Civil aviation airport security method and system based on multi-source perception and dynamic risk assessment

Abstract

The invention discloses a civil aviation airport security method and system based on multi-source perception and dynamic risk assessment, and the method comprises the steps: building a personnel-article-scene ternary relation model through collecting multi-source data such as video monitoring, face recognition, luggage scanning and sound signals; based on an abnormal matching mode of the model, dynamic risk assessment is carried out by combining real-time people flow density and historical event thermodynamic distribution, and a dynamic risk calibration map is generated; an identification strategy parameter is configured according to the risk level, a self-adaptive strategy set is formed and applied to the intelligent terminal, target identification and behavior tracking are achieved, and a security and protection system is automatically linked when a triggering condition is met; according to the method, the behavior-article-scene ternary relation model is constructed, so that multi-source data collaborative analysis is realized, the limitation of traditional single-mode detection is overcome, the abnormal behavior recognition accuracy in a complex scene is remarkably improved, and the false alarm rate is effectively reduced.

Description

Technical Field

[0001] The present invention belongs to the field of information security technology, and in particular relates to a civil aviation airport security method and system based on multi-source perception and dynamic risk assessment. Background Art

[0002] As national transportation hubs, civil aviation airports carry highly concentrated flows of people, goods, and information, making them crucial locations for public security and control. With the rapid development of the air transportation industry, airport operations are expanding, and security requirements are becoming increasingly complex. Currently, airports widely utilize multimodal sensing technologies, including video surveillance, facial recognition, baggage X-ray scanning, and sound detection, to achieve multi-dimensional monitoring of people, objects, and the environment. Furthermore, the introduction of intelligent analysis algorithms and automated security equipment has further enhanced the technical level and response efficiency of airport security systems.

[0003] Existing airport security systems typically utilize a layered architecture, consisting of a front-end perception layer, a data transmission layer, and a central control layer. The front-end perception layer consists of independent monitoring devices deployed in different areas, such as high-definition cameras, facial recognition terminals, and X-ray security scanners, responsible for collecting real-time data. The data transmission layer aggregates this perception data via wired or wireless networks to a central control platform, where it is stored and analyzed by back-end servers. The central control layer utilizes rule-based or fixed threshold alarm mechanisms, such as triggering alarms for facial recognition anomalies and initiating contact with security personnel when dangerous items are detected by baggage X-rays, to ensure rapid response to security incidents.

[0004] Although existing security systems have certain monitoring capabilities, they still have obvious shortcomings. Various perception systems (such as video, facial recognition, and X-rays) usually operate independently and lack a unified data fusion mechanism, making it difficult to collaboratively analyze cross-modal information, which can easily lead to security blind spots or misjudgments. In addition, existing systems rely on static rules or fixed threshold alarms and are unable to dynamically adjust monitoring strategies based on real-time crowd density, historical event distribution, and environmental changes. This leads to delayed identification of high-risk areas or irrational resource allocation. These problems limit the intelligence level and emergency response efficiency of airport security systems. Summary of the Invention

[0005] Purpose of the invention: The purpose of the present invention is to provide a civil aviation airport security method based on multi-source perception and dynamic risk assessment that can integrate cross-modal information, dynamically adjust monitoring strategies, and improve emergency response efficiency; on the other hand, to provide a civil aviation airport security system based on multi-source perception and dynamic risk assessment.

[0006] Technical solution: The civil aviation airport security method of the present invention comprises the following steps:

[0007] (1) By deploying sensing terminals in various areas of the airport to collect multi-source raw sensing data, including video surveillance data, face recognition data, baggage scanning data, and sound signal data, and combining them with the airport structure diagram to build a personnel-object-scene ternary relationship model, we can achieve a deep fusion analysis of personnel behavior, carried objects, and scene information in the airport environment, breaking through the limitations of traditional single data source detection and providing a more comprehensive data foundation for subsequent risk assessment;

[0008] (2) Based on the abnormal matching pattern in the ternary relationship model, combined with the real-time passenger flow density and the thermal distribution of historical events, dynamic risk assessment is performed in the airport space to generate a dynamic risk calibration map. By combining the real-time passenger flow status and the distribution of historical events, an accurate quantitative assessment of airport security risks is achieved, which can dynamically reflect the risk changes in different areas and provide a scientific basis for the formulation of differentiated security strategies;

[0009] (3) According to the risk level output by the dynamic risk calibration map, corresponding identification strategy parameters are configured for different areas, including identification frequency, identification algorithm selection and false alarm tolerance threshold, forming a scene-aware adaptive strategy configuration set, which realizes the intelligent scheduling of security resources, strengthens monitoring in high-risk areas, and optimizes resource utilization in low-risk areas, significantly improving the operating efficiency of the overall security system;

[0010] (4) The adaptive policy configuration set is applied to the on-site intelligent terminal to perform target recognition and behavior tracking tasks in real time. When the trigger conditions are detected, the control center, security personnel and on-site broadcasting system are automatically linked to issue disposal instructions and record response logs, realizing closed-loop management from risk identification to emergency disposal. Through automated instruction issuance and response recording, the timeliness and traceability of security incident disposal are greatly improved.

[0011] Preferably, step 1 comprises:

[0012] (11) Sensing terminal deployment: Multiple types of sensing terminal devices are deployed in the airport’s waiting areas, security checkpoints, baggage claim areas, and boarding gate areas. The sensing terminal devices include high-definition surveillance cameras, facial recognition terminals, X-ray baggage imaging equipment, and array-type sound acquisition units.

[0013] (12) Edge data preprocessing: Collecting the multi-source original perception data stream output by the perception terminal device, and performing unified format conversion and timestamp alignment on the images, audio, and baggage maps in the multi-source original perception data stream through the edge preprocessing node to construct a standardized multi-source data set;

[0014] (13) Spatial semantic mapping and feature extraction: The standardized multi-source dataset is spatially mapped with the airport structure diagram. By establishing an association index between personnel location information, item path tracking data, and scene functional zoning, the personnel behavior characteristics, item type, and environment scene at each moment are extracted;

[0015] (14) Construction of ternary relationship model: A ternary relationship model of person-object-scene is established based on the graph data structure, in which the connection edge weights between nodes are calculated through historical data and used to quantify the matching probability and anomaly confidence between nodes.

[0016] Through the rational deployment of multi-source sensing terminals and edge data preprocessing, the standardized collection and spatiotemporal alignment of heterogeneous data in various areas of the airport are achieved; combined with spatial semantic mapping technology, personnel behavior, item flow and scene functions are accurately associated; the final constructed ternary relationship model quantifies the matching probability and anomaly confidence between nodes, providing a structured analysis basis for the integration of multi-dimensional information for subsequent risk assessment, significantly improving the accuracy and reliability of abnormal behavior identification in complex scenarios.

[0017] Preferably, the edge data preprocessing is achieved by edge computing nodes deployed near each functional area of the airport. The edge computing nodes are used to perform at least one processing operation of data format standardization, multi-source time synchronization, feature information extraction and data compression encoding on the collected multi-source original perception data before the data is uploaded to the central control platform.

[0018] By deploying edge computing nodes near the functional areas of the airport, local preprocessing of multi-source perception data is achieved, including operations such as data format standardization, time synchronization, feature extraction and compression encoding. This effectively reduces the bandwidth pressure of data transmission to the central platform, while ensuring the temporal and spatial consistency of multi-source data, providing a high-quality data foundation for subsequent ternary relationship modeling and risk assessment, and significantly improving the real-time performance and processing efficiency of the system.

[0019] Preferably, step 2 includes:

[0020] (21) Abnormal pattern extraction: identifying abnormal edge relationships between nodes with high confidence but low historical correlation from the ternary relationship model, which is used to determine potential abnormal security behavior pairs;

[0021] (22) Real-time density analysis: The current crowd density information of each area of the airport is calculated by the people counting module deployed near the scene node, and the node aggregation degree in the ternary relationship model is normalized to generate the current crowd density map M. cur ;

[0022] (23) Historical thermal generation: Call the historical safety event database, extract the frequency of abnormal events mapped under the airport structure diagram within the historical time period, and generate the spatial thermal distribution map H hist , where the thermal value is normalized according to the event density;

[0023] (24) Dynamic risk assessment: Comprehensive abnormal matching pattern output, current crowd density map M cur and spatial thermal distribution map H hist , a risk fusion calculation based on the regional grid is performed in the airport structure diagram to obtain the risk value of each spatial unit, and it is divided into three risk levels: high, medium and low according to the preset level standards, and finally a dynamic risk calibration map is formed.

[0024] By integrating abnormal behavior patterns, real-time crowd density distribution, and historical event heat maps in the ternary relationship model, a dynamic and refined assessment of airport security risks is achieved. Abnormal edge relationship analysis can accurately capture potential threats, real-time density monitoring ensures the timeliness of risk assessment, and historical thermal data provides empirical reference. Finally, a dynamically updated risk calibration map is generated through grid fusion calculation, enabling the security system to intelligently identify risk hotspots and implement graded early warnings, greatly improving the accuracy and response efficiency of airport security.

[0025] Preferably, the dynamic risk calibration map is updated based on a sliding time window, and a time decay factor is used to weight the impact of historical events.

[0026] The sliding time window mechanism is used to achieve continuous dynamic updating of the risk calibration map, and the time decay factor is combined to adaptively adjust the impact of historical events, ensuring that recent security events occupy a more important weight in risk assessment, while the impact of long-term events gradually decreases. This design not only maintains the real-time and dynamic adaptability of risk assessment, but also effectively avoids the interference of outdated historical data on the current security situation, thereby significantly improving the system's sensitivity to sudden abnormal events, while maintaining the consistency and accuracy of risk trend analysis, and ultimately achieving more intelligent airport security situation awareness and response.

[0027] Preferably, step 3 includes:

[0028] (31) According to the risk level value R(x,y) corresponding to each spatial region (x,y) in the dynamic risk calibration map, the region is mapped to the preset policy configuration level interval Corresponding to the three perception levels of low, medium and high;

[0029] (32) For those at high perception level In areas with high accuracy, the recognition frequency is increased to the minimum interval; multimodal collaborative recognition algorithms are enabled, including image-audio joint models; a lower false alarm tolerance threshold is set, and the triggering mechanism for sensitive events is strengthened;

[0030] (33) For those at the medium perception level For areas with a medium frequency, configure the recognition interval, adopt a single main modality recognition model, and set the default false alarm tolerance threshold;

[0031] (34) For those at low perception level area, reduce the recognition frequency, select a lightweight fast classification algorithm, and relax the false alarm tolerance threshold. The recognition frequency, recognition algorithm selection, and false alarm tolerance threshold are uniformly packaged according to the regional address and perception level to form a scene perception adaptive strategy configuration set, which is automatically loaded and executed by the terminal recognition module.

[0032] By mapping regional risk levels in the dynamic risk calibration map to three perception levels (low, medium, and high), and intelligently matching differentiated recognition strategies (including recognition frequency, algorithm selection, and false alarm tolerance threshold) to different levels, precise dynamic allocation of security resources is achieved. High-sensitivity multimodal collaborative recognition is used in high-risk areas, while balanced detection capabilities are maintained in medium-risk areas. Computing resources are optimized in low-risk areas. This effectively reduces the overall system computing load while ensuring close monitoring of key areas, significantly improving the intelligence level and resource utilization efficiency of the security system.

[0033] Preferably, the multimodal collaborative recognition algorithm includes an image recognition module for analyzing video surveillance data, an audio recognition module for processing sound signal data, and a structural information recognition module for parsing luggage scanning data; the multimodal collaborative recognition algorithm is configured to fuse the output results of the image recognition module, the audio recognition module, and the structural information recognition module to generate a comprehensive abnormality determination result.

[0034] By integrating the multimodal data of the three modules of image recognition, audio recognition and structural information recognition, and adopting a collaborative analysis mechanism to cross-validate and comprehensively analyze the output results of each module, the accuracy and robustness of abnormal behavior detection have been significantly improved, effectively overcoming the limitations of single-modality detection that is susceptible to environmental interference or data missing, enabling the system to comprehensively perceive potential security threats from multiple dimensions and significantly reducing the probability of missed reports and false alarms.

[0035] Preferably, step 4 includes:

[0036] (41) Mapping and distributing the scene-aware adaptive strategy configuration set according to the area number and the terminal device identifier, and writing it into the intelligent identification terminal deployed at each monitoring point on the airport site;

[0037] (42) Each intelligent terminal performs target recognition and behavior tracking tasks in real time based on the received configuration parameters, including personnel behavior trajectory extraction, object-behavior matching detection and scene consistency judgment, and continuously calculates event trigger weight values;

[0038] (43) If the event trigger weight value exceeds the preset response threshold, the local terminal automatically triggers the linkage action.

[0039] By precisely distributing adaptive policy configuration sets to smart terminals in each region, localized real-time identification and dynamic tracking are achieved. Based on an intelligent evaluation mechanism based on event trigger weights, preset linkage responses are automatically executed when an anomaly is detected, forming a closed-loop security system from risk perception to emergency response. This significantly improves the timeliness and automation level of handling airport security incidents, while reducing the computational pressure on the central control system.

[0040] The civil aviation airport security system of the present invention comprises:

[0041] A multi-source data acquisition module is used to collect multi-source raw perception data through perception terminals deployed in various areas of the airport. The multi-source raw perception data includes video surveillance data, facial recognition data, luggage scanning data, and sound signal data. The data is then combined with the airport structure diagram to construct a ternary relationship model between people, objects, and scenes.

[0042] A ternary relationship modeling module is used to conduct dynamic risk assessment in the airport space based on the abnormal matching pattern in the ternary relationship model, combined with the real-time passenger flow density and the thermal distribution of historical events, and generate a dynamic risk calibration map;

[0043] The policy configuration module is used to configure corresponding recognition policy parameters for different areas based on the risk level output by the dynamic risk calibration map, including recognition frequency, recognition algorithm selection, and false alarm tolerance threshold, forming a scene-aware adaptive policy configuration set;

[0044] The intelligent execution module is used to apply the adaptive policy configuration set to the on-site intelligent terminal to perform target recognition and behavior tracking tasks in real time. When the trigger conditions are detected, it automatically links the control center, security personnel and the on-site broadcasting system to issue disposal instructions and record response logs.

[0045] A computer device, characterized in that it includes a memory and a processor, wherein the memory stores a computer program that can be loaded and executed by the processor for the civil aviation airport security method based on multi-source perception and dynamic risk assessment.

[0046] A computer-readable storage medium having a computer program stored thereon, characterized in that when the computer program is executed by a processor, the civil aviation airport security method based on multi-source perception and dynamic risk assessment is implemented.

[0047] Beneficial effects: Compared with the existing technology, the present invention has the following significant advantages: 1. By constructing a behavior-object-scene ternary relationship model, multi-source data collaborative analysis is realized, overcoming the limitations of traditional single-modal detection, significantly improving the accuracy of abnormal behavior recognition in complex scenarios, and effectively reducing the false alarm rate; 2. Based on real-time crowd density and historical event thermal distribution, it can dynamically divide risk levels in the airport space and adaptively adjust monitoring strategies, thereby improving the overall operation efficiency of the system while ensuring security effects; 3. Automatically matching response strategies according to risk levels to achieve accurate early warning and rapid disposal, significantly improving emergency response efficiency, and reducing interference with normal passengers; 4. By integrating time attenuation factors and spatial grid calculations, it can dynamically update risk maps and analyze the evolution trend of abnormal behaviors, providing forward-looking support for security decisions and enhancing the active defense capabilities of airport safety management. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 It is a schematic diagram of the process of the present invention;

[0049] Figure 2 This is a schematic diagram of constructing the ternary relationship model of the present invention. DETAILED DESCRIPTION

[0050] The technical solution of the present invention will be further described below with reference to the accompanying drawings.

[0051] like Figure 1 As shown, the civil aviation airport security method of the present invention includes the following steps:

[0052] S1. Establish a behavior-object-scene ternary relationship model:

[0053] Collect multi-source raw sensory data from multiple areas of the airport, including video surveillance, facial recognition, baggage scanning, and sound signals. Combined with the airport structure diagram, a ternary relationship model is constructed to describe the matching status between personnel behavior characteristics, the type of personal belongings, and the current scene environment.

[0054] S2. Perform dynamic risk area calibration:

[0055] Based on the abnormal matching pattern in the ternary relationship model, combined with real-time passenger flow density and historical event thermal distribution, the airport space is dynamically divided into risk level areas to form a dynamic risk calibration map;

[0056] S3. Generate an adaptive recognition strategy configuration set:

[0057] According to the risk level output by the dynamic risk calibration map, corresponding recognition strategy parameters are configured for different areas, including recognition frequency, recognition algorithm selection and false alarm tolerance threshold, forming a scene perception adaptive strategy configuration set;

[0058] S4. Triggering linkage response decision-making mechanism:

[0059] Apply the adaptive recognition policy configuration set to on-site intelligent terminals to perform target recognition and behavior tracking tasks in real time. When the trigger conditions are detected, the control center, security personnel and on-site broadcasting system are automatically linked to issue disposal instructions and record response logs.

[0060] like Figure 2 As shown, S1 specifically includes:

[0061] S11. Sensing Terminal Deployment: Various sensing terminal devices will be deployed in the airport's waiting areas, security checkpoints, baggage claim areas, and boarding gates. These sensing terminal devices include high-definition surveillance cameras, facial recognition terminals, X-ray baggage imaging equipment, and array-type sound acquisition units.

[0062] S12, Edge Data Preprocessing: Collect multi-source raw sensory data streams output by sensory terminal devices, and perform unified format conversion and timestamp alignment on images, audio, and baggage maps through edge preprocessing nodes to construct a standardized multi-source dataset;

[0063] S13, Spatial Semantic Mapping and Feature Extraction: This involves spatially mapping standardized multi-source datasets with airport structure diagrams. By establishing an index linking personnel location information, item path tracking data, and scene functional zoning, we extract the characteristics of personnel behavior, the type of belongings, and the surrounding environment at each moment.

[0064] S14. Ternary relationship model construction: Construct a ternary relationship model based on the graph data structure, connect the personnel behavior nodes, item nodes and scene nodes through matching edges, and assign edge weights based on historical data to characterize their matching probability and anomaly confidence, which can be called by subsequent risk assessment and identification strategy generation modules.

[0065] S12's edge pre-processing node is a computing module deployed near various functional areas of the airport. It is used to standardize the format, synchronize time, extract features, and perform compression encoding operations on multi-source raw perception data collected on-site before uploading the data to the central control platform. The functional modules of the edge pre-processing node include data synchronization module, format conversion module, behavior clue extraction module, scene label matching module, and data summary and encryption module.

[0066] S2 specifically includes:

[0067] S21. Extracting abnormal matching patterns from the constructed behavior-object-scene ternary relationship graph. Abnormal matching patterns are edge relationships between nodes that indicate high confidence but low historical correlation, which are used to determine potential security abnormal behavior pairs.

[0068] S22. Collect the current crowd density information of each area of the airport. The crowd density is calculated by the people counting module deployed near the scene node, and normalized by the node aggregation degree in the ternary graph to generate the current density map.

[0069] S23. Call the historical security event database, extract the frequency of abnormal events mapped under the airport structure diagram within the historical time period, generate a spatial thermal distribution map, and normalize the thermal value according to the event density;

[0070] S24. Based on the output of the integrated abnormal matching pattern, the current crowd density map and the historical thermal distribution map, a risk fusion calculation based on the regional grid is performed in the airport structure map to obtain the risk value of each spatial unit. The risk value is divided into three risk levels: high, medium and low according to the preset level standards, and finally a dynamic risk calibration map is formed. The dynamic risk calibration map is updated based on a sliding time window and uses a time attenuation factor to weight the impact of historical events.

[0071] S3 specifically includes:

[0072] S31, according to the risk level value R(x,y) corresponding to each spatial area (x,y) in the dynamic risk calibration map, map the area to the preset policy configuration level interval Corresponding to the three perception levels of low, medium and high;

[0073] S32, for those at high perception level For areas with high-definition video, configure the following recognition strategy parameters: increase the recognition frequency to a minimum interval (e.g., 1 second); enable multimodal collaborative recognition algorithms, including image-audio joint models; set a lower false alarm tolerance threshold (e.g., false alarm rate ≤ 5%), and strengthen the sensitive event triggering mechanism;

[0074] S33, for those at the medium perception level For areas with medium frequency, configure the recognition interval, use a single main modality recognition model (such as image recognition only), and set the default false alarm tolerance threshold (such as ≤ 10%).

[0075] S34, for those at low perception level For areas with high recognition accuracy, reduce the recognition frequency (e.g., once every 5 seconds), select a lightweight and fast classification algorithm, and relax the false alarm tolerance threshold (≤20e). The recognition frequency, recognition algorithm selection, and false alarm tolerance threshold are uniformly packaged according to the area address and perception level to form a scene perception adaptive policy configuration set, which is automatically loaded and executed by the terminal recognition module.

[0076] S32's multimodal collaborative recognition algorithm is used to improve recognition accuracy and anomaly detection robustness in high-risk areas. Its core idea is to integrate recognition results from multiple modalities such as image, audio, and structural information in the smart terminal to collaboratively determine whether the target entity has abnormal behavior or scene inconsistency. It specifically includes the following modules:

[0077] Image modality recognition module (main modality): Uses convolutional neural networks (such as YOLOv5) to perform target detection and behavior recognition on surveillance video streams and outputs a set of behavior labels At the same time, output the confidence vector P of the corresponding behavior label img ={p i ∣p i ∈[0,1],i=1,…,n}, where p i represents the confidence of the i-th behavior label, with a value range of [0,1]; n represents the number of labels output by image recognition;

[0078] Audio modality recognition module (auxiliary modality): performs short-time Fourier transform and Mel-frequency cepstral coefficient extraction on the audio signal collected by the array microphone; inputs it into the pre-trained audio event classification model to identify the audio event label Output the confidence vector P of the audio event aud ={q j ∣q j ∈

[0079] 0,1],j=1,…,m}; where q j represents the confidence of the jth audio tag, with a value range of [0,1]; m represents the number of tags output by audio recognition;

[0080] Ternary structure semantic verification module (decision support): For the identified behavior tags and audio tags, check whether they have associated matching history in the behavior-object-scene ternary relationship model; if a behavior tag has no semantic association with the current scene, the behavior is marked as a potential anomaly; the structure matching output abnormal structure factor η struct ∈[0,1];

[0081] Multimodal collaborative fusion mechanism: Constructing fusion scoring function S trig =λ1·max(P img )+λ2·max(P aud )+λ3·η struct , where λ1, λ2, and λ3 represent modality fusion weights, satisfying λ1+λ2+λ3=1, and are generally set to image-dominant type (e.g., λ1=0.5, λ2=0.3, λ3=0.2); max(P img) represents the behavior label score with the highest confidence in the image modality; max(P aud ) represents the audio event score with the highest confidence in the audio modality; δ res Indicates the preset threshold for triggering response. When S trig ≥δ res (Trigger threshold), execute the response linkage operation.

[0082] S4 specifically includes:

[0083] S41. Mapping and distributing the scene-aware adaptive policy configuration set to the terminal device identifier according to the area number, and writing it to the intelligent recognition terminal deployed at each monitoring point on the airport site, where the intelligent recognition terminal includes a video analysis unit, an audio recognition unit, and a behavior modeling module;

[0084] S42, each intelligent terminal performs target recognition and behavior tracking tasks in real time according to the received configuration parameters, including personnel behavior trajectory extraction, object-behavior matching detection and scene consistency judgment, and continuously calculates the event trigger weight value S trig ;

[0085] S43, if S trig The value exceeds the preset response threshold δ res , the local terminal automatically triggers the following linkage actions: push linkage instruction package to the security control center, including identification target information, event type, risk level and spatial coordinates; start the directional broadcast module to output guidance prompt voice or warning notice; start the mobile terminal pop-up prompt of security personnel in the nearby area, displaying the event summary and recommended disposal method;

[0086] S44. The terminal also writes key information in the event response process, including identification results, response timestamps, broadcast execution status, and security check confirmation, into the local response log and regularly uploads it to the central log aggregation server for subsequent traceability analysis and model optimization.

[0087] The invention also discloses a computer device.

[0088] Specifically, the computer device can be a computer device such as a desktop computer, a laptop computer, a handheld computer, and a cloud server. The computer device may include, but is not limited to, a processor and a memory. The processor and the memory may be connected via a bus or otherwise. The processor may be a central processing unit (CPU). The processor may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, graphics processing units (GPU), embedded neural network processors (NPU) or other dedicated deep learning coprocessors, discrete gate or transistor logic devices, discrete hardware components and other chips, or a combination of the above-mentioned various chips.

[0089] As a non-transient computer-readable storage medium, the memory can be used to store non-transient software programs, non-transient computer executable programs and modules. The processor executes various functional applications and data processing of the processor by running the non-transient software programs, instructions and modules stored in the memory. The memory may include a program storage area and a data storage area, wherein the program storage area may store a control unit, an application required for at least one function; the data storage area may store data created by the processor, etc. In addition, the memory may include a high-speed random access memory and may also include a non-transient memory. In some embodiments, the memory may optionally include a memory remotely located relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network and a combination thereof.

[0090] The invention also discloses a computer-readable storage medium.

[0091] Specifically, a computer-readable storage medium is used to store a computer program, and when the computer program is executed by a processor, the method in the above-mentioned method implementation is implemented. Those skilled in the art will understand that the implementation of all or part of the process in the above-mentioned embodiment method of the present application can be completed by instructing the relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium. When the program is executed, it may include the process of the implementation of each of the above-mentioned methods. Among them, the storage medium may be a magnetic disk, an optical disk, a read-only memory (ROM), a random access memory (RAM), a flash memory (Flash Memory), a hard disk (Hard Disk Drive, abbreviated: HDD) or a solid-state drive (SSD), etc.; the storage medium may also include a combination of the above-mentioned types of memory.

Claims

1. A civil aviation airport security method based on multi-source perception and dynamic risk assessment, characterized by: The following steps are involved: (1) Using sensing terminals deployed in various areas of the airport to collect multi-source raw sensing data, including video surveillance data, face recognition data, baggage scanning data, and sound signal data, a personnel-object-scene ternary relationship model is constructed in combination with the airport structure diagram; (2) Based on the abnormal matching pattern in the ternary relationship model, combined with the real-time passenger flow density and the thermal distribution of historical events, dynamic risk assessment is performed in the airport space to generate a dynamic risk calibration map; (3) According to the risk level output by the dynamic risk calibration map, the corresponding recognition strategy parameters are configured for different areas, including recognition frequency, recognition algorithm selection and false alarm tolerance threshold, forming a scene perception adaptive strategy configuration set; (4) The adaptive policy configuration set is applied to the on-site intelligent terminal to perform target recognition and behavior tracking tasks in real time. When the trigger conditions are detected, the control center, security personnel and on-site broadcasting system are automatically linked to issue disposal instructions and record response logs.

2. The civil aviation airport security method according to claim 1, characterized in that: Step 1 includes: (11) Sensing terminal deployment: Multiple types of sensing terminal devices are deployed in the airport’s waiting areas, security checkpoints, baggage claim areas, and boarding gate areas. The sensing terminal devices include high-definition surveillance cameras, facial recognition terminals, X-ray baggage imaging equipment, and array-type sound acquisition units. (12) Edge data preprocessing: Collecting the multi-source original perception data stream output by the perception terminal device, and performing unified format conversion and timestamp alignment on the images, audio, and baggage maps in the multi-source original perception data stream through the edge preprocessing node to construct a standardized multi-source data set; (13) Spatial semantic mapping and feature extraction: The standardized multi-source dataset is spatially mapped with the airport structure diagram. By establishing an association index between personnel location information, item path tracking data, and scene functional zoning, the personnel behavior characteristics, item type, and environment scene at each moment are extracted; (14) Construction of ternary relationship model: A ternary relationship model of person-object-scene is established based on the graph data structure, in which the connection edge weights between nodes are calculated through historical data and used to quantify the matching probability and anomaly confidence between nodes.

3. The civil aviation airport security method according to claim 2, characterized in that: The edge data preprocessing is achieved through edge computing nodes deployed near the functional areas of the airport. The edge computing nodes are used to perform at least one processing operation of data format standardization, multi-source time synchronization, feature information extraction and data compression encoding on the collected multi-source original perception data before the data is uploaded to the central control platform.

4. The civil aviation airport security method according to claim 1, characterized in that: Step 2 includes: (21) Abnormal pattern extraction: identifying abnormal edge relationships between nodes with high confidence but low historical correlation from the ternary relationship model, which is used to determine potential abnormal security behavior pairs; (22) Real-time density analysis: The current crowd density information of each area of the airport is calculated by the people counting module deployed near the scene node, and the node aggregation degree in the ternary relationship model is normalized to generate the current crowd density map M. cur ; (23) Historical thermal generation: Call the historical safety event database, extract the frequency of abnormal events mapped under the airport structure diagram within the historical time period, and generate the spatial thermal distribution map H hist , where the thermal value is normalized according to the event density; (24) Dynamic risk assessment: Comprehensive abnormal matching pattern output, current crowd density map M cur and spatial thermal distribution map H hist , a risk fusion calculation based on the regional grid is performed in the airport structure diagram to obtain the risk value of each spatial unit, and it is divided into three risk levels: high, medium and low according to the preset level standards, and finally a dynamic risk calibration map is formed.

5. The civil aviation airport security method according to claim 4, characterized in that: The dynamic risk calibration map is updated based on a sliding time window and uses a time decay factor to weight the impact of historical events.

6. The civil aviation airport security method according to claim 1, characterized in that: Step 3 includes: (31) According to the risk level value R(x,y) corresponding to each spatial region (x,y) in the dynamic risk calibration map, the region is mapped to the preset policy configuration level interval Corresponding to the three perception levels of low, medium and high; (32) For those at high perception level In areas with high accuracy, the recognition frequency is increased to the minimum interval; multimodal collaborative recognition algorithms are enabled, including image-audio joint models; a lower false alarm tolerance threshold is set, and the triggering mechanism for sensitive events is strengthened; (33) For those at the medium perception level For areas with a medium frequency, configure the recognition interval, adopt a single main modality recognition model, and set the default false alarm tolerance threshold; (34) For those at low perception level area, reduce the recognition frequency, select a lightweight fast classification algorithm, and relax the false alarm tolerance threshold. The recognition frequency, recognition algorithm selection, and false alarm tolerance threshold are uniformly packaged according to the regional address and perception level to form a scene perception adaptive strategy configuration set, which is automatically loaded and executed by the terminal recognition module.

7. The civil aviation airport security method according to claim 6, characterized in that: The multimodal collaborative recognition algorithm includes an image recognition module for analyzing video surveillance data, an audio recognition module for processing sound signal data, and a structural information recognition module for parsing luggage scanning data; the multimodal collaborative recognition algorithm is configured to fuse the output results of the image recognition module, audio recognition module, and structural information recognition module to generate a comprehensive anomaly determination result.

8. The civil aviation airport security method according to claim 1, characterized in that: Step 4 includes: (41) Mapping and distributing the scene-aware adaptive strategy configuration set according to the area number and the terminal device identifier, and writing it into the intelligent identification terminal deployed at each monitoring point on the airport site; (42) Each intelligent terminal performs target recognition and behavior tracking tasks in real time based on the received configuration parameters, including personnel behavior trajectory extraction, object-behavior matching detection and scene consistency judgment, and continuously calculates event trigger weight values; (43) If the event trigger weight value exceeds the preset response threshold, the local terminal automatically triggers the linkage action.

9. A civil aviation airport security system based on multi-source perception and dynamic risk assessment, characterized by: include: A multi-source data acquisition module is used to collect multi-source raw perception data through perception terminals deployed in various areas of the airport. The multi-source raw perception data includes video surveillance data, facial recognition data, luggage scanning data, and sound signal data. The data is then combined with the airport structure diagram to construct a ternary relationship model between people, objects, and scenes. A ternary relationship modeling module is used to conduct dynamic risk assessment in the airport space based on the abnormal matching pattern in the ternary relationship model, combined with the real-time passenger flow density and the thermal distribution of historical events, and generate a dynamic risk calibration map; The policy configuration module is used to configure corresponding recognition policy parameters for different areas based on the risk level output by the dynamic risk calibration map, including recognition frequency, recognition algorithm selection, and false alarm tolerance threshold, forming a scene-aware adaptive policy configuration set; The intelligent execution module is used to apply the adaptive policy configuration set to the on-site intelligent terminal to perform target recognition and behavior tracking tasks in real time. When the trigger conditions are detected, it automatically links the control center, security personnel and the on-site broadcasting system to issue disposal instructions and record response logs.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, it implements the civil aviation airport security method based on multi-source perception and dynamic risk assessment according to any one of claims 1 to 8.

Images