of violations and the ability to predict their development trends. This application constructs a chain of violations and references the chain, encompassing typical development paths. This allows for real-time identification of the evolution of violation intent and trends. It also predicts whether violations are deteriorating, proactively providing early warning intervention before clear safety risks emerge, and reducing the probability of unexpected accidents.

[0116] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0117] The above describes the embodiments of the present application in conjunction with the accompanying drawings, but the present application is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of this application, ordinary technicians in this field can also make many forms without departing from the purpose and scope of protection of this application, all of which are protected by this application.

Claims

1. A dynamic early warning system for subway station safety management and control, characterized by: It includes monitoring module, identification module, boundary evolution module, violation prediction module, and early warning module; among them: The monitoring module is used to collect on-site data of each control area in the subway depot; An identification module identifies hard violations and soft violations in each control area based on the field data; The boundary evolution module learns the discrimination boundary of the soft violation behavior based on historical soft violation records; The violation prediction module constructs a violation chain for each control area based on the hard violation, the soft violation, and the discrimination boundary, and predicts the security control risk of each control area based on the violation chain; The early warning module performs additional identification of security control risks based on the illegal behavior chain, and issues risk early warnings based on the security control risks.

2. A dynamic early warning system for subway station safety management and control according to claim 1, characterized in that: The on-site data includes restricted area boundary range, personnel location data, and monitoring screen data; The identification module includes a first identification unit and a second identification unit; The first recognition unit is used to identify hard violations in each control area; the hard violations include at least not wearing a safety helmet or not wearing prescribed work clothes; The first recognition unit is configured with an image recognition algorithm; the first recognition unit processes the monitoring screen data based on the image recognition algorithm to identify the hard violation behavior; The second identification unit identifies soft violations in each controlled area based on the restricted area boundary range and personnel positioning data, and records the violation indicators of each soft violation; the soft violation behaviors include at least restricted area approaching behavior and frequent entry and exit behavior; wherein, the violation indicator corresponding to the restricted area approaching behavior is the duration of the restricted area approaching behavior; the violation indicator corresponding to the frequent entry and exit behavior is the number of times frequent entry and exit of the designated area is detected per unit time.

3. A dynamic early warning system for subway station safety management and control according to claim 2, characterized in that: The boundary evolution module includes a database unit; the database unit is configured with historical soft violation records of each control area; each historical soft violation record includes a violation indicator and a soft violation label of a soft violation behavior; the soft violation label includes explicit soft violations and implicit soft violations; The explicit soft violation indicates that the violation index of the soft violation behavior is greater than or equal to the corresponding judgment boundary; The implicit soft violation indicates that the violation index of the soft violation behavior is smaller than the corresponding judgment boundary; Any piece of historical soft violation data also includes the job task type corresponding to the soft violation behavior.

4. A dynamic early warning system for subway station safety management and control according to claim 3, characterized in that: The boundary evolution module further includes a dynamic boundary unit; the dynamic boundary unit is configured with a boundary evolution strategy for learning the discrimination boundary of the soft violation behavior; the boundary evolution strategy specifically includes: Obtain the current job task type of each control area; select historical soft violation records with the same job task type as the current one for each control area from the historical soft violation records, and use them as reference soft violation records for the corresponding control area; For any control area, classify the reference soft violation records based on the type of soft violation to obtain a reference violation record for each soft violation; For any soft violation, further classify the reference violation records based on the soft violation labels; count all violation indicators corresponding to explicit soft violations in the reference violation records of the soft violation, and form a first boundary range of the corresponding soft violation indicators; count all violation indicators corresponding to implicit soft violations in the reference violation records of the soft violation, and form a second boundary range of the corresponding soft violation indicators; A determination boundary for soft violation behavior is determined based on the first boundary range and the second boundary range.

5. A dynamic early warning system for subway station safety management and control according to claim 4, characterized in that: The boundary evolution strategy further includes determining a judgment boundary for any soft violation based on the first boundary range and the second boundary range, as follows: Setting a false positive weight value and a missed positive weight value; the false positive weight value and the missed positive weight value are both greater than 0 and less than 1, and the sum is 1; When there is no intersection between the first boundary range and the second boundary range, the judgment boundary is determined based on the false positive weight value and the missed judgment weight value; specifically, if the false positive weight value is greater than the missed judgment weight value, the judgment boundary is the maximum value of the violation indicator in the second boundary range; otherwise, the judgment boundary is the minimum value of the violation indicator in the first boundary range.

6. A dynamic early warning system for subway station safety management and control according to claim 5, characterized in that: The boundary evolution strategy further includes: if the first boundary range and the second boundary range have an intersection, obtaining the values ​​of all violation indicators in the intersection as candidate judgment boundaries; any candidate judgment boundary is greater than or equal to the minimum value of the violation indicator in the first boundary range and less than or equal to the maximum value of the violation indicator in the second boundary range; Calculate the false positive rate and missed positive rate of any candidate decision boundary as the decision boundary, and perform weighted summation of the false positive rate and missed positive rate to obtain the confidence level of the corresponding candidate decision boundary, wherein the weight of the false positive rate is the false positive weight value, and the weight of the missed positive rate is the missed positive weight value; The candidate decision boundary with the highest confidence is taken as the decision boundary.

7. A dynamic early warning system for subway station safety management and control according to claim 6, characterized in that: The violation prediction module includes a behavior modeling unit and a violation matching unit; The behavior modeling unit is configured with a violation modeling strategy for constructing a violation behavior chain in any control area; The violation modeling strategy specifically includes: Specifying the order of each hard violation and soft violation, and initializing a violation chain; each element in the violation chain corresponds to a hard violation or a soft violation; Based on the identification results of hard violations, the corresponding elements in the violation chain are assigned values. If any hard violation exists in the control area, the corresponding element is assigned a value of 1; otherwise, the corresponding element is assigned a value of 0. Based on the identification results of soft violations, violation indicators, and corresponding judgment boundaries, the corresponding elements in the violation chain are assigned values, as follows: if the control area does not identify any hard violations, the corresponding element is assigned a value of 0; if the control area identifies any hard violations, the soft violation label is determined based on the violation indicators and the corresponding judgment boundary. If the soft violation label is an explicit soft violation, the corresponding element is assigned a value of 1; otherwise, the corresponding element is assigned a value of r, where r is a positive number less than 1.

8. A dynamic early warning system for subway station safety management and control according to claim 7, characterized in that: The violation matching unit is configured with a reference violation chain for each control area; any reference violation chain is constructed based on the violation modeling strategy; The violation matching unit is also configured with a risk prediction strategy for predicting the security control risk of any control area; the risk prediction strategy specifically includes: Calculate the similarity between the violation chain and each reference violation chain in turn; Continuously collecting the field data and updating the violation chain in real time; when the value of at least one element in the violation chain changes, triggering a recalculation of the similarity between the violation chain and each reference violation chain; The security control risk is predicted based on the similarity between the violation chain and the reference violation chain.

9. A dynamic early warning system for subway station safety management and control according to claim 8, characterized in that: Predict security control risks based on the similarity between the violation chain and the reference violation chain, specifically including: The violation matching unit is further configured with a first similarity threshold and a second similarity threshold; If it is detected that the similarity between the violation chain and any reference violation chain is greater than the first similarity threshold, there is a security control risk, and a first risk alert is sent to the early warning module; If it is detected that the similarity between the violation chain and any reference violation chain is greater than the second similarity threshold, and the similarity between the violation chain and the corresponding reference violation chain continues to increase in at least m consecutive similarity calculations, there is a security control risk, and a first risk alert is sent to the early warning module; m is a positive integer.

10. A dynamic early warning system for subway station safety management and control according to claim 9, characterized in that: The early warning module includes a risk identification unit and a risk early warning unit; The risk identification unit is configured with a risk identification strategy for supplementary identification of security management and control risks; The risk identification strategy specifically includes: performing a weighted summation on each element in the violation chain to obtain a risk value for the corresponding control area; The risk identification unit is further configured with a risk value threshold. If the risk value of any control area is greater than the risk value threshold, there is a security control risk, and a second risk alert is sent to the risk warning unit. The risk warning unit performs a risk warning based on the security management and control risk, specifically including: if at least one of the first risk warning and the second risk warning is received, performing a risk warning.

Images