Method and system for dynamically calibrating large passenger flow early warning threshold and electronic equipment
By constructing an accident profile database and establishing mapping rules between the combination of triggers and the critical conditions of passenger flow, the early warning threshold for large passenger flow is dynamically calibrated, which solves the shortcomings of the fixed threshold mode in the existing technology and achieves high adaptability and accurate early warning for different scenarios.
Patent Information
- Application Number
- CN202610543556.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-04-23
- Publication Date
- 2026-07-24
AI Technical Summary
Existing large passenger flow early warning technology uses a fixed threshold mode, which cannot be dynamically adapted to the characteristics of accident risks. This results in untimely warnings in high-risk scenarios and frequent false alarms in low-risk scenarios. The accuracy of the warnings and the risk adaptability are insufficient, and it cannot meet the safety management needs in complex scenarios.
Based on the accident profile database, risk factor combinations are constructed, and mapping rules between risk factor combinations and passenger flow critical conditions are established. By using the association rule mining model, the passenger flow critical conditions corresponding to different risk factor combinations are learned, and the early warning threshold is dynamically calibrated to adapt to risk changes in different scenarios.
It achieves dynamic calibration of the large passenger flow warning threshold, improves the accuracy and reliability of the warning, can respond to risks in complex scenarios in a timely manner, and provides highly adaptable technical support.
Smart Images

Figure CN122453134A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of urban intelligent traffic management technology, and more specifically, to a method, system, and electronic device for dynamic calibration of large passenger flow early warning thresholds. Background Technology
[0002] With the continuous advancement of networked and integrated urban public transportation operations, scenarios such as large-scale event gatherings, severe weather, and passenger congestion in key areas are becoming increasingly frequent. This has led to increasing pressure on passenger flow safety management and emergency response, placing higher demands on the accuracy, timeliness, and risk adaptability of large passenger flow early warning systems. Large passenger flow early warning technology, as a core means of ensuring public transportation operational safety, can identify high-risk conditions such as passenger congestion and stampedes in advance, providing crucial information for on-site traffic management, capacity adjustments, and emergency control. It has become an important component of intelligent public transportation and safe operation systems.
[0003] Existing large passenger flow early warning technologies generally adopt a fixed threshold warning mode. This mode uses preset criteria such as passenger flow density and passenger flow aggregation scale to trigger a corresponding level of warning when real-time passenger flow data reaches the preset threshold. These technologies typically set uniform warning thresholds based on historical, normal passenger flow data. In scenarios with stable daily passenger flow, they can achieve basic passenger flow anomaly monitoring and meet general safety management needs under normal operating conditions. This is currently the mainstream implementation method in the field of large passenger flow early warning for public transportation.
[0004] In the process of realizing this invention, the inventors discovered that the existing large passenger flow early warning method using fixed thresholds has obvious limitations. The early warning thresholds cannot be dynamically adapted to the characteristics of accident risks and can only be judged based on the normal passenger flow status. It is difficult to match the differences in passenger flow risks in different scenarios, resulting in untimely early warnings in high-risk scenarios and frequent false alarms in low-risk scenarios. The accuracy and risk adaptability of the early warning are insufficient, and it is impossible to provide accurate and reliable early warning support for the safety management and emergency response of large passenger flows in public transportation. Summary of the Invention
[0005] Therefore, to address the aforementioned issues, this invention provides a method, system, and electronic device for dynamic calibration of large passenger flow warning thresholds. Based on a constructed accident profile database, risk trigger combinations are formed, and mapping rules are established between these combinations and passenger flow critical conditions. This allows for the determination of passenger flow critical conditions based on real-time risk trigger combinations corresponding to the current passenger flow scenario, and the updating of warning thresholds. This approach facilitates dynamic calibration of large passenger flow warning thresholds, enhances the adaptability of warning thresholds to different passenger flow risk scenarios, and improves the accuracy and reliability of large passenger flow warnings, providing highly adaptable technical support for the safe management of large passenger flows in public transportation.
[0006] In a first aspect, the present invention provides a method for dynamically calibrating a large passenger flow early warning threshold. The method includes: constructing an accident profile library based on multi-dimensional features of historical accidents; extracting primary and secondary risk factors from the accident profile library to form multiple risk factor combinations; establishing a mapping rule between risk factor combinations and passenger flow critical conditions based on an association rule mining model; matching the passenger flow critical conditions corresponding to the real-time risk factor combination corresponding to the current passenger flow scenario from the mapping rule; and updating the existing early warning threshold based on the corresponding passenger flow critical conditions.
[0007] Optionally, in this embodiment of the invention, an accident profile database is constructed based on multi-dimensional features of historical accidents, including: collecting temporal features, spatial features, passenger flow features, environmental features, and handling features of historical accidents; normalizing and structuring the multi-dimensional features to form accident feature entries in a unified format; and classifying and storing the accident feature entries according to scenario type to form the accident profile database. By collecting multi-dimensional features of historical accidents and performing normalization, structuring, and classification, a standardized, complete, and searchable accident profile database can be formed. This provides a comprehensive and standardized data foundation for subsequent extraction of primary and secondary risk factors, construction of risk factor combinations, and establishment of mapping rules, ensuring the accuracy and stability of the subsequent dynamic threshold calibration process.
[0008] Optionally, in this embodiment of the invention, primary and secondary risk factors are extracted from the accident profile database to form multiple risk factor combinations. This includes: selecting primary risk factors with weights higher than a first threshold from the accident profile database based on preset risk weight rules; wherein the first threshold is a weight threshold used to distinguish between high-impact factors and ordinary-impact factors; selecting secondary risk factors with weights lower than the first threshold but higher than a second threshold from the accident profile database; wherein the second threshold is a weight threshold used to distinguish between ordinary-impact factors and low-impact factors; and binding the primary and secondary risk factors according to scenarios to form multiple risk factor combinations. By distinguishing primary and secondary risk factors according to preset risk weight rules and binding them according to scenarios to form risk factor combinations, the risk impact of different factors can be accurately quantified, providing a clear, layered, and scenario-matched basis for subsequent establishment of mapping rules and dynamic calibration of early warning thresholds.
[0009] In the above implementation process, multiple risk factor combinations are formed, including: merging and deduplicating risk factor combinations corresponding to a single passenger flow scenario across scenarios; and classifying the merged risk factor combinations according to risk level, number of factors, and type of factors to generate multi-dimensional risk factor combinations covering different passenger flow scenarios and risk levels. By merging and deduplicating risk factor combinations for a single passenger flow scenario across scenarios, and classifying them according to risk level, number of factors, and type of factors, a comprehensive and clearly defined multi-dimensional risk factor combination system can be formed. This provides a complete, standardized, and efficiently matching data foundation for subsequently establishing mapping rules between factor combinations and passenger flow critical conditions.
[0010] Optionally, in this embodiment of the invention, a mapping rule between risk factor combinations and passenger flow critical conditions is established based on an association rule mining model. This includes: using risk factor combinations as model inputs and passenger flow critical conditions as model outputs; learning the distribution patterns of passenger flow critical conditions corresponding to different risk factor combinations through the association rule mining model to generate standardized mapping rules between risk factor combinations and passenger flow critical conditions. By using risk factor combinations as inputs and passenger flow critical conditions as outputs, and utilizing the association rule mining model to learn their inherent distribution patterns, scientific, objective, and standardized mapping rules can be established, providing a reliable and reusable basis for subsequent matching of passenger flow critical conditions based on real-time risk factor combinations.
[0011] Optionally, in this embodiment of the invention, the passenger flow critical condition corresponding to the real-time risk trigger combination corresponding to the current passenger flow scenario is obtained by matching from the mapping rules, including: real-time collection of scenario feature data of the current passenger flow scenario; identification of trigger type in the scenario feature data, and selection of primary risk trigger features and secondary risk trigger features corresponding to the current passenger flow scenario; combination of the selected primary risk trigger features and secondary risk trigger features to determine the real-time risk trigger combination corresponding to the current passenger flow scenario; comparison of the real-time risk trigger combination with the risk trigger combination in the mapping rules item by item; when the real-time risk trigger combination is consistent with the risk trigger combination in the mapping rules, the passenger flow critical condition corresponding to the consistent risk trigger combination is determined as the passenger flow critical condition corresponding to the current passenger flow scenario. By collecting scene feature data in real time and identifying primary and secondary risk factors to generate real-time risk factor combinations, and then accurately matching them with mapping rules item by item, the current scene's passenger flow critical conditions can be quickly and accurately determined. This provides an objective basis for subsequent early warning threshold updates that fits the on-site risk status, ensuring the timeliness and accuracy of dynamic calibration of early warning thresholds.
[0012] Optionally, in this embodiment of the invention, updating the existing warning threshold based on the corresponding passenger flow critical condition includes: calculating the difference between the existing initial warning threshold and the corresponding passenger flow critical condition to obtain a numerical difference; determining whether the existing initial warning threshold needs to be adjusted based on the numerical difference; when the numerical difference meets the preset adjustment conditions, updating the existing initial warning threshold according to the corresponding passenger flow critical condition to obtain a target warning threshold adapted to the current scenario; when the numerical difference does not meet the preset adjustment conditions, directly using the existing initial warning threshold as the target warning threshold for the current scenario. By calculating the difference between the initial warning threshold and the passenger flow critical condition and dynamically adjusting it according to preset conditions, a target warning threshold adapted to the current scenario can be accurately and reasonably generated, ensuring that the warning threshold is highly matched with the on-site risk level, further improving the accuracy, stability, and risk adaptability of large passenger flow warnings.
[0013] Optionally, in this embodiment of the invention, the trigger accuracy, missed alarm rate, and false alarm rate of existing warning thresholds in actual passenger flow monitoring are collected; the trigger accuracy, missed alarm rate, and false alarm rate are fed back to the accident profile database and the association rule mining model; the risk factor combination, mapping rules, and passenger flow critical conditions are iteratively optimized for dynamic calibration. By collecting warning trigger effect data and feeding it back to the accident profile database and the association rule mining model, the risk factor combination, mapping rules, and passenger flow critical conditions are iteratively optimized, which can continuously improve the threshold calibration basis, so that the warning threshold can continuously self-correct with actual operation, maintain high adaptability and high accuracy in the long term, and provide stable and reliable technical support for the safety management of large passenger flows.
[0014] Secondly, this invention also provides a dynamic calibration system for large passenger flow early warning thresholds. The system includes: an accident profile database construction module for constructing an accident profile database based on multi-dimensional features of historical accidents; a trigger extraction module for extracting primary and secondary risk triggers from the accident profile database to form multiple risk trigger combinations; a rule establishment module for establishing mapping rules between risk trigger combinations and passenger flow critical conditions based on an association rule mining model; a matching calculation module for matching the passenger flow critical conditions corresponding to the real-time risk trigger combinations in the mapping rules based on the current passenger flow scenario; and a threshold update module for updating existing early warning thresholds based on the corresponding passenger flow critical conditions.
[0015] Thirdly, embodiments of the present invention provide an electronic device, the electronic device including a memory and a processor, the memory storing program instructions, and the processor reading and running the program instructions, executing the steps in any of the above implementations.
[0016] The present invention provides a method, system, and electronic device for dynamic calibration of large passenger flow warning thresholds. Based on a constructed accident profile database, risk trigger combinations are formed, and mapping rules between trigger combinations and passenger flow critical conditions are established. This allows for the determination of passenger flow critical conditions based on the real-time risk trigger combinations corresponding to the current passenger flow scenario, and the updating of warning thresholds. This approach helps to achieve dynamic calibration of large passenger flow warning thresholds, improves the adaptability of warning thresholds to different passenger flow risk scenarios, thereby improving the accuracy and reliability of large passenger flow warnings and providing highly adaptable technical support for the safety management of large passenger flows in public transportation. Attached Figure Description
[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the conventional technology, the drawings used in the description of the embodiments or the conventional technology will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.
[0018] Figure 1 This is a flowchart of a method for dynamically calibrating a large passenger flow early warning threshold provided in one embodiment of the present invention; Figure 2 This is a flowchart of an accident profile database construction method provided in one embodiment of the present invention; Figure 3 This is a flowchart of a risk factor combination generation method provided in one embodiment of the present invention; Figure 4 This is a flowchart of a risk factor combination classification method provided in one embodiment of the present invention; Figure 5 This is a flowchart of a mapping rule establishment method provided in one embodiment of the present invention; Figure 6 This is a flowchart of a passenger flow critical condition matching method provided in one embodiment of the present invention; Figure 7 This is a flowchart of a pre-warning threshold update method provided in one embodiment of the present invention; Figure 8 This is a flowchart of an iterative optimization method for early warning thresholds provided in one embodiment of the present invention; Figure 9 This is a schematic diagram of the structure of a dynamic calibration system for a large passenger flow early warning threshold provided in another embodiment of the present invention. Detailed Implementation
[0019] The technical solutions of the embodiments of the present invention will now be described with reference to the accompanying drawings. For example, the flowcharts and block diagrams in the drawings illustrate the architecture, functions, and operations of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in the flowchart or block diagram may represent a module, program segment, or part of code, which contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagram and / or flowchart, and combinations of blocks in the block diagram and / or flowchart, can be implemented using a dedicated hardware-based system that performs the specified function or action, or can be implemented using a combination of dedicated hardware and computer instructions. In addition, the functional modules in the various embodiments of the present invention may be integrated together to form an independent part, or each module may exist separately, or two or more modules may be integrated to form an independent part.
[0020] As the core carrier of urban operation and residents' travel, urban public transportation is facing increasing challenges due to the integrated development of commerce, culture, sports, and tourism, the regular hosting of large-scale events, and the continuous advancement of integrated ground public transportation operations. This has led to increasingly frequent large passenger flows around key stations, transportation hubs, cultural and tourism venues, and passenger distribution channels. Safety hazards such as overcrowding, congestion, and stampedes have significantly increased, placing unprecedentedly stringent demands on public transportation operational safety, passenger flow management efficiency, and emergency response capabilities. Large passenger flow early warning technology, a key component of the public transportation safety assurance system, can identify passenger congestion risks in real time and issue early warning information. It provides decision-making support for on-site guidance, flow control, capacity reinforcement, and emergency response, serving as a core technological support for ensuring passenger safety, improving operational management, and building a smart public transportation system. Its early warning accuracy and risk adaptability directly affect operational safety and service quality.
[0021] The existing early warning technologies for large passenger flows in public transportation generally adopt a fixed threshold early warning scheme, which is the mainstream implementation method in the industry. Its core working principle is as follows: Based on historical regular passenger flow data and operation experience, a unified and unchanged passenger flow density early warning threshold is set, and the mild, moderate, and severe early warning thresholds are fixed as specific values and deployed in the passenger flow monitoring system. During actual operation, passenger flow density data at bus stops, transfer hubs, waiting areas, and集散通道 (aggregation and dispersion channels) are collected in real time through devices such as cameras and passenger flow counters, and the real-time data is compared with the preset fixed threshold. When the real-time passenger flow density reaches or exceeds the corresponding level threshold, the system automatically triggers the corresponding level of passenger flow early warning and pushes it to the operation control platform and on-site disposal personnel to implement control measures such as evacuation and flow restriction. This scheme has a simple implementation logic and convenient deployment, and can complete basic abnormal passenger flow monitoring in scenarios such as off-peak passenger flow and regular and stable passenger flow, and can meet the general safety control requirements under normal daily operation conditions.
[0022] The inventors' research found that the existing fixed threshold early warning technology has a fundamental defect of being unable to adapt to changes in risk scenarios. The early warning rules are rigid and there is no dynamic optimization mechanism, making it difficult to match the passenger flow risk differences in different scenarios and no longer meeting the large passenger flow safety control requirements in complex scenarios. The specific drawbacks are as follows: First, the early warning threshold is fixed throughout the process and not dynamically adjusted in combination with key factors such as accident risks, environmental changes, regional characteristics, and time period attributes. The same judgment standard is used for both high and low risks, resulting in late early warnings in high-risk scenarios and frequent false alarms in low-risk scenarios. Second, there is no correlation mechanism between accident data and early warning thresholds, and the causes and critical conditions of historical accidents and congestion events cannot feed back to the early warning rules, lacking scientificity and pertinence in the early warning basis. Third, the threshold setting depends on experience, and the main risk causes, secondary risk causes, and the superposition effect of causes are not distinguished. When facing complex scenarios such as bad weather, the end of large-scale events, and areas with high evacuation difficulty, the early warning accuracy drops significantly. Fourth, there is no early warning effect feedback and iterative optimization mechanism, and the early warning accuracy rate, missed warning rate, and false warning rate cannot be used to optimize the threshold, and the technical solution is long-term fixed, making it difficult to continuously improve the early warning ability.
[0023] Taking the scenario of the end of a large-scale cultural and tourism event at a key urban public transportation hub plus rainy weather as an example, a comprehensive bus hub in a certain city is adjacent to large-scale cultural and tourism venues. During rainy days, when there is a concentrated end of the event at the venue, an instantaneous large passenger flow is generated. A large number of passengers quickly rush into the waiting area of the hub and the surrounding passages, and the ground is slippery, resulting in a reduced evacuation efficiency. When implementing the large passenger flow warning using the existing fixed threshold warning scheme, the specific application process is as follows: The system still makes judgments according to the fixed passenger flow density threshold set during normal sunny days and flat peak hours, without making any adjustments to the superimposed risks such as rainy days, the end of the event, and high evacuation difficulty; When the passenger flow density quickly rises to the conventional severe warning threshold, the warning is triggered. However, at this time, the passenger flow has been highly concentrated and the evacuation conditions are restricted. The warning is significantly lagged, and the on-site guidance force is too late to be arranged, which is extremely likely to cause hidden dangers such as congestion, blockage, and even trampling; If the fixed threshold is simply reduced, false alarms will occur frequently during risk-free periods, increasing unnecessary control costs. After analysis, the core reason for the above problems is that the existing technology uses a fixed threshold mode, which cannot dynamically adjust the warning standard according to risk incentives, and does not establish a threshold calibration mechanism driven by accident portraits. The threshold is seriously out of touch with the actual scenario risks, and the fixed threshold scheme has become difficult to adapt to the complex and changeable large passenger flow safety control requirements.
[0024] Based on this, a large passenger flow warning threshold dynamic calibration method, system and electronic device provided by an embodiment of the present invention form a risk incentive combination based on the constructed accident portrait library, and establish a mapping rule between the incentive combination and the passenger flow critical condition. Thus, the passenger flow critical condition can be determined according to the real-time risk incentive combination corresponding to the current passenger flow scenario, and the warning threshold can be updated. This method helps to realize the dynamic calibration of the large passenger flow warning threshold, improve the adaptability of the warning threshold to different passenger flow risk scenarios, and thus improve the accuracy and reliability of the large passenger flow warning, providing highly adaptable technical support for the safety control of public transportation large passenger flows.
[0025] Please refer to Figure 1 , Figure 1 which is the flowchart of the large passenger flow warning threshold dynamic calibration method provided by an embodiment of the present invention; The large passenger flow warning threshold dynamic calibration method includes: Step S100: Construct an accident portrait library based on multi-dimensional features of historical accidents.
[0026] In the above step S100, taking historical safety accidents such as congestion, overcrowding, and stampedes in the public transportation passenger flow scenario as the analysis object, comprehensively collect multi-dimensional raw data related to the accidents, including time characteristics, spatial characteristics, passenger flow characteristics, environmental characteristics, and disposal characteristics, etc., completely covering typical passenger flow scenario-related information such as regular flat peak periods, severe weather near the sea, dispersal of cultural and tourism events and sports competitions, equipment failures, bus connection delays, and inadequate on-site management; normalize, structurally process, and label the collected multi-dimensional characteristic data, eliminate abnormal and redundant data, and form standardized accident characteristic entries in a unified format; then classify and centrally store the accident characteristic entries according to scenario types, risk levels, and cause attributes, and construct an accident portrait library that includes complete accident causes, passenger flow critical conditions, and disposal result information, and is retrievable, matchable, and extensible, providing comprehensive, standardized, and reliable data support for subsequent extraction of primary and secondary risk causes, generation of risk cause combinations, and establishment of mapping rules.
[0027] Step S200: Extract the primary risk causes and secondary risk causes from the accident portrait library to form multiple risk cause combinations.
[0028] In the above step S200, based on a preset risk weight rule, quantitatively assign values according to the degree of influence of the cause on the passenger flow accident. Among them, core factors such as severe weather near the sea, special periods of cultural and tourism / sports events, and high evacuation difficulty, which directly induce high risks and significantly reduce the evacuation efficiency, are set as high weight scores, and factors such as equipment failures, bus connection delays, and inadequate on-site management, which only play an auxiliary aggravating role, are set as medium weight scores; based on this risk weight rule, screen out the primary risk causes with weights higher than the first threshold, where the first threshold is the weight critical value for distinguishing high-influence causes from ordinary-influence causes; at the same time, screen out the secondary risk causes with weights lower than the first threshold and higher than the second threshold, where the second threshold is the weight critical value for distinguishing ordinary-influence causes from low-influence causes; bind and combine the extracted primary risk causes and secondary risk causes according to a single passenger flow scenario, then perform cross-scenario merging and duplicate removal processing on the cause combinations in each single scenario, and further classify and grade them according to risk levels, the number of causes, and cause types, and finally form multi-dimensional risk cause combinations covering regular scenarios, single primary cause scenarios, multi-primary cause combination scenarios, and primary cause plus secondary cause scenarios, providing layered and clear cause data support that fits the actual operation scenario for subsequent establishment of mapping rules.
[0029] Step S300: Based on the association rule mining model, establish a mapping rule between the risk cause combination and the passenger flow critical condition.
[0030] In the above step S300, the generated multi-dimensional risk incentive combinations are used as the input items of the association rule mining model, and the passenger flow critical conditions corresponding to historical accidents under different incentive combinations are used as the model output items. The Apriori algorithm can be used to perform association learning on the incentive combinations and passenger flow critical conditions in the accident portrait library, mine the distribution rules of the passenger flow critical conditions corresponding to different risk incentive combinations, and then generate a one-to-one standardized mapping rule between the risk incentive combinations and the passenger flow critical conditions, so that different scenarios such as single incentives, superposition of multiple incentives, and combination of primary and secondary incentives can be matched with the corresponding passenger flow critical conditions, providing a standard and reliable judgment basis for subsequent real-time scenario matching and threshold dynamic calibration.
[0031] Step S400: According to the real-time risk incentive combination corresponding to the current passenger flow scenario, match the passenger flow critical condition corresponding to the real-time risk incentive combination from the mapping rules.
[0032] In the above step S400, the scenario feature data of the current passenger flow scenario can be collected in real time, the incentive type of the scenario feature data can be identified, the primary risk incentive features and secondary risk incentive features corresponding to the current passenger flow scenario can be screened respectively, and the screened primary risk incentive features and secondary risk incentive features can be combined to determine the real-time risk incentive combination corresponding to the current passenger flow scenario; compare the real-time risk incentive combination item by item with the risk incentive combinations in the pre-established mapping rules. When the real-time risk incentive combination is consistent with a certain risk incentive combination in the mapping rules, the passenger flow critical condition corresponding to the consistent risk incentive combination is determined as the passenger flow critical condition corresponding to the current passenger flow scenario, so as to quickly and accurately obtain the judgment benchmark adapted to the on-site risk status.
[0033] Step S500: Update the existing warning threshold based on the corresponding passenger flow critical condition.
[0034] In the above step S500, the difference between the existing initial warning threshold and the matched passenger flow critical condition can be calculated to obtain a numerical difference, and it can be judged whether the existing initial warning threshold needs to be adjusted according to the numerical difference; when the numerical difference meets the preset adjustment condition, the existing initial warning threshold is increased or decreased according to the corresponding passenger flow critical condition to obtain the target warning threshold adapted to the current risk scenario; when the numerical difference does not meet the preset adjustment condition, the existing initial warning threshold is directly used as the target warning threshold for the current scenario, so as to realize the dynamic adjustment of the warning threshold with the on-site risk level.
[0035] It can be seen that the dynamic calibration method for the large passenger flow warning threshold provided by the embodiment of the present invention forms a risk incentive combination based on the constructed accident portrait library, and establishes a mapping rule between the incentive combination and the passenger flow critical condition. Thus, the passenger flow critical condition can be determined according to the real-time risk incentive combination corresponding to the current passenger flow scenario, and the warning threshold can be updated. This method helps to achieve the dynamic calibration of the large passenger flow warning threshold, improve the adaptability of the warning threshold to different passenger flow risk scenarios, and thus improve the accuracy and reliability of the large passenger flow warning, providing highly adaptable technical support for the safety control of public transportation large passenger flows.
[0036] Please refer to Figure 2 , Figure 2 which is a flowchart of the accident portrait library construction method provided in an embodiment of the present invention; this accident portrait library construction method includes: Step S20: Collect the time characteristics, spatial characteristics, passenger flow characteristics, environmental characteristics, and handling characteristics of historical accidents.
[0037] In the above step S20, it is possible to collect the time characteristics, spatial characteristics, passenger flow characteristics, environmental characteristics, and handling characteristics related to historical congestion, crowding, and safety accidents in the urban large passenger flow scenario, focusing on covering typical scenario information such as sunny days, severe weather near the sea, cultural and tourism and event periods, areas with high evacuation difficulty, equipment failures, bus connection delays, and inadequate on-site management, providing complete and true original data support for the subsequent structured processing and classification storage of the accident portrait library.
[0038] Step S21: Normalize and structure the multi-dimensional characteristics to form accident feature entries in a unified format.
[0039] In the above step S21, it is possible to perform normalization and structured processing on the multi-dimensional data such as the collected time characteristics, spatial characteristics, passenger flow characteristics, environmental characteristics, and handling characteristics, eliminate abnormal, redundant, and invalid data, and unify the field format and annotation specifications to form accident feature entries with a unified standard that can be directly used for analysis and storage.
[0040] Step S22: Classify and store the accident feature entries according to the scenario type to form an accident portrait library.
[0041] In the above step S22, the standardized accident feature entries are classified and sorted according to scenario types such as normal scenarios, severe weather, cultural and tourism events, areas with high evacuation difficulty, connection delays, and insufficient management, and then uniformly stored to form an urban traffic accident portrait library with comprehensive coverage, clear classification, and convenient retrieval and call.
[0042] It can be seen that the accident portrait library construction method provided by the embodiment of the present invention can form a standardized, complete, and retrievable accident portrait library by collecting multi-dimensional features of historical accidents, normalizing, structuring, and classifying them into the library, providing a comprehensive and standard data basis for subsequent extraction of primary and secondary risk factors, construction of risk factor combinations, and establishment of mapping rules, and ensuring the accuracy and stability of the subsequent threshold dynamic calibration process.
[0043] Please refer to Figure 3 , Figure 3 which is the flowchart of the risk factor combination generation method provided by the embodiment of the present invention; the risk factor combination generation method includes: Step S30: Based on a preset risk weight rule, screen out the primary risk factors with weights higher than the first threshold from the accident portrait library.
[0044] In the above step S30, various risk factors in the accident portrait library are quantitatively scored according to the preset risk weight rule, and corresponding weight scores are assigned according to the degree of inducing and aggravating urban large passenger flow congestion, crowding, and safety accidents by each factor. The first threshold is the weight critical value for distinguishing high-impact factors from ordinary-impact factors. The factors with weights greater than this first threshold are determined as primary risk factors. The primary risk factors include three core factors: severe weather near the sea, special periods of cultural tourism / events, and areas with high evacuation difficulty.
[0045] Step S31: Screen out the secondary risk factors with weights lower than the first threshold and higher than the second threshold from the accident portrait library.
[0046] In the above step S31, continue to determine the weights of the remaining factors in the accident portrait library according to the preset risk weight rule. The second threshold is the weight critical value for distinguishing ordinary-impact factors from low-impact factors. The factors with weights between the first threshold and the second threshold are determined as secondary risk factors. The secondary risk factors include auxiliary factors such as equipment failures, inadequate management and control, and bus connection delays.
[0047] Step S32: Bind the primary risk factors and the secondary risk factors according to the scenarios to form multiple risk factor combinations.
[0048] In the above step S32, the screened primary risk factors and secondary risk factors can be associated and bound according to the corresponding actual urban traffic scenarios, so that the primary and secondary risk factors in the same passenger flow scenario form a corresponding relationship, and then multiple risk factor combinations adapted to different scenarios are generated.
[0049] It can be seen that the risk incentive combination generation method provided by the embodiment of the present invention can accurately quantify the risk impact degrees of different incentives by distinguishing the primary risk incentives and secondary risk incentives according to the preset risk weight rules, and binding the primary and secondary risk incentives by scenario to form a risk incentive combination, providing an incentive basis with clear stratification and matching the real scenario for establishing mapping rules and dynamically calibrating warning thresholds subsequently.
[0050] Please refer to Figure 4 , Figure 4 which is the flowchart of the risk incentive combination grading and classification method provided by the embodiment of the present invention; the risk incentive combination grading and classification method includes: Step S40: Cross-scene merge and de-duplicate the risk incentive combinations corresponding to a single passenger flow scenario.
[0051] In the above step S40, the risk incentive combinations formed under each single passenger flow scenario are summarized and integrated, the incentive combinations with the same content and similar characteristics in different scenarios are cross-scene merged, and the duplicate and redundant incentive combination entries are removed to ensure the simplicity and effectiveness of subsequent classification processing.
[0052] Step S41: Grade and classify the merged risk incentive combinations according to the risk level, the number of incentives, and the incentive type to generate a multi-dimensional risk incentive combination covering different passenger flow scenarios and different risk levels.
[0053] In the above step S41, based on the high or low risk level, the number of included incentives, and the difference in the categories to which the incentives belong, a systematic grading and classification arrangement is carried out on the merged and de-duplicated risk incentive combinations, and the combinations are divided into four major categories: regular scenario, single primary incentive scenario, multi-primary incentive combination scenario, and primary incentive + secondary incentive auxiliary scenario, finally forming a multi-dimensional risk incentive combination system covering various urban traffic passenger flow scenarios and adapting to different safety risk levels.
[0054] It can be seen that the risk incentive combination grading and classification method provided by the embodiment of the present invention can form a multi-dimensional risk incentive combination system with comprehensive coverage and clear dimensions by cross-scene merging, de-duplicating the risk incentive combinations of a single passenger flow scenario, and grading and classifying according to the risk level, the number of incentives, and the incentive type, providing a complete, standardized, and efficiently matchable incentive data basis for establishing mapping rules between incentive combinations and passenger flow critical conditions subsequently.
[0055] Please refer to Figure 5 , Figure 5 which is the flowchart of the mapping rule establishment method provided by the embodiment of the present invention; the mapping rule establishment method includes: Step S50: Use the risk incentive combination as the model input item and the passenger flow critical condition as the model output item.
[0056] In the above step S50, the multi-dimensional risk factor combinations that have completed hierarchical classification are used as input items for the association rule mining model, and the passenger flow critical conditions that trigger passenger flow congestion and crowding risks in the corresponding scenarios are used as model output items to complete the structured construction of the model training data; in the specific implementation process, for the conventional scenario, the basic warning thresholds can be preset as the benchmark, with a mild warning of 8 people / m², a moderate warning of 12 people / m², and a severe warning of 16 people / m².
[0057] Step S51: The association rule mining model is used to learn the distribution law of the passenger flow critical conditions corresponding to different risk factor combinations, so as to generate a standardized mapping rule between the risk factor combination and the passenger flow critical condition.
[0058] In the above step S51, the association rule mining model can be used to conduct in-depth association analysis and law mining on the input risk factor combination and the output passenger flow critical condition. This model learns from a large number of sample data in the historical accident portrait library to identify and refine the internal association and distribution law between different risk factor combinations and the corresponding passenger flow critical conditions; in the specific implementation process, combined with the characteristics of low regular passenger flow, strong instantaneous tourism aggregation, significant impact of coastal weather, and large differences in station evacuation capabilities in the Lingang New Area, a standardized mapping rule between the risk factor combination and the passenger flow critical condition that is unified and can be directly called can be generated, that is, the threshold for the single main cause scenario is moderately lowered; for the scenario of multiple main causes superimposed, it is further lowered on the basis of a single cause; for the scenario of main cause superimposed with secondary causes, it is slightly assisted to be lowered on the basis of the main cause threshold; for the conventional scenario, the basic warning threshold remains unchanged.
[0059] It can be seen that the mapping rule establishment method provided by the embodiment of the present invention, by using the risk factor combination as the input and the passenger flow critical condition as the output, and using the association rule mining model to learn its internal distribution law, can establish a scientific, objective, and standardized mapping rule, providing a reliable and reusable judgment basis for subsequent matching of passenger flow critical conditions according to real-time factor combinations.
[0060] Please refer to Figure 6 , Figure 6 which is the flowchart of the passenger flow critical condition matching method provided by the embodiment of the present invention; this passenger flow critical condition matching method includes: Step S60: Real-time collect the scenario feature data of the current passenger flow scenario.
[0061] In the above step S60, the scenario feature data such as the environment, passenger flow, area control, and feeder operation in the current urban traffic passenger flow scenario can be real-time collected through front-end perception, environmental monitoring, and traffic operation monitoring devices, providing a real-time data basis for subsequent cause identification.
[0062] Step S61: Identify the cause types of the scenario feature data, and separately screen out the main risk cause features and secondary risk cause features corresponding to the current passenger flow scenario.
[0063] In the above step S61, perform feature analysis and cause matching on the collected real-time scenario feature data, distinguish the influence degree according to the preset weight rules, and identify whether there are main causes such as severe coastal weather, cultural and tourism activity periods, and high evacuation difficulty in the current scenario, as well as secondary causes such as equipment failures, transfer delays, and insufficient management and control, to form a real-time risk cause combination for the current scenario.
[0064] Step S62: Compare the real-time risk cause combination item by item with the risk cause combination in the mapping rule; when the real-time risk cause combination is consistent with the risk cause combination in the mapping rule, determine the passenger flow critical condition corresponding to the consistent risk cause combination as the passenger flow critical condition corresponding to the current passenger flow scenario.
[0065] In the above step S62, perform item-by-item feature comparison between the constructed real-time risk cause combination and various risk cause combinations in the pre-established mapping rule. If there is a completely matching combination, the passenger flow critical condition corresponding to this combination can be directly called as the determination standard for the current passenger flow scenario. For example, when severe weather appears alone: 14 persons / ㎡ for severe degree, 10 persons / ㎡ for moderate degree; when cultural and tourism / event periods appear alone: 13 persons / ㎡ for severe degree, 11 persons / ㎡ for moderate degree; when high evacuation difficulty areas appear alone: 14 persons / ㎡ for severe degree, 10 persons / ㎡ for moderate degree; when two main causes are superimposed: 11 - 12 persons / ㎡ for severe degree, 8 - 9 persons / ㎡ for moderate degree; when three main causes are superimposed: 10 persons / ㎡ for severe degree, 7 persons / ㎡ for moderate degree; when main cause + secondary cause: reduce by 1 person / ㎡ based on the main cause threshold.
[0066] It can be seen that the passenger flow critical condition matching method provided by the embodiment of the present invention can quickly and accurately determine the passenger flow critical condition corresponding to the current scenario by collecting real-time scenario feature data, identifying the main and secondary risk causes to generate a real-time risk cause combination, and then performing precise item-by-item matching with the mapping rule, providing an objective basis that conforms to the on-site risk status for the subsequent update of the warning threshold, and ensuring the timeliness and accuracy of the dynamic calibration of the warning threshold.
[0067] Please refer to Figure 7 , Figure 7 which is the flowchart of the warning threshold update method provided by the embodiment of the present invention; this warning threshold update method includes: Step S70: Calculate the difference between the existing initial warning threshold and the corresponding passenger flow critical condition to obtain a numerical difference.
[0068] In the above step S70, a numerical subtraction operation is performed on the passenger flow critical condition obtained by matching the preset initial passenger flow warning threshold with the current scenario to obtain the numerical difference between the two, providing a quantitative basis for subsequent threshold adjustment judgment.
[0069] Step S71: Determine whether it is necessary to adjust the existing initial warning threshold according to the numerical difference.
[0070] In the above step S71, first, if the difference in the normal scenario is zero, there is no adjustment; subsequently, the calculated numerical difference is compared with the preset allowable range of the difference, and according to the comparison result, it is judged whether the current initial warning threshold needs to be adjusted accordingly based on the passenger flow critical condition.
[0071] Step S72: When the numerical difference meets the preset adjustment condition, update the existing initial warning threshold according to the corresponding passenger flow critical condition, and perform a downward update on the existing initial warning threshold according to the mapping rule. Moreover, the secondary inducement is only slightly reduced by 1 person / ㎡ additionally when the primary inducement exists, so as to obtain the target warning threshold adapted to the current scenario.
[0072] In the above step S72, if the numerical difference exceeds the preset allowable range, that is, it meets the adjustment condition, the initial warning threshold is corrected and updated based on the passenger flow critical condition to generate the target warning threshold that fits the current risk scenario.
[0073] Step S73: When the numerical difference does not meet the preset adjustment condition, directly use the existing initial warning threshold as the target warning threshold for the current scenario.
[0074] In the above step S73, if the numerical difference is within the preset allowable range, that is, it does not meet the adjustment condition, keep the original warning threshold unchanged, and the initial warning threshold can be directly used as the target warning threshold for the current scenario.
[0075] It can be seen that the warning threshold update method provided by the embodiment of the present invention can accurately and reasonably generate the target warning threshold adapted to the current scenario by calculating the difference between the initial warning threshold and the passenger flow critical condition and dynamically adjusting according to the preset conditions, ensuring a high degree of matching between the warning threshold and the on-site risk level, and further improving the accuracy, stability and risk adaptation ability of the large passenger flow warning.
[0076] Please refer to Figure 8 , Figure 8 which is the flowchart of the warning threshold iterative optimization method provided by the embodiment of the present invention; this warning threshold iterative optimization method includes: Step S80: Collect the trigger accuracy rate, missed warning rate, and false warning rate of the existing warning threshold in actual passenger flow monitoring.
[0077] In the above step S80, continuously collect the actual operation data of the current warning threshold in the on-site monitoring of urban traffic passenger flow. The warning accuracy rate, missed warning rate, and false warning rate can be statistically calculated, and the warning effects in high-frequency scenarios such as cultural and tourism activities, bad weather, and high evacuation areas are key recorded to form a quantitative evaluation index for the application effect of the threshold.
[0078] Step S81: Feed back the trigger accuracy rate, missed warning rate, and false warning rate to the accident portrait library and the association rule mining model.
[0079] In the above step S81, use the above accuracy rate, missed warning rate, and false warning rate indicators as feedback data, and transmit them back to the historical accident portrait library and the association rule mining model. At the same time, incorporate the actual accident data such as newly occurring congestion, blocked evacuation, and passenger flow aggregation to provide a verification basis for model optimization in real scenarios.
[0080] Step S82: Iteratively optimize the risk incentive combination, mapping rule, and passenger flow critical condition for dynamic calibration.
[0081] In the above step S82, based on the feedback data, iteratively correct and optimize the existing risk incentive combination, mapping rule, and corresponding passenger flow critical condition. The whole process is automatically executed without manual intervention, which helps to quickly adapt to the scenario changes such as the instantaneous fluctuation of passenger flow of cultural and tourism visitors, sudden bad weather, and diverse transfer modes, and continuously improve the warning accuracy and control applicability.
[0082] It can be seen that the warning threshold iterative optimization method provided by the embodiment of the present invention can continuously improve the threshold calibration basis by collecting warning trigger effect data and feeding it back to the accident portrait library and the association rule mining model, and iteratively optimize the risk incentive combination, mapping rule, and passenger flow critical condition, so that the warning threshold can continuously self-correct according to the actual operation situation, maintain high adaptability and high accuracy in the long term, and provide stable and reliable technical support for the safety control of large passenger flows.
[0083] Please refer to Figure 9 , Figure 9Schematic diagram of the large passenger flow early warning threshold dynamic calibration system provided by the embodiment of the present invention. The large passenger flow early warning threshold dynamic calibration system 9 includes: an accident portrait library construction module 10, an inducement extraction module 20, a rule establishment module 30, a matching calculation module 40, and a threshold update module 50; wherein, the accident portrait library construction module 10 is used to construct an accident portrait library based on multi-dimensional features of historical accidents; the inducement extraction module 20 is used to extract primary risk inducements and secondary risk inducements from the accident portrait library to form multiple risk inducement combinations; the rule establishment module 30 is used to establish a mapping rule between the risk inducement combination and the passenger flow critical condition based on the association rule mining model; the matching calculation module 40 is used to match and obtain the passenger flow critical condition corresponding to the real-time risk inducement combination from the mapping rule according to the real-time risk inducement combination corresponding to the current passenger flow scenario; the threshold update module 50 is used to update the existing early warning threshold based on the corresponding passenger flow critical condition.
[0084] Based on the same inventive concept, the embodiment of the present invention also provides an electronic device, which includes a memory and a processor. When the processor reads and runs the program instructions stored in the memory, the steps in any of the above implementation manners are executed.
[0085] The above are only the embodiments of the present invention and are not used to limit the protection scope of the present invention. For those skilled in the art, the present invention can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.
Claims
1. A method for dynamically calibrating a large passenger flow early warning threshold, characterized in that, The method includes: Construct an accident profile database based on multi-dimensional features of historical accidents; Primary and secondary risk factors are extracted from the accident profile database to form multiple risk factor combinations. Based on the association rule mining model, a mapping rule between the risk factor combination and the passenger flow critical condition is established. Based on the combination of real-time risk factors corresponding to the current passenger flow scenario, the passenger flow critical condition corresponding to the combination of real-time risk factors is obtained from the mapping rules. Based on the corresponding passenger flow critical conditions, update the existing early warning thresholds.
2. The method according to claim 1, characterized in that, The accident profile database constructed based on multi-dimensional features of historical accidents includes: Collect historical accident data on temporal, spatial, passenger flow, environmental, and handling characteristics; The multi-dimensional features are normalized and structured to form accident feature entries in a unified format. The accident feature entries are categorized and stored in the database according to scenario type to form an accident profile database.
3. The method according to claim 1, characterized in that, The process of extracting primary and secondary risk factors from the accident profile database to form multiple risk factor combinations includes: Based on preset risk weighting rules, the main risk factors with weights higher than a first threshold are selected from the accident profile database; wherein, the first threshold is used as a weight threshold to distinguish between high-impact factors and ordinary-impact factors. The accident profile database is used to select secondary risk factors with weights lower than a first threshold and higher than a second threshold; wherein, the second threshold is used as a weight threshold to distinguish between factors with ordinary impact and factors with low impact. The primary risk factors and secondary risk factors are bound together according to the scenario to form multiple risk factor combinations.
4. The method according to claim 3, characterized in that, The combination of multiple risk factors includes: The risk factors corresponding to a single passenger flow scenario are combined and deduplicated across scenarios. The merged risk factor combinations are classified and categorized according to risk level, number of triggers, and type of triggers to generate multi-dimensional risk factor combinations covering different passenger flow scenarios and risk levels.
5. The method according to claim 1, characterized in that, The association rule mining model establishes mapping rules between the risk factor combination and passenger flow critical conditions, including: The risk factors combination is used as the model input and the passenger flow critical condition is used as the model output. By using association rule mining models to learn the distribution patterns of passenger flow critical conditions corresponding to different combinations of risk factors, standardized mapping rules between risk factor combinations and passenger flow critical conditions are generated.
6. The method according to claim 1, characterized in that, The step of matching the passenger flow critical condition corresponding to the real-time risk factor combination based on the current passenger flow scenario from the mapping rules includes: Real-time collection of scene characteristic data of current passenger flow scenarios; The scenario feature data is used to identify the type of risk, and the primary risk risk features and secondary risk risk features corresponding to the current passenger flow scenario are selected respectively. The primary risk factor features and secondary risk factor features obtained from the screening are combined to determine the real-time risk factor combination corresponding to the current passenger flow scenario; The real-time risk factor combination is compared item by item with the risk factor combination in the mapping rule; when the real-time risk factor combination is consistent with the risk factor combination in the mapping rule, the passenger flow critical condition corresponding to the consistent risk factor combination is determined as the passenger flow critical condition corresponding to the current passenger flow scenario.
7. The method according to claim 1, characterized in that, The step of updating the existing early warning threshold based on the corresponding passenger flow critical condition includes: The difference between the existing initial warning threshold and the corresponding passenger flow critical condition is calculated to obtain the numerical difference. Determine whether the existing initial warning threshold needs to be adjusted based on the numerical difference. When the numerical difference meets the preset adjustment conditions, the existing initial warning threshold is updated according to the corresponding passenger flow critical conditions to obtain the target warning threshold that is adapted to the current scenario. When the numerical difference does not meet the preset adjustment conditions, the existing initial warning threshold is directly used as the target warning threshold for the current scene.
8. The method according to claim 1, characterized in that, The method further includes: The accuracy rate of triggering existing early warning thresholds, the rate of missed alarms, and the rate of false alarms in actual passenger flow monitoring are collected. The trigger accuracy, false alarm rate, and false alarm rate are fed back to the accident profile database and the association rule mining model. The risk factor combination, mapping rules, and passenger flow critical conditions are iteratively optimized for dynamic calibration.
9. A dynamic calibration system for a large passenger flow early warning threshold, characterized in that, include: The accident profile database construction module is used to build an accident profile database based on the multi-dimensional features of historical accidents. The trigger extraction module is used to extract primary and secondary risk triggers from the accident profile database to form multiple risk trigger combinations; The rule establishment module is used to establish mapping rules between the risk factor combination and the passenger flow critical conditions based on the association rule mining model. The matching calculation module is used to match the passenger flow critical condition corresponding to the real-time risk factor combination from the mapping rules based on the real-time risk factor combination corresponding to the current passenger flow scenario. The threshold update module is used to update the existing warning threshold based on the corresponding passenger flow critical conditions.
10. An electronic device, characterized in that, The electronic device includes a memory and a processor, the memory storing program instructions, and the processor executing the steps of the method according to any one of claims 1-8 when running the program instructions.
