Method and system for monitoring dynamic congestion degree of subway station based on multi-source data fusion, and medium

Abstract

The invention discloses a multi-source data fusion subway station dynamic congestion degree monitoring method and system and a medium, and relates to the technical field of subway safety monitoring, and the method comprises the steps: obtaining the number of persons collected by each Internet of Things sensor, carrying out the weighted fusion calculation of the collected number of persons according to the type of each Internet of Things sensor, and obtaining the number of persons; obtaining an estimated personnel number of each key area; calculating an environment correction parameter estimated value of each key area according to the initial personnel quantity, and verifying the environment correction parameter estimated value and the collected environment correction parameter value; and dynamically adjusting the weight of each Internet of Things sensor during weighted fusion calculation according to a verification result, finally outputting the number of people in each key area, integrating the crowding degree of each key area, and generating the crowded space distribution condition of the whole subway station. Personnel data are collected through a non-vision-oriented multi-source sensing network, and a multi-source data fusion result is corrected through environment correction parameters, so that a more reliable congestion degree detection result is obtained.

Description

Technical Field

[0001] This invention relates to the field of subway safety monitoring technology, specifically to a method, system, and medium for monitoring dynamic congestion in subway stations using multi-source data fusion. Background Technology

[0002] With the continuous growth of passenger flow in urban rail transit, overcrowding in subway stations has become increasingly prominent, posing serious challenges to operational safety, passenger experience, and emergency response. Existing subway station crowding monitoring schemes have significant shortcomings, mainly in the following aspects: 1. Limited data, incomplete perception, and poor accuracy: Existing solutions largely rely on gate counting or video surveillance analysis from automated fare collection (AFC) systems. Gate counting only obtains the total number of people entering and exiting the station, and cannot perceive the dynamic distribution of people in key areas such as the station hall, passageways, and elevators; video analysis technology suffers from significant recognition errors when faced with complex scenarios such as crowd obstruction, changes in lighting, and limited viewing angles, and there is a continuous risk of privacy leaks.

[0003] 2. High latency and untimely response: Most systems adopt a model of transmitting all data to a cloud processing center for unified calculation. This model has a long data transmission chain and a large computational load, resulting in high response latency and an inability to provide real-time decision support for passenger flow control and management within the station. Often, by the time the cloud calculates the congestion situation, congestion has already occurred within the station.

[0004] 3. Lack of predictive and tracing capabilities: Existing technologies are mainly limited to reflecting historical or current momentary congestion conditions and cannot effectively predict congestion trends in the next few minutes. Furthermore, the system struggles to analyze the causes and spread paths of congestion; for example, it cannot determine which specific train arrival triggered the congestion or which entrance / exit experienced a concentrated influx of passengers.

[0005] 4. Poor system scalability and flexibility: The centralized architecture, which relies on large central servers, has high construction and maintenance costs and makes it difficult to flexibly deploy or update algorithm models according to site characteristics, thus limiting system scalability.

[0006] Therefore, there is an urgent need for a solution that can achieve comprehensive, real-time, accurate, and predictable dynamic congestion calculation for subway stations. Summary of the Invention

[0007] The purpose of this invention is to provide a method, system, and medium for monitoring the dynamic congestion of subway stations by fusing multi-source data. The multi-source sensing network composed of non-visual IoT sensors protects privacy and covers blind spots at the data source. Environmental correction parameters are used to correct the multi-source data fusion results, resulting in more reliable congestion detection results.

[0008] To achieve the above objectives, this application proposes the following solutions: On the one hand, this invention provides a method for monitoring dynamic congestion in subway stations using multi-source data fusion. Multiple complementary IoT sensors are deployed in key areas of the subway station to collect data on passenger numbers and environmental correction parameters. Specifically, the method includes the following steps: S1. Obtain the number of people collected by each IoT sensor deployed in each key area, and perform weighted fusion calculation on the number of people collected according to the type of each IoT sensor to obtain the estimated number of people in each key area. S2. Calculate the estimated value of environmental correction parameters for each key area based on the initial number of personnel, and verify the estimated value of environmental correction parameters with the collected environmental correction parameter values. S3. Based on the verification results, dynamically adjust the weights of each IoT sensor during each weighted fusion calculation, and finally output the number of people in each key area and calculate the congestion level of each key area. S4. Integrate the congestion levels of each key area to generate the congestion spatial distribution of the entire subway station.

[0009] In some specific implementation schemes, the process of obtaining the estimated number of personnel in each key area in step S1 is as follows: S11. Assign an initial weight to the number of people collected by each IoT sensor based on its type and installation location. S12. Select the corresponding multi-source data fusion algorithm based on the proportion of various IoT sensors deployed in the current key areas; S13. Using a multi-source data fusion algorithm and initial weights, the number of people collected by each IoT sensor is weighted and fused to obtain the estimated number of people in each key area.

[0010] In some specific implementation plans, the process of verifying the estimated environmental correction parameters with the collected environmental correction parameter values ​​is as follows: S21. Call the number of people-environment correction parameter correlation function, substitute the initial number of people into the correlation function, and output the estimated value of the environment correction parameter; S22. Calculate the dynamic range of environmental correction parameters based on the estimated values ​​of environmental correction parameters; S23. Input the estimated value of the environmental correction parameter, the dynamic range of the environmental correction parameter, and the collected environmental correction parameter values ​​into the environmental correction parameter verification model, and output the verification results.

[0011] In some specific implementation schemes, the correlation function between the number of people and the environment correction parameter is as follows: C_expected = C_base + k × n; Where C_expected is the estimated value of the environmental correction parameter, C_base is the baseline value of the environmental correction parameter collected during the unmanned period, k represents the influence coefficient per unit number of people, and n represents the initial number of people.

[0012] In some specific implementation schemes, the environment correction parameter validation model env_consistency is: env_consistency = 1 - |C_measured - C_expected| / C_expected_range Where C_measured represents the collected environmental correction parameter value, C_expected represents the estimated value of the environmental correction parameter, and C_expected_range represents the dynamic range of the environmental correction parameter.

[0013] In some specific implementation plans, the process of dynamically adjusting based on verification results is as follows: When the verification results are consistent, the estimated number of personnel will be used as the final number of personnel in each key area; When the verification results are inconsistent and the verification result is greater than the preset threshold, it is determined that the initial number of personnel is too small. The weight of the IoT sensor with the fewest personnel is reduced, and the weighted fusion calculation is re-performed to estimate the number of personnel. When the verification results are inconsistent and the verification results are less than the preset threshold, it is determined that the initial number of personnel is too large. The weight of the IoT sensor with the most personnel is reduced, and the weighted fusion calculation is performed again to estimate the number of personnel.

[0014] In some specific implementation schemes, environmental correction parameters include CO2 concentration. IoT sensors are also installed in key areas to collect environmental impact parameters that assist in determining the environmental correction parameters. These environmental impact parameters include temperature, humidity, and particulate matter concentration. A further step is included between steps S1 and S2: Based on the collected environmental impact parameters, determine whether the environmental correction parameters are reliable; If the environmental correction parameters are reliable, proceed to steps S2-S4; If the environmental correction parameters are unreliable, the estimated number of people in each key area is used as the final number of people, the congestion level of each key area is calculated, and step S4 is executed.

[0015] In some specific implementation schemes, the process for determining whether environmental correction parameters are reliable is as follows: Based on the collected CO2 concentration and various environmental impact parameters, determine whether the trend of CO2 concentration change is consistent with the trend of each environmental impact parameter. If the trends are consistent, the environmental correction parameters are considered reliable; if the trends are inconsistent, the environmental correction parameters are considered unreliable.

[0016] Secondly, this application provides a multi-source data fusion-based dynamic congestion monitoring system for subway stations, comprising: The perception layer includes multiple complementary IoT sensors deployed in each key area of ​​the subway station. Each IoT sensor is used to collect the number of people in the key area and environmental correction parameters. The edge layer includes edge computing gateways deployed in each key area. These edge computing gateways include: The multi-source data fusion calculation module is used to perform weighted fusion calculation on the number of people collected by each IoT sensor in the key area, according to the type of each IoT sensor, to obtain the estimated number of people in each key area. The environmental correction parameter correction module is used to calculate the estimated value of environmental correction parameters for each key area based on the initial number of personnel, verify the estimated value of environmental correction parameters with the collected environmental correction parameter values, dynamically adjust the weight of each IoT sensor during each weighted fusion calculation based on the verification results, and finally output the number of personnel in each key area and calculate the congestion degree of each key area. The platform layer is used to integrate the congestion levels of various key areas and generate the congestion spatial distribution of the entire subway station.

[0017] Thirdly, this application provides a computer-readable storage medium, comprising: One or more processors; A storage unit is used to store one or more programs that, when executed by one or more processors, enable the one or more processors to implement a multi-source data fusion method for monitoring dynamic congestion in subway stations as described in the first aspect.

[0018] The beneficial effects of this invention are as follows: This invention collects the number of people through a multi-source sensing network composed of non-visual IoT sensors. It does not collect information about people, thus protecting pedestrians' privacy at the data source and covering blind spots that cannot be captured by video. In order to improve the fusion accuracy of multi-source data, environmental correction parameters are used to correct the fusion results of multi-source data, resulting in more reliable congestion detection results. Attached Figure Description

[0019] Figure 1 A flowchart of a multi-source data fusion method for monitoring dynamic congestion in subway stations provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the dynamic adjustment process provided in an embodiment of the present invention; Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The following description of at least one exemplary embodiment is merely illustrative and is in no way intended to limit the present invention or its application or use. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Unless otherwise specifically stated, the relative arrangement, numerical expressions, and values ​​of the components and steps described in these embodiments do not limit the scope of the invention.

[0022] At the same time, it should be understood that, for ease of description, the dimensions of the various parts shown in the accompanying drawings are not drawn according to actual scale.

[0023] Furthermore, for clarity and brevity, descriptions of well-known structures, functions, and configurations may have been omitted. Those skilled in the art will recognize that various changes and modifications can be made to the examples described herein without departing from the spirit and scope of this disclosure.

[0024] Techniques, methods, and equipment known to those skilled in the art may not be discussed in detail, but where appropriate, such techniques, methods, and equipment should be considered part of the specification.

[0025] In all examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not as limitations. Therefore, other examples of exemplary embodiments may have different values.

[0026] Example 1 like Figure 1 As shown, this embodiment provides a method for monitoring dynamic congestion in subway stations using multi-source data fusion. Multiple complementary IoT sensors are deployed in key areas of the subway station to collect data on passenger numbers and environmental correction parameters. The method specifically includes the following steps: S1. Obtain the number of people collected by each IoT sensor deployed in each key area, and perform weighted fusion calculation on the number of people collected according to the type of each IoT sensor to obtain the estimated number of people in each key area. Understandably, key areas can be divided according to specific zones within the subway station, such as platforms, concourses, passageways, elevators, and turnstiles. Within each key area, various complementary IoT sensors can be deployed (i.e., the type and number of IoT sensors in each area can be set according to actual conditions), forming a multi-source data acquisition network. These IoT sensors include: 1) Millimeter-wave radar sensor: Deployed on the ceiling in key areas to detect the number, speed, and direction of human movement within those areas. Its advantages include complete privacy protection, immunity to ambient light, penetration of non-metallic obstacles, and high accuracy even in crowded or obstructed environments.

[0027] 2) Wi-Fi / Bluetooth sniffing sensor: By anonymously collecting the MAC addresses of smart devices (and after de-identification processing such as hash encryption), the number of devices is estimated based on signal strength, and the movement trajectory and gathering pattern of the crowd are analyzed.

[0028] 3) LiDAR: Deployed in key areas such as entry and exit points where precise counting is required, it achieves accurate counting of people passing through through high-resolution contour scanning.

[0029] 4) Environmental sensors: used to collect environmental correction parameters such as temperature, humidity, and carbon dioxide concentration, as supplementary evidence for assessing crowding (because dense crowds will cause characteristic changes in environmental correction parameters).

[0030] 5) Existing system data interface: Access real-time passenger flow data for entering and exiting the subway's existing automatic fare collection (AFC) system, as well as information such as arrival timetables and carriage occupancy forecasts from the train operation system.

[0031] With the deployment of the aforementioned IoT sensors, the process for obtaining the estimated number of people in each key area is as follows: S11. Assign an initial weight to the number of people collected by each IoT sensor based on its type and installation location. The initial weights of the data collected by each IoT sensor can be adaptively adjusted based on the location of the IoT sensor, ensuring excellent sensing accuracy from the initial stage. Specific initialization strategies include: 1) Sensor Type-Based Basic Reliability Grading: Based on the built-in, technology-based basic reliability level of the sensor. Typically, LiDAR (Light Detection and Ranging) has near-enumeration accuracy in bottleneck areas with good field of view and can be assigned the highest initial weight (e.g., 0.5); millimeter-wave radar has high accuracy and strong anti-interference capabilities in non-extremely congested environments and is assigned a relatively high initial weight (e.g., 0.3-0.4); Wi-Fi / Bluetooth sniffing, whose accuracy is affected by device prevalence and signal stability, is assigned a medium initial weight as an important reference (e.g., 0.1-0.2). Environmental sensors do not directly output the number of people and therefore do not participate in the weighting allocation but serve as verifiers.

[0032] 2) Adaptive fine-tuning based on installation location and field of view: For the same type of sensor, its initial weight will be fine-tuned according to the effectiveness of its deployment location. Sensors deployed at key passenger flow bottlenecks with unobstructed views (such as turnstiles and narrow passages) will be given a higher initial weight than similar sensors deployed in open areas (such as the center of the station hall) due to their high data certainty.

[0033] S12. Select the corresponding multi-source data fusion algorithm based on the proportion of various IoT sensors deployed in the current key areas; Multi-source data fusion algorithms include those based on Kalman filtering, Bayesian networks, density estimation models, and complex population statistics models. Edge gateways in different key areas can load the most suitable, highly customized fusion algorithm model based on the proportion of sensor types deployed and the characteristics of the scene. For example: If the current key areas, such as stations, are dominated by radar-based IoT sensors, then a complex people counting model based on point cloud clustering and tracking should be selected.

[0034] If the current key area is located in a commercial district and the deployed IoT sensors mainly use Wi-Fi sniffing, then choose to load a density estimation model based on device signal attenuation and mobility pattern analysis.

[0035] S13. Using a multi-source data fusion algorithm and initial weights, the number of people collected by each IoT sensor is weighted and fused to obtain the estimated number of people in each key area.

[0036] Specifically, to avoid the impact of abnormal environmental correction parameters on congestion detection, environmental correction parameters include CO2 concentration. IoT sensors are also installed in key areas to collect environmental impact parameters that assist in the judgment of environmental correction parameters. These environmental impact parameters include temperature, humidity, and particulate matter concentration. The determination of whether to use environmental correction parameters to adjust the weights is made, and the specific process is as follows: Based on the collected environmental impact parameters, determine whether the environmental correction parameters are reliable; Determine whether the trend of CO2 concentration change is consistent with the trend of change of various environmental impact parameters. If the trends are consistent, the environmental correction parameters are considered reliable; if the trends are inconsistent, the environmental correction parameters are considered unreliable.

[0037] If the environmental correction parameters are reliable, proceed to steps S2-S4; If the environmental correction parameters are unreliable, the estimated number of people in each key area is used as the final number of people, the congestion level of each key area is calculated, and step S4 is executed.

[0038] It is easy to understand that because the movement of people affects the environment, environmental parameters in key areas will also change with the movement of people. These environmental parameters include: CO2 concentration: Based on changes in CO2 concentration produced by human respiration, this is the most direct and effective indicator of population density.

[0039] Temperature: Based on the effect of increased microenvironment temperature caused by human body heat dissipation.

[0040] Humidity: Based on the increase in air humidity caused by human respiration and perspiration.

[0041] Particulate matter concentration correction factor: based on the change in inhalable particulate matter concentration caused by the stirring up of ground dust by human activities.

[0042] Because the reliability, sensitivity, and anti-interference ability of different environmental parameters vary greatly in different scenarios, among these environmental parameters, the one that best reflects the impact of personnel movement on the environment is generally selected as the most reliable environmental correction parameter, while other environmental parameters are used as environmental impact parameters to assist in verifying the reliability of the environmental correction parameter.

[0043] Generally, during personnel movement, CO2 is directly generated by human respiration, strongly correlated with the number of people, and less affected by other non-human factors (such as weather, equipment start-up and shutdown), making it the most direct and reliable indicator among the four parameters. Therefore, CO2 concentration is chosen as the environmental correction parameter to adjust the weight of personnel numbers. Environmental impact parameters include temperature, humidity, and particulate matter. When the changing trends of the environmental correction parameter (CO2) and environmental impact parameters (such as temperature and humidity) are consistent, the system's confidence in the "environmental evidence" will be greatly improved. For example, CO2 concentration rises, while temperature and humidity also rise slightly. This forms a complete chain of evidence that "more people lead to environmental changes," allowing the system to adjust sensor weights more decisively. However, when the environmental correction parameter contradicts the environmental impact parameter, diagnostic logic is triggered. For example, CO2 concentration rises sharply, but temperature drops. This is contrary to common sense. At this point, it is necessary to determine whether the CO2 sensor is malfunctioning or whether the ventilation system has introduced a large amount of cold air. In this case, environmental correction will be used cautiously, or not corrected temporarily, and an "abnormal environmental data" alarm will be reported.

[0044] S2. Calculate the estimated value of environmental correction parameters for each key area based on the initial number of personnel, and verify the estimated value of environmental correction parameters with the collected environmental correction parameter values. The specific process for verifying the estimated environmental correction parameters with the collected environmental correction parameter values ​​is as follows: S21. Call the number of people-environment correction parameter correlation function, substitute the initial number of people into the correlation function, and output the estimated value of the environment correction parameter; The environmental parameter correction parameter is essentially a mechanism for verifying the population aggregation effect based on environmental physical changes. Its establishment process includes: Baseline learning phase: During initial deployment or periodic calibration, the system establishes a basic correlation model between environmental parameters and the number of users through the following methods: 1) During periods of low population density (such as after nighttime shutdown), record baseline values ​​for each environmental parameter, including baseline values ​​for CO2 concentration, temperature, and humidity.

[0045] 2) Establish a correlation function between the number of people and environmental parameters through controlled experiments or long-term data accumulation, for example: C_expected = C_base + k × n. Where C_expected is the estimated value of the environmental correction parameter, C_base is the baseline value of the environmental correction parameter collected during the unmanned period, k represents the influence coefficient per unit number of people, and n represents the initial number of people.

[0046] Based on the correlation function, a reasonable fluctuation range for environmental parameters is set for each estimated number of people [C_expected - Δ, C_expected + Δ], where Δ is the normal fluctuation threshold obtained through historical statistics. The dynamic range of the environmental correction parameter is C_expected_range = Δ; S22. Calculate the dynamic range of environmental correction parameters based on the estimated values ​​of environmental correction parameters; S23. Input the estimated value of the environmental correction parameter, the dynamic range of the environmental correction parameter, and the collected environmental correction parameter values ​​into the environmental correction parameter verification model, and output the verification results.

[0047] The environment-corrected parameter validation model env_consistency is: env_consistency = 1 – (C_measured - C_expected) / C_expected_range Where C_measured represents the collected environmental correction parameter value, C_expected represents the estimated value of the environmental correction parameter, and C_expected_range represents the dynamic range of the environmental correction parameter.

[0048] S3. Based on the verification results, dynamically adjust the weights of each IoT sensor during each weighted fusion calculation, and finally output the number of people in each key area and calculate the congestion level of each key area. Environmental parameters are not directly involved in population calculations, but rather serve as a "real-time monitor" of the reliability of sensor data, such as... Figure 2 As shown, the specific process of dynamically adjusting based on the verification results is as follows: When the verification results are consistent, maintain the initial weight calculation for each sensor and use the estimated number of personnel as the final number of personnel in each key area; When the verification results are inconsistent and the verification results are greater than the preset threshold, if the difference between the estimated and actual values ​​of environmental parameters is within the fluctuation range, it indicates that the number of people should be more and the estimated initial number of people is too low. Therefore, the weight of the IoT sensor with the fewest people is reduced and the weighted fusion calculation is performed again to estimate the number of people. When the verification results are inconsistent and the verification results are less than the preset threshold, the difference between the estimated and actual values ​​of environmental parameters exceeds the fluctuation range, indicating that the number of people should be less and the estimated initial number of people is too high. Therefore, the weight of the IoT sensor with the most people should be reduced and the estimated number of people should be recalculated using weighted fusion.

[0049] In some dynamic adjustment methods, the weights can be further adjusted based on the current detection time period, dynamically adjusting the sensor's reliability and historical accuracy in the current time period and environment. For example, during morning and evening rush hours, when smartphone penetration is high and Wi-Fi sniffing data quality is good, its weight is increased accordingly; at night or when there are few people, millimeter-wave radar data is more stable, so its weight is increased.

[0050] Once the number of people in each key area is obtained, the congestion level is calculated based on multiple factors, including the ratio of the number of people to the maximum capacity that the key area can accommodate during its design, pedestrian flow rate, and environmental components. congestion_level = w_density·D + w_flow·F + w_comfort·C Where w_density, w_flow, and w_comfort represent the weights of each factor, which are adaptively set; D represents the density component, D = min(1, estimated_people / N_dynamic), where estimated_people represent the number of people in each key area, and N_dynamic represents the dynamic effective capacity.

[0051] F represents the flow velocity component, F=1- min(1,current_flow_speed / free_flow_speed); C represents the comfort component, C = w_temp·ΔT + w_humidity·ΔH + w_CO2·ΔCO2 Where current_flow_speed represents the current flow rate, and free_flow_speed represents the maximum flow rate; w_temp represents the weight of the temperature change ΔT, w_humidity represents the weight of the humidity change ΔH, and w_CO2 represents the weight of the CO2 concentration change ΔCO2.

[0052] Obtain the real-time congestion level (congestion_status) for each key area. For example, when congestion_level < 0.3, congestion_status = unobstructed. When 0.3 ≤ congestion_level < 0.7, congestion_status = relatively crowded; 0.7 ≤ congestion_level < 0.9, congestion_status = crowded; congestion_level ≥ 0.9, congestion_status = full.

[0053] S4. Integrate the congestion levels of each key area to generate the congestion spatial distribution of the entire subway station.

[0054] The method described in this embodiment can also be deployed in edge gateways in various key areas for computation. A failure in a single gateway will not affect normal monitoring in other areas, resulting in extremely strong overall system robustness. Employing a multi-source data fusion algorithm, it performs real-time cross-validation and weight correction on key sensor data using environmental parameters. This eliminates reliance on camera devices for headcount counting, protecting personal privacy from the source. All raw data potentially involving personal privacy (such as identifiable human silhouettes and device MAC addresses) is processed and analyzed within the edge gateway and immediately discarded, never leaving the local network. Anonymization is performed during data processing. For example, MAC addresses are converted into irreversible hash values; personnel trajectories are represented as abstract movement vectors. Edge gateways in different areas can load the most suitable, highly customized fusion algorithm model based on the type of sensors deployed and the characteristics of the scenario. This avoids the waste of computing resources caused by a "one-size-fits-all" universal model in the cloud, achieving efficient resource utilization.

[0055] Example 2 This embodiment provides a multi-source data fusion dynamic congestion monitoring system for subway stations, consisting of a perception layer, an edge layer, a platform layer, and an application layer. It executes the method described in Embodiment 1 above, utilizing edge computing nodes for real-time data fusion and processing to achieve low-latency, regionally accurate calculations. A cloud platform is used for station-wide situation fusion, macro-analysis, dynamic prediction, and command issuance. Finally, the application layer achieves a closed loop from early warning and coordination to passenger services. Specifically: The perception layer includes multiple complementary IoT sensors deployed in each key area of ​​the subway station. Each IoT sensor is used to collect the number of people in the key area and environmental correction parameters. At the edge layer, an edge computing gateway is deployed in each critical area (e.g., a standalone station or a main corridor). The edge computing gateway includes: The multi-source data fusion calculation module is used to perform weighted fusion calculation on the number of people collected by each IoT sensor in the key area, according to the type of each IoT sensor, to obtain the estimated number of people in each key area. The environmental correction parameter correction module is used to calculate the estimated value of environmental correction parameters for each key area based on the initial number of personnel, verify the estimated value of environmental correction parameters with the collected environmental correction parameter values, dynamically adjust the weight of each IoT sensor during each weighted fusion calculation based on the verification results, and finally output the number of personnel in each key area and calculate the congestion degree of each key area. The edge computing gateway is responsible for receiving raw data uploaded by all sensors within its area. A multi-source data fusion algorithm is run within the edge gateway, assigning dynamic weights to different sensor data and performing weighted fusion to calculate the real-time, high-precision number of people in the area. The edge gateway directly outputs the real-time congestion level of the area and reports this result to the cloud platform.

[0056] Edge computing gateways are not merely data forwarding stations, but rather "regional intelligent brains" with powerful processing capabilities. Their functional characteristics include: 1. Lightweight data and bandwidth optimization 1) Raw data processing: The gateway directly receives and processes the raw data streams from the sensors locally (such as point cloud data from millimeter-wave radar, MAC address lists from Wi-Fi sniffing, and scan frames from LiDAR).

[0057] 2) Feature extraction and aggregation: Inside the gateway, sophisticated AI algorithms perform real-time analysis of the raw data, extract core feature information, and aggregate massive amounts of raw data into a simplified structured result.

[0058] 3) Transmission optimization comparison: Traditional mode: Requires continuous uploading of several megabytes of raw video stream or radar point cloud data per second.

[0059] This invention's mode requires uploading only tens of bytes of structured data every few seconds, for example: { "area_id": "Platform_A", "timestamp":"specific time of year-month-day", "people_count": 142, "congestion_level": "crowded", "congestion_status":0.92 Technical benefits: Network bandwidth requirements are reduced by more than 99%, greatly alleviating the burden on network infrastructure and operating costs.

[0060] 2. Source privacy protection mechanism 1) Sensitive data does not leave the domain: All raw data that may involve personal privacy (such as identifiable human body contours and the device's real MAC address) is processed and analyzed within the edge gateway and then immediately discarded, never leaving the local network.

[0061] 2) Anonymization: Anonymize data immediately during processing. For example, convert MAC addresses into irreversible hash values; represent personnel trajectories as abstract movement vectors.

[0062] Technical effect: It eliminates the risk of privacy data leakage from the root of the technical architecture and perfectly complies with strict domestic and international data regulations such as GDPR and the Personal Information Protection Law.

[0063] 3. High system reliability and self-regulation capability during network outages 1) Advantages of distributed architecture: The system consists of multiple edge gateways. The failure of a single gateway will not affect the normal monitoring of other areas, and the overall system is extremely robust.

[0064] 2) Weak Network / Outage Autonomy: When the network connection between the edge gateway and the cloud platform is interrupted, the gateway has an "island operation mode": a) Continuous monitoring: Continue to perform data collection, fusion, and congestion calculation in this area.

[0065] b) Local linkage: Based on preset rules, automatically trigger local control measures, such as: Control the broadcasting equipment in this area to play pre-stored traffic guidance messages.

[0066] Adjust the intensity of the ventilation system in this area.

[0067] Change the status of indicator lights or the content of electronic signage in this area.

[0068] c) Data caching and synchronization: Historical data during network outages can be temporarily stored locally and automatically uploaded to the cloud after the network is restored, ensuring data integrity.

[0069] Technical effect: It ensures the continuous operation of the monitoring and control system under extreme conditions in the critical public transportation scenario of the subway.

[0070] 4. Heterogeneous computing and specialized processing 1) Customized Algorithm Loading: Edge gateways in different regions can load the most suitable, highly customized fusion algorithm model based on the types of sensors deployed and the characteristics of the scene. For example: For radar-based stations: a complex people counting model based on point cloud clustering and tracking is loaded.

[0071] Commercial areas primarily using Wi-Fi sniffing: Load density estimation models based on device signal attenuation and mobility pattern analysis.

[0072] 2) Efficient use of resources: It avoids the waste of computing resources caused by the "one-size-fits-all" general model in the cloud and realizes professional and efficient processing.

[0073] The platform layer is used to integrate the congestion levels of various key areas and generate the congestion spatial distribution of the entire subway station.

[0074] The cloud platform receives and aggregates regional congestion data reported by all edge gateways, performing macro-level analysis and in-depth mining, including: 1) Station-wide situation integration: Integrate data from various areas to generate a dynamic heat map of the entire subway station, intuitively displaying the distribution of congested spaces throughout the station.

[0075] 2) Dynamic Prediction: Based on historical big data and real-time inflow data, time series models (such as ARIMA) or machine learning models are used to predict the trend of congestion in various areas within the next 5-15 minutes. For example, the output prediction is: "It is expected that in 3 minutes, due to the arrival of train XX on Line 7, the congestion of platform 4 will rise to 85%".

[0076] 3) Source analysis: By analyzing the temporal and spatial correlation of passenger flow, the source of the current congestion can be traced. For example, the analysis concludes that "the current congestion in the station hall is mainly caused by the convergence of passenger flow entering through Exit A and passenger flow disembarking at Platform 1".

[0077] After obtaining the above dynamic congestion chart, the analysis results can be transformed into specific control measures and service information at the application layer, forming a closed loop for problem-solving. 1) Real-time early warning and intelligent linkage: The system sends early warning information to station staff's handheld terminals and the control center's large screen. It can automatically or semi-automatically link with existing subsystems. Trigger the public address system to play customized crowd control prompts.

[0078] Adjust the environmental control system to increase ventilation in crowded areas.

[0079] Control indicator lights, intelligent flow guidance barriers, and other facilities to dynamically change the direction of passenger flow.

[0080] It provides data support to the train dispatching system, providing a basis for dispatching decisions such as "skipping" trains and adding extra trains.

[0081] 2) Passenger Information Service: Real-time station congestion information is released to the public through official apps, mini-programs and other channels, and passengers are intelligently guided to choose relatively empty entrances or waiting areas to achieve a balanced distribution of passenger flow.

[0082] Example 3 A computer-readable storage medium is provided, comprising: One or more processors; A storage unit is used to store one or more programs, which, when executed by one or more processors, enable the one or more processors to implement the multi-source data fusion method for monitoring dynamic congestion in subway stations as described in Embodiment 1.

[0083] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Based on the technical essence of the present invention, any simple modifications, equivalent substitutions, and improvements made to the above embodiments within the spirit and principles of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A method for monitoring dynamic congestion degree of a subway station by multi-source data fusion, characterized in that, Multiple complementary IoT sensors are deployed in key areas of the subway station to collect data on the number of people and environmental correction parameters. The specific steps include: S1. Obtain the number of people collected by each IoT sensor deployed in each key area, and perform weighted fusion calculation on the number of people collected according to the type of each IoT sensor to obtain the estimated number of people in each key area. S2. Calculate the estimated value of environmental correction parameters for each key area based on the initial number of personnel, and verify the estimated value of environmental correction parameters with the collected environmental correction parameter values. S3. Based on the verification results, dynamically adjust the weights of each IoT sensor during each weighted fusion calculation, and finally output the number of people in each key area and calculate the congestion level of each key area. S4. Integrate the congestion levels of each key area to generate the congestion spatial distribution of the entire subway station. 2.The method according to claim 1, wherein, The process of obtaining the estimated number of personnel in each key area in step S1 is as follows: S11. Assign an initial weight to the number of people collected by each IoT sensor based on its type and installation location. S12. Select the corresponding multi-source data fusion algorithm based on the proportion of various IoT sensors deployed in the current key areas; S13. Using a multi-source data fusion algorithm and initial weights, the number of people collected by each IoT sensor is weighted and fused to obtain the estimated number of people in each key area. 3.The metro station dynamic congestion degree monitoring method of claim 1, wherein, The specific process for verifying the estimated environmental correction parameters with the collected environmental correction parameter values ​​is as follows: S21. Call the number of people-environment correction parameter correlation function, substitute the initial number of people into the correlation function, and output the estimated value of the environment correction parameter; S22. Calculate the dynamic range of environmental correction parameters based on the estimated values ​​of environmental correction parameters; S23. Input the estimated value of the environmental correction parameter, the dynamic range of the environmental correction parameter, and the collected environmental correction parameter values ​​into the environmental correction parameter verification model, and output the verification results.

4. The multi-source data fusion subway station dynamic congestion degree monitoring method according to claim 3, characterized in that, The correlation function between the number of people and the environmental correction parameter is: C_expected = C_base + k × n; Where C_expected is the estimated value of the environmental correction parameter, C_base is the baseline value of the environmental correction parameter collected during the unmanned period, k represents the influence coefficient per unit number of people, and n represents the initial number of people.

5. The multi-source data fusion subway station dynamic congestion degree monitoring method according to claim 3, characterized in that, The environment-corrected parameter validation model env_consistency is: env_consistency = 1 - |C_measured - C_expected| / C_expected_range Where C_measured represents the collected environmental correction parameter value, C_expected represents the estimated value of the environmental correction parameter, and C_expected_range represents the dynamic range of the environmental correction parameter. 6.The multi-source data fusion subway station dynamic congestion degree monitoring method according to claim 1, wherein, The specific process for dynamically adjusting based on the verification results is as follows: When the verification results are consistent, the estimated number of personnel will be used as the final number of personnel in each key area; When the verification results are inconsistent and the verification result is greater than the preset threshold, it is determined that the initial number of personnel is too small. The weight of the IoT sensor with the fewest personnel is reduced, and the weighted fusion calculation is re-performed to estimate the number of personnel. When the verification results are inconsistent and the verification results are less than the preset threshold, it is determined that the initial number of personnel is too large. The weight of the IoT sensor with the most personnel is reduced, and the weighted fusion calculation is performed again to estimate the number of personnel.

7. The multi-source data fusion subway station dynamic congestion degree monitoring method according to claim 1, characterized in that, Environmental correction parameters include CO2 concentration. IoT sensors are also installed in key areas to collect environmental impact parameters that assist in determining environmental correction parameters. These environmental impact parameters include temperature, humidity, and particulate matter concentration. Between steps S1 and S2, there is also a step: Based on the collected environmental impact parameters, determine whether the environmental correction parameters are reliable; If the environmental correction parameters are reliable, proceed to steps S2-S4; If the environmental correction parameters are unreliable, the estimated number of people in each key area is used as the final number of people, the congestion level of each key area is calculated, and step S4 is executed.

8. The method for monitoring dynamic congestion in subway stations using multi-source data fusion according to claim 7, characterized in that, The specific process for determining whether environmental correction parameters are reliable is as follows: Based on the collected CO2 concentration and various environmental impact parameters, determine whether the trend of CO2 concentration change is consistent with the trend of each environmental impact parameter. If the trends are consistent, the environmental correction parameters are considered reliable; if the trends are inconsistent, the environmental correction parameters are considered unreliable. 9.A metro station dynamic congestion degree monitoring system based on multi-source data fusion, characterized in that, include: The perception layer includes multiple complementary IoT sensors deployed in each key area of ​​the subway station. Each IoT sensor is used to collect the number of people in the key area and environmental correction parameters. The edge layer includes edge computing gateways deployed in each key area. These edge computing gateways include: The multi-source data fusion calculation module is used to perform weighted fusion calculation on the number of people collected by each IoT sensor in the key area, according to the type of each IoT sensor, to obtain the estimated number of people in each key area. The environmental correction parameter correction module is used to calculate the estimated value of environmental correction parameters for each key area based on the initial number of personnel, verify the estimated value of environmental correction parameters with the collected environmental correction parameter values, dynamically adjust the weight of each IoT sensor during each weighted fusion calculation based on the verification results, and finally output the number of personnel in each key area and calculate the congestion degree of each key area. The platform layer is used to integrate the congestion levels of various key areas and generate the congestion spatial distribution of the entire subway station.

10. A computer-readable storage medium, characterized in that, include: One or more processors; A storage unit is used to store one or more programs, which, when executed by one or more processors, enable the one or more processors to implement a multi-source data fusion method for monitoring dynamic congestion in subway stations as described in any one of claims 1-8.

Images