Urban rail transit security check system

Abstract

The invention discloses an urban rail transit security check system, and the system comprises a passenger flow dynamics prediction module which constructs a modified Navier-Stokes equation passenger flow model based on the passenger flow pressure distribution and a motion vector field, and outputs a congestion entropy value and a risk particle flow; the adaptive millimeter wave scanning array switches a mobile focusing scanning mode or a wide-area screening mode according to double thresholds of risk particle flow intensity and congestion entropy, and the wide-area screening mode generates a basic threat index and marks a to-be-rechecked target; the feature analysis module generates a real-time threat level through contour feature comparison in a mobile focusing mode, and starts lightweight analysis on a to-be-rechecked target in a wide-domain mode to generate a threat assessment report; the immune decision-making center fuses multi-source data and dynamically generates a decision-making antibody through a clone selection algorithm, and a channel topology instruction contained in the decision-making antibody is executed according to a rule. According to the invention, through dynamic scanning decision and intelligent channel regulation and control, collaborative optimization of security check efficiency and security of a large passenger flow scene is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of urban rail transit safety, and more particularly to an urban rail transit security inspection system. Background Art

[0002] Urban rail transit security inspection systems are specialized facilities that conduct security checks on passengers and their belongings at subway and light rail stations, aiming to prevent prohibited items from entering the transportation system. Current security inspection systems face a structural contradiction in the coordinated optimization of efficiency and safety in high-passenger flow scenarios. Existing security inspection systems generally use fixed scanning modes, which are unable to dynamically adjust detection strategies based on real-time passenger flow. This leads to increased channel congestion caused by high-precision scanning during peak hours, or a significant increase in missed detection rates in rapid screening modes. Furthermore, static channel designs lack the physical ability to respond to sudden risks, and their reliance on manual identification and guidance results in response lags, making it difficult to isolate high-risk targets in a timely manner. The lack of accuracy in passenger flow prediction models for security inspection systems is a core constraint. Traditional methods make it difficult to construct an effective real-time assessment system, primarily due to the difficulty in quantifying and modeling the strongly nonlinear characteristics of crowd flow and the inefficiency of fusion processing of multi-source heterogeneous data (such as pressure distribution and motion vectors). The lack of coordination mechanism among the functional modules of the existing security inspection system leads to a disconnect between decision-making and execution, and the channel resource allocation cannot adapt to passenger flow fluctuations, especially under the impact of instantaneous large passenger flow, which can easily cause system disorder. When trying to introduce automated control, it is difficult to balance traffic efficiency and threat interception accuracy due to the lack of objective indicators for quantitative risk propagation and adaptive channel topology technology. Specifically, the scanning mode switching relies on manual experience judgment and lacks a quantitative trigger mechanism based on risk propagation characteristics; the threat handling process is rigid and cannot implement gradient physical isolation according to risk level differences; the channel structure is static and fixed, and cannot respond to changes in congestion entropy for dynamic topology optimization. These The above defects have jointly led to the rail transit security inspection system being forced to sacrifice security inspection accuracy or traffic efficiency under the pressure of large passenger flow, forming a long-term unresolved technical bottleneck. Summary of the Invention

[0003] An object of the present invention is to solve at least the above problems and to provide at least the advantages which will be described hereinafter.

[0004] Another object of the present invention is to provide an urban rail transit security inspection system to solve the technical problem of balancing security inspection efficiency and safety in large passenger flow scenarios.

[0005] In order to achieve these purposes and other advantages according to the present invention, an urban rail transit security inspection system is provided, which includes: The passenger flow dynamics prediction module builds a passenger flow microenvironment model based on the modified Navier-Stokes equations based on the collected passenger flow pressure distribution and motion vector field, and outputs congestion entropy values ​​and risk particle flows in real time; An adaptive millimeter-wave scanning array receives the congestion entropy value and the risk particle flow. When the risk particle flow intensity is greater than a predetermined intensity threshold, a mobile focus scanning mode is activated. When the risk particle flow intensity is less than or equal to the predetermined intensity threshold and the congestion entropy value is greater than the predetermined entropy threshold, a wide-area screening mode is activated and a basic threat index is generated. When a passenger's basic threat index is identified as greater than a safety threshold, the passenger is marked as a target for re-inspection. Otherwise, the wide-area screening mode is executed. The feature analysis module, in mobile focus scanning mode, receives and analyzes millimeter-wave scanning data, extracts the outline features of the focused passenger's items, compares these features with a pre-stored database of prohibited items, and generates a real-time threat level for the passenger. In wide-area screening mode, upon receiving a target identification, it activates a lightweight analysis model to analyze the millimeter-wave data of the target and generate a threat assessment report for the target. The immune decision center receives the congestion entropy value, real-time threat level and threat assessment report of the target to be re-inspected, and dynamically generates a decision antibody through a clonal selection algorithm. The decision antibody contains a channel topology instruction and a device parameter set; the generation rules of the channel topology instruction include: if the threat information is a high-risk real-time threat level, an instruction containing a 30° acute angle diverter is forcibly generated; if the threat information is a non-high-risk real-time threat level, the corresponding channel topology instruction is generated according to the predetermined entropy value interval of the congestion entropy value; when the target to be re-inspected identifier and the corresponding threat assessment report of the target to be re-inspected are received, a channel topology instruction pointing to the secondary diversion area is generated.

[0006] Preferably, the urban rail transit security inspection system further includes: The topology-variable channel group receives and executes channel topology instructions and equipment parameter sets, and dynamically adjusts the physical structure of the channel; forms an acute-angle diverter structure at the corresponding position according to instructions containing a 30° acute-angle diverter; guides the target to be re-inspected to the secondary diverter area preset therein according to instructions pointing to the secondary diverter area; within the secondary diverter area, further constructs the channel structure according to the channel topology instructions generated by the threat assessment report for the target to be re-inspected; when the threat assessment report indicates high risk, constructs a π-type polymer channel to guide the target to be re-inspected to the re-inspection area; collects real-time data on changes in passenger flow pressure in the channel, and feeds it back to the passenger flow dynamics prediction module for calibrating the passenger flow microenvironment dynamics model.

[0007] Preferably, in the urban rail transit security inspection system, the intensity value of the risk particle flow is dynamically generated by the following steps: S1.1. Real-time access to the risk path spatial distribution data output by the passenger flow dynamics prediction module, synchronous acquisition of mobile device signaling density distribution data provided by mobile communication base stations, and the vector directions of gesture skeleton key points captured by the security inspector's wearable device; S1.2. Based on the mobile device signaling density distribution data, identify the boundary coordinates of the abnormal cluster area; based on the risk path spatial distribution data, extract the path vector of the main diffusion direction; S1.3. Calculate the overlap ratio between the boundary coordinates of the abnormal clustering area and the spatial distribution area of ​​the risk path as the first correction coefficient. Calculate the cosine value of the angle between the direction of the gesture skeleton key point vector and the path vector of the main diffusion direction as the second correction coefficient. S1.4. Use the historical frequency of prohibited items stored in the security inspection event database as a base weight, combined with the first and second correction coefficients, and input them into a pre-trained risk propagation tree model. The risk propagation tree model is pre-trained and generated based on the spatiotemporal distribution patterns of the historical security inspection event database. S1.5. The risk propagation tree model outputs the updated risk particle flow intensity value and feeds it back to the passenger flow dynamics prediction module.

[0008] Preferably, the urban rail transit security inspection system further includes: The dynamic path calibration engine has its input connected to the initial path output of the risk particle flow of the passenger flow dynamics prediction module and is connected in parallel to the multi-spectral interference sensing array deployed in the channel vault. The multi-spectral interference sensing array simultaneously collects the physiological agitation index in the thermal infrared band and the motion mutation trajectory in the visible light band. When the physiological agitation index exceeds the preset agitation threshold and lasts for more than 5 seconds, or the cumulative azimuth deviation between the motion mutation trajectory and the initial path is greater than or equal to 30 degrees, the nonlinear compensator is activated; When activated, the nonlinear compensator performs the following operations: S3.1. Map the physiological agitation index to a path deviation probability density function, and generate a first compensation vector using a convolution kernel; S3.2. Decompose the tangential and normal acceleration components of the motion mutation trajectory, and generate a second compensation vector using the amplitude of the normal component as a weight; S3.3. Fuse the first compensation vector and the second compensation vector to the initial path of the risk particle flow, and output the anti-interference calibration path to the risk path input end of the adaptive millimeter wave scanning array.

[0009] Preferably, the urban rail transit security inspection system continuously performs antibody activity monitoring after the immune decision center generates decision antibodies: S4.1. Real-time acquisition of the threat level update flag of the feature analysis module and the instruction execution progress value of the topology variable channel group; S4.2. When the progress value of the channel group executing the high-risk acute angle flow divider instruction is monitored to be less than 50%, if the threat level update flag is switched to medium or low risk, the antibody apoptosis protocol is activated; S4.3. The antibody apoptosis protocol uses the instruction execution progress value as the decay base and combines it with the updated risk particle flow intensity value to generate the antibody half-life; S4.4. When the instruction execution progress value exceeds the progress threshold corresponding to the antibody half-life, the clone selection algorithm is triggered to reorganize the decision antibody. The reorganization rule uses the non-acute angle diversion instruction generated by the entropy value interval to overwrite the original high-risk instruction.

[0010] Preferably, the urban rail transit security inspection system further comprises a real-time video compensation unit, which continuously collects passenger flow video streams through a wide-angle camera array deployed on the side wall of the channel; The real-time video compensation unit performs the following operations: S5.1. Extract the passenger velocity distribution matrix and direction change rate in the video stream using the optical flow method; S5.2, when the standard deviation of the moving speed exceeds the preset speed tolerance and the direction change rate is greater than the direction threshold, an emergency topology pre-instruction is generated; S5.3. Inputting the emergency topology pre-commands in parallel to the immune decision-making center and the topology variable channel group; Among them, when the output of the passenger flow dynamics prediction module is delayed, the immune decision center uses the emergency topology pre-instruction as an antigen input substitute to drive the clone selection algorithm to generate decision antibodies containing channel topology instructions; after receiving the emergency topology pre-instruction, the topology variable channel group generates channel topology instructions according to the entropy value interval.

[0011] Preferably, in the urban rail transit security inspection system, the real-time video compensation unit further performs the following operations: S5.4. Continuously monitor the acquisition status of mobile communication base station signaling density distribution data and gesture skeleton key point vectors; S5.5. When the duration of missing signaling density distribution data exceeds the first tolerance or the invalid count of gesture skeleton key point vectors exceeds the second tolerance, activate the video compensation enhancement protocol and perform the following operations: ) Using the video stream captured by the wide-angle camera array, we use spatiotemporal slicing analysis to generate a passenger flow concentration distribution map to replace the missing signaling density distribution data; ) Analyze the passenger's upper limb motion trajectory through a posture estimation algorithm and generate a simulated gesture skeleton key point vector; S5.7. Input the passenger flow concentration distribution map and the simulated gesture skeleton key point vector into the risk particle flow intensity generation step.

[0012] Preferably, the urban rail transit security inspection system further comprises a multimodal data redundancy engine, the input end of which is connected to the real-time signaling data interface of the existing monitoring camera group and the mobile communication base station of the station; The multimodal data redundancy engine performs the following operations: S6.1. Continuously monitor the status codes of the ground pressure sensor network and the 3D LiDAR. When the status code remains abnormal for a period greater than the fault tolerance, activate the data reconstruction protocol. S6.2. After the data reconstruction protocol is activated, perform the following operations: a) Analyze the video streams from the surveillance camera group and generate a passenger flow density distribution heat map using the background subtraction method; b) Extract the signaling hopping trajectory of the mobile communication base station and construct the passenger movement direction probability matrix; S6.3. Map the passenger flow density distribution heat map to simulated pressure distribution data, convert the passenger movement direction probability matrix into a simulated motion vector field, and input it into the passenger flow dynamics prediction module to replace the original data source; S6.4. When the device status code returns to normal, it automatically switches back to the original data collection mode.

[0013] Preferably, the urban rail transit security inspection system, when synchronously acquiring the mobile communication base station signaling density distribution data and the gesture skeleton key point vector direction in step S1.1, activates the multi-source biometric feature fusion protocol when the effective coverage rate of the signaling density distribution data is less than a preset coverage threshold or the confidence level of the gesture skeleton key point vector is less than a confidence threshold; The multi-source biometric fusion protocol performs the following operations: ) An infrared thermal imaging array deployed at the top of the channel collects the phase difference of the passenger gait cycle to generate a gait aggregation density distribution map, which is normalized to replace the signaling density distribution data; ) The micro-Doppler effect of the millimeter wave scanning array is used to analyze the passenger's trunk swing direction and generate a biomechanical direction vector, which replaces the gesture skeleton key point vector when the swing amplitude is greater than the sensitivity threshold; The gait cluster density distribution map is input into the abnormal cluster area identification process of step S1.2, and the biomechanical direction vector is input into the second correction coefficient calculation process of step S1.3.

[0014] The present invention has at least the following beneficial effects: 1. The urban rail transit security inspection system provided by the present invention constructs a passenger flow microenvironment model based on the modified Navier-Stokes equations through a passenger flow dynamics prediction module, and outputs quantitative congestion entropy values ​​and risk particle flows in real time, solving the problem of traditional passenger flow prediction models being difficult to accurately model due to their nonlinear characteristics. The adaptive millimeter-wave scanning array dynamically switches between mobile focus scanning and wide-area screening modes based on dual thresholds, enabling precise allocation of scanning resources according to risk levels, while ensuring basic traffic efficiency for large passenger flows while significantly reducing the risk of missed detections. The immune decision-making center uses a clonal selection algorithm to generate decision antibodies. For 30° acute angle flow instructions forcibly triggered by high-risk real-time threat levels, the physical topology structure achieves second-level isolation and blocking of high-risk targets, overcoming the control loopholes caused by manual response delays. Combined with the dynamic path calibration engine's real-time anti-interference calibration of the risk particle flow path, and the antibody activity monitoring mechanism's dynamic correction of the instruction execution process, a "prediction-decision-execution-feedback" closed-loop system is formed, ultimately breaking through the technical bottleneck of the long-standing difficulty in coordinating and optimizing security inspection efficiency and safety in large passenger flow scenarios. 2. The present invention uses the passenger flow dynamics prediction module to output congestion entropy values ​​and risk particle flows in real time based on the modified Navier-Stokes equations, providing a quantitative basis for scanning decisions; the adaptive millimeter wave scanning array dynamically switches between mobile focus scanning mode and wide-area screening mode based on dual thresholds of risk particle flow intensity and congestion entropy values, achieving precise allocation of scanning resources according to risk levels; the feature analysis module outputs real-time threat levels and threat assessment reports in a hierarchical manner; the immune decision-making center generates differentiated channel topology instructions based on the clonal selection algorithm, solving the problem of imbalance between efficiency and safety under large passenger flows caused by fixed scanning strategies and static channels in traditional systems.

[0015] 3. The topology-variable channel group of the present invention executes the 30° acute-angle diverter command triggered by high-risk targets, forming a physical isolation structure to block the high-risk target's movement path; for targets to be re-inspected, a π-shaped aggregate channel is constructed within the secondary diversion area to guide them to the re-inspection area, achieving gradient management and control of threatening targets; and through real-time feedback of crowd pressure data to calibrate the prediction model, a closed-loop optimization mechanism is formed to enhance the system's adaptability.

[0016] 4. The present invention uses mobile signaling data to identify the boundaries of abnormal aggregation areas, combines the main diffusion direction of the risk path with the skeleton vector of the security inspector's gesture to calculate the dual correction coefficient, and uses the risk propagation tree model to dynamically generate the risk particle flow intensity value, thereby solving the problem of insufficient accuracy of traditional risk prediction relying on a single data source and improving the accuracy of high-risk area identification.

[0017] 5. The dynamic path calibration engine of the present invention activates the nonlinear compensator based on the thermal infrared physiological agitation index and the visible light motion mutation trajectory, and compensates the initial path of the risk particle flow by generating a vector, thereby overcoming the path deviation defect caused by environmental interference and ensuring the positioning accuracy of the adaptive millimeter wave scanning array.

[0018] 6. When the instruction execution progress does not reach 50%, the antibody activity monitoring mechanism of the present invention dynamically triggers the antibody apoptosis protocol according to the threat level degradation, generates the antibody half-life with the progress value as the decay base, and reorganizes the non-acute angle diversion instruction to cover the original high-risk instruction, avoiding the waste of channel resources caused by excessive control.

[0019] 7. The real-time video compensation unit of the present invention uses the optical flow method to extract the passenger flow movement speed and direction mutation characteristics to generate emergency topology pre-instructions, and drives the immune decision center and channel group emergency response when the passenger flow dynamics prediction module is delayed to prevent the system from making decisions due to data delays.

[0020] 8. When signaling or gesture data fails, the video compensation enhancement protocol of the present invention generates a passenger flow concentration distribution map through spatiotemporal slicing analysis, combines posture estimation to simulate gesture skeleton vectors, ensures the data integrity of the risk particle flow intensity calculation process, and enhances the system fault tolerance.

[0021] 9. When the pressure sensing network or lidar fails, the multimodal data redundancy engine of the present invention uses monitoring video to generate passenger flow density heat maps and signaling jump trajectories to construct simulated motion vector fields, maintain the data input stability of the passenger flow dynamics prediction module, and ensure the continuous operation of the system.

[0022] 10. The multi-source biometric fusion protocol of the present invention generates a normalized gait cluster density map through infrared thermal imaging gait phase difference to replace signaling data, and extracts the trunk swing direction vector (when the swing amplitude meets the standard) based on the millimeter wave micro-Doppler effect to replace gesture data, thereby solving the data reliability problem when the biometric collection environment is restricted.

[0023] Other advantages, objectives and features of the present invention will be reflected in part through the following description, and in part will be understood by those skilled in the art through study and practice of the present invention. DETAILED DESCRIPTION

[0024] The present invention is further described in detail below with reference to the embodiments so that those skilled in the art can implement the invention with reference to the description.

[0025] It should be understood that terms such as “having”, “including” and “comprising” used herein do not preclude the existence or addition of one or more other elements or combinations thereof.

[0026] It should be noted that the experimental methods described in the following embodiments are conventional methods unless otherwise specified, and the reagents and materials can be obtained from commercial channels unless otherwise specified.

[0027] The present invention provides an urban rail transit security inspection system, which includes: The passenger flow dynamics prediction module constructs a passenger flow microenvironment model based on the modified Navier-Stokes equations based on the collected passenger flow pressure distribution and motion vector field, and outputs congestion entropy and risk particle flow in real time. The congestion entropy is a scalar indicator that quantifies the degree of regional congestion, and the risk particle flow is a vector field that characterizes the direction and intensity of risk propagation. The mobile focus scanning mode refers to the high-precision scanning of specific passengers by the millimeter wave array, and the wide-area screening mode refers to the rapid scanning of the crowd to generate a basic threat index. An adaptive millimeter-wave scanning array receives the congestion entropy value and the risk particle flow. When the risk particle flow intensity is greater than a predetermined intensity threshold, a mobile focus scanning mode is activated. When the risk particle flow intensity is less than or equal to the predetermined intensity threshold and the congestion entropy value is greater than the predetermined entropy threshold, a wide-area screening mode is activated and a basic threat index is generated. When a passenger's basic threat index is identified as greater than a safety threshold, the passenger is marked as a target for re-inspection. Otherwise, the wide-area screening mode is executed. The feature analysis module, in mobile focus scanning mode, receives and analyzes millimeter-wave scanning data, extracts the outline features of the focused passenger's items, compares these features with a pre-stored database of prohibited items, and generates a real-time threat level for the passenger. In wide-area screening mode, upon receiving a target identification, it activates a lightweight analysis model to analyze the millimeter-wave data of the target and generate a threat assessment report for the target. The immune decision center receives the congestion entropy value, real-time threat level and threat assessment report of the target to be re-inspected, and dynamically generates a decision antibody through a clonal selection algorithm. The decision antibody contains a channel topology instruction and a device parameter set; the generation rules of the channel topology instruction include: if the threat information is a high-risk real-time threat level, an instruction containing a 30° acute angle diverter is forcibly generated; if the threat information is a non-high-risk real-time threat level, the corresponding channel topology instruction is generated according to the predetermined entropy value interval of the congestion entropy value; when the target to be re-inspected identifier and the corresponding threat assessment report of the target to be re-inspected are received, a channel topology instruction pointing to the secondary diversion area is generated.

[0028] In the above technical solution, the passenger flow dynamics prediction module of the urban rail transit security inspection system provided by the present invention continuously collects the pressure distribution and motion vector data of the passenger flow in the channel through ground pressure sensors and three-dimensional laser radars, simulates the passenger flow fluid dynamics behavior through the modified Navier-Stokes equations, and outputs in real time the congestion entropy value that quantifies the degree of regional congestion and the risk particle flow that marks the risk propagation path.

[0029] The adaptive millimeter-wave scanning array makes dynamic decisions based on the received congestion entropy value and risk particle flow parameters: when the risk particle flow intensity exceeds the predetermined intensity threshold, it immediately locks onto the high-risk propagation direction and starts the mobile focus scanning mode, driving the rotating pan-tilt head to lock onto the high-risk target for millimeter-wave precision scanning; if the risk particle flow intensity is equal to or lower than the predetermined intensity threshold (not exceeded), and the congestion entropy value exceeds the entropy threshold, it switches to wide-area screening mode, quickly scans the crowd through the wide-beam transmitter to generate a basic threat index, and marks passengers with an index exceeding the standard as targets for re-inspection; when both the congestion entropy value and the risk particle flow parameters are within the standard, basic wide-area screening is maintained to ensure traffic efficiency.

[0030] The feature analysis module analyzes the scanning data (millimeter wave scanning data) in the mobile focus mode and generates a real-time threat level by comparing the contour features with the pre-stored contraband library; in the wide-area mode, a lightweight model is activated for the target to be re-inspected to output a threat assessment report.

[0031] The immune decision-making center integrates congestion entropy values, real-time threat levels, and threat assessment reports, and uses a clonal selection algorithm to generate dynamic decision antibodies containing channel topology instructions and device parameter sets. Channel topology instructions are generated according to rules: high-risk real-time threat levels trigger mandatory isolation instructions with a 30° acute angle diverter; non-high-risk threats adapt the channel layout based on the congestion entropy range; and targets awaiting re-inspection uniformly trigger guidance instructions to secondary diversion areas.

[0032] Traditional security inspection systems, due to their fixed scanning patterns and static channel designs, face a conflict between efficiency and safety under high passenger flow conditions. High-precision scanning leads to congestion, rapid screening increases the risk of missed detections, and reliance on manual response delays. This solution uses passenger flow modeling to predict risk areas and dynamically allocate scanning resources. A clonal selection algorithm generates topological instructions to respond to threats in seconds, and a 30° sharp angle splitter physically blocks high-risk targets. Entropy intervals adapt channel layouts to continuously optimize traffic efficiency. Pressure data fed back by channel groups is combined with real-time calibration of the prediction model to form a closed-loop optimization loop, ultimately achieving a synergistic improvement in security inspection efficiency and safety under high passenger flow conditions.

[0033] The above technical solution achieves dynamic adaptation of scanning strategies to passenger flow conditions: a dual-threshold trigger mechanism precisely switches scanning modes, quickly locking onto high-risk targets when risk particle flow exceeds the standard, initiating wide-area screening to avoid channel congestion when congestion entropy values ​​rise, and maintaining basic traffic efficiency when both parameters are normal. A hierarchical analysis mechanism ensures a balance between in-depth detection of high-threat targets and rapid classification of targets to be inspected. The differentiated channel instructions generated by the clone selection algorithm achieve three-level control: physical isolation of high-risk targets, entropy adaptation and diversion of non-high-risk targets, and directional guidance of targets to be inspected. This comprehensively resolves the core contradiction between missed detection rate and traffic efficiency under fixed scanning strategies.

[0034] In another technical solution, the urban rail transit security inspection system further includes: The topology-variable channel group receives and executes channel topology instructions and equipment parameter sets, and dynamically adjusts the physical structure of the channel; forms an acute-angle diverter structure at the corresponding position according to instructions containing a 30° acute-angle diverter; guides the target to be re-inspected to the secondary diverter area preset therein according to instructions pointing to the secondary diverter area; within the secondary diverter area, further constructs the channel structure according to the channel topology instructions generated by the threat assessment report for the target to be re-inspected; when the threat assessment report indicates high risk, constructs a π-type polymer channel to guide the target to be re-inspected to the re-inspection area; collects real-time data on changes in passenger flow pressure in the channel, and feeds it back to the passenger flow dynamics prediction module for calibrating the passenger flow microenvironment dynamics model.

[0035] Traditional urban rail transit security inspection systems utilize fixed physical channel designs, making it impossible to dynamically adjust their layout based on risk levels. When high-risk targets are detected, security inspectors are required to manually guide them to the re-inspection area. Delayed manual responses lead to control blind spots. Targets awaiting inspection share channels with regular passengers, causing cross-congestion and efficiency losses. More critically, channel resource allocation lacks real-time passenger flow data, making it impossible to dynamically optimize in response to changes in congestion entropy, creating a disconnect between decision-making and execution. For example, static channels are forced to have longer queues during peak hours, while channel resources remain idle during off-peak periods, making it difficult to synergistically improve resource utilization and traffic efficiency.

[0036] This solution achieves dynamic response at the physical layer through a topologically variable channel group. When a high-risk instruction is received, the hydraulic drive mechanism pushes the baffle to form a 30° acute-angle diversion structure, creating a physical isolation zone in front of the target passenger's path, forcibly preventing them from contacting other passengers. After closing, the baffle automatically locks until re-inspection is completed. Targets to be inspected are guided to the internal secondary diversion area, where they are graded according to the threat assessment report. High-risk targets trigger the construction of a π-shaped closed channel with three movable barriers to achieve point-to-point closed transfer. Non-high-risk targets automatically adjust the channel width according to the real-time congestion entropy value. For example, high entropy values ​​shrink the channel to accelerate diversion, while low entropy values ​​expand the width to reduce density. The pressure sensor embedded in the channel collects real-time changes in passenger pressure, maps it into a pressure gradient matrix, and feeds it back to the prediction module. The viscosity coefficient parameters in the passenger flow model are continuously calibrated, forming a closed-loop enhancement of execution effect and model optimization.

[0037] Compared to existing technologies, this solution achieves physical isolation of high-risk targets within seconds through a dynamic channel structure, increasing the response speed to several times that of manual guidance and completely eliminating blind spots in management and control. The π-shaped closed channel prevents the intersection of targets to be inspected and normal passenger flow, reducing the incidence of secondary risks during re-inspection. The channel width is dynamically adjusted according to the entropy value to accurately match the traffic capacity with the passenger flow load, and the turnover efficiency during peak hours is improved. The closed-loop optimization mechanism driven by pressure feedback continuously improves the accuracy of congestion entropy calculation over time. Data from a subway pilot showed that the prediction error rate decreased by 40% after 30 days. These substantial improvements unify risk control and resource scheduling at the physical execution layer for the first time, providing reliable technical support for the security inspection system.

[0038] This technical solution establishes closed-loop control at the physical execution layer. A 30° acute-angle diverter utilizes a rigid baffle structure to isolate high-risk targets, preventing them from contacting other passengers. A π-shaped aggregate channel forms a closed transfer channel for high-risk targets awaiting inspection, preventing secondary risks during re-inspection. Channel widths within the secondary diversion zone are dynamically adjusted based on entropy, optimizing traffic efficiency in real time. Pressure feedback data continuously calibrates the predictive model parameters, continuously improving the calculation accuracy of congestion entropy and risk particle flow as the system operates, forming a self-reinforcing mechanism of "command execution-environmental feedback-model optimization."

[0039] The specific operating steps of the earthly rail transit security inspection system provided by the present invention are as follows: Step 1: Passenger flow dynamic prediction The pressure distribution and motion vector data of pedestrian flow in the channel are collected in real time. The passenger flow model is constructed by modifying the Navier-Stokes equation, and two key parameters are calculated and output: the congestion entropy value - which quantifies the degree of congestion in the area; and the risk particle flow - which marks the direction and intensity of risk propagation.

[0040] Step 2: Dynamic Scanning Decision Switch the scanning mode according to the parameters output in step 1: If the risk particle flow intensity is greater than the predetermined intensity threshold: the mobile focus scanning mode is activated to perform a millimeter wave deep scan on the high-risk pointing passengers; If the risk particle flow intensity is less than or equal to the intensity threshold and the congestion entropy value is greater than the entropy threshold: the wide-area screening mode is activated, and the crowd is quickly scanned to generate a basic threat index. Passengers with an index greater than the safety threshold are marked as targets for re-inspection. If both parameters are ≤ threshold: maintain basic broad-band screening mode.

[0041] Step 3: Threat Level Analysis In mobile focus mode: the millimeter wave data of the targeted passenger is analyzed, the outline features of the item are extracted and compared with the prohibited items library, and a real-time threat level (high risk / medium risk / low risk) is generated; In wide-area screening mode: a lightweight model is activated for the target to be re-inspected, and a threat assessment report is quickly output.

[0042] Step 4: Intelligent channel decision Use the clonal selection algorithm to generate decision instructions: If the real-time threat level = high risk: force the generation of channel instructions containing a 30° sharp angle flow divider; If the real-time threat level is ≠ high risk: generate an adaptive channel instruction according to the interval to which the congestion entropy value belongs; If there is a target to be re-inspected: generate guidance instructions pointing to the secondary diversion area.

[0043] Step 5: Physical channel execution High-risk command response: Form a 30° acute angle flow divider structure at the designated location to physically block the target path; Response to the recheck command: Guide the target to the secondary diversion area and build a channel within the area based on the threat assessment report: The report indicates high risk: generate a π-type polymerization channel to close and guide to the re-inspection area; In other cases: adjust the channel layout (such as widening / shrinking the path) according to the entropy value instruction.

[0044] Step 6: Closed-loop model calibration The data on passenger flow pressure changes after channel structure adjustment is collected in real time and fed back to the passenger flow dynamics prediction module to dynamically calibrate the fluid model parameters.

[0045] In another technical solution, in the urban rail transit security inspection system, the intensity value of the risk particle flow is dynamically generated through the following steps: S1.1. Real-time access to the risk path spatial distribution data output by the passenger flow dynamics prediction module, synchronous acquisition of mobile device signaling density distribution data provided by mobile communication base stations, and the vector directions of gesture skeleton key points captured by the security inspector's wearable device; S1.2. Based on the mobile device signaling density distribution data, identify the boundary coordinates of the abnormal cluster area; based on the risk path spatial distribution data, extract the path vector of the main diffusion direction; S1.3. Calculate the overlap ratio between the boundary coordinates of the abnormal clustering area and the spatial distribution area of ​​the risk path as the first correction coefficient. Calculate the cosine value of the angle between the direction of the gesture skeleton key point vector and the path vector of the main diffusion direction as the second correction coefficient. S1.4. Use the historical frequency of prohibited items stored in the security inspection event database as a base weight, combined with the first and second correction coefficients, and input them into a pre-trained risk propagation tree model. The risk propagation tree model is pre-trained and generated based on the spatiotemporal distribution patterns of the historical security inspection event database. S1.5. The risk propagation tree model outputs the updated risk particle flow intensity value and feeds it back to the passenger flow dynamics prediction module.

[0046] Traditional risk prediction models typically use video analysis or Wi-Fi location tracking to determine crowd density distribution, triggering alerts based on preset density thresholds. These methods fail to distinguish between normal queues and abnormal crowds. For example, Beijing West Railway Station mistakenly triggered 12 high-risk alerts due to excessive queue density during the morning rush hour, resulting in frequent channel closures. More critically, these models fail to consider the synergy between risk transmission and human intervention. Predicted paths often deviate from actual response requirements, forcing security inspectors to manually correct system instructions and resulting in response delays of up to minutes.

[0047] This solution reconstructs the risk particle flow intensity calculation process through multi-source data fusion: It accesses the risk path spatial distribution data output by the passenger flow prediction module in real time, simultaneously obtains the signaling density distribution of mobile base stations and the gesture skeleton vectors captured by the security inspector's wearable device, identifies the boundary coordinates of the abnormal cluster area based on the signaling density, and extracts the main diffusion direction vector in combination with the risk path data. The first correction coefficient is calculated as the proportion of the overlapping area between the abnormal area boundary and the risk path space to reflect the spatial matching degree; the cosine value of the directional angle between the gesture skeleton vector and the main path vector is calculated as the second correction coefficient to quantify the consistency between manual intervention and system prediction. The historical frequency of prohibited items is used as the basic weight, and the two correction coefficients are fused and input into the pre-trained risk propagation tree model, which is trained and generated based on the spatiotemporal patterns of 100,000 historical security inspection events. The final output is the dynamically calibrated risk particle flow intensity value and fed back to the prediction module.

[0048] Compared to existing technologies, this solution overcomes the limitations of a single data dimension. It utilizes spatial overlap coefficients to address the problem of misidentifying ordinary queues as high-risk areas. Direction cosine coefficients enable instant alignment of human intervention intentions with system predictions. The risk propagation tree model integrates static historical patterns with dynamic environmental corrections, ensuring that risk intensity values ​​consistently adapt to the real-time scenario. The intervention dimension introduced by the security inspector's gesture vectors enables human-machine co-evolution, reducing the false alarm rate to one-fifth of traditional methods. Ultimately, this approach ensures accurate identification of high-risk areas while avoiding the waste of channel resources caused by false alarms.

[0049] This technical solution improves the reliability of multi-source fusion risk prediction. Mobile signaling data identifies macro-anomaly clusters, risk path data provides the primary diffusion direction, and security inspector gesture vectors introduce a manual intervention dimension. The triple data sources are quantitatively corrected using the overlapping area ratio (spatial consistency) and direction cosines (behavioral consistency). Historical contraband frequency weights establish a baseline for risk propagation probability, and the pre-trained model integrates dynamic correction coefficients to output intensity values. This addresses the problem of misjudging high-risk areas caused by missing data dimensions in traditional single-sensor predictions, significantly reducing the risk location error rate.

[0050] In another technical solution, the urban rail transit security inspection system further includes: The dynamic path calibration engine has its input connected to the initial path output of the risk particle flow of the passenger flow dynamics prediction module and is connected in parallel to the multi-spectral interference sensing array deployed in the channel vault. The multi-spectral interference sensing array simultaneously collects the physiological agitation index in the thermal infrared band and the motion mutation trajectory in the visible light band. When the physiological agitation index exceeds the preset agitation threshold and lasts for more than 5 seconds, or the cumulative azimuth deviation between the motion mutation trajectory and the initial path is greater than or equal to 30 degrees, the nonlinear compensator is activated; When activated, the nonlinear compensator performs the following operations: S3.1. Map the physiological agitation index to a path deviation probability density function, and generate a first compensation vector using a convolution kernel; S3.2. Decompose the tangential and normal acceleration components of the motion mutation trajectory, and generate a second compensation vector using the amplitude of the normal component as a weight; S3.3. Fuse the first compensation vector and the second compensation vector to the initial path of the risk particle flow, and output the anti-interference calibration path to the risk path input end of the adaptive millimeter wave scanning array.

[0051] Existing risk path calibration systems typically rely on a single sensor (such as a lidar) to detect position deviations and use a preset angle threshold to trigger mechanical corrections. For example, an airport security system, failing to detect the localized congestion caused by a passenger's sudden fainting, mistakenly identified a physiological stagnation as the end of the path, causing subsequent scans to focus on the wrong area for three minutes. More critically, traditional linear compensators are unable to handle the acceleration vector decomposition associated with sudden motion changes. Calibration simply translates the path line, ignoring the differential effects of tangential and normal acceleration on the offset direction.

[0052] This solution utilizes multispectral interference sensing and nonlinear compensation to reconstruct and calibrate the process. A thermal infrared camera deployed in the tunnel vault continuously collects the rate of human skin temperature rise and outputs a physiological agitation index. A visible light camera simultaneously captures sudden motion trajectories, such as sudden stops and reverse movements. The dynamic path calibration engine activates the nonlinear compensator in two scenarios: when the physiological agitation index exceeds a threshold and persists for 5 seconds, it indicates disordered group behavior caused by emotional contagion; when the azimuth deviation of the sudden motion trajectory from the initial path reaches 30 degrees, it indicates that the behavior pattern has deviated from the predicted framework. The compensator maps the physiological agitation index into a probability density function and generates a first compensation vector through spatial convolution to quantify the spatial displacement effect of emotional fluctuations on the path. It then decomposes the tangential and normal acceleration components of the sudden motion trajectory and generates a second compensation vector, weighted by the amplitude of the normal component, to capture the centrifugal disturbance intensity caused by sudden turns. These two vectors are then combined to form the initial path of the risky particle flow, outputting the anti-disturbance calibration path to the millimeter-wave scanning array.

[0053] Compared to traditional methods, this solution overcomes the limitations of rigid and single-dimensional compensation. Multispectral synergy using thermal infrared and visible light links physiological agitation to path deviation for the first time, preventing the misinterpretation of emotional gatherings as risk transmission. A nonlinear compensator precisely quantifies the impact of behavioral mutations through acceleration vector decomposition, aligning the calibration path with the actual risk diffusion situation. The millimeter-wave scanning array significantly reduces the false lock rate caused by interference, achieving high-risk target focus accuracy comparable to that of human experts, fundamentally curbing systemic resource waste.

[0054] This technical solution enhances the interference resistance of risky paths. The thermal infrared physiological agitation index (temperature rise >0.5°C / s) reflects sudden emotional changes, while visible light motion trajectory changes (sudden stops / reversing) capture behavioral anomalies. These two serve as indicators of environmental interference to trigger the compensation mechanism. A path deviation probability function converts physiological parameters into spatial offsets, while normal acceleration weighting reflects the intensity of behavioral changes. Vector fusion compensates for initial path deviations. Ultimately, the millimeter-wave scanning array maintains focus positioning accuracy above 90% even in high-interference environments, preventing false or missed scans caused by ambient noise.

[0055] In another technical solution, the urban rail transit security inspection system continuously monitors antibody activity after the immune decision center generates decision antibodies: S4.1. Real-time acquisition of the threat level update flag of the feature analysis module and the instruction execution progress value of the topology variable channel group; S4.2. When the progress value of the channel group executing the high-risk acute angle flow divider instruction is monitored to be less than 50%, if the threat level update flag is switched to medium or low risk, the antibody apoptosis protocol is activated; S4.3. The antibody apoptosis protocol uses the instruction execution progress value as the decay base and combines it with the updated risk particle flow intensity value to generate the antibody half-life; S4.4. When the instruction execution progress value exceeds the progress threshold corresponding to the antibody half-life, the clone selection algorithm is triggered to reorganize the decision antibody. The reorganization rule uses the non-acute angle diversion instruction generated by the entropy value interval to overwrite the original high-risk instruction.

[0056] The existing security inspection system uses a one-way command triggering model. Once a high-risk channel command is generated, it is executed to the end. A misjudgment occurred at a subway station: metal cutlery in a passenger's bag triggered a high-risk command. The sharp-angle diverter activated a blockade, but re-inspection confirmed it was not a threat. However, because the system was unable to automatically release the command, the channel blockade persisted until manual intervention, wasting up to 15 minutes of traffic resources during peak hours. This rigid execution model caused cascading congestion in high-traffic scenarios and failed to respond to real-time risk mitigation.

[0057] This solution reconstructs the instruction lifecycle management through antibody activity monitoring, and obtains the threat level update status of the feature analysis module and the channel group instruction execution progress value in real time. When the execution progress of the high-risk sharp-angle diversion flow instruction does not reach 50% and the threat level is downgraded to medium or low risk, the antibody apoptosis protocol is activated. This protocol uses the current execution progress value as the decay calculation base, and combines the updated risk particle flow intensity to generate the antibody half-life. When the execution progress exceeds the threshold corresponding to the half-life, the clone selection algorithm is forced to trigger the reorganization of the decision antibody, and a non-sharp-angle diversion instruction is generated to cover the original high-risk instruction. For example, when the initial execution progress is 40%, the threat is downgraded. The system calculates the half-life based on the progress value and the risk intensity. If the progress increases to 60% within the next three seconds, the instruction reorganization is triggered, the sharp-angle baffle stops closing and switches to a serpentine diversion channel.

[0058] Compared to traditional approaches, this solution transcends the limitations of rigid command execution. The real-time coupling of progress values ​​and threat levels establishes a dynamic command optimization mechanism for the first time, precisely matching channel resource occupancy duration to the actual risk duration. A half-life model adaptively adjusts reorganization timing based on risk intensity, avoiding residual risk caused by prematurely lifting controls during early stages of downgrade. This increases system resource utilization to a level that aligns with the risk landscape in real time, approaches channel turnover efficiency nearing its theoretical optimal value, and reduces the need for manual intervention to a fraction of what is required under traditional models.

[0059] This technical solution establishes a dynamic instruction optimization mechanism, using instruction execution progress as the decay factor (progress value P < 50%) and combining it with the real-time risk particle flow intensity (R) to calculate the half-life (T ∝ P / R). When progress exceeds the threshold corresponding to T, instructions are reorganized. This ensures that if the threat level degrades before physical isolation is completed (the baffle is not fully closed), non-sharp angle diversion instructions can be promptly switched to avoid inefficient use of channel resources. Testing has shown a significant reduction in the over-execution rate of high-risk instructions and a substantial improvement in channel turnover efficiency.

[0060] In another technical solution, the urban rail transit security inspection system further includes a real-time video compensation unit, which continuously collects passenger flow video streams through a wide-angle camera array deployed on the side wall of the channel; The real-time video compensation unit performs the following operations: S5.1. Extract the passenger velocity distribution matrix and direction change rate in the video stream using the optical flow method; S5.2, when the standard deviation of the moving speed exceeds the preset speed tolerance and the direction change rate is greater than the direction threshold, an emergency topology pre-instruction is generated; S5.3. Inputting the emergency topology pre-commands in parallel to the immune decision-making center and the topology variable channel group; Among them, when the output of the passenger flow dynamics prediction module is delayed, the immune decision center uses the emergency topology pre-instruction as an antigen input substitute to drive the clone selection algorithm to generate decision antibodies containing channel topology instructions; after receiving the emergency topology pre-instruction, the topology variable channel group generates channel topology instructions according to the entropy value interval.

[0061] Existing systems typically employ static contingency plans when prediction data is delayed, such as forcing a switch to low-precision scanning mode or full manual takeover. During the New Year's Day passenger flow peak at a certain city hub, a prediction calculation timeout forced the system to downgrade to full-channel, wide-area screening. As a result, three passengers carrying suspicious items were not promptly identified and were only intercepted after manual review. These contingency plans lack the ability to perceive instantaneous passenger flow dynamics and are unable to distinguish between normal congestion and sudden, risky gatherings.

[0062] This solution builds an emergency response link through a real-time video compensation unit. The wide-angle camera array deployed on the side walls of the channel continuously collects passenger flow video streams, and the optical flow processing algorithm extracts the passenger movement speed distribution matrix and directional change rate. When the standard deviation of the movement speed exceeds the preset tolerance and the directional change rate exceeds the directional threshold, it is determined to be a sudden passenger flow disturbance and an emergency topology pre-instruction is generated. This instruction contains the channel contraction ratio and diversion angle parameters, and is input in parallel to the emergency interface of the immune decision center and the execution interface of the topology variable channel group. When the decision center predicts the output delay of the module, it uses the pre-instruction as an alternative to the antigen input to drive the clonal selection algorithm to generate decision antibodies; the channel group synchronously executes the instruction according to the preset rules of the entropy value range. For example, high entropy values ​​trigger serpentine channel contraction, and low entropy values ​​maintain linear diversion.

[0063] Compared to traditional emergency response plans, this solution transcends the limitations of passive degradation. Optical flow feature analysis enables emergency commands to precisely match actual passenger flow patterns, avoiding the risk of missed detections caused by system-wide degradation. Pre-commands drive both the decision-making and execution layers, establishing a temporary response link within 200 milliseconds. The system automatically switches back to master control mode upon recovery of predicted data, eliminating the need for manual intervention and ensuring uninterrupted risk management continuity.

[0064] This technical solution creates an emergency response channel. It detects sudden changes in passenger flow based on velocity standard deviation (>1.2 m / s²) and directional change rate (>45° / s) extracted using optical flow. The generated emergency pre-commands include channel width contraction rate and diversion angle parameters. When the predicted data is delayed, the immune decision center directly uses it as an antigen input to generate antibody commands. The channel group synchronously executes the preset actions within the entropy range (for example, activating a serpentine channel when the entropy value is >0.8). This enables the system to respond to sudden passenger flow shocks within 200 ms, significantly improving the response speed compared to traditional systems.

[0065] In another technical solution, the urban rail transit security inspection system, the real-time video compensation unit further performs the following operations: S5.4. Continuously monitor the acquisition status of mobile communication base station signaling density distribution data and gesture skeleton key point vectors; S5.5. When the duration of missing signaling density distribution data exceeds the first tolerance or the invalid count of gesture skeleton key point vectors exceeds the second tolerance, activate the video compensation enhancement protocol and perform the following operations: ) Using the video stream captured by the wide-angle camera array, we use spatiotemporal slicing analysis to generate a passenger flow concentration distribution map to replace the missing signaling density distribution data; ) Analyze the passenger's upper limb motion trajectory through a posture estimation algorithm and generate a simulated gesture skeleton key point vector; S5.7. Input the passenger flow concentration distribution map and the simulated gesture skeleton key point vector into the risk particle flow intensity generation step.

[0066] Existing systems typically respond to data source anomalies by using threshold alarms followed by manual intervention or simple interpolation to fill in missing values. In a smart station project, a mobile base station failure caused signaling data to be interrupted. The system then used historical mean values ​​for interpolation, mistakenly identifying morning rush hour queues as abnormal clusters. This erroneously triggered a high-risk command, resulting in a two-hour channel closure. More seriously, when the gesture capture device is obstructed, the system defaults to a vector value of zero, rendering the direction correction coefficient ineffective and completely deviating from the risk path's actual handling direction.

[0067] This solution enhances the protocol's data resilience mechanism through video compensation, continuously monitoring the acquisition status of signaling density data and gesture skeleton vectors. When the signaling data loss persists beyond the first tolerance or the gesture vector invalid count exceeds the second tolerance, the compensation protocol is automatically activated. Spatiotemporal slicing analysis is performed on the wide-angle camera video stream, and passenger positions are superimposed along the time axis to generate a passenger flow concentration distribution map, replacing the invalid signaling density data. The posture estimation algorithm analyzes the motion trajectory of the passenger's upper limb joints and generates simulated gesture skeleton key point vectors through fitting to replace abnormal gesture data. The concentration distribution map is input into the abnormal area identification link of the risk particle flow generation process, simulating the gesture vector input direction correction coefficient calculation process.

[0068] Compared to traditional methods, this solution overcomes the limitations of passive infill. The clustering distribution map generated by spatiotemporal slicing analysis restores the spatiotemporal evolution of passenger flow, preventing normal queues from being misidentified as abnormal clusters. The skeleton vectors fitted by posture estimation preserve the continuity of behavioral intent, ensuring that the manual intervention dimension is not lost due to equipment anomalies. The system maintains the integrity of its risk prediction functionality even when some data sources fail, with a false alarm rate close to normal operating conditions, and improves decision-making reliability in fault scenarios to a practical level.

[0069] This technical solution enables seamless data failure recovery. Spatiotemporal slicing analysis converts the video stream into a spatiotemporal cube, projecting it along the time axis to generate a passenger flow density distribution map (resolution 0.5m / pixel), replacing missing signaling density data. Gesture estimation generates 17-dimensional gesture skeleton vectors by fitting joint point trajectories. These vectors replace the original data when the confidence level exceeds 0.85. This effectively controls the error rate in calculating the risk particle flow intensity in the event of signaling interruptions or gesture acquisition failures, significantly enhancing system robustness.

[0070] In another technical solution, the urban rail transit security inspection system further includes a multimodal data redundancy engine, the input end of which is connected to the real-time signaling data interface of the existing surveillance camera group and the mobile communication base station at the station; The multimodal data redundancy engine performs the following operations: S6.1. Continuously monitor the status codes of the ground pressure sensor network and the 3D LiDAR. When the status code remains abnormal for a period greater than the fault tolerance, activate the data reconstruction protocol. S6.2. After the data reconstruction protocol is activated, perform the following operations: a) Analyze the video streams from the surveillance camera group and generate a passenger flow density distribution heat map using the background subtraction method; b) Extract the signaling hopping trajectory of the mobile communication base station and construct the passenger movement direction probability matrix; S6.3. Map the passenger flow density distribution heat map to simulated pressure distribution data, convert the passenger movement direction probability matrix into a simulated motion vector field, and input it into the passenger flow dynamics prediction module to replace the original data source; S6.4. When the device status code returns to normal, it automatically switches back to the original data collection mode.

[0071] Existing systems often use threshold alarm shutdown strategies or switch to lower-precision backup sensors for sensor failures. At the Shanghai Hongqiao hub, a short circuit in the pressure sensor network caused the system to switch to a backup infrared counter, which only provided a count of the number of people passing through but lost its density distribution. This caused the prediction module to misjudge the aisle as idle and fail to warn of the risk of a stampede caused by localized congestion. More seriously, when the lidar fails, the system fills the motion data with a fixed directional vector, causing the risk particle flow to deviate from the actual main passenger flow path.

[0072] This solution builds a fault-tolerant system through a multimodal data redundancy engine: it continuously monitors the device status codes of the pressure sensing network and three-dimensional lidar, and activates the data reconstruction protocol when the anomaly persists beyond the fault tolerance limit. It analyzes the video stream from the station's existing surveillance camera system, separates moving foreground pixels using background subtraction, and generates a heat map whose grayscale values ​​are proportional to passenger flow density. It also simultaneously extracts the real-time signaling transition trajectories of mobile communication base stations and constructs a passenger movement direction probability matrix based on a Markov chain model. The heat map is converted into simulated pressure distribution data through linear mapping, and the direction probability matrix is ​​converted into a motion vector field with weighted vectors. This is then input into the passenger flow dynamics prediction module to replace the original fault data source. Once the device status recovers, it automatically switches back to the main sensing mode.

[0073] Compared to traditional shutdown strategies, this solution overcomes the limitations of data substitution distortion: the heat map generated by background subtraction preserves the spatial gradient of passenger density, preventing local high density from being misinterpreted as global idleness; the directional probability matrix constructed by signaling transitions restores actual movement trends, ensuring that the direction of risk particle flow aligns with the true propagation path. The system maintains its prediction function uninterrupted during core sensor failures, and its output accuracy meets basic risk management requirements, substantially overcoming hardware reliability shortcomings.

[0074] This technical solution establishes a fault-tolerant system. A background difference method is used to extract foreground pixels from surveillance video to generate a heat map (grayscale value is proportional to density). Signaling transition trajectories are then transformed into a directional probability matrix using a Markov chain. The heat map is converted into pressure distribution data through linear mapping (grayscale values ​​0-255 correspond to pressures 0-10 kPa), and the directional matrix is ​​converted into a motion vector field (with probabilities > 0.7 being considered the primary direction). This maintains data input continuity to the prediction module during sensor failures, minimizing system downtime.

[0075] In another technical solution, the urban rail transit security inspection system, when synchronously acquiring the mobile communication base station signaling density distribution data and the gesture skeleton key point vector direction in step S1.1, activates the multi-source biometric feature fusion protocol when the effective coverage rate of the signaling density distribution data is less than a preset coverage threshold or the confidence level of the gesture skeleton key point vector is less than a confidence threshold; The multi-source biometric fusion protocol performs the following operations: ) An infrared thermal imaging array deployed at the top of the channel collects the phase difference of the passenger gait cycle to generate a gait aggregation density distribution map, which is normalized to replace the signaling density distribution data; ) The micro-Doppler effect of the millimeter wave scanning array is used to analyze the passenger's trunk swing direction and generate a biomechanical direction vector, which replaces the gesture skeleton key point vector when the swing amplitude is greater than the sensitivity threshold; The gait cluster density distribution map is input into the abnormal cluster area identification process of step S1.2, and the biomechanical direction vector is input into the second correction coefficient calculation process of step S1.3.

[0076] Existing systems lack effective redundancy mechanisms for biometric feature collection anomalies. For example, a metal shield at an international exhibition's security checkpoint disrupted mobile signaling, forcing the system to abandon risk calculations for that area and fail to identify an unusual concentration of contraband in the exhibit transport corridor. This was later manually discovered. Similarly, when a security inspector's raincoat obscured the gesture device during heavy rain, the system used a default forward vector, causing the risk path to deviate by more than 60 degrees from the inspector's actual direction of action, misleading the scanning array into focusing on the wrong area.

[0077] This solution establishes an environmental adaptability mechanism through a multi-source biometric fusion protocol. The fusion protocol is activated when signaling coverage is insufficient or gesture confidence is too low. An infrared thermal imaging array at the top of the tunnel collects temperature fluctuations from the passenger's lower limbs. Gait cycle phase difference analysis generates a normalized gait cluster density map. Regions with excessive phase differences are marked as high-density areas, which replace missing signaling density data and are input into the abnormal cluster identification module. A millimeter-wave scanning array simultaneously captures the micro-Doppler frequency shift reflected by the passenger's torso. When the swing amplitude meets the standard, the swing direction vector is analyzed and used to replace obscured gesture skeletal key points for direction correction calculations.

[0078] Compared to traditional abandonment strategies, this solution overcomes environmental constraints. Gait thermal imaging uses physiological rhythmic characteristics to restore passenger flow patterns, eliminating signaling blind spots and monitoring gaps. Millimeter-wave micro-Doppler's penetrating power captures torso swing intent, maintaining continuity across the human intervention dimension. The system maintains full risk prediction functionality even in harsh data collection environments, and its output reliability reaches levels comparable to conventional operating conditions, effectively addressing security shortcomings in specialized scenarios.

[0079] This technical solution overcomes the limitations of biometric data collection. Gait phase difference is calculated using the temperature fluctuation period of the lower limbs from infrared thermal images, generating a normalized cluster density map (a phase difference >15° is defined as a high-density area). Millimeter-wave micro-Doppler frequency shift is used to analyze trunk sway (frequency shift >50Hz for amplitude >15°), generating biomechanical vectors to replace gesture data. This improves the accuracy of identifying abnormal areas and reduces the error rate in calculating the directional correction coefficient in scenarios with insufficient signaling coverage or gesture obstruction.

[0080] The number of devices and processing scales described herein are intended to simplify the description of the present invention. Applications, modifications, and variations of the present invention will be readily apparent to those skilled in the art.

[0081] Although the embodiments of the present invention have been disclosed above, they are not limited to the applications listed in the description and implementation methods. They can be fully applied to various fields suitable for the present invention. For those familiar with the art, additional modifications can be easily implemented. Therefore, without departing from the general concept defined by the claims and the scope of equivalents, the present invention is not limited to specific details.

Claims

1. An urban rail transit security inspection system, characterized in that: include: The passenger flow dynamics prediction module builds a passenger flow microenvironment model based on the modified Navier-Stokes equations based on the collected passenger flow pressure distribution and motion vector field, and outputs congestion entropy values ​​and risk particle flows in real time; An adaptive millimeter wave scanning array receives a congestion entropy value and a risk particle flow, and starts a mobile focus scanning mode when the intensity of the risk particle flow is greater than a predetermined intensity threshold; When the risk particle flow intensity is less than or equal to the predetermined intensity threshold and the congestion entropy value is greater than the predetermined entropy threshold, the wide-area screening mode is activated and a basic threat index is generated. When a passenger's basic threat index is identified as greater than the safety threshold, the passenger is marked as a target for re-inspection. Otherwise, the wide-area screening mode is executed; A feature analysis module receives and analyzes millimeter-wave scanning data in mobile focus scanning mode, extracts the outline features of the items carried by the focused passenger, compares the outline features with a pre-stored database of prohibited items, and generates a real-time threat level for the passenger; When receiving the target identification for review in the wide-area screening mode, the lightweight analysis model is activated to analyze the millimeter wave data of the target for review and generate a threat assessment report for the target for review; The immune decision-making center receives the congestion entropy value, real-time threat level, and threat assessment report of the target to be reviewed, and dynamically generates a decision antibody through a clonal selection algorithm. The decision antibody contains a channel topology instruction and a set of device parameters. The generation rule of the channel topology instruction includes: if the threat information indicates a high-risk real-time threat level, a command containing a 30° acute angle flow is forcibly generated; If the threat information is a non-high-risk real-time threat level, the corresponding channel topology instruction is generated according to the predetermined entropy value interval of the congestion entropy value; when the target identification to be reviewed and the corresponding threat assessment report of the target to be reviewed are received, the channel topology instruction pointing to the secondary diversion area is generated.

2. The urban rail transit security inspection system according to claim 1, characterized in that: Also includes: The topology-variable channel group receives and executes channel topology instructions and equipment parameter sets, dynamically adjusting the physical structure of the channel; forms an acute-angle diverter structure at the corresponding position according to instructions containing a 30° acute-angle diverter; guides the target to be re-inspected to the secondary diverter area preset within it according to instructions pointing to the secondary diverter area; within the secondary diverter area, further constructs the channel structure according to the channel topology instructions generated by the threat assessment report for the target to be re-inspected; when the threat assessment report indicates a high risk, constructs a π-type polymer channel to guide the target to be re-inspected to the re-inspection area; collects real-time pedestrian pressure change data in the channel and feeds it back to the passenger flow dynamics prediction module for calibrating the passenger flow microenvironment dynamics model.

3. The urban rail transit security inspection system according to claim 2, characterized in that: The intensity value of the risk particle flow is dynamically generated through the following steps: S1.

1. Real-time access to the risk path spatial distribution data output by the passenger flow dynamics prediction module, synchronous acquisition of mobile device signaling density distribution data provided by mobile communication base stations, and the vector directions of gesture skeleton key points captured by the security inspector's wearable device; S1.

2. Based on the mobile device signaling density distribution data, identify the boundary coordinates of the abnormal cluster area; based on the risk path spatial distribution data, extract the path vector of the main diffusion direction; S1.

3. Calculate the overlap ratio between the boundary coordinates of the abnormal cluster area and the spatial distribution area of ​​the risk path as the first correction coefficient; Calculate the cosine value of the angle between the direction of the gesture skeleton key point vector and the path vector of the main diffusion direction as the second correction coefficient; S1.

4. Use the historical frequency of prohibited items stored in the security inspection event database as a base weight, combined with the first and second correction coefficients, and input them into a pre-trained risk propagation tree model. The risk propagation tree model is pre-trained and generated based on the spatiotemporal distribution patterns of the historical security inspection event database. S1.

5. The risk propagation tree model outputs the updated risk particle flow intensity value and feeds it back to the passenger flow dynamics prediction module.

4. The urban rail transit security inspection system according to claim 2, characterized in that: Also includes: The dynamic path calibration engine has its input connected to the initial path output of the risk particle flow of the passenger flow dynamics prediction module and is connected in parallel to the multi-spectral interference sensing array deployed in the channel vault; The multi-spectral interference sensing array simultaneously collects physiological agitation index in the thermal infrared band and motion mutation trajectories in the visible light band; When the physiological agitation index exceeds the preset agitation threshold and lasts for more than 5 seconds, or the cumulative azimuth deviation between the motion mutation trajectory and the initial path is greater than or equal to 30 degrees, the nonlinear compensator is activated; When activated, the nonlinear compensator performs the following operations: S3.

1. Map the physiological agitation index to a path deviation probability density function, and generate a first compensation vector using a convolution kernel; S3.

2. Decompose the tangential and normal acceleration components of the motion mutation trajectory, and generate a second compensation vector using the amplitude of the normal component as a weight; S3.

3. Fuse the first compensation vector and the second compensation vector to the initial path of the risk particle flow, and output the anti-interference calibration path to the risk path input end of the adaptive millimeter wave scanning array.

5. The urban rail transit security inspection system according to claim 4, characterized in that: After the immune decision-making center produces decision-making antibodies, continue to monitor antibody activity: S4.

1. Real-time acquisition of the threat level update flag of the feature analysis module and the instruction execution progress value of the topology variable channel group; S4.

2. When the progress value of the channel group executing the high-risk acute angle flow divider instruction is monitored to be less than 50%, if the threat level update flag is switched to medium or low risk, the antibody apoptosis protocol is activated; S4.

3. The antibody apoptosis protocol uses the instruction execution progress value as the decay base and combines it with the updated risk particle flow intensity value to generate the antibody half-life; S4.

4. When the instruction execution progress value exceeds the progress threshold corresponding to the antibody half-life, the clone selection algorithm is triggered to reorganize the decision antibody. The reorganization rule uses the non-acute angle diversion instruction generated by the entropy value interval to overwrite the original high-risk instruction.

6. The urban rail transit security inspection system according to claim 3, characterized in that: It also includes a real-time video compensation unit that continuously collects passenger flow video streams through a wide-angle camera array deployed on the side walls of the channel; The real-time video compensation unit performs the following operations: S5.

1. Extract the passenger velocity distribution matrix and direction change rate in the video stream using the optical flow method; S5.2, when the standard deviation of the moving speed exceeds the preset speed tolerance and the direction change rate is greater than the direction threshold, an emergency topology pre-instruction is generated; S5.

3. Inputting the emergency topology pre-commands in parallel to the immune decision-making center and the topology variable channel group; Among them, when the output of the passenger flow dynamics prediction module is delayed, the immune decision center uses the emergency topology pre-instruction as an antigen input substitute to drive the clone selection algorithm to generate decision antibodies containing channel topology instructions; after receiving the emergency topology pre-instruction, the topology variable channel group generates channel topology instructions according to the entropy value interval.

7. The urban rail transit security inspection system according to claim 6, characterized in that: The real-time video compensation unit further performs the following operations: S5.

4. Continuously monitor the acquisition status of mobile communication base station signaling density distribution data and gesture skeleton key point vectors; S5.

5. When the duration of missing signaling density distribution data exceeds the first tolerance or the invalid count of gesture skeleton key point vectors exceeds the second tolerance, activate the video compensation enhancement protocol and perform the following operations: ) Using the video stream captured by the wide-angle camera array, we use spatiotemporal slicing analysis to generate a passenger flow concentration distribution map to replace the missing signaling density distribution data; ) Analyze the passenger's upper limb motion trajectory through a posture estimation algorithm and generate a simulated gesture skeleton key point vector; S5.

7. Input the passenger flow concentration distribution map and the simulated gesture skeleton key point vector into the risk particle flow intensity generation step.

8. The urban rail transit security inspection system according to claim 1, characterized in that: It also includes a multimodal data redundancy engine, whose input end is connected to the real-time signaling data interface of the station's existing surveillance camera group and mobile communication base station; The multimodal data redundancy engine performs the following operations: S6.

1. Continuously monitor the status codes of the ground pressure sensor network and the 3D LiDAR. When the status code remains abnormal for a period greater than the fault tolerance, activate the data reconstruction protocol. S6.

2. After the data reconstruction protocol is activated, perform the following operations: a) Analyze the video streams from the surveillance camera group and generate a passenger flow density distribution heat map using the background subtraction method; b) Extract the signaling hopping trajectory of the mobile communication base station and construct the passenger movement direction probability matrix; S6.

3. Map the passenger flow density distribution heat map to simulated pressure distribution data, convert the passenger movement direction probability matrix into a simulated motion vector field, and input it into the passenger flow dynamics prediction module to replace the original data source; S6.

4. When the device status code returns to normal, it automatically switches back to the original data collection mode.

9. The urban rail transit security inspection system according to claim 3, characterized in that: When simultaneously acquiring the mobile communication base station signaling density distribution data and the gesture skeleton key point vector direction in step S1.1, when the effective coverage of the signaling density distribution data is less than the preset coverage threshold or the confidence of the gesture skeleton key point vector is less than the confidence threshold, activating the multi-source biometric fusion protocol; The multi-source biometric fusion protocol performs the following operations: ) An infrared thermal imaging array deployed at the top of the channel collects the phase difference of the passenger gait cycle to generate a gait aggregation density distribution map, which is normalized to replace the signaling density distribution data; ) The micro-Doppler effect of the millimeter wave scanning array is used to analyze the passenger's trunk swing direction and generate a biomechanical direction vector, which replaces the gesture skeleton key point vector when the swing amplitude is greater than the sensitivity threshold; The gait cluster density distribution map is input into the abnormal cluster area identification process of step S1.2, and the biomechanical direction vector is input into the second correction coefficient calculation process of step S1.3.

Images