Railway and subway security sharing method and system

Abstract

The present application provides a kind of railway and subway security sharing method and system, the method includes: receiving the passenger security data uploaded by subway security terminal, including multimodal security data package and identity information, after preprocessing, input pre-trained multimodal fusion analysis model, output security evaluation result and store with identity information associated;Receive the verification request of railway verification terminal, including the identity information to be verified, ticket information and third timestamp;Based on identity information, query associated security evaluation result and subway in-out time stamp;Verify identity and ticket consistency and security of security result, and combine three time stamps, train timetable and transfer time threshold to determine whether it constitutes uninterrupted station transfer;When all verification is passed, generate free inspection pass instruction, otherwise generate secondary security instruction.The present application realizes the intelligent sharing mutual recognition of railway and subway security process, and overcomes the problem of low efficiency and resource waste caused by repeated security check.

Description

Technical Field

[0001] This invention relates to the field of intelligent transportation and public safety technology, and in particular to a method and system for sharing security checks between railways and subways. Background Technology

[0002] With the rapid development of urban integrated transportation networks, the demand for transfers between railways and subways at hub stations is increasing daily. However, currently, railways and subways generally adopt independent security check models, requiring passengers to undergo repeated security checks during transfers. This not only easily leads to congestion and significantly reduces travel efficiency during peak hours, but also increases inconvenience and time costs for passengers. Furthermore, the two independent security check systems result in redundant investment and waste of equipment and manpower resources, increasing overall operating costs. Therefore, achieving security check sharing between railways and subways and building an efficient "single security check, all-area access" model has become a critical issue that urgently needs to be addressed in the field of intelligent transportation.

[0003] In existing technologies, some regions have attempted to promote mutual recognition of security checks through administrative coordination to address this issue. However, due to the different management systems of railways and subways, their security check standards, procedures, and equipment have inherent differences, and such attempts often have limited effectiveness. A deeper reason lies in the lack of an effective technical solution for achieving unified management and collaborative analysis of cross-system security check data. Traditional security check methods mainly rely on manual interpretation of X-ray images and simple threshold alarm devices, lacking a high degree of intelligence and making it difficult to deeply integrate and accurately analyze massive amounts of multi-source, heterogeneous security check data. Although some theoretical research exists on security check information sharing, most of it fails to solve the problems of secure data transmission, multimodal information fusion, and verification difficulties caused by standard differences, resulting in a lack of practical operability.

[0004] Therefore, there is an urgent need in this field for an innovative technical solution that can break down data silos and achieve systematic integration and intelligent upgrading of railway and subway security inspection processes while ensuring security and privacy. Summary of the Invention

[0005] In view of this, embodiments of the present invention provide a method and system for sharing security checks between railways and subways, in order to solve the problems existing in the prior art, such as the inability to share security check data between railways and subways, difficulties in cross-system collaboration, and low passage efficiency caused by passengers having to undergo repeated security checks when transferring.

[0006] One aspect of the present invention provides a method for sharing security checks between railways and subways, the method comprising the following steps: The system receives passenger security check data uploaded from the subway security check terminal. The passenger security check data includes multimodal security check data packets and identity identification information associated with the multimodal security check data packets. The multimodal security check data packets include at least baggage image data collected by an X-ray machine, a first entry timestamp and a second exit timestamp recorded by the turnstile. After preprocessing the multimodal security inspection data packet, it is input into a pre-trained multimodal fusion analysis model for feature extraction and fusion analysis to output a security inspection assessment result. The security inspection assessment result includes at least a security inspection status identifier and a risk level identifier. The security inspection assessment result is then associated and stored with the corresponding identity identifier information. Receive a verification request sent from a railway verification terminal, the request containing at least the identity information of the passenger to be verified, railway ticketing itinerary information, and a third timestamp when the verification request occurred; Based on the received identity information, query the associated security assessment result and the first timestamp and the second timestamp; Verify whether the identity information and the railway ticket itinerary information point to the same identity; verify whether the security status indicator and risk level indicator in the security assessment result both indicate safety; based on the first timestamp, the second timestamp, and the third timestamp, and in conjunction with the subway train timetable and the preset transfer time threshold, determine whether the passenger's journey within the subway and railway hub constitutes an uninterrupted transfer within a closed station; when all the above verifications and judgments are valid, it is determined that the verification is passed, and an exemption passage instruction is generated and sent to the railway verification terminal; if any verification fails, a second security check instruction is generated.

[0007] In some embodiments of the present invention, the preprocessing includes: The luggage image data is subjected to denoising and image enhancement processing to improve image quality and highlight key features; the identity information is standardized and encoded to achieve a unified structured representation of the identity information; the first timestamp and the second timestamp are normalized and converted into a time difference sequence to construct spatiotemporal features that meet the input requirements of the multimodal fusion analysis model.

[0008] In some embodiments of the present invention, the pre-trained multimodal fusion analysis model is a model based on the Transformer architecture.

[0009] In some embodiments of the present invention, the training process of the multimodal fusion analysis model includes: Construct a training sample set containing multiple samples. Each sample includes the multimodal security inspection data packet and its corresponding label. The label includes: defining a danger indicator as a first label, defining a risk level indicator as a second label, and defining a violation of security inspection regulations as a third label. After preprocessing the data of the multimodal security inspection data packet, it is input into the initial multimodal fusion analysis model for feature extraction and cross-modal fusion, and based on the fused features, the predicted values ​​for the danger label, the risk level label, and the violation of security inspection regulations label are output synchronously. A first loss function is constructed by minimizing the deviation between the first label and the predicted value of the danger indicator; a second loss function is constructed by minimizing the deviation between the second label and the predicted value of the risk level indicator; a third loss function is constructed by minimizing the deviation between the third label and the predicted value of the violation of security check regulations; and a joint loss function is constructed by weighted summation of the first loss, the second loss, and the third loss. The initial multimodal fusion analysis model is updated by minimizing the joint loss using the gradient descent algorithm to obtain the multimodal fusion analysis model.

[0010] In some embodiments of the present invention, the first loss is a binary cross-entropy loss; the second loss function is a multi-class cross-entropy loss; and the third loss function is a multi-label cross-entropy loss.

[0011] In some embodiments of the present invention, determining whether a passenger's journey within a subway and railway hub constitutes an uninterrupted transfer within a closed station includes: Based on the time period and passenger type when the verification request occurs, select the corresponding travel time reference value and transfer time reference value from multiple pre-stored transfer time reference values; verify whether the difference between the second timestamp and the first timestamp is within the first time interval formed by the travel time determined according to the subway train timetable, the travel time reference value, and the first buffer margin; verify whether the difference between the third timestamp and the second timestamp is within the second time interval formed by the transfer time reference value and the second buffer margin.

[0012] In some embodiments of the present invention, the set of reference values ​​is constructed in the following manner: Samples of passengers’ effective walking time within the station were collected by time period and by passenger type. For each time period and passenger type combination, after removing samples with abnormal stops, the arithmetic mean of the remaining valid walking time samples is calculated as the baseline value for that combination.

[0013] In some embodiments of the present invention, the multimodal security inspection data packet further includes at least one of the following: Passenger video frames captured by surveillance cameras; facial feature image data captured by facial recognition devices; passenger historical security check records associated with the identity information; real-time passenger flow data at the subway security checkpoint; and subway line information where the subway security checkpoint is located.

[0014] On the other hand, the present invention also provides a railway and subway security check sharing system, the system comprising: The data collection subsystem installed at the subway security checkpoint includes an X-ray security inspection machine, an identity recognition device, and a turnstile, and is used to collect multimodal security inspection data packets from passengers. The verification subsystem installed at railway stations includes an identity verification terminal and a gate control unit; A central processing subsystem, comprising at least one server configured with a processor and a memory; and a communication network connecting the data acquisition subsystem, the verification subsystem, and the central processing subsystem; wherein the data acquisition subsystem, the verification subsystem, and the central processing subsystem are configured to perform the method as described above.

[0015] This invention provides a method and system for sharing security checks between railways and subways. The method includes: receiving passenger security check data uploaded by a subway security check terminal, which includes multimodal security check data packets and identity identification information; preprocessing the data and inputting it into a pre-trained multimodal fusion analysis model for fusion analysis, outputting security check assessment results and storing them in association with the identity information; receiving verification requests from railway verification terminals, including identity information to be verified, ticket information, and a third timestamp; querying the associated security check assessment results and the first timestamp for subway entry / exit and the second timestamp for exit based on the identity identification information; verifying the consistency between identity and ticket and the security of the security check results, and determining whether an uninterrupted in-station transfer is constituted by combining the three timestamps, train timetable, and transfer time threshold; generating a pass-through instruction when all verifications pass, otherwise generating a secondary security check instruction. This invention realizes intelligent sharing and mutual recognition of security check processes between railways and subways, overcoming the problems of low passage efficiency and resource waste caused by repeated security checks.

[0016] Furthermore, when judging the continuity of transfers within the station, the system uses time intervals composed of measured transfer time benchmarks and buffer margins based on time periods and different groups of people for verification. This effectively avoids misjudgments caused by individual differences in walking speed or slight travel delays, and improves the rationality of the verification pass rate.

[0017] Additional advantages, objects, and features of the invention will be set forth in part in the description which follows, and will also become apparent in part to those skilled in the art upon studying the description, or may be learned by practice of the invention. The objects and other advantages of the invention can be realized and obtained by means of the structures specifically pointed out in the description and drawings.

[0018] Those skilled in the art will understand that the objectives and advantages achievable with the present invention are not limited to those specifically described above, and that the above and other objectives achievable with the present invention will become clearer from the following detailed description. Attached Figure Description

[0019] The accompanying drawings, which are provided to further illustrate the invention and form part of this application, are not intended to limit the scope of the invention.

[0020] Figure 1 This is a flowchart illustrating a method for sharing security checks between railways and subways according to an embodiment of the present invention.

[0021] Figure 2 This is a flowchart illustrating the workflow of a railway and subway security check sharing system according to an embodiment of the present invention.

[0022] Figure 3 This is a schematic diagram of the overall architecture of the multimodal fusion analysis model according to an embodiment of the present invention.

[0023] Figure 4 This is a core structural diagram of the Transformer model encoder-decoder according to an embodiment of the present invention. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the embodiments and accompanying drawings. Here, the illustrative embodiments and descriptions of this invention are used to explain the invention, but are not intended to limit the invention.

[0025] It should also be noted that, in order to avoid obscuring the invention with unnecessary details, only the structures and / or processing steps closely related to the solution according to the invention are shown in the accompanying drawings, while other details that are not closely related to the invention are omitted.

[0026] Currently, independent security check systems are commonly used in integrated transportation hubs where railways and subways coexist. This situation directly leads to the following problems: 1) Low throughput: Repeated security check processes significantly extend passenger transfer times, easily creating congestion bottlenecks during peak hours and severely impacting the hub's evacuation capacity and operational efficiency. 2) Duplicated investment of public resources: To maintain two independent security check processes, similar security equipment and human resources need to be deployed repeatedly within the same hub, failing to achieve optimal allocation of infrastructure and personnel efficiency. 3) Lack of technological collaboration: Railway and subway security check systems are independent of each other, with different data standards and formats. For example, multi-source information such as X-ray images, identity information, and spatiotemporal data cannot be shared, forming data silos. Existing technology lacks a complete technical solution that can effectively integrate and intelligently analyze the aforementioned multimodal security check data and support cross-system security verification.

[0027] In view of this, one aspect of the present invention provides a method for sharing security checks between railways and subways, such as... Figure 1 As shown, the method includes the following steps S101~S106: S101: Receive passenger security check data uploaded from the subway security check terminal. The passenger security check data includes multimodal security check data packets and identity identification information associated with the multimodal security check data packets. Among them, the multimodal security check data packets include at least baggage image data collected by the X-ray machine, the first timestamp of entry and the second timestamp of exit recorded by the turnstile. S102: After preprocessing the multimodal security inspection data packet, it is input into the pre-trained multimodal fusion analysis model for feature extraction and fusion analysis to output the security inspection assessment result. The security inspection assessment result includes at least the security inspection status identifier and the risk level identifier; the security inspection assessment result is associated with and stored with the corresponding identity information. S103: Receive a verification request sent from the railway verification terminal. The request shall include at least the identity information of the passenger to be verified, the railway ticketing itinerary information, and a third timestamp when the verification request was made. S104: Based on the received identity information, query the associated security assessment results and the first and second timestamps; S105: Verify whether the identity information and railway ticket travel information point to the same identity; verify whether the security status indicator and risk level indicator in the security assessment results both indicate safety; based on the first, second, and third timestamps, and combined with the subway train timetable and the preset transfer time threshold, determine whether the passenger's journey within the subway and railway hub constitutes an uninterrupted transfer within a closed station; when all the above verifications and judgments are valid, the verification is deemed successful, and an exemption from inspection instruction is generated and sent to the railway verification terminal; if any verification fails, a second security inspection instruction is generated.

[0028] In step S101, the central server deployed at the back end of the system receives passenger security check data uploaded from the subway security check terminal in real time or near real time through its built-in data transmission unit and communication network.

[0029] Specifically, passenger security check data is acquired by a data collection subsystem deployed at subway security checkpoints. This subsystem integrates a series of automated detection and identification devices, forming a multi-source heterogeneous data collection terminal. Its core components include X-ray security scanners, turnstile systems, and identity recognition devices.

[0030] X-ray security scanners are used to scan passengers' luggage and generate X-ray images of the luggage. This image data can reveal the shape, density, and structure of the items inside the luggage, and is a key basis for determining whether there are contraband items.

[0031] The turnstile system is used to accurately record the first timestamp of entry when passengers swipe their tickets to pass through the subway entrance and exit turnstiles. and the second timestamp of exiting the station These two timestamps together define the closed-loop travel time for passengers within the subway network.

[0032] Identification devices, such as ID card readers or facial recognition terminals, are used to collect passengers' identification information. This identification information serves as a core association key, uniquely binding the passenger to all their multimodal security check data.

[0033] In some embodiments, the identification information may be a de-identified ID card number.

[0034] In some embodiments, the identity information can be a feature vector extracted from biometrics. This feature vector is a numerical representation formed through mathematical transformation, and its characteristic is that it cannot be reversibly derived from the original biometric image. This satisfies the requirements for the protection of biometric privacy information while achieving unique identity association.

[0035] The multimodal security inspection data packet contains at least the aforementioned X-ray image data, a first timestamp, and a second timestamp.

[0036] In some embodiments, the multimodal security check data package can be further extended to provide a more comprehensive analysis context, including: passenger video frames captured by surveillance cameras for analyzing passenger behavior and posture; facial feature image data captured by facial recognition devices for cross-verification with identity information; passenger historical security check records associated with identity information for risk trend analysis; real-time passenger flow data at the subway security checkpoint; and subway line information as a spatiotemporal characteristic for assessing the overall security environment.

[0037] In some embodiments, to ensure the security and privacy of data during transmission and to prevent the leakage and tampering of sensitive information, the data acquisition subsystem communicates with the central server through a secure encrypted channel.

[0038] Specifically, the data is encrypted using the AES-256 algorithm, and a transmission link is established based on the SSL / TLS protocol. Meanwhile, to meet the real-time requirements of high concurrency and low latency, 5G communication technology can be used to transmit data with edge computing nodes, ensuring that data transmission latency is controlled within 200 milliseconds under typical operating conditions, thereby achieving rapid, stable, and secure reporting of security inspection data.

[0039] In some embodiments, the central server receives data in real time through a Kafka message queue to ensure that data can be received and buffered efficiently, asynchronously, and reliably in scenarios with a large number of concurrent passengers, thus avoiding data loss or system blockage.

[0040] In step S102, the central server performs preprocessing and intelligent analysis on the received multimodal security inspection data packets, specifically including sub-steps S1021~S1023: Step S1021: The central server performs preprocessing operations on the various types of raw data in the multimodal security inspection data packets to improve data quality, unify data format, and adapt to the input requirements of the subsequent multimodal fusion analysis model.

[0041] In some embodiments, the preprocessing operation includes: Image enhancement operations, including contrast enhancement and noise suppression, are performed on baggage image data to highlight key outlines and structural features of items inside the baggage.

[0042] Perform standardized encoding operations on identity information to convert it into a unified, structured data format.

[0043] Perform normalization processing on the timestamp data, and set the first timestamp... With the second timestamp The data is converted into a time difference sequence to construct a spatiotemporal feature characterizing passenger journey duration that meets the input requirements of a multimodal fusion analysis model.

[0044] Step S1022: The central server inputs the preprocessed multimodal data into a pre-trained multimodal fusion analysis model. This model simultaneously processes and correlates image, text, and spatiotemporal data, performs feature extraction and fusion analysis, and outputs a structured security assessment result. This result includes at least a binary security status label and a multi-class risk level label.

[0045] In some embodiments, the security assessment results may also include a violation detail identifier to indicate the specific source of risk.

[0046] In some embodiments, the output of the multimodal fusion analysis model is a structured data array, the output structure of which is [anomaly detection, risk level, violation details].

[0047] For example, the model's output is Where "1" indicates the security check status is "requires secondary security check", "2" indicates the risk level is "high risk", and "2" indicates the risk level is "high risk". The violation rule triggered is "liquid standard difference". The overall meaning of this output is that the passenger was detected as abnormal, the risk level is high, and the violation rule "liquid standard difference" was triggered.

[0048] In some embodiments, the multimodal fusion analysis model is built on the Transformer architecture. Preprocessed data of various types are projected into a unified implicit space and concatenated into a sequence; after this sequence is input into the model, deep understanding and fusion of cross-modal information are achieved through its internal self-attention mechanism.

[0049] In some embodiments, the training process of the multimodal fusion analysis model includes: Construct a training sample set containing multiple samples. Each sample includes a multimodal security inspection data packet and its corresponding label. The labels include: defining whether it is dangerous as the first label, defining the risk level as the second label, and defining the violation of security inspection regulations as the third label. After preprocessing the multimodal security inspection data packets, the data is input into the initial multimodal fusion analysis model for feature extraction and cross-modal fusion. Based on the fused features, the predicted values ​​for whether a hazard label is present, the predicted values ​​for risk level labels, and the predicted values ​​for labels indicating violations of security inspection regulations are output synchronously. To construct a first loss function, minimize the deviation between the first label and the predicted value of whether a hazard is detected; to construct a second loss function, minimize the deviation between the second label and the predicted value of the risk level; to construct a third loss function, minimize the deviation between the third label and the predicted value of the violation of security check regulations; and to construct a joint loss function by weighted summing of the first, second, and third losses. The initial multimodal fusion analysis model is updated by minimizing the joint loss using the gradient descent algorithm to obtain the multimodal fusion analysis model.

[0050] In some embodiments, the first loss function is a binary cross-entropy loss, which is applicable to binary classification of whether a risk is identified; the second loss function is a multi-class cross-entropy loss, which is applicable to multi-class risk level identification; and the third loss function is a multi-label classification cross-entropy loss, which is applicable to detailed rule identification that may violate multiple regulations simultaneously.

[0051] Step S1023: The central server associates the security assessment results generated in step S1022 with the passenger identification information that triggered this analysis and writes it into the system database.

[0052] In step S103, the central server receives a verification request sent from the railway verification terminal.

[0053] In some embodiments, the verification request is automatically generated and sent by the railway verification terminal when a passenger performs an identity verification operation. This request is transmitted to a central server via a secure communication link.

[0054] The verification request must contain at least three types of key information: the passenger's identification information, railway ticketing and travel information, and a third-party timestamp when the verification request was made.

[0055] The identity information of passengers to be verified is collected on-site by the railway verification terminal, and its format is consistent with that collected at the subway terminal, such as ID card number or biometric vector. Railway ticketing and travel information is obtained by reading the passenger's ticket voucher or by querying the railway ticketing system in real time in conjunction with the identity information; this includes train number, seat number, and departure time. A third timestamp is precisely generated and marked by the railway verification terminal when the request is initiated.

[0056] In step S104, the central server performs a data query operation based on the identity information in the received verification request.

[0057] Specifically, the central server uses the identity information it receives as a unique query key to retrieve data from its system database.

[0058] In some embodiments, the central server retrieves from the database the security assessment result associated with the passenger, generated and stored in step S102, and the first and second timestamps recorded and stored in step S101, using the passenger's identification information. The security assessment result includes at least a security status identifier and a risk level identifier. The first and second timestamps together define the passenger's travel time within the subway network.

[0059] In step S105, the central server executes preset verification rules based on the data queried in step S104, and makes a final decision accordingly, specifically including sub-steps S1051~S1054: Step S1051: Verify whether the identity information and the railway ticket itinerary information point to the same identity.

[0060] Specifically, the central server verifies whether the identity information received from the verification request and the railway ticket travel information point to the same legitimate passenger. This verification ensures that the travel ticket matches the identity of the person entering the station, which is a fundamental prerequisite for authorization.

[0061] Step S1052: Verify whether the security status label and risk level label in the security assessment results both indicate that the security is safe.

[0062] Specifically, the central server verifies the security assessment results written in step S102 by querying the database.

[0063] In some embodiments, verification is considered successful only if both of the following conditions are met: The security check status indicator is either "Safe" or "Security Check Passed"; The risk level is indicated as "low risk" or at an acceptable level.

[0064] Furthermore, if the security assessment results contain any violations of regulations, the list must be empty. If the list contains any specific violations, the verification will automatically fail.

[0065] Step S1053: The central server, based on the first, second, and third timestamps, and in conjunction with the subway train timetable and the preset transfer time threshold, comprehensively determines whether the passenger's journey within the subway and railway hub constitutes an uninterrupted transfer within a closed station.

[0066] In some embodiments, determining whether a passenger's journey within a subway and railway hub constitutes an uninterrupted transfer within a closed station includes: Based on the time period and passenger type when the verification request occurs, select the corresponding travel time base value and transfer time base value from multiple pre-stored transfer time base values; Verify whether the difference between the second timestamp and the first timestamp is within the first time interval determined by the travel time, the travel time base value, and the first buffer margin based on the subway train timetable; Verify whether the difference between the third timestamp and the second timestamp falls within the second time interval formed by the transfer time base value and the second buffer margin.

[0067] The pre-stored multiple sets of transfer time baseline values ​​are constructed in the following way: Samples of passengers’ effective walking time within the station were collected by time period and by passenger type. For each time period and passenger type combination, after removing samples with abnormal stops, the arithmetic mean of the remaining valid walking time samples is calculated as the baseline value for that combination.

[0068] Step S1054: The central server integrates all the above verification results and executes the final decision, specifically: When all verifications and judgments in steps S1051 to S1053 are successful, the central server determines that the verification is passed and generates an exemption pass instruction, which is then sent to the railway verification terminal. Upon receiving this instruction, the railway verification terminal can control the gate to allow passage or provide a passage prompt, allowing passengers to directly enter the railway waiting area.

[0069] If any of the verification steps S1051 to S1053 fails, the central server generates a secondary security check instruction and sends it to the railway verification terminal. Upon receiving the instruction, the terminal should guide or instruct the passenger to proceed to the secondary security check area for targeted inspection.

[0070] On the other hand, the present invention also provides a railway and subway security check sharing system, comprising: The data acquisition subsystem installed at the subway security checkpoint includes an X-ray security inspection machine, an identity recognition device, and a turnstile, and is used to collect multimodal security inspection data packets from passengers. The verification subsystem is set up in railway stations and includes an identity verification terminal and a gate control unit. A central processing subsystem includes at least one server equipped with a processor and memory; and a communication network connecting the data acquisition subsystem, the verification subsystem, and the central processing subsystem; wherein the data acquisition subsystem, the verification subsystem, and the central processing subsystem are configured to perform the methods described above.

[0071] On the other hand, the present invention also provides a computer-readable storage medium having a computer program or instructions stored thereon, which, when executed by a processor, implement the steps of any of the methods described above.

[0072] The present invention will now be described with reference to a specific embodiment: This embodiment will describe the three stages of subway security check, railway verification and result processing in the natural sequence of the security check sharing business.

[0073] 1. Subway security check stage Passengers undergo security checks at the subway security checkpoints, where subway X-ray security equipment scans passengers' luggage to obtain luggage image data.

[0074] Identity recognition devices identify passenger identity information, such as ID card numbers and facial features from videos.

[0075] The data acquisition module deployed at the front end of the subway transmits passenger security check data to the back-end security check information sharing platform through a secure and encrypted channel. The security check data includes baggage image data, metal detection data, identity information, etc.

[0076] Timestamp marker: such as Figure 2 As shown, when passengers enter and exit subway station gates, the system marks the first time stamp for entry and the second time stamp for exit; when passengers swipe their ID cards at railway station gates, it marks the third time stamp.

[0077] Subway security check result storage: The security check results of passengers are returned by the fusion analysis module based on the Transformer multimodal large model. The results are then associated with the identity information and stored in the security check sharing system. Passengers with no risk are marked as "passed security check"; other cases are marked as "requires secondary security check".

[0078] 2. Railway security check verification: After arriving at the subway station, passengers will proceed to the railway entrance via a dedicated passage, where they will be spatially separated from passengers arriving by other modes of transportation. The verification process is as follows: Identity information and security check verification results confirmation: The identity recognition equipment at the railway station entrance re-identifies the passenger's identity information and compares it with the identity information stored in the platform. After successful verification, the system retrieves the passenger's stored subway security check result from the platform, such as "Security check passed" or "Second security check required."

[0079] Timestamp Confirmation: The system verifies that the time of passengers entering and exiting the station conforms to the subway timetable, and adds the average walking transfer time of passengers to ensure that passengers do not leave the station during the journey.

[0080] The average walking transfer time for passengers is calculated as follows: 1) Sample selection criteria Covering all time periods: morning peak (7:30~9:00), off-peak (10:00~17:00), evening peak (18:00~19:30), and low-peak (20:00~last train), samples were collected during each time period. 50 copies; Covering the entire population: divided by age into youth (18-40 years old), middle-aged (41-60 years old), and elderly (60+ years old), and by items carried into no luggage, small handbag, and large luggage, to ensure a balanced sample distribution; Exclude outliers: Remove measurement data from unexpected situations to avoid interference, such as passenger congestion or equipment failure.

[0081] 2) Measurement tools and nodes Tools: Use a stopwatch accurate to 1 second or professional passenger flow analysis equipment, such as video surveillance AI timing; Key milestone timing: Starting point: Passengers have completely exited the previous train's doors (feet touching the ground); Destination: Passengers arrive at the transfer platform and wait by the target train door (in stable condition). Record: Only the "effective walking time" is recorded, excluding time spent stopping along the way, such as checking signs or avoiding others. If a stop exceeds 10 seconds, the sample is marked as "abnormal" and not included in the statistics.

[0082] 3) Average transfer calculation For each scenario, after removing outliers, the average transfer time is calculated using the arithmetic mean method: Average transfer time The sum of all valid sample times Number of valid samples.

[0083] 3. Verification result processing: Based on the above verification results, the system performs the following processing: If the verification result is "passed," passengers can directly enter the waiting area; if the verification result is "failed," railway security personnel will conduct a targeted secondary security check on the passengers, such as a detailed inspection of suspicious items. Passengers can only enter the waiting area after passing the secondary security check.

[0084] The above verification will "pass" only if all of the following conditions are met simultaneously: 1) Passenger identity information should be consistent with the travel information on the high-speed rail ticket.

[0085] 2) The security check result should be "Security check passed".

[0086] 3) The passenger only entered and exited the subway station once during the trip and did not exit the station midway.

[0087] 4) The time difference should meet the following requirements: The difference between the second timestamp and the first timestamp conforms to the time difference between passengers boarding and alighting in the subway train timetable, plus the passenger's transfer time within the station, plus a 2-minute margin. The difference between the third timestamp and the second timestamp corresponds to the passenger's in-station transfer time plus a 2-minute buffer.

[0088] Corresponding to the above-described security check sharing method, this embodiment also provides a railway and subway multi-modal transportation security check sharing system for implementing this method. The specific implementation scheme of this system is as follows: 1. Security Check Information Sharing Platform: The security check information sharing platform is the core of the system, and it specifically includes the following units and modules: Data transmission unit: Employing 5G+edge computing technology, it transmits subway and railway security inspection data to the central processing platform in real time via an AES-256 encrypted channel, minimizing transmission latency. 200ms.

[0089] Multimodal data acquisition module: Responsible for accessing and managing multi-source data. Data collected from the subway includes X-ray image data, metal detection data, video surveillance data, facial feature data, timestamp data, and subway train information; data collected from the railway includes X-ray package image data from secondary security checks, video data, and liquid explosive detection data, etc. All collected data is stored on the platform server.

[0090] Data Interface: Through high-throughput message queues such as Kafka, data from the railway ticketing system and the public security identity authentication system are accessed in real time to achieve data fusion with external systems.

[0091] 2. Data Preprocessing Module: This module performs standardization processing on the collected multi-source heterogeneous security inspection data, specifically including: X-ray images undergo preprocessing operations such as denoising, enhancement, and normalization. Image enhancement algorithms are then used to highlight key features in the images, facilitating subsequent target detection.

[0092] The detection data from security gates and other equipment are filtered to remove noise interference.

[0093] Data from different sources is formatted and standardized so that it can be uniformly input into the subsequent large model for processing.

[0094] 3. Fusion Analysis Module Based on Transformer Multimodal Large Model Multimodal feature fusion involves inputting the aforementioned subway and railway security inspection image data, security inspection regulations and texts, and spatiotemporal features into the Transformer large model to generate a fused feature vector.

[0095] The aforementioned fusion analysis module based on the Transformer multimodal large model is the core of intelligent discrimination, and its overall architecture is as follows: Figure 3 As shown, the core Transformer structure is as follows: Figure 4 As shown, the model structure and training process are as follows: (1) Model structure analysis ① Multimodal input layer: responsible for receiving image features, text features, and spatiotemporal features; The image features include X-ray security inspection images, surveillance video frames, and facial recognition images; Text features include passenger information, baggage description, history, and real-time notifications; Spatiotemporal features include timestamps, station locations, pedestrian traffic data, and route information.

[0096] ② Feature preprocessing and projection layer: Image path: High-level features are extracted from the original image using a CNN convolutional neural network, and then transformed into an image embedding vector with dimensions of [dimensionality missing]. .

[0097] Text path: The contextual representation of the input text is obtained through the BERT encoder, and then transformed into a text embedding vector with dimension 1. .

[0098] Spatiotemporal path: Structured spatiotemporal data is processed through a fully connected network, and then transformed into a spatiotemporal embedding vector with dimensions of [missing information]. .

[0099] ③ Modal alignment and positional encoding: Modal type encoding is added to the above images, texts and spatiotemporal embedding vectors to distinguish the information source, and positional encoding is added to provide sequence position information. Finally, all multimodal features are concatenated into a unified sequence.

[0100] ④ Transformer Multimodal Fusion Core: The core of the multimodal fusion analysis model is built on the Transformer architecture, which consists of an encoder and a decoder.

[0101] In this embodiment, the encoder is responsible for deep feature encoding and fusion of the input multimodal fusion sequence, while the decoder performs cross-modal attention calculation based on the encoder's output to generate structured security inspection assessment results.

[0102] The Transformer encoder consists of N identical layers stacked together (N is a positive integer, for example, 6). Each layer contains two sub-layers: a multi-head self-attention mechanism and a feed-forward neural network (FFN).

[0103] Each sublayer employs residual connectivity and layer normalization techniques.

[0104] The workflow of a Transformer encoder is as follows: Input representation: Each word in the input sequence is first converted into a word embedding vector, and then positional encoding is added to inject positional information.

[0105] Self-attention layer: For each word, the self-attention mechanism is used to calculate its relevance to all other words, thereby capturing contextual information. The multi-head mechanism allows the model to learn information in different representation subspaces.

[0106] Feedforward network: performs a nonlinear transformation on the output of the self-attention layer, and processes each position independently.

[0107] Residual connections and layer normalization: The output of each sublayer is added to the input (residual connection), and then layer normalization is performed, which helps in training deep networks.

[0108] The Transformer decoder consists of N identical layers stacked together. Each decoder layer contains three core sub-layers: Masked Multi-Head Self-Attention, Encoder-Decoder Attention, and Feed-Forward Neural Network.

[0109] Each sublayer also uses residual connections and layer normalization.

[0110] The workflow of a Transformer decoder is as follows: Input representation: The decoder's input is the word embeddings of the target sequence plus positional encoding. During training, the entire target sequence is used (but masking ensures that the current position can only see previous words); during inference, the generation process is autoregressive, generating one word at a time.

[0111] Masked self-attention layer: ensures that when generating each lexical unit, only previously generated lexical units are seen, preventing information leakage.

[0112] Encoder-decoder attention layer: The decoder's representation is used as a query, and the encoder's output is used as a key and value, thus enabling the decoder to focus on relevant parts of the input sequence.

[0113] Feedforward network: Similar to encoder, it performs nonlinear transformation.

[0114] Residual connectivity and layer normalization: applied after each sublayer.

[0115] ⑤ Feature Fusion Output Layer: Extracts the beginning of the sequence output by the encoder. The hidden states corresponding to the labels are used as the fused representation of the entire sequence, and the output is a unified fused feature vector. Specifically, this layer performs the following operations in parallel: CLS tag characterization: Extract the hidden state corresponding to the special tag [CLS] at the beginning of the sequence. This state is designed to carry the aggregate information of the entire sequence.

[0116] Average pooling: The hidden states at all positions in the sequence are averaged to obtain the overall distribution characteristics of the sequence.

[0117] Max pooling: Maximizes the hidden states at all positions in the sequence to extract the most salient features. Attention pooling: Introduces a learnable attention weight vector to sum the weights of each position in the sequence, thereby focusing on the key information most relevant to the security check task.

[0118] Subsequently, the feature vectors obtained from the above four strategies are concatenated to form a fused feature vector that is more comprehensive in information and has stronger representational capabilities. ,in The total dimension of the concatenated features will serve as the input for the subsequent security check shared task layer.

[0119] ⑥ Security Check Shared Task Layer: Based on the fused feature vectors, the model synchronously outputs the final security check assessment result, in the format of [anomaly detection, risk level, violation details], as follows: Anomaly detection: Perform binary classification. Output "0" to indicate no anomaly found (No), and output "1" to indicate an anomaly exists (Yes).

[0120] Risk Level: When performing multi-classification tasks, risks are assessed into discrete levels. The specific mapping is as follows: 0 corresponds to 'low risk', 1 corresponds to 'medium risk', 2 corresponds to 'high risk', and 3 corresponds to 'critical risk'.

[0121] Violation of Detailed Rules: Perform a multi-label classification task to identify the specific reasons that triggered the risk alert. Its output is a list that may contain zero or more predefined violation identifiers, such as: 'liquid_difference': This indicates that the liquid in the carried items does not meet the railway security inspection requirements.

[0122] 'item_restriction': Indicates that the items you carry belong to the category that are allowed on the subway but prohibited on the railway.

[0123] 'transfer_check': Indicates that the passenger's transfer behavior pattern is abnormal, triggering the transfer check rule. 'special_screening': Indicates that the passenger meets other special screening criteria and requires further examination.

[0124] This structured output provides a precise and interpretable basis for subsequent decisions on whether to allow inspection-free passage.

[0125] (2) Model training process Training sample data: Image data: from X-ray security inspection machines, surveillance cameras, etc., labeled with the type and location of hazardous materials.

[0126] Text data includes passenger information, baggage descriptions, and security check records.

[0127] Spatiotemporal data: including timestamps, sites, and visitor traffic.

[0128] Tag data includes whether it is dangerous, risk level, and details of violations of security check regulations.

[0129] Training set: Test set: Validation set .

[0130] Training process: For binary classification tasks that determine "whether it is dangerous", binary cross-entropy loss is used; for multi-classification tasks such as determining "risk level" and "violation of security check regulations", cross-entropy loss is used.

[0131] First, the model is pre-trained on a large multimodal dataset, and then fine-tuned end-to-end using security inspection data. The AdamW optimizer is used, with a batch size of 8; the training cycle is adjusted based on the performance on the validation set.

[0132] Evaluation metrics: Precision, recall, and F1 score are used as evaluation metrics for model performance.

[0133] 4. Cross-system collaborative decision-making: Mutual recognition of security check results: The railway security check system retrieves the subway security check results. If the verification is successful, the passenger can pass directly; if the verification fails, a second security check process is triggered.

[0134] Dynamic resource scheduling: Based on real-time passenger flow and risk distribution, the central platform automatically adjusts the number of security checkpoints, personnel allocation, and equipment operating parameters at each hub.

[0135] 5. Privacy Protection and Data Security: AES-256 encryption and SSL / TLS protocol are used during data transmission to ensure data security.

[0136] During model training, all participants only share the model parameter gradients, and the original data does not leave the local machine, thus protecting data privacy.

[0137] In summary, this invention provides a method and system for sharing security checks between railways and subways. The method includes: receiving passenger security check data uploaded by a subway security check terminal, which includes multimodal security check data packets and identity identification information; preprocessing the data and inputting it into a pre-trained multimodal fusion analysis model for fusion analysis, outputting security check assessment results and storing them in association with the identity information; receiving verification requests from railway verification terminals, including identity information to be verified, ticket information, and a third timestamp; querying the associated security check assessment results and the first timestamp for subway entry / exit and the second timestamp for exit based on the identity identification information; verifying the consistency between identity and ticket and the security of the security check results, and determining whether an uninterrupted in-station transfer is constituted by combining the three timestamps, train timetable, and transfer time threshold; generating a pass-through instruction when all verifications pass, otherwise generating a secondary security check instruction. This invention realizes intelligent sharing and mutual recognition of security check processes between railways and subways, overcoming the problems of low passage efficiency and resource waste caused by repeated security checks.

[0138] Furthermore, when judging the continuity of transfers within the station, the system uses time intervals composed of measured transfer time benchmarks and buffer margins based on time periods and different groups of people for verification. This effectively avoids misjudgments caused by individual differences in walking speed or slight travel delays, and improves the rationality of the verification pass rate.

[0139] Those skilled in the art will understand that the exemplary components, systems, and methods described in conjunction with the embodiments disclosed herein can be implemented in hardware, software, or a combination of both. Whether implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this invention. When implemented in hardware, it can be, for example, electronic circuits, application-specific integrated circuits (ASICs), appropriate firmware, plug-ins, function cards, etc. When implemented in software, the elements of this invention are programs or code segments used to perform the desired tasks. The programs or code segments can be stored in a machine-readable medium or transmitted over a transmission medium or communication link via data signals carried in a carrier wave.

[0140] It should be clarified that the present invention is not limited to the specific configurations and processes described above and shown in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of the present invention is not limited to the specific steps described and shown. Those skilled in the art can make various changes, modifications, and additions, or change the order of steps, after understanding the spirit of the present invention.

[0141] In this invention, features described and / or illustrated for one embodiment may be used in the same or similar manner in one or more other embodiments, and / or combined with or in place of features of other embodiments.

[0142] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, various modifications and variations of the embodiments of the present invention are possible. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for sharing security checks between railways and subways, characterized in that, The method includes the following steps: The system receives passenger security check data uploaded from the subway security check terminal. The passenger security check data includes multimodal security check data packets and identity identification information associated with the multimodal security check data packets. The multimodal security check data packets include at least baggage image data collected by an X-ray machine, a first entry timestamp and a second exit timestamp recorded by the turnstile. After preprocessing the multimodal security inspection data packet, it is input into a pre-trained multimodal fusion analysis model for feature extraction and fusion analysis to output a security inspection assessment result. The security inspection assessment result includes at least a security inspection status identifier and a risk level identifier. The security inspection assessment result is then associated and stored with the corresponding identity identifier information. Receive a verification request sent from a railway verification terminal, the request containing at least the identity information of the passenger to be verified, railway ticketing itinerary information, and a third timestamp when the verification request occurred; Based on the received identity information, query the associated security assessment result and the first timestamp and the second timestamp; Verify whether the identity information and the railway ticket itinerary information point to the same identity; verify whether the security status indicator and risk level indicator in the security assessment result both indicate safety; based on the first timestamp, the second timestamp, and the third timestamp, and in conjunction with the subway train timetable and the preset transfer time threshold, determine whether the passenger's journey within the subway and railway hub constitutes an uninterrupted transfer within a closed station; when all the above verifications and judgments are valid, it is determined that the verification is passed, and an exemption passage instruction is generated and sent to the railway verification terminal; if any verification fails, a second security check instruction is generated.

2. The railway and subway security check sharing method according to claim 1, characterized in that, The preprocessing includes: The luggage image data is subjected to denoising and image enhancement processing to improve image quality and highlight key features; the identity information is standardized and encoded to achieve a unified structured representation of the identity information; the first timestamp and the second timestamp are normalized and converted into a time difference sequence to construct spatiotemporal features that meet the input requirements of the multimodal fusion analysis model.

3. The railway and subway security check sharing method according to claim 1, characterized in that, The pre-trained multimodal fusion analysis model is a model based on the Transformer architecture.

4. The railway and subway security check sharing method according to claim 3, characterized in that, The training process of the multimodal fusion analysis model includes: Construct a training sample set containing multiple samples. Each sample includes the multimodal security inspection data packet and its corresponding label. The label includes: defining a danger indicator as a first label, defining a risk level indicator as a second label, and defining a violation of security inspection regulations as a third label. After preprocessing the data of the multimodal security inspection data packet, it is input into the initial multimodal fusion analysis model for feature extraction and cross-modal fusion, and based on the fused features, the predicted values ​​for the danger label, the risk level label, and the violation of security inspection regulations label are output synchronously. A first loss function is constructed by minimizing the deviation between the first label and the predicted value of the danger indicator; a second loss function is constructed by minimizing the deviation between the second label and the predicted value of the risk level indicator; a third loss function is constructed by minimizing the deviation between the third label and the predicted value of the violation of security check regulations; and a joint loss function is constructed by weighted summation of the first loss, the second loss, and the third loss. The initial multimodal fusion analysis model is updated by minimizing the joint loss using the gradient descent algorithm to obtain the multimodal fusion analysis model.

5. The railway and subway security check sharing method according to claim 4, characterized in that, The first loss function uses binary cross-entropy loss; the second loss function uses multi-class cross-entropy loss; and the third loss function uses multi-label cross-entropy loss.

6. The railway and subway security check sharing method according to claim 1, characterized in that, The determination of whether a passenger's journey within a subway and railway hub constitutes an uninterrupted transfer within a closed station includes: Based on the time period and passenger type when the verification request occurs, select the corresponding travel time reference value and transfer time reference value from multiple pre-stored transfer time reference values; verify whether the difference between the second timestamp and the first timestamp is within the first time interval formed by the travel time determined according to the subway train timetable, the travel time reference value, and the first buffer margin; verify whether the difference between the third timestamp and the second timestamp is within the second time interval formed by the transfer time reference value and the second buffer margin.

7. The railway and subway security check sharing method according to claim 6, characterized in that, The set of benchmark values ​​is constructed in the following manner: Samples of passengers’ effective walking time within the station were collected by time period and by passenger type. For each time period and passenger type combination, after removing samples with abnormal stops, the arithmetic mean of the remaining valid walking time samples is calculated as the baseline value for that combination.

8. The railway and subway security check sharing method according to claim 1, characterized in that, The multimodal security inspection data packet also includes at least one of the following: Passenger video frames captured by surveillance cameras; facial feature image data captured by facial recognition devices; passenger historical security check records associated with the identity information; real-time passenger flow data at the subway security checkpoint; and subway line information where the subway security checkpoint is located.

9. A shared security check system for railways and subways, characterized in that, The system includes: The data collection subsystem installed at the subway security checkpoint includes an X-ray security inspection machine, an identity recognition device, and a turnstile, and is used to collect multimodal security inspection data packets from passengers. The verification subsystem installed at railway stations includes an identity verification terminal and a gate control unit; A central processing subsystem, comprising at least one server configured with a processor and a memory; and a communication network connecting the data acquisition subsystem, the verification subsystem, and the central processing subsystem; wherein the data acquisition subsystem, the verification subsystem, and the central processing subsystem are configured to perform the method as described in any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program or instructions stored thereon, characterized in that, When the computer program or instructions are executed by a processor, they implement the steps of the method as described in any one of claims 1 to 8.

Images