Airport safety management method and system

Abstract

The invention relates to the technical field of airport safety management, in particular to an airport safety management method and system, and the method comprises the steps: obtaining various data of airport operation in real time through a multi-source heterogeneous data collection module, carrying out the feature fusion processing through a dynamic risk assessment model, generating a three-dimensional risk thermodynamic diagram, and outputting a risk level and a type identifier; triggering a hierarchical response mechanism, generating a resource configuration scheme through a resource scheduling optimization algorithm, and sending a scheduling instruction; in the emergency disposal stage, a multi-terminal collaborative platform is used for monitoring the disposal effect, when the risk diffusion coefficient exceeds a preset threshold value, a cross-department collaborative disposal protocol is started, and the resource allocation weight is dynamically adjusted. According to the method, the risk detection precision is improved through multi-source data fusion and space-time correlation analysis, multi-target optimization configuration is realized by adopting a dynamic resource scheduling matrix, a closed-loop mechanism of hierarchical response and cross-department co-processing is constructed, the traceability of the process is enhanced in combination with digital twin and block chain technologies, and the airport safety management level is effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of airport security management, and in particular to an airport security management method and system. Background Art

[0002] With the rapid development of the global aviation industry, airports, as the core hub of air transportation, are becoming more and more important in terms of safety management. Airport operations are currently facing a complex and ever-changing risk environment, and traditional safety management methods are unable to meet actual needs in many aspects, as shown below:

[0003] Limitations of traditional monitoring methods: In the past, airports mainly relied on a single type of data or independent monitoring systems for risk detection, which was unable to capture all potential risks in airport operations. For example, single video surveillance and security equipment detection could not fully cover all types of risk scenarios.

[0004] Lack of multi-source data integration: Each monitoring system is independent of each other, and the data is not effectively integrated, which makes it impossible to give full play to the complementary role of different types of data. For example, there is no correlation analysis between flight dynamic data, security inspection data, and personnel location data, which affects the accurate assessment and response of risks.

[0005] Disadvantages of the fixed plan model: Traditional resource scheduling is based on fixed plans, which are formulated based on past experience and typical scenarios. It is difficult to cope with complex and changeable risk events in actual airport operations, and it is easy to lead to unreasonable resource allocation;

[0006] Insufficient real-time situation response: lack of ability to dynamically adjust resource allocation based on real-time situation. Airport operation risk conditions change rapidly, and traditional scheduling models cannot perceive and respond in time, resulting in a disconnect between resource scheduling and actual needs.

[0007] The problem of information islands is serious: during emergency response, there are information barriers between airport departments, independent information systems are used, data formats and standards are not unified, and information is difficult to share and interact;

[0008] Imperfect coordinated command system: There is a lack of an efficient coordinated command system, and the responsibilities and collaboration processes of various departments in emergency response are unclear. When faced with complex risk events, uncoordinated actions, duplication of work or buck-passing are likely to occur.

[0009] Therefore, in view of the above problems, an airport safety management method and system are proposed. Summary of the invention

[0010] The purpose of the present invention is to provide an airport security management method and system to solve the problems raised in the above background technology.

[0011] To achieve the above object, the present invention provides the following technical solutions:

[0012] An airport safety management method comprises the following steps:

[0013] S1. Real-time acquisition of airport operation data through multi-source heterogeneous data acquisition modules, including: video surveillance data, millimeter-wave radar data, security equipment detection data, flight dynamic data, personnel location tag data and environmental sensor data;

[0014] S2. Establish a dynamic risk assessment model, perform feature fusion processing on the input data, generate a three-dimensional risk heat map and output risk level and type identification; the risk types at least include the risk of carrying prohibited items, the risk of gathering of people, the risk of equipment abnormality, the risk of flight conflict and the risk of sudden environment;

[0015] S3. Triggering a hierarchical response mechanism based on risk level and type, generating a resource allocation plan through a resource scheduling optimization algorithm, and sending scheduling instructions to security inspection channels, ground handling vehicles, emergency response units, and medical rescue units;

[0016] S4. During the emergency response phase, the multi-terminal collaborative platform monitors the response effect in real time. When the risk diffusion coefficient exceeds the preset threshold, the cross-departmental collaborative response agreement is initiated and the resource allocation weight is dynamically adjusted;

[0017] S5. Generate a disposal process traceability report, update the risk assessment model parameters, and store the optimized scheduling strategy in the strategy knowledge base.

[0018] As a preferred solution, the data acquisition module in step S1 includes:

[0019] Multispectral camera arrays deployed at key points in the terminal to collect data on personnel behavior and baggage status;

[0020] A distributed meteorological sensor group deployed in the runway area monitors visibility, wind speed and pavement friction coefficient in real time;

[0021] The Beidou positioning terminal integrated in the shuttle bus and service vehicles obtains real-time location and operation trajectory data.

[0022] As a preferred solution, in step S2, the dynamic risk assessment model adopts:

[0023] The modified spatiotemporal convolutional neural network M-STCNN processes video stream data and extracts abnormal behavior feature vectors; assuming that the video stream data is ,In M-STCNN, after the convolution layer operation, the feature map The calculation formula is: ,in, Represents a location at a specific spatial and temporal index The feature map elements at Respectively represent the index of the feature map in the spatial and temporal dimensions, is the size of the convolution kernel, is the index of the convolution kernel in the first direction of the spatial dimension, is the index of the convolution kernel in the second direction of the spatial dimension, is the index of the convolution kernel in the time dimension, Represents the input video stream data exist The value at position, is the convolution kernel weight, is the bias term; the temporal and spatial features in the video stream data are extracted through this formula, and then the abnormal behavior feature vector is obtained;

[0024] The modified graph neural network M-GNN constructs the topological graph of personnel-equipment-environment and identifies the potential risk propagation path; ,in, is a collection of nodes, is a set of edges; nodes The eigenvector of , in M-GNN, the node The updated feature vector The calculation formula is: ,in, Is a node The set of neighbor nodes of Is a node and The edge weights between is the weight matrix, is a nonlinear activation function, is the bias term; Representation Node The node feature vector is updated through this formula to identify the potential risk propagation path;

[0025] Based on the fuzzy logic decision engine, the global risk coefficient is calculated by integrating the risk indicators of each dimension. Suppose the risk indicator set of each dimension is , the weight of each risk indicator is , global risk factor The calculation formula is: ,in, Represents the number of risk indicators, and the global risk coefficient is calculated through this formula.

[0026] As a preferred solution, in step S3, the resource scheduling optimization algorithm includes:

[0027] Establish a multi-objective optimization function, taking into account response timeliness, resource utilization and disposal cost; suppose response timeliness is , the resource utilization is The disposal cost is , multi-objective optimization function for: ,in, , , is the weight coefficient, which is used to balance the relationship between different objectives, and ;

[0028] The improved genetic algorithm M-GA is used to solve the optimal scheduling scheme, in which the fitness function includes the risk attenuation rate, resource coverage and collaborative operation efficiency index; the risk attenuation rate is set to , the resource coverage is , the collaborative work efficiency index is , the fitness function Fitness is: ,in, , , is the weight coefficient, which is used to adjust the importance of different indicators in the fitness function, and ;

[0029] The output includes a three-dimensional scheduling matrix of personnel grouping, equipment configuration, and path planning.

[0030] As a preferred solution, in step S4, the cross-departmental collaborative disposal agreement includes:

[0031] Build a blockchain-based emergency command chain to record the response time and disposal effect of each unit;

[0032] When a runway incursion is detected, an augmented reality situation map is simultaneously pushed to the air traffic control tower, ground control center and security department;

[0033] In medical rescue scenarios, a rescue plan is automatically generated that includes the optimal path, casualty classification, and material requirements.

[0034] As a preferred solution, in step S5, the disposal process traceability report includes:

[0035] Spatiotemporal evolution map of risk events;

[0036] Resource scheduling response delay analysis;

[0037] Disposal strategy effectiveness assessment matrix;

[0038] Model parameter adaptive adjustment records.

[0039] As a preferred solution, abnormal behavior feature vector extraction adopts:

[0040] Abnormal posture recognition algorithm based on skeleton key point detection;

[0041] The optical flow method is used to analyze the conflict probability of personnel movement trajectories;

[0042] The detection sensitivity of suspicious objects in hidden areas is enhanced through the attention mechanism.

[0043] As a preferred solution, the process of generating the three-dimensional scheduling matrix includes:

[0044] Divide the airport area into honeycomb grid cells and calculate the resource demand density of each grid;

[0045] Conduct scheduling simulation based on the digital twin platform;

[0046] Introduce an elastic resource pool mechanism to dynamically allocate resources to meet sudden demands.

[0047] As a preferred solution, it also includes:

[0048] Implement a graded inspection strategy in the passenger security check process and dynamically adjust the inspection intensity of the security check channel based on the risk level;

[0049] Establish equipment health prediction models to trigger maintenance warnings and adjust spare parts inventory in advance;

[0050] Use the digital twin system to simulate contingency plans for major risk scenarios.

[0051] An airport security management system, which is equipped with an airport security management method to manage airport security.

[0052] It can be seen from the technical solutions provided by the present invention that the airport safety management method and system provided by the present invention have the following beneficial effects:

[0053] Multi-source data fusion: Through the multi-source heterogeneous data acquisition module, multiple types of data such as video surveillance, millimeter-wave radar, security equipment detection, flight dynamics, personnel location tags and environmental sensors are obtained, and feature fusion processing is performed; different types of data reflect the airport operation status from different dimensions and complement each other, changing the limitations of traditional single-dimensional risk monitoring; for example, video surveillance data can intuitively present personnel behavior, and millimeter-wave radar data can accurately monitor the movement trajectory of objects. The fusion of the two can more comprehensively and accurately identify potential risks;

[0054] Spatiotemporal correlation analysis: Use spatiotemporal convolutional neural networks to process video stream data and extract abnormal behavior feature vectors; use graph neural networks to build a correlation topology of personnel, equipment, and environment to identify potential risk transmission paths; this spatiotemporal correlation analysis of data fully considers the evolution of events in time and space dimensions, greatly improves the accuracy and timeliness of risk detection, and can detect potential risks in advance;

[0055] Multi-objective optimization configuration: Adopt resource scheduling optimization algorithm, establish multi-objective optimization function, and consider response timeliness, resource utilization and disposal cost at the same time; change the traditional fixed plan scheduling mode, dynamically generate resource allocation plan according to real-time situation, and realize the optimal allocation of resources; for example, when facing different types and levels of risks, it can quickly allocate the appropriate number and type of security inspection channels, ground handling vehicles, emergency disposal units and medical rescue units and other resources, which can not only ensure rapid response, but also improve resource utilization efficiency and reduce disposal costs;

[0056] Dynamic resource scheduling matrix: Generate a three-dimensional scheduling matrix including personnel grouping, equipment configuration, and path planning to make resource scheduling more scientific and accurate; by dividing the airport area into honeycomb grid units, calculating the resource demand density of each grid, and dynamically allocating resources based on simulation and deduction on the digital twin platform and combined with the elastic resource pool mechanism, it can flexibly adjust resource allocation according to real-time changes in risks and efficiently respond to various complex situations;

[0057] Graded response mechanism: The graded response mechanism is triggered based on the risk level and type, and different response measures of different intensity are taken for different risk levels to ensure that a quick and effective response can be made when risks occur, and resources are allocated reasonably to avoid over-response or under-response;

[0058] Cross-departmental collaborative disposal: Building a closed-loop mechanism of hierarchical response and cross-departmental collaborative disposal solves the problem of information islands in emergency disposal; by building a blockchain-based emergency command chain, ensure that the response time and disposal effect of each unit can be recorded and traced; in different risk scenarios, such as runway intrusion incidents and medical rescue scenarios, achieve real-time information sharing and collaborative operations among relevant departments, and improve the efficiency and success rate of emergency disposal. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 The figure is a flow chart of an airport safety management method according to the present invention. DETAILED DESCRIPTION

[0060] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0061] In order to better understand the above technical solution, the above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0062] like Figure 1 As shown, an embodiment of the present invention provides an airport security management method, comprising the following steps:

[0063] S1. Real-time acquisition of airport operation data through multi-source heterogeneous data acquisition modules, including: video surveillance data, millimeter-wave radar data, security equipment detection data, flight dynamic data, personnel location tag data and environmental sensor data;

[0064] S2. Establish a dynamic risk assessment model, perform feature fusion processing on the input data, generate a three-dimensional risk heat map and output risk level and type identification; the risk types at least include the risk of carrying prohibited items, the risk of gathering of people, the risk of equipment abnormality, the risk of flight conflict and the risk of sudden environment;

[0065] The dynamic risk assessment model uses:

[0066] The modified spatiotemporal convolutional neural network M-STCNN processes video stream data and extracts abnormal behavior feature vectors; assuming that the video stream data is ,In M-STCNN, after the convolution layer operation, the feature map The calculation formula is: ,in, Represents a location at a specific spatial and temporal index The feature map elements at Respectively represent the index of the feature map in the spatial and temporal dimensions, is the size of the convolution kernel, is the index of the convolution kernel in the first direction of the spatial dimension, is the index of the convolution kernel in the second direction of the spatial dimension, is the index of the convolution kernel in the time dimension, Represents the input video stream data exist The value at position, is the convolution kernel weight, is the bias term; the temporal and spatial features in the video stream data are extracted through this formula, and then the abnormal behavior feature vector is obtained;

[0067] The modified graph neural network M-GNN constructs the topological graph of personnel-equipment-environment and identifies the potential risk propagation path; ,in, is a collection of nodes, is a set of edges; nodes The eigenvector of , in M-GNN, the node The updated feature vector The calculation formula is: ,in, Is a node The set of neighbor nodes of Is a node and The edge weights between is the weight matrix, is a nonlinear activation function, is the bias term; Representation Node The node feature vector is updated through this formula to identify the potential risk propagation path;

[0068] Based on the fuzzy logic decision engine, the global risk coefficient is calculated by integrating the risk indicators of each dimension. Suppose the risk indicator set of each dimension is , the weight of each risk indicator is , global risk factor The calculation formula is: ,in, Represents the number of risk indicators, and the global risk coefficient is calculated through this formula;

[0069] S3. Triggering a hierarchical response mechanism based on risk level and type, generating a resource allocation plan through a resource scheduling optimization algorithm, and sending scheduling instructions to security inspection channels, ground handling vehicles, emergency response units, and medical rescue units;

[0070] Resource scheduling optimization algorithms include:

[0071] Establish a multi-objective optimization function, taking into account response timeliness, resource utilization and disposal cost; suppose response timeliness is , the resource utilization is The disposal cost is , multi-objective optimization function for: ,in, , , is the weight coefficient, which is used to balance the relationship between different objectives, and ;

[0072] The improved genetic algorithm M-GA is used to solve the optimal scheduling scheme, in which the fitness function includes the risk attenuation rate, resource coverage and collaborative operation efficiency index; the risk attenuation rate is set to , the resource coverage is , the collaborative work efficiency index is , the fitness function Fitness is: ,in, , , is the weight coefficient, which is used to adjust the importance of different indicators in the fitness function, and ;

[0073] Output a three-dimensional scheduling matrix including personnel grouping, equipment configuration, and path planning;

[0074] S4. During the emergency response phase, the multi-terminal collaborative platform monitors the response effect in real time. When the risk diffusion coefficient exceeds the preset threshold, the cross-departmental collaborative response agreement is initiated and the resource allocation weight is dynamically adjusted;

[0075] S5. Generate a disposal process traceability report, update the risk assessment model parameters, and store the optimized scheduling strategy in the strategy knowledge base.

[0076] In this embodiment, the detailed steps of step S1 are:

[0077] Step S1-1: Data collection module planning and deployment

[0078] Determine collection requirements:

[0079] According to the overall goal of airport security management and the specific needs of each business link, the type of data to be collected should be clarified; for example, to ensure passenger safety and security inspection efficiency, security inspection equipment detection data needs to be collected; to monitor the airport operation status, flight dynamic data, personnel location tag data, and environmental sensor data need to be obtained;

[0080] Analyze the role of different data in risk assessment, resource scheduling and emergency response, and provide guidance for subsequent data collection and processing;

[0081] Planning the collection location:

[0082] Based on the video surveillance data, the installation locations of the multispectral camera arrays should be reasonably planned at the key nodes of the flow of people and luggage, such as the terminal entrance, security inspection area, waiting hall, and boarding gate; ensure that the cameras can cover key areas without blind spots, and can clearly capture the behavioral characteristics of people and the status data of luggage; for example, in the security inspection area, the cameras should be able to clearly capture the inspection process of passengers' luggage, the operations of security personnel, and the relevant behaviors of passengers;

[0083] For millimeter-wave radar data, millimeter-wave radars are rationally deployed around the runway, apron and other areas to accurately monitor the position, speed, attitude and other information of the aircraft during takeoff and landing, as well as the movement of surrounding vehicles and personnel;

[0084] The collection of security inspection equipment detection data requires that the security inspection equipment (such as X-ray security inspection machines, metal detectors, etc.) be installed at appropriate locations in the passenger security inspection channel to ensure that they can effectively detect luggage and items carried by passengers;

[0085] Flight dynamic data is mainly obtained from the airport's flight information management system, which is connected to the system through a network interface to obtain real-time flight take-off and landing times, flight status, route information and other data;

[0086] To collect data from personnel location tags, it is necessary to equip staff and passengers in specific areas of the airport with location tags, and deploy signal receiving base stations in the corresponding areas. The layout of the base stations must ensure that the real-time location information of personnel can be accurately obtained, such as reasonably distributing base stations in important areas such as each floor of the terminal building, runway area, and parking lot.

[0087] In terms of environmental sensor data collection, in the runway area, according to the meteorological monitoring needs and the characteristics of the runway structure, a distributed meteorological sensor group is reasonably arranged to monitor meteorological parameters such as visibility, wind speed and pavement friction coefficient in real time; in the terminal building, environmental sensors such as temperature and humidity sensors and air quality sensors are installed to monitor indoor environmental conditions;

[0088] Deploy collection equipment:

[0089] Install and debug the multispectral camera array according to the planned location; ensure that the camera's installation angle, height, focal length and other parameters are set reasonably to obtain clear and complete image data; at the same time, configure the camera network so that it can transmit the collected video data to the data processing center in real time;

[0090] Deploy millimeter-wave radar equipment and ensure that it is firmly installed and can adapt to the complex environmental conditions of the airport, such as strong winds and dust; calibrate radar parameters, including distance resolution and speed resolution, to improve the accuracy of radar data; connect radar equipment to the data transmission network to achieve real-time data upload;

[0091] Install and debug security inspection equipment to ensure that its detection accuracy and stability meet relevant standards; integrate security inspection equipment with data acquisition systems to achieve automatic collection and transmission of detection data;

[0092] Establish a data interface with the flight information management system to ensure stable and real-time acquisition of flight dynamic data through a secure network connection; conduct regular testing and maintenance of the data interface to prevent data transmission interruptions or errors;

[0093] Distribute and activate personnel positioning tags, and debug and optimize signal receiving base stations; ensure that the signal coverage of the base stations meets the design requirements, can accurately receive the signals sent by the positioning tags, and transmit personnel location data to the data processing system;

[0094] Install distributed meteorological sensor groups and environmental sensors in the terminal building, ensuring that the sensor installation locations meet meteorological monitoring specifications and environmental monitoring requirements; calibrate and initialize the sensors to ensure the accuracy and reliability of the collected data; connect the sensors to the data transmission network via wired or wireless means to achieve real-time collection and transmission of environmental data;

[0095] Step S1-2: Data collection and preliminary processing

[0096] Video surveillance data collection and processing:

[0097] The multispectral camera array collects video at a preset frame rate and resolution. At the same time, the frame rate and resolution can be adjusted dynamically according to the needs of different scenarios. For example, in crowded areas or when important events occur, the frame rate can be increased to capture more detailed behavioral information.

[0098] Real-time preprocessing of the collected video data, including image enhancement to improve image clarity and readability; noise reduction, using methods based on wavelet transform or Gaussian filtering to remove noise interference in the video and improve image quality;

[0099] Millimeter wave radar data acquisition and processing:

[0100] The millimeter-wave radar scans the target area at a set frequency and collects information such as the distance, speed, and angle of the target object in real time;

[0101] Filter the collected radar echo data, using methods such as Kalman filter or Wiener filter to remove noise signals caused by environmental interference and improve data accuracy;

[0102] Convert the polar coordinate data collected by the radar into Cartesian coordinate data for fusion processing with other types of data; at the same time, classify and identify the targets detected by the radar, distinguish different types of targets such as people, vehicles, and aircraft, and mark their relevant attribute information;

[0103] Security inspection equipment detection data collection and processing:

[0104] During the passenger security check, security inspection equipment collects detection data in real time; for example, X-ray security inspection machines collect image data of items inside luggage, and metal detectors detect information about metal items carried by people or luggage;

[0105] Perform preliminary analysis and processing on the collected security inspection data; for X-ray image data, use image recognition algorithms (such as deep learning-based target detection algorithms) to identify the outline and category of items in the luggage, determine whether there are prohibited items, and mark the location and type of prohibited items; for metal detector data, record the location, strength and other information of the detected metal objects;

[0106] The processed security inspection data is formatted and stored for subsequent data transmission and analysis; at the same time, the security inspection data is encrypted to ensure passenger privacy and data security;

[0107] Flight dynamic data collection and processing:

[0108] Through the real-time data interface established with the flight information management system, flight dynamic data is obtained regularly; the data obtained includes the flight's planned take-off and landing time, actual take-off and landing time, flight status (normal, delayed, cancelled, etc.), flight number, route information, etc.;

[0109] Clean and verify the acquired flight dynamic data to remove duplicate, erroneous or invalid data records; for example, check whether the flight time format is correct and whether the flight status is consistent with the actual situation;

[0110] The processed flight dynamic data is stored in a certain format (such as JSON or XML format) to facilitate subsequent data query and analysis; at the same time, the flight dynamic data is associated with other data (such as personnel location data, equipment status data, etc.) to provide more comprehensive information support for airport safety management;

[0111] Personnel positioning tag data collection and processing:

[0112] The personnel positioning tag sends a location signal to the signal receiving base station at a set time interval (such as once per second or once every 5 seconds); after receiving the signal, the base station calculates the personnel's real-time location coordinates (such as longitude, latitude, floor, etc.) through a positioning algorithm (such as a positioning algorithm based on RSSI, TOA or TDOA);

[0113] Smooth the collected personnel location data to remove position jumps caused by signal fluctuations or interference; use methods such as moving average filtering or Kalman filtering to improve the stability and accuracy of location data;

[0114] Associate personnel location data with personnel identity information, and query the corresponding personnel identity (such as passengers, staff, security personnel, etc.), department, job position, etc. in the personnel information database by reading the ID information of the positioning tag;

[0115] The processed personnel location tag data is stored and transmitted to the data processing center in real time to provide real-time location information support for personnel behavior analysis, resource scheduling, etc.

[0116] Environmental sensor data acquisition and processing:

[0117] The distributed meteorological sensor group collects real-time meteorological parameters such as visibility, wind speed, wind direction, and road surface friction coefficient in the runway area. The environmental sensors in the terminal building monitor environmental data such as temperature and humidity, air quality (such as PM2.5, PM10, carbon dioxide concentration, etc.) in real time;

[0118] Calibrate and correct the collected environmental sensor data, and adjust the original data according to the characteristics of the sensor and calibration parameters to improve the accuracy of the data;

[0119] Detect and process outliers in environmental data, using statistical analysis methods (such as the 3σ criterion) or machine learning algorithms (such as the IsolationForest algorithm) to identify and remove abnormal data points;

[0120] The processed environmental sensor data is stored, and summarized and counted at certain time intervals (such as every minute or every 5 minutes) to generate environmental data reports to provide data support for airport environmental monitoring and risk assessment;

[0121] Step S1-3: Data transfer and integration

[0122] Data transmission:

[0123] Video surveillance data is transmitted to the data processing center through a wired network (such as fiber-optic Ethernet) or a wireless network (such as 5G, Wi-Fi). For video stream data with high real-time requirements, efficient transmission protocols (such as RTSP, RTMP) are used to ensure smooth transmission of video data and reduce delays and packet loss.

[0124] Millimeter-wave radar data, security equipment detection data, personnel positioning tag data, and environmental sensor data are transmitted to the data processing center via wired or wireless methods. For radar data and positioning tag data with small data volume but high real-time requirements, low-power, high-reliability wireless transmission protocols (such as LoRa, ZigBee) can be used. For security equipment detection data and environmental sensor data, wired or wireless networks can be selected for transmission according to actual conditions.

[0125] Flight dynamic data is transmitted from the flight information management system to the data processing center through a dedicated data interface and network channel. To ensure the security and stability of data transmission, data is encrypted using encryption transmission protocols (such as SSL / TLS) to prevent data leakage and tampering.

[0126] Data Integration:

[0127] In the data processing center, a data receiving module is set up to receive data from different data sources; the received data is classified and identified, and marked according to the data type (such as video data, radar data, security inspection data, etc.) and source (such as specific cameras, radar equipment, security inspection channels, etc.) to facilitate subsequent data management and processing;

[0128] The received data is stored in the corresponding database or data storage system; for structured data (such as flight dynamic data, personnel location tag data, environmental sensor data, etc.), it is stored in a relational database (such as MySQL, Oracle) for efficient query and analysis; for unstructured data (such as video surveillance data, X-ray image data, etc.), it is stored in a distributed file system (such as Ceph, GlusterFS) or object storage system (such as AmazonS3, Alibaba Cloud OSS), and indexes are established to associate with relevant metadata information to facilitate data retrieval and call;

[0129] Establish a data integration mechanism to associate and fuse different types of data; match and integrate data from different data sources through key attributes such as data timestamp, geographic location information, and related identifiers; for example, associate video surveillance data, personnel location tag data, and environmental sensor data in the same time and area to provide a basis for subsequent multi-source data fusion analysis.

[0130] In this embodiment, the specific operation steps of step S2 include:

[0131] Step S2-1: Data preprocessing

[0132] Video surveillance data: The video surveillance data obtained from step S1 is firstly processed in a unified format, and the various formats of videos collected by different multispectral cameras are converted into a standard format that is easy to process, such as the H.264 encoding format; then, the video is subjected to noise reduction processing, and a noise reduction algorithm based on wavelet transform is used to remove noise interference in the video and improve the image quality; then, the video frame is cropped and normalized, and the video frame is cropped according to the area of ​​concern (such as the security inspection channel, the waiting area, etc.), and the image size is normalized to a specific size for subsequent neural network processing;

[0133] Millimeter-wave radar data: Filter the millimeter-wave radar data to remove clutter signals caused by environmental interference, and use the Kalman filter to process the radar echo data to improve the accuracy of target detection. Then, convert the radar data from polar coordinates to Cartesian coordinates to facilitate fusion with other data. At the same time, classify the targets detected by the radar to distinguish whether they are people, vehicles or other objects, and mark their location, speed, direction of movement and other information.

[0134] Security inspection equipment detection data: Analyze various types of data detected by security inspection equipment, such as image data from X-ray security inspection equipment, identify the outline and category of items in luggage through image recognition algorithms, and determine whether there are prohibited items; for metal detector data, record the location and strength information of detected metal items; organize these data, extract key features, such as the type and location of prohibited items, and perform standardization processing so that data from different security inspection equipment are on the same scale for easy subsequent fusion;

[0135] Flight dynamic data: Obtain flight dynamic data from the flight information system, including flight take-off and landing times, estimated arrival / departure times, and flight status (normal, delayed, canceled, etc.); cleanse this data to remove duplicate or erroneous information, such as inconsistent flight status updates, and make corrections based on authoritative data sources; at the same time, calibrate the flight dynamic data with the airport's time system to ensure the time accuracy of the data;

[0136] Personnel positioning tag data: The real-time location data sent by the personnel positioning tag is solved, and the precise location coordinates of the personnel in the airport are determined based on the positioning technology of the tag (such as RFID, UWB, etc.); the location data is smoothed to remove the position jump caused by signal fluctuations, and the continuous location data is processed using a moving average filter algorithm; at the same time, the personnel identity information is associated with the location data, such as matching the tag ID with the personnel information database to obtain the identity category of the personnel (passenger, staff, etc.);

[0137] Environmental sensor data: Integrate data from the distributed meteorological sensor group in the runway area and environmental sensors in the terminal building; perform outlier detection and correction on meteorological data, such as visibility, wind speed, and road surface friction coefficient. For example, if the wind speed data deviates significantly from the historical mean and does not conform to the current weather conditions, use interpolation to correct it; perform normalization on environmental sensor data in the terminal building, such as temperature, humidity, and air quality, and convert data from different sensors to the same value range for easy comprehensive analysis;

[0138] Step S2-2: Feature extraction

[0139] Processing video stream data based on modified spatiotemporal convolutional neural network (M-STCNN): The preprocessed video stream data is input into M-STCNN; first, after passing through multiple convolutional layers, according to the formula Convolution operation, where is the input video stream data. Represents a location at a specific spatial and temporal index The feature map elements at Respectively represent the index of the feature map in the spatial and temporal dimensions, is the size of the convolution kernel, is the index of the convolution kernel in the first direction of the spatial dimension, is the index of the convolution kernel in the second direction of the spatial dimension, is the index of the convolution kernel in the time dimension, Represents the input video stream data exist The value at position, is the convolution kernel weight, is a bias term; the spatiotemporal features in the video, such as the actions and behavior patterns of people, are extracted through convolution operations; then, the feature map is subjected to dimensionality reduction processing through the pooling layer to reduce the amount of data while retaining key features; finally, the extracted features are converted into abnormal behavior feature vectors through the fully connected layer, which contain the difference characteristics between the behavior of people in the video and the normal behavior pattern, and are used for subsequent risk assessment;

[0140] Construct an association topology graph based on a modified graph neural network (M-GNN): different entities such as personnel, equipment, and environment are used as nodes, and the relationships between them are used as edges to construct a graph structure ,in, is a collection of nodes, is an edge collection, for each node , assign its corresponding eigenvector For example, the feature vector of a person node can include the person's identity information, location information, etc., and the feature vector of a device node can include the device type, operating status, etc. In M-GNN, according to the formula Update node feature vector, where Representation Node The updated feature vector, Is a node The set of neighbor nodes of Is a node and The edge weights between is the weight matrix, is a nonlinear activation function, is the bias term; Representation Node By updating the node feature vectors through multiple iterations, the potential risk propagation path can be identified;

[0141] Other data feature extraction: For security inspection equipment detection data, extract the type, quantity, frequency of occurrence and other features of prohibited items; for flight dynamic data, extract flight delay duration, number of flight conflicts and other features; for personnel location tag data, extract personnel gathering density, personnel flow speed in a specific area and other features; for environmental sensor data, extract meteorological condition change trends, abnormal fluctuations in environmental parameters and other features;

[0142] Step S2-3: Feature Fusion

[0143] Selection of data fusion method: A feature fusion method based on weighted average is used to assign corresponding weights to each feature vector according to the importance of different types of data in risk assessment. For example, feature vectors of security equipment detection data and video surveillance data that are directly related to safety are given higher weights. For flight dynamic data and environmental sensor data, the weights are reasonably adjusted according to specific scenarios and risk types.

[0144] Fusion process: The feature vectors extracted from different data sources are fused in a weighted average manner; let the feature vector from the video surveillance data be , the eigenvector of millimeter wave radar data is , the feature vector of the detection data of the security inspection equipment is , the eigenvector of flight dynamic data is , the eigenvector of the personnel positioning label data is , the eigenvector of environmental sensor data is , the corresponding weights are , , , , , The fused eigenvector The calculation formula is: , in, ,Through feature fusion, the features of multi-source heterogeneous data are integrated into a comprehensive feature vector, which can more comprehensively reflect the ,operation status of the airport;

[0145] Step S2-4: Risk Assessment and Output

[0146] The decision engine based on fuzzy logic calculates the global risk coefficient: the fused feature vector is input into the decision engine based on fuzzy logic; the risk index set of each dimension is ,These risk indicators are further extracted from the fused feature vector and are ,parameters related to different risk types, e.g. It can indicate the risk indicator of contraband carrying, Represents the risk index of personnel gathering, etc. The weight of each risk index is , determine the weight value according to the importance and relevance of the risk type; according to the formula Calculate the global risk coefficient ,in, Indicates the number of risk indicators;

[0147] Risk level and type identification: Based on the calculated global risk coefficient , determine the risk level according to the preset risk level classification standards, for example, when The risk level is between 0-0.3, the risk level is between 0.3-0.6, and the risk level is between 0.6-1. The risk level is high. At the same time, according to the contribution of the risk indicators of each dimension in the fusion feature vector, the risk type identification is determined, such as the risk of carrying prohibited items, the risk of gathering of people, the risk of equipment abnormality, the risk of flight conflict, and the risk of sudden environment.

[0148] Generate a three-dimensional risk heat map: Use the spatial position (two-dimensional plane) and time of the airport area as three-dimensional coordinates, map the risk level and type information into the three-dimensional space, and generate a three-dimensional risk heat map; in the heat map, different colors and brightness represent different risk levels, and use animation to display the changes in risk in the time dimension, intuitively presenting the risk distribution and evolution trend in different areas of the airport at different times; finally output the risk level, type identification, and three-dimensional risk heat map to provide a decision-making basis for the subsequent graded response mechanism.

[0149] In this embodiment, the specific operation steps of step S3 are as follows:

[0150] Step S3-1: Risk Level and Type Judgment

[0151] Receive the risk level and type identification output from step S2; clarify the risk level, such as low risk, medium risk, and high risk; determine the risk type, such as the risk of carrying prohibited items, the risk of gathering of people, the risk of equipment abnormality, the risk of flight conflict, and the risk of sudden environmental events; this step is the basis for the subsequent triggering of the hierarchical response mechanism and the generation of resource allocation plans, ensuring that the system can make targeted decisions based on different risk conditions;

[0152] Step S3-2: Trigger the hierarchical response mechanism

[0153] Low-risk response: If it is determined to be a low-risk level, basic response measures will be initiated; for example, in response to the risk of carrying prohibited items, the frequency of random inspections of security checkpoints will be appropriately increased; for the risk of crowd gathering, a small number of security personnel will be arranged to go to the gathering area for guidance, and the public address system will be used to remind people to keep a proper distance; in terms of equipment abnormality risks, equipment maintenance personnel will be notified to pay attention to the status of related equipment and conduct preliminary inspections; for flight conflict risks, the air traffic control department will strengthen the monitoring of flight dynamics and adjust the flight queue order in a timely manner; for sudden environmental risks, such as mild weather anomalies, environmental warning information will be issued to remind relevant operating personnel to pay attention to safety;

[0154] Medium-risk response: When the risk level is medium, more active response measures will be taken; for the risk of carrying prohibited items, the inspection intensity of security checkpoints will be comprehensively strengthened, security personnel will be increased, and detailed inspections will be carried out on passenger luggage of key flights or areas; for the risk of crowd gathering, more security forces will be deployed to the scene, isolation areas will be set up, and people will be guided to evacuate in an orderly manner; in terms of equipment abnormality risks, the equipment emergency maintenance process will be immediately initiated, a professional maintenance team will be arranged for emergency repairs, and spare equipment will be called to ensure that airport operations are not greatly affected; for the risk of flight conflicts, the air traffic control department will coordinate with the relevant flight crews and airlines to formulate detailed flight adjustment plans, which may include flight delays, route changes and other measures; for sudden environmental risks, such as moderate meteorological disasters, some affected areas will be closed, and relevant operating procedures will be adjusted, such as closing some areas of the runway to deal with severe weather;

[0155] High-risk response: If it is determined to be a high-risk level, a comprehensive emergency response will be implemented; for the risk of carrying prohibited items, the relevant security inspection areas will be immediately blocked, all personnel and luggage will be fully checked, and the emergency plan will be activated at the same time, and multiple departments will be jointly dealt with; in terms of the risk of crowd gathering, a large-scale personnel evacuation plan will be activated, and all security forces and emergency rescue personnel in the airport will be coordinated to guide personnel to evacuate quickly according to the predetermined evacuation route to ensure the safety of personnel lives; for equipment abnormality risks, if a serious failure occurs in key equipment, the operation activities in the relevant area will be stopped immediately, the surrounding personnel will be evacuated, and an expert team will be organized to carry out emergency repairs. At the same time, the impact on the overall operation of the airport will be evaluated and a comprehensive recovery plan will be formulated; for the risk of flight conflict, the air traffic control department will issue emergency instructions, suspend the take-off and landing of some flights, coordinate with surrounding airports to provide temporary alternate landing services, and make every effort to ensure the safety of airport operations; for sudden environmental risks, such as severe meteorological disasters or other major environmental events, the airport will enter a full emergency state, stop most non-essential operations, and activate the airport's overall emergency shelter plan to ensure the safety of personnel and important facilities;

[0156] Step S3-3: Create a multi-objective optimization function

[0157] Response timeliness objective: Considering the importance of rapid response in different risk scenarios, response timeliness is incorporated into the optimization function; the time from risk identification to completion of resource allocation is set as In order to minimize the response time, the part of the objective function related to response timeliness is ;

[0158] Resource utilization rate target: In order to rationally utilize the limited resources of the airport, resource utilization rate is set as a target. Assume that the total amount of resources invested is , the actual amount of resources to be effectively utilized is , then the resource utilization rate is , the part about resource utilization in the objective function is ;

[0159] Disposal cost target: Consider the cost factors in the resource allocation and disposal process, such as labor costs, equipment usage costs, etc. Suppose the disposal cost is , the part of the objective function regarding the disposal cost is ;

[0160] Comprehensive multi-objective optimization function: Combining the above three objectives, construct a multi-objective optimization function , by weight coefficient , , to balance the relationship between different goals, and , multi-objective optimization function for: ;

[0161] Step S3-4: Use improved genetic algorithm to solve the optimal scheduling solution

[0162] Coding: Encode the resource scheduling problem and convert the information such as personnel grouping, equipment configuration, and path planning into chromosomes in the genetic algorithm; for example, encode the number and combination of different types of security personnel, ground handling vehicles, emergency response units, and medical rescue units, and each gene represents a specific resource allocation decision;

[0163] Initialize the population: randomly generate a set of initial chromosomes as the population. The population size is determined according to the complexity of the actual problem and the computing resources. Each chromosome represents a possible resource scheduling solution. These initial solutions form the basis for the genetic algorithm to search for the optimal solution.

[0164] Fitness function calculation: Based on the multi-objective optimization function in step S3-3 , combining the risk attenuation rate, resource coverage and collaborative operation efficiency indicators, construct the fitness function Fitness; assuming the risk attenuation rate is , the resource coverage is , the collaborative work efficiency index is , by weight coefficient , , Adjust the importance of different indicators in the fitness function, and ;The fitness function Fitness is: ;

[0165] For each chromosome (i.e., each resource scheduling scheme), calculate its fitness value. The higher the fitness value, the closer the scheme is to the optimal solution.

[0166] Selection operation: According to the fitness value, the roulette wheel selection method or tournament selection method is used to select the better chromosomes from the current population to enter the next generation population; the selection process simulates the "survival of the fittest" principle in natural selection, so that chromosomes with high fitness have a greater probability of being selected, thereby retaining the characteristics of excellent resource scheduling schemes;

[0167] Crossover operation: Perform crossover operation on selected chromosomes to simulate the gene exchange process in biological genetics; randomly select two chromosomes, exchange some genes between them, and generate new chromosomes; for example, for chromosomes with two different personnel groupings and equipment configuration plans, exchange some genes related to the deployment of security personnel and the allocation of ground handling vehicles to generate a new combination of resource scheduling plans; crossover operation helps to generate new potential excellent plans and expand the search space;

[0168] Mutation operation: Perform mutation operation on chromosomes with a certain probability, randomly changing the values ​​of certain genes in the chromosomes; for example, randomly adjusting the number of deployments of a certain emergency response unit or changing the path planning of a medical rescue unit to prevent the algorithm from falling into a local optimal solution; mutation operation introduces new gene features to increase the diversity of the population;

[0169] Iterative optimization: Repeat the selection, crossover and mutation operations, and continuously iterate and update the population until the preset termination conditions are met, such as reaching the maximum number of iterations or the fitness value converges to a certain degree; the resource scheduling plan corresponding to the optimal chromosome finally obtained is the optimal scheduling plan solved by the improved genetic algorithm;

[0170] Step S3-5: Generate a three-dimensional scheduling matrix and output scheduling instructions

[0171] Generate a three-dimensional scheduling matrix: Divide the airport area into honeycomb grid units and calculate the resource demand density of each grid. According to the optimal scheduling plan, integrate the information such as personnel grouping, equipment configuration, and path planning into the three-dimensional space to generate a three-dimensional scheduling matrix containing personnel, equipment, and path information. In this matrix, each element corresponds to a specific resource allocation decision, such as the number of security personnel, ground handling vehicle types, and driving paths that should be allocated to a grid unit at a certain point in time.

[0172] Output dispatch instructions: According to the three-dimensional dispatch matrix, dispatch instructions are sent to security check channels, ground handling vehicles, emergency response units and medical rescue units; these instructions clarify the specific tasks, action time and location of each resource unit; for example, instruct the security check channel to increase the number of security personnel and adjust the inspection process, inform ground handling vehicles to go to designated locations to perform specific tasks, and notify emergency response units and medical rescue units to arrive at the corresponding areas for emergency response according to the planned path and time; through precise dispatch instructions, effective allocation of airport resources is achieved to respond to different risk scenarios and ensure the safe operation of the airport.

[0173] Step S4-1: Build a multi-terminal collaborative platform

[0174] Hardware deployment: Deploy terminal devices with high reliability and real-time communication capabilities in key areas of the airport, such as the command center, security checkpoint, ground operation area, and medical emergency station. These terminal devices include but are not limited to industrial tablet computers, smart handheld devices, and vehicle-mounted terminals to ensure that relevant departments and personnel can obtain information and receive instructions in real time.

[0175] Software integration: Develop unified multi-terminal collaboration software with functions such as real-time data display, interactive operation, and task management; install and adapt the software to various terminal devices to achieve seamless collaboration between different types of terminal devices; at the same time, establish a stable connection between the terminal devices and the central data processing system through network communication protocols to ensure fast and accurate data transmission;

[0176] Permission setting: According to the responsibilities of different personnel and departments, corresponding operation permissions are set for each terminal device; for example, the command center personnel have the highest permissions and can view all risk information and issue global dispatch instructions; security personnel can only view risk and disposal information related to security inspection work and perform corresponding security inspection tasks; medical rescue personnel focus on information and tasks related to medical rescue; through reasonable permission setting, the security of information and the standardization of operations are guaranteed;

[0177] Step S4-2: Real-time monitoring of the treatment effect

[0178] Data collection and transmission: During the emergency response process, each terminal device collects data related to the response effect in real time; for example, the security check channel records the number of contraband seized and the execution time of the security check process through equipment; ground handling vehicles use on-board sensors to feedback the time of arrival at the designated location and the completion of the task; emergency response units and medical rescue units report on-site conditions and the progress of treatment of the wounded through the handheld devices of on-site personnel; these data are transmitted to the central data processing system in real time through the network;

[0179] Data processing and analysis: The central data processing system processes and analyzes the large amount of collected data in real time; uses data analysis algorithms to compare the actual disposal data with the preset standards and goals to evaluate the disposal effect; for example, compare the number of contraband seized with the expected target to determine whether the security inspection work has achieved the expected effect; analyze the task execution time and planned time of ground handling vehicles to evaluate their execution efficiency; and determine whether the medical rescue work is effective based on the progress of the treatment of the wounded and the standard treatment process;

[0180] Visual display: The processed and analyzed disposal effect data is displayed in an intuitive and visual way on each terminal device; the disposal progress, effect evaluation results and existing problems of each link are presented in real time through charts, graphs, texts and other forms; for example, on the large screen of the command center, the location and task execution status of each emergency disposal unit are displayed in the form of a map, and the number of contraband seized in different security inspection channels is compared with a bar chart, so that relevant personnel can quickly understand the overall situation;

[0181] Step S4-3: Calculate the risk diffusion coefficient

[0182] Determine risk diffusion indicators: Determine risk diffusion indicators by comprehensively considering multiple factors, including but not limited to the scope of personnel flow, the area affected by equipment failure, the scope of flight delays, the area affected by environmental factors, etc. For example, for personnel gathering risks, the expansion of personnel flow may mean the spread of risks; for equipment abnormality risks, if equipment failure affects other surrounding equipment or areas, it indicates that the risk has a tendency to spread;

[0183] Data collection and quantification: Through the multi-source data collection system, data related to risk diffusion indicators are collected and quantified; for example, the real-time location and movement trajectory of personnel are obtained using personnel location tag data, and the flow range of personnel within a certain period of time is calculated; the impact range and degree of equipment failure are obtained through the equipment monitoring system; the impact of flight delays on subsequent flights is counted based on flight dynamic data; the impact area of ​​sudden environmental risks is determined with the help of environmental sensor data, etc.

[0184] Calculate the risk diffusion coefficient: Use a specific mathematical model or algorithm to calculate the risk diffusion coefficient based on the quantified risk diffusion indicator data; this coefficient can comprehensively reflect the degree of risk diffusion in the current disposal process; for example, the weighted average method can be used to assign different weights to each indicator according to different risk types, and then calculate the risk diffusion coefficient. :

[0185] ,in, The number of risk spread indicators, For the The weight of the indicator, For the Quantitative value of each indicator; weight The determination of the risk factors needs to be scientifically analyzed and adjusted according to the characteristics of different risk types and actual conditions;

[0186] Step S4-4: Determine whether to initiate a cross-departmental collaborative disposal agreement

[0187] Set preset thresholds: According to historical data, risk assessment models and actual airport operations, set reasonable preset thresholds for the risk diffusion coefficient; this threshold is a key indicator for determining whether to initiate a cross-departmental collaborative disposal agreement; for example, different thresholds can be set for different types of risks, such as the risk diffusion coefficient threshold for carrying prohibited items is , The threshold of the risk diffusion coefficient of personnel aggregation is The setting of these thresholds should take into account both the timely activation of the coordination mechanism to effectively control risks and the avoidance of unnecessary waste of resources and increased coordination costs due to excessive sensitivity;

[0188] Comparison and judgment: Compare the calculated risk diffusion coefficient with the corresponding preset threshold; if the risk diffusion coefficient is less than the preset threshold, it indicates that the current emergency response measures can effectively control the risk, and there is no need to initiate the cross-departmental collaborative disposal agreement, and continue to monitor and deal with it according to the existing disposal plan and process; if the risk diffusion coefficient exceeds the preset threshold, it means that the risk has a tendency to get out of control, and it is necessary to immediately initiate the cross-departmental collaborative disposal agreement to integrate more resources, strengthen cooperation between departments, and jointly deal with risks;

[0189] Step S4-5: Initiate cross-departmental collaborative disposal agreements and dynamically adjust resource allocation weights

[0190] Constructing a blockchain-based emergency command chain: When it is determined that a cross-departmental collaborative disposal agreement needs to be initiated, the blockchain technology is used to construct an emergency command chain. On the blockchain platform, each participating department and relevant personnel serve as nodes, and key information in the emergency disposal process, such as the response time of each unit, disposal effect data, resource allocation records, etc., are recorded in an encrypted manner on the blockchain. The blockchain's immutable and traceable characteristics ensure the authenticity and credibility of the information, providing reliable data support for collaborative command.

[0191] Information sharing and situation push: In response to different types of risk events, corresponding information sharing and situation push mechanisms are initiated. For example, when a runway incursion is detected, an augmented reality (AR) situation map is simultaneously pushed to the air traffic control tower, ground control center and security department through a multi-terminal collaborative platform. The situation map can intuitively display the real-time situation of the runway, including the location of the intruding object, the runway status, the distribution of surrounding flights and vehicles, etc., so that various departments can make collaborative decisions and actions based on the same accurate information.

[0192] Generate rescue plans for specific scenarios: In medical rescue scenarios, the system automatically generates rescue plans that include the optimal route, classification of the injured, and material needs based on information such as the on-site injured situation, distribution of medical resources, and traffic conditions. It uses intelligent algorithms to plan the optimal driving route for rescue vehicles and send the injured to the appropriate medical treatment point as quickly as possible. It classifies the injured according to the severity of their injuries and allocates medical resources reasonably. At the same time, it allocates corresponding medical supplies in a timely manner based on material demand forecasts.

[0193] Dynamically adjust resource allocation weights: Based on the real-time situation of risk spread and the handling feedback from various departments, use optimization algorithms to dynamically adjust resource allocation weights. For example, if it is found that the risk of gathering of people in a certain area is spreading rapidly, it is necessary to increase the investment in security personnel and emergency evacuation equipment, and increase the weight of these resources in the overall resource allocation accordingly. By continuously adjusting resource allocation weights in real time, resources can be more accurately invested in where they are most needed, improving the efficiency and effectiveness of cross-departmental collaborative handling, and effectively controlling the further spread of risks.

[0194] In this embodiment, the specific operation steps of step S5 are as follows:

[0195] Step S5-1: Collect disposal process data

[0196] Risk event-related data: Collect the starting time and location of risk events from multi-source data collection modules and records of each link, and record the type and evolution of risk events in detail; for example, if it is a risk of gathering of people, it is necessary to record the starting location of the gathering, the change in the number of people, the spread of the gathering area, etc.; for equipment abnormality risks, it is necessary to record the time of equipment failure, equipment name, failure type, and the impact of the failure on surrounding equipment and business processes, etc.

[0197] Resource scheduling data: Obtain the initial resource allocation plan generated by the resource scheduling optimization algorithm, as well as the resource scheduling records adjusted due to various situations during the emergency response process; including the deployment time, quantity, destination, actual arrival time and commissioning time of resources such as security inspection channels, ground handling vehicles, emergency response units and medical rescue units, etc., for analyzing resource scheduling response delays;

[0198] Disposal strategy execution data: collect specific operations and effect data of each department according to the disposal strategy; such as the number of contraband seized by security personnel at different inspection intensities, the response measures taken by the emergency disposal unit and the effectiveness of risk control, the treatment process and treatment effect of the wounded by the medical rescue unit, etc., in order to evaluate the effectiveness of the disposal strategy;

[0199] Model parameter adjustment related data: records the relevant information about the adjustment of dynamic risk assessment model parameters due to the emergence of new situations or the update of risk assessment during the disposal process; including the name of the adjusted parameter, the value before and after the adjustment, the reason and time of the adjustment, etc., to fully present the model parameter adaptive adjustment record;

[0200] Step S5-2: Generate a spatiotemporal evolution graph of risk events

[0201] Data collation and annotation: The collected risk event-related data are collated in chronological order, and key information such as the risk event status, impact scope, and location of relevant personnel and equipment at each time node are annotated; for example, for flight conflict risks, the planned location, actual location, and adjusted route of each flight at different time points are annotated;

[0202] Mapping: Using geographic information system (GIS) technology and time series analysis tools, based on airport maps, the development process of risk events is visualized in spatial and temporal dimensions. In the map, different colors, lines, icons and other elements are used to represent different types of risk events, the movement trajectories of related personnel and equipment, as well as the spread of risks and the changes in the impact area over time. For example, red lines are used to represent the diffusion path of areas where people gather, and green icons are used to represent emergency equipment that has arrived at the designated location and is put into disposal.

[0203] Step S5-3: Perform resource scheduling response delay analysis

[0204] Calculate the response delay time: Calculate the response delay time of each resource unit based on the time when the resource allocation instruction is issued and the time when the resources actually arrive and are put into use in the resource scheduling data; for example, for a ground handling vehicle, the response delay time of the vehicle is obtained by subtracting the time when the scheduling instruction is issued from the time when the vehicle actually arrives at the designated work location and is ready to perform the task;

[0205] Analyze the reasons for delay: Combined with the actual operation of the airport, such as traffic congestion, equipment failure, poor communication between personnel, etc., conduct a detailed analysis of the reasons for the response delay of each resource unit;

[0206] Summarize delay patterns: Summarize and analyze the response delay time and causes of all resource units to find out the general patterns and main influencing factors of resource scheduling response delay;

[0207] Step S5-4: Conduct a treatment strategy effectiveness assessment

[0208] Determine the evaluation indicators: According to different risk types and disposal objectives, determine a series of indicators for evaluating the effectiveness of disposal strategies; for example, for the risk of carrying prohibited items, the evaluation indicators may include the rate of seizure of prohibited items and the rate of missed detection; for the risk of gathering of people, the evaluation indicators may include the speed of evacuation of people, the time to complete evacuation, the rate of secondary gathering, etc.; for the risk of equipment abnormality, the evaluation indicators may include the time to repair equipment, the rate of recurrence of faults, the degree of recovery of business processes, etc.;

[0209] Calculate the evaluation index value: Calculate the value of each evaluation index based on the collected disposal strategy execution data; for example, calculate the contraband seizure rate by counting the number of contraband actually seized during the security inspection and the total number of items to be inspected; calculate the evacuation speed and evacuation completion time based on the personnel evacuation records;

[0210] Construct an evaluation matrix: organize the values ​​of various evaluation indicators of different risk types into a matrix form; the rows of the matrix represent different risk types, and the columns represent the corresponding evaluation indicators; through this evaluation matrix, the execution effects of disposal strategies under different risk types can be intuitively compared, and the effectiveness of disposal strategies can be comprehensively evaluated;

[0211] Step S5-5: Update the risk assessment model parameters

[0212] Analyze data and determine adjustment direction: Based on the information reflected by the spatial-temporal evolution of risk events, resource scheduling response delay analysis, and disposal strategy effectiveness evaluation matrix, analyze the deviations between the current risk assessment model and the actual situation; if the assessment finds that the assessment of abnormal risk of a certain equipment is inaccurate, the parameters related to the equipment need to be adjusted; determine the parameters that need to be adjusted and the direction of adjustment (increase or decrease the parameter value);

[0213] Parameter adjustment and verification: adjust the parameters of the risk assessment model according to the determined adjustment direction and amplitude; then, verify the adjusted model using historical data or simulation data to ensure that the adjusted model can more accurately reflect the actual risk situation; if the verification result is not ideal, further analyze the cause and fine-tune the parameters until the model can meet the accuracy requirements;

[0214] Step S5-6: Storage optimized scheduling strategy

[0215] Strategy collation and summary: Combined with the resource scheduling response delay analysis and the effectiveness evaluation results of the disposal strategy, the dispatch strategy formed during this emergency disposal process is comprehensively collated and summarized; the effective parts of the strategy are extracted, such as efficient resource allocation methods under specific risk scenarios, and the coordination and cooperation mode between resource units, etc., and the existing problems are reflected and improved;

[0216] Store in the policy knowledge base: The optimized scheduling strategy is stored in the policy knowledge base in a standardized format. During the storage process, the strategy is annotated in detail, including applicable risk types, application scenarios, advantages and limitations of the strategy, etc., so that it can be quickly retrieved and called when encountering similar risk events in the future, providing more effective decision-making support for airport safety management.

[0217] In this embodiment, the abnormal behavior feature vector is extracted by:

[0218] Abnormal posture recognition algorithm based on skeleton key point detection:

[0219] Video frame preprocessing: Extract video frames from the input video stream data at a fixed frame rate; grayscale each frame of the extracted image, converting the color image into a grayscale image to simplify subsequent calculations; then, use Gaussian filtering to smooth the image, remove noise interference in the image, improve image quality, and provide a clearer image for subsequent skeleton key point detection;

[0220] Skeleton key point detection: Use a skeleton key point detection model based on deep learning, such as the OpenPose model; input the preprocessed image into the model, and the model predicts the coordinate position of each key point on the human skeleton through a series of convolution, pooling and other operations; these key points usually include the head, shoulders, elbows, wrists, hips, knees and ankles, etc. Different models may detect different numbers and types of key points; each key point is represented in the image in the form of two-dimensional coordinates (x, y);

[0221] Posture feature extraction: Based on the detected skeleton key point coordinates, the relative position relationship and angle information between the various parts of the human body are calculated as posture features; these posture features can reflect the posture state of the human body and are an important basis for judging abnormal postures;

[0222] Abnormal posture recognition: The extracted posture features are compared with the pre-set normal posture feature range; the normal posture feature range can be obtained by analyzing and counting a large amount of normal behavior video data; if the posture feature exceeds the normal range, it is determined to be an abnormal posture; the identified abnormal posture information is encoded as part of the feature vector for subsequent abnormal behavior feature vector construction;

[0223] The optical flow method is used to analyze the conflict probability of personnel movement trajectories:

[0224] Optical flow calculation: For continuous video frames, the optical flow method (such as the Lucas-Kanade optical flow method or the Horn-Schunck optical flow method) is used to calculate the motion vector of each pixel between two adjacent frames, that is, the optical flow. The optical flow reflects the motion information of objects in the image, and the movement trajectory of people in the video frames can be tracked through the optical flow. Taking the Lucas-Kanade optical flow method as an example, this method assumes that the movement of objects in a small neighborhood is consistent, and calculates the optical flow by solving a set of linear equations.

[0225] Trajectory tracking: Track the trajectory of people in the video based on the calculated optical flow; starting from the starting frame of the video, by matching the optical flow information of the same person in adjacent frames, determine the position of the person in different frames, and thus obtain the movement trajectory of the person; in order to improve the accuracy of tracking, other information, such as the color characteristics and shape characteristics of the person, can be combined to optimize the tracking process;

[0226] Calculation of conflict probability: Analyze the intersection and proximity between different personnel movement trajectories and calculate the conflict probability. When two or more trajectories are too close or intersecting in space, it indicates the possibility of personnel conflict. Set certain distance thresholds and angle thresholds to determine whether the trajectories conflict. For example, if the minimum distance between two trajectories is less than the set distance threshold and their angle is within a certain range, it is considered that the two trajectories may conflict and the conflict probability is calculated. The conflict probability can be quantified based on factors such as the intersection degree of the trajectories and the proximity time.

[0227] Feature vector encoding: The calculated conflict probability information is encoded as part of the feature vector; the higher the conflict probability, the larger the corresponding value in the feature vector, so as to highlight the conflict risk in the movement trajectory of personnel and provide an important basis for subsequent abnormal behavior analysis;

[0228] The attention mechanism is used to enhance the detection sensitivity of suspicious objects in hidden areas:

[0229] Region division and feature extraction: The video frame is divided into multiple sub-regions, including visible areas and hidden areas where hidden objects may exist (such as the inside of luggage, parts blocked by people's clothing, etc.). For each sub-region, a convolutional neural network (CNN) is used to extract its features. CNN automatically learns the feature representation of the image through operations such as convolutional layers and pooling layers to obtain the feature vector of each sub-region.

[0230] Application of attention mechanism: Introduce the attention mechanism to assign different weights to the feature vectors of different sub-regions; the attention mechanism enables the model to pay more attention to the features of hidden regions; the attention weight of each sub-region is obtained by calculating the similarity between the feature vector of each sub-region and a learnable attention vector; the higher the similarity, the greater the attention weight obtained by the corresponding sub-region, indicating that the model pays more attention to the region;

[0231] Feature fusion and enhancement: The sub-region feature vectors assigned with attention weights are fused to obtain the enhanced feature vector of the entire video frame. During the fusion process, the features of the hidden region contribute more to the enhanced feature vector due to the higher attention weights, thereby strengthening the feature representation of the hidden region.

[0232] Suspicious object detection: The enhanced feature vector is input into a trained suspicious object detection model, which can be a target detection model based on deep learning, such as FasterR-CNN, YOLO, etc. The model determines whether there is a suspicious object in the video frame based on the information in the enhanced feature vector. If a suspicious object is detected, the relevant information (such as the category and location of the suspicious object) is encoded as part of the feature vector to further enrich the information of the abnormal behavior feature vector and improve the detection accuracy of abnormal behavior.

[0233] Through the specific operation steps in the above three aspects, abnormal behavior information of different dimensions is extracted respectively and encoded into feature vectors, which are finally combined to form a complete abnormal behavior feature vector, providing comprehensive and accurate feature information for subsequent risk assessment and abnormal behavior analysis.

[0234] In this embodiment, the process of generating the three-dimensional scheduling matrix includes:

[0235] Divide the airport area into honeycomb grid cells and calculate the resource demand density of each grid:

[0236] Regional division:

[0237] According to the actual layout of the airport, including different functional areas such as terminal buildings, runways, aprons, and parking lots, the entire airport area is divided into honeycomb grid units of uniform size using geographic information system (GIS) technology or coordinate-based division methods; each grid unit has a specific geographic coordinate range, and its size setting needs to comprehensively consider factors such as airport scale, resource scheduling accuracy requirements, and computational complexity;

[0238] Data Collection:

[0239] Collect various data related to resource demand in each grid unit, including but not limited to the number of personnel, equipment distribution, business activity types and frequencies, etc.; for example, in the terminal waiting area grid, count the number of waiting passengers, the number of service personnel, and the number of various service equipment (such as seats, charging facilities, etc.) in the grid; in the grid near the runway, record the distribution of runway operation equipment (such as snowplows, de-icing vehicles, etc.) and the business activity intensity in the area when flights take off and land;

[0240] Requirements Analysis:

[0241] Analyze the demand characteristics of business activities in each grid unit for different types of resources (such as security personnel, ground handling vehicles, emergency response equipment, etc.);

[0242] Density calculation:

[0243] Based on the collected data and demand analysis results, the demand density of each grid unit for different resources is calculated; the resource demand density can be expressed by the ratio of the demand for a specific resource to the area of ​​the grid unit; by calculating the resource demand density of each grid, basic data support is provided for subsequent resource scheduling;

[0244] Scheduling plan simulation based on the digital twin platform:

[0245] Digital twin model construction:

[0246] On the digital twin platform, a virtual model corresponding to the actual airport is constructed; the model should not only accurately reflect the geographical layout and facilities and equipment of the airport, but also simulate the operating logic of various business activities in the airport; each element in the model (such as buildings, vehicles, personnel, etc.) has attributes and behavioral characteristics corresponding to the actual one, and the operating data of the actual airport is obtained in real time through the data interface to keep the virtual model updated synchronously with the actual situation;

[0247] Scheduling scheme input:

[0248] Input the scheduling plan initially formulated based on resource demand density into the digital twin model; the scheduling plan includes personnel grouping, equipment configuration, path planning, etc.

[0249] Simulation run:

[0250] Start the simulation operation on the digital twin platform, and simulate the operation of the airport under the scheduling plan according to the set time step. During the simulation process, the model dynamically updates the status of each element, such as the location movement of personnel, the working status of equipment, and the advancement of business processes, based on the input scheduling plan and the real-time airport operation data. At the same time, various possible interference factors are considered, such as weather changes, equipment failures, and personnel emergencies, so as to simulate the actual operation scenario more realistically.

[0251] Effect evaluation:

[0252] After the simulation operation is completed, the implementation effect of the scheduling scheme is evaluated according to the preset evaluation indicators (such as response timeliness, resource utilization, disposal cost, etc.); for example, the time when resources are in place for each grid unit is counted to evaluate the response timeliness; the ratio of resources actually put into use to total resources is calculated to measure resource utilization; the human and material costs generated during the scheduling process are counted to evaluate the disposal cost; through the quantitative analysis of these indicators, the feasibility and effectiveness of the scheduling scheme in actual operation are fully understood;

[0253] Solution optimization:

[0254] According to the evaluation results, the scheduling plan is optimized and adjusted. If it is found that the resource response time in a certain area is too long, the resource allocation in that area can be increased or the path planning can be adjusted. If the resource utilization rate is low, the rationality of resource allocation can be re-evaluated and optimized. The optimized scheduling plan is input into the digital twin model again for simulation and deduction, and repeated iterations are carried out until the optimal scheduling plan that meets the requirements is obtained.

[0255] Introduce an elastic resource pool mechanism to dynamically allocate resources to cope with sudden demands:

[0256] Resource pool construction:

[0257] A portion of the airport's existing resources is allocated as a flexible resource pool; these resources may include spare security personnel, ground handling vehicles, emergency rescue equipment, etc., and their quantity and type are determined based on the airport's historical operating data and risk assessment results; the resources in the flexible resource pool are usually on standby but can be put into use at any time;

[0258] Requirement monitoring:

[0259] Establish a real-time monitoring system to continuously monitor changes in resource demand in various areas of the airport; timely discover possible sudden resource demand situations by integrating multi-source data, including real-time personnel location data, equipment operation status data, flight dynamic data, etc.;

[0260] Trigger mechanism:

[0261] Set the conditions that trigger the allocation of elastic resource pools. Generally, the resource demand exceeds a certain threshold of normal supply capacity. Once the trigger condition is met, the system automatically starts the resource allocation process.

[0262] Resource provisioning:

[0263] According to the type and location of sudden demand, the corresponding resources are quickly allocated from the elastic resource pool; the allocation process follows certain priorities and strategies; at the same time, the three-dimensional scheduling matrix is ​​updated to include the allocated resource information to ensure real-time adjustment and consistency of the scheduling plan;

[0264] Dynamic adjustment and recycling:

[0265] After resource allocation, the demand changes are continuously monitored, and the amount of allocated resources and the time of use are dynamically adjusted according to actual demand. When the sudden demand is alleviated and the resource demand is reduced to a certain level, the excess resources are recovered from the site and re-incorporated into the elastic resource pool for next use. Through this dynamic adjustment and recovery mechanism, efficient use and flexible allocation of resources are achieved, and the airport's ability to respond to emergencies is improved.

[0266] Through the above three steps, a three-dimensional scheduling matrix containing information such as personnel grouping, equipment configuration, and path planning is generated to achieve scientific and efficient scheduling of airport resources to cope with various complex operating conditions and sudden demands.

[0267] In this embodiment, the airport safety management method further includes:

[0268] Implement a graded inspection strategy in the passenger security check process and dynamically adjust the inspection intensity of the security check channel based on the risk level:

[0269] Risk Level Assessment:

[0270] According to the output result of the dynamic risk assessment model in step S2, the risk level related to the passenger is obtained; the risk level comprehensively considers multiple sources of data such as the passenger's identity information, flight information, security equipment detection data, and personnel location tag data;

[0271] The risk level is divided into three levels: low, medium and high, and each level corresponds to different risk characteristics and probabilities;

[0272] Check strength standard settings:

[0273] According to different risk levels, corresponding security inspection channel inspection intensity standards are pre-set;

[0274] For passengers with low risk levels, the inspection intensity is relatively low;

[0275] The inspection intensity for passengers at the medium risk level is moderate. In addition to the conventional X-ray and metal detection, manual inspection of some items in the luggage is increased, the sampling inspection ratio is increased to 30%, and the questioning of passengers is also more detailed to understand their travel purpose, items carried, etc.

[0276] High-risk passengers will be subject to high-intensity inspections; their luggage will be thoroughly and carefully checked manually, including opening all suspicious items for inspection; the passengers' bodies will be fully metal-detected, and even more advanced detection technologies such as millimeter-wave body imaging inspections may be used; at the same time, the inspection of passengers and their luggage will be 100% covered, and more in-depth background checks and inquiries may be conducted;

[0277] Real-time adjustment of inspection intensity:

[0278] At the security check site, the security check system obtains the risk level information of passengers in real time; based on the risk level, it automatically adjusts the inspection process and resource allocation of the security check channel to achieve dynamic adjustment of the inspection intensity;

[0279] For example, when the system identifies a passenger as high-risk, it immediately sends a reminder to the staff at the security checkpoint, deploys more security personnel to the channel, and adds manual inspection equipment and technical means to ensure a comprehensive and detailed inspection of the passenger and his / her luggage;

[0280] For groups of passengers who pass through security checks continuously, the security system will dynamically adjust the inspection intensity of each channel according to the overall risk level. For example, if the risk level of passengers from a specific flight is generally high during a certain period of time, the security system will automatically assign the passengers of this flight to channels with higher inspection intensity and increase the resource allocation of these channels accordingly.

[0281] Establish equipment health prediction model to trigger maintenance warning and adjust spare parts inventory in advance:

[0282] Data Collection:

[0283] Collect operation data of various equipment, including basic information of equipment, real-time operation parameters, historical maintenance records and equipment failure records;

[0284] These data come from multiple data sources such as the equipment’s own sensors, monitoring systems, and maintenance management systems;

[0285] Feature Engineering:

[0286] Clean and preprocess the collected data to remove noise data, outliers and duplicate data to ensure data quality and consistency;

[0287] Extract features related to equipment health from raw data; extract features such as maintenance intervals and maintenance costs from historical maintenance records; determine features such as the frequency and severity of faults based on fault records;

[0288] Normalize the extracted features and convert feature values ​​of different ranges and magnitudes to the same numerical interval to facilitate model training and comparison;

[0289] Model selection and training:

[0290] According to the characteristics of the equipment and data features, select the appropriate prediction model, such as support vector machine (SVM), random forest (RandomForest) based on machine learning, or long short-term memory network (LSTM) based on deep learning;

[0291] Divide the processed data into training set and test set, usually in a ratio of 70%-30% or 80%-20%; use the training set to train the selected model, and continuously adjust the model parameters so that the model can learn the relationship between equipment operation data and health;

[0292] During the training process, methods such as cross-validation are used to evaluate the performance of the model and select the model parameters with the best performance;

[0293] Model evaluation and optimization:

[0294] Use the test set to evaluate the trained model, and use indicators such as accuracy, recall, and mean square error (MSE) to measure the prediction accuracy of the model;

[0295] If the model performance does not meet the requirements, analyze the reasons and optimize the model; possible optimization methods include adjusting the model structure, adding more features, using ensemble learning methods, etc. For example, for the LSTM model, you can try to increase the number of hidden layers or adjust the number of hidden units to improve the model's expressiveness;

[0296] Health Prediction and Maintenance Warning:

[0297] Input the real-time collected equipment operation data into the optimized model to predict the health of the equipment; the model outputs a value or status indicating the health of the equipment;

[0298] Set a health threshold. When the predicted equipment health falls below a certain threshold, a maintenance warning is triggered.

[0299] At the same time, based on the equipment health prediction results and historical fault data, the possible fault types and probabilities are analyzed, and the corresponding spare parts are prepared in advance for maintenance work;

[0300] Spare parts inventory adjustment:

[0301] Adjust spare parts inventory based on maintenance warnings and fault analysis results combined with current spare parts inventory conditions;

[0302] If a certain equipment is predicted to have a certain failure and the spare parts required for the failure are insufficient, the system will automatically generate a spare parts purchase order and notify the purchasing department to replenish the spare parts in time; at the same time, the spare parts inventory is dynamically managed and the inventory quantity and location information are updated in real time;

[0303] Regularly review the usage and inventory turnover of spare parts, optimize spare parts inventory strategy according to actual conditions, and ensure that inventory costs are minimized while meeting equipment maintenance needs;

[0304] Use the digital twin system to simulate the disposal plan for major risk scenarios:

[0305] Definition and modeling of major risk scenarios:

[0306] Identify major risk scenarios that the airport may face, such as severe meteorological disasters, major equipment failures, etc.;

[0307] In the digital twin system, a detailed virtual model is built for each major risk scenario. The model not only includes the physical facilities and layout of the airport, but also simulates various factors when the risk occurs, such as the intensity and impact range of meteorological disasters, the specific manifestations of equipment failures, etc. At the same time, the impact of risks on personnel, equipment, flight operations, etc. in the airport is considered. For example, when simulating a hurricane scenario, the model must reflect the hurricane's wind force, path, and impact on the terminal building structure, apron equipment, and flight takeoffs and landings.

[0308] Developing a disposal plan:

[0309] Develop corresponding emergency response plans based on different major risk scenarios; the plans include emergency response procedures, division of responsibilities among departments, resource allocation plans, personnel evacuation routes, etc.

[0310] Convert the response plan into instructions and parameters that can be recognized by the digital twin system, including scheduling instructions for personnel and equipment, action time nodes, resource allocation quantities, etc.;

[0311] Initial condition setting:

[0312] In the digital twin system, initial conditions are set for the simulation of each major risk scenario; these initial conditions include specific parameters such as the time, location, and intensity of the risk, as well as the operating status of the airport before the risk occurs, such as the number of flights, passenger distribution, and equipment operation status. For example, when simulating a runway lighting system failure scenario, the specific time of the failure is set to be before a flight is about to land, the flight queue on the runway at that time, and the operating status of surrounding equipment.

[0313] Plan deduction execution:

[0314] Start the simulation of the disposal plan in the digital twin system; the system simulates a series of events and response actions after the risk occurs according to the set initial conditions and disposal plan;

[0315] During the simulation, the digital twin system updates the status of the airport virtual model in real time, including the movement trajectory of personnel, changes in the operating status of equipment, adjustments to flights, etc. For example, when simulating the evacuation of personnel, the system dynamically displays the evacuation process of personnel in the airport based on the set evacuation route and personnel movement speed;

[0316] Record key data during the simulation, such as response time, resource consumption, number of casualties, flight delay duration, etc. These data will be used to evaluate the effectiveness of the response plan;

[0317] Effective evaluation and plan optimization:

[0318] Evaluate the effectiveness of the response plan based on the data recorded in the simulation; use a series of evaluation indicators, such as casualty rate, property loss level, and time for flights to resume normal operations, to measure the effectiveness of the response plan in dealing with major risk scenarios;

[0319] Analyze the evaluation results to identify problems and deficiencies in the response plan. For example, if it is found that the evacuation time is too long, it may be that the evacuation route is unreasonable or the guidance measures are not in place. If the resource consumption is too large, it may be necessary to optimize the resource allocation plan.

[0320] According to the analysis results, the disposal plan is optimized and adjusted; the optimized plan is deduced again in the digital twin system, and it is iterated repeatedly until the disposal plan can achieve better results, effectively deal with major risk scenarios, and reduce losses and impacts;

[0321] Plan storage and update:

[0322] The optimized disposal plan is stored in the plan library of the digital twin system, so that it can be quickly called up when facing major risks;

[0323] As the airport develops and changes in operations, as well as the emergence of new risk factors, major risk scenarios are regularly reassessed and modeled, corresponding disposal plans are updated, and re-executive verification is performed in the digital twin system to ensure the effectiveness and adaptability of the plans;

[0324] An airport security management system, which is equipped with an airport security management method to manage airport security.

[0325] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An airport safety management method, characterized in that: The following steps are involved: S1. Real-time acquisition of airport operation data through multi-source heterogeneous data acquisition modules, including: video surveillance data, millimeter-wave radar data, security equipment detection data, flight dynamic data, personnel location tag data and environmental sensor data; S2. Establish a dynamic risk assessment model, perform feature fusion processing on the input data, generate a three-dimensional risk heat map and output risk level and type identification; the risk types at least include the risk of carrying prohibited items, the risk of gathering of people, the risk of equipment abnormality, the risk of flight conflict and the risk of sudden environment; S3. Triggering a hierarchical response mechanism based on risk level and type, generating a resource allocation plan through a resource scheduling optimization algorithm, and sending scheduling instructions to security inspection channels, ground handling vehicles, emergency response units, and medical rescue units; S4. During the emergency response phase, the multi-terminal collaborative platform monitors the response effect in real time. When the risk diffusion coefficient exceeds the preset threshold, the cross-departmental collaborative response agreement is initiated and the resource allocation weight is dynamically adjusted; S5. Generate a disposal process traceability report, update the risk assessment model parameters, and store the optimized scheduling strategy in the strategy knowledge base.

2. An airport security management method according to claim 1, characterized in that: The data acquisition module in step S1 includes: Multispectral camera arrays deployed at key points in the terminal to collect data on personnel behavior and baggage status; A distributed meteorological sensor group deployed in the runway area monitors visibility, wind speed and pavement friction coefficient in real time; The Beidou positioning terminal integrated in the shuttle bus and service vehicles obtains real-time location and operation trajectory data.

3. The airport security management method according to claim 1, characterized in that: In step S2, the dynamic risk assessment model adopts: The modified spatiotemporal convolutional neural network M-STCNN processes video stream data and extracts abnormal behavior feature vectors; assuming that the video stream data is ,In M-STCNN, after the convolution layer operation, the feature map The calculation formula is: ,in, Represents a location at a specific spatial and temporal index The feature map elements at Respectively represent the index of the feature map in the spatial and temporal dimensions, is the size of the convolution kernel, is the index of the convolution kernel in the first direction of the spatial dimension, is the index of the convolution kernel in the second direction of the spatial dimension, is the index of the convolution kernel in the time dimension, Represents the input video stream data exist The value at position, is the convolution kernel weight, is the bias term; the temporal and spatial features in the video stream data are extracted through this formula, and then the abnormal behavior feature vector is obtained; The modified graph neural network M-GNN constructs the topological graph of personnel-equipment-environment and identifies the potential risk propagation path; ,in, is a collection of nodes, is a set of edges; nodes The eigenvector of , in M-GNN, the node The updated feature vector The calculation formula is: ,in, Is a node The set of neighbor nodes of Is a node and The edge weights between is the weight matrix, is a nonlinear activation function, is the bias term; Representation Node The node feature vector is updated through this formula to identify the potential risk propagation path; Based on the fuzzy logic decision engine, the global risk coefficient is calculated by integrating the risk indicators of each dimension. Suppose the risk indicator set of each dimension is , the weight of each risk indicator is , global risk factor The calculation formula is: ,in, Represents the number of risk indicators, and the global risk coefficient is calculated through this formula.

4. The airport security management method according to claim 1, characterized in that: In step S3, the resource scheduling optimization algorithm includes: Establish a multi-objective optimization function, taking into account response timeliness, resource utilization and disposal cost; suppose response timeliness is , the resource utilization is The disposal cost is , multi-objective optimization function for: ,in, , , is the weight coefficient, which is used to balance the relationship between different objectives, and ; The improved genetic algorithm M-GA is used to solve the optimal scheduling scheme, in which the fitness function includes the risk attenuation rate, resource coverage and collaborative operation efficiency index; the risk attenuation rate is set to , the resource coverage is , the collaborative work efficiency index is , the fitness function Fitness is: ,in, , , is the weight coefficient, which is used to adjust the importance of different indicators in the fitness function, and ; The output includes a three-dimensional scheduling matrix of personnel grouping, equipment configuration, and path planning.

5. The airport security management method according to claim 1, characterized in that: In step S4, the cross-departmental collaborative disposal agreement includes: Build a blockchain-based emergency command chain to record the response time and disposal effect of each unit; When a runway incursion is detected, an augmented reality situation map is simultaneously pushed to the air traffic control tower, ground control center and security department; In medical rescue scenarios, a rescue plan is automatically generated that includes the optimal path, casualty classification, and material requirements.

6. The airport security management method according to claim 1, characterized in that: In step S5, the disposal process traceability report includes: Spatiotemporal evolution of risk events; Resource scheduling response delay analysis; Disposal strategy effectiveness assessment matrix; Model parameter adaptive adjustment records.

7. The airport security management method according to claim 3, characterized in that: The abnormal behavior feature vector extraction adopts: Abnormal posture recognition algorithm based on skeleton key point detection; The optical flow method is used to analyze the conflict probability of personnel movement trajectories; The detection sensitivity of suspicious objects in hidden areas is enhanced through the attention mechanism.

8. The airport security management method according to claim 4, characterized in that: The generation process of the three-dimensional scheduling matrix includes: Divide the airport area into honeycomb grid cells and calculate the resource demand density of each grid; Conduct scheduling simulation based on the digital twin platform; Introduce an elastic resource pool mechanism to dynamically allocate resources to meet sudden demands.

9. The airport security management method according to claim 1, characterized in that: Also includes: Implement a graded inspection strategy in the passenger security check process and dynamically adjust the inspection intensity of the security check channel based on the risk level; Establish equipment health prediction models to trigger maintenance warnings and adjust spare parts inventory in advance; Use the digital twin system to simulate contingency plans for major risk scenarios.

10. An airport security management system, characterized in that: The airport security management system is equipped with the airport security management method described in any one of claims 1-9 to manage airport security.

Images