An Airport Perimeter Monitoring and Early Warning Method and System

Abstract

This application provides an airport perimeter monitoring and early warning method and system, relating to the field of airport security technology. The method includes: an edge node responding to a prediction command sent by a central server, acquiring monitoring video of the airport perimeter and performing behavior prediction on target objects in the image frames of the monitoring video to obtain behavior prediction results; if the behavior prediction result indicates that the target object is engaging in intrusion behavior, generating a video generation strategy for the target object; a P2P network, based on the video generation strategy, acquiring target video containing image frames of the target object from the edge node, caching the target video to a target storage node, and sending it to a non-target storage node; and the central server generating early warning information based on the behavior prediction results and sending it to the client. This provides early warning and allows users to view monitoring video without waiting for video loading, improving the smoothness of viewing monitoring video and thus increasing the efficiency of handling intrusion behavior.

Description

Technical Field

[0001] This application relates to the field of airport security technology, and in particular to an airport perimeter monitoring and early warning method and system. Background Technology

[0002] To ensure the security of all facilities and overall operations within the airport area, an intrusion monitoring system can be installed within the airport perimeter to monitor various potential intrusion activities and generate alarm information upon detection. After an alarm is triggered, users can view the monitoring video to understand the specific details of the intrusion, including the location of the intrusion, the characteristics of the person or object, and the trajectory of the movement, and then make appropriate handling decisions based on this information.

[0003] However, current intrusion monitoring systems have certain limitations. They typically only activate alarm mechanisms after an intrusion has already occurred, meaning they cannot provide effective early warnings before an intrusion happens, leading to security risks and delayed responses. Furthermore, due to the wide coverage of airport perimeters and frequent data exchanges between various internal devices and systems, network throughput is high. In such situations, when network conditions are poor, the transmission of surveillance video is prone to stuttering. Video footage may experience pauses, delayed loading, blurry images, or even brief interruptions, making it difficult for users to quickly and continuously obtain complete intrusion information. This impacts their ability to accurately and efficiently handle intrusions, reducing processing efficiency and threatening the airport's secure operations. Summary of the Invention

[0004] To address the aforementioned issues, this application provides an airport perimeter monitoring and early warning method and system, which can resolve the problems of untimely alarms for intrusion behavior and video transmission lag after alarms are generated, leading to low efficiency in handling intrusion behavior.

[0005] To achieve the above objectives, in a first aspect, this application provides an airport perimeter monitoring and early warning method, applied to an airport perimeter monitoring and early warning system. This system includes a central server, edge nodes, and a P2P network. The P2P network includes multiple storage nodes and multiple clients. The airport perimeter monitoring and early warning method includes: sending a prediction command to the edge nodes through the central server; the edge nodes responding to the prediction command by acquiring monitoring video of the airport perimeter; the edge nodes performing behavior prediction on target objects in the image frames of the monitoring video to obtain behavior prediction results for the target objects. The behavior prediction results include whether the target object exhibits intrusion behavior or not; if the behavior prediction result indicates that the target object exhibits intrusion behavior... Edge nodes generate a video generation strategy for the target object; the edge nodes send the video generation strategy to the P2P network and the behavior prediction results to the central server; the P2P network, based on the video generation strategy, obtains the target video containing the image frame where the target object is located from the edge nodes; the P2P network caches the target video to the target storage node and sends the target video to a non-target storage node, wherein the target storage node is the storage node closest to the edge node among multiple storage nodes, and the non-target storage node is the storage node that is not closest to the edge node among multiple storage nodes; the central server generates warning information based on the behavior prediction results and sends the warning information to the client.

[0006] The airport perimeter monitoring and early warning method provided in this application can acquire monitoring video of the airport perimeter through edge nodes and perform behavior prediction on target objects in the image frames of the monitoring video. When intrusion behavior is predicted, the behavior prediction result is sent to the central server, which generates early warning information and sends it to the client for early warning, thus providing early warning before the intrusion behavior occurs. In addition, the edge nodes can also generate video generation strategies based on the behavior prediction results. The P2P network stores the monitoring images of the image frames containing the target object containing intrusion behavior according to the video generation strategy. Users can obtain the cached video when they need to view the monitoring video without waiting for the video to load, improving the smoothness of the monitoring video viewing and thus improving the efficiency of intrusion behavior processing.

[0007] In an optional implementation, the airport perimeter monitoring and early warning method further includes: acquiring behavioral sample data, including input images and result labels, through a central server; performing feature extraction on the input images to obtain behavioral features of the target objects in the input images; inputting the behavioral features into a behavior prediction model to be trained to obtain the prediction results output by the behavior prediction model; calculating the training loss based on the prediction results and result labels; adjusting the model parameters of the behavior prediction model to be trained based on the training loss if the training loss is greater than or equal to a preset loss threshold; outputting the model parameters of the behavior prediction model to be trained if the training loss is less than the preset loss threshold to obtain the behavior prediction model; and sending the behavior prediction model to edge nodes. This allows the behavior prediction model to be trained in the central server, improving the accuracy of intrusion behavior prediction.

[0008] In one optional implementation, the edge node performs behavior prediction on target objects in image frames of the surveillance video to obtain behavior prediction results. This includes: determining whether the target object exhibits abnormal behavior based on behavioral indicators; if the target object exhibits abnormal behavior, performing preprocessing and feature extraction on the image frame containing the target object to obtain abnormal behavior features; and inputting the abnormal behavior features into the behavior prediction model to obtain the behavior prediction result. Thus, before judging intrusion behavior through the behavior prediction model, abnormal behavior is first preliminarily screened using behavioral indicators to improve the accuracy of the behavior prediction results.

[0009] In one optional implementation, the edge node generates a video generation strategy for the target object, including: obtaining the time point of the image frame containing the target object in the monitoring video; and marking video segments of the monitoring video from a preset time before the time point to a preset time after the time point as the target video, thereby obtaining the video generation strategy. In this way, by using the image frame containing the target object as a reference and using the video of the time period before and after that image frame as the target video, it is convenient for users to view the specific content of the intrusion behavior.

[0010] In one optional implementation, the airport perimeter monitoring and early warning method further includes: if the behavior prediction result indicates that the target object is engaging in intrusion behavior, the edge node generates behavioral features corresponding to the intrusion behavior of the target object; the edge node sends the behavioral features to the central server; and the central server updates the feature library in the edge node based on the behavioral features. In this way, the central server can update the feature library in the edge node based on the intrusion behavior features, increasing the behavior prediction model's ability to identify multiple intrusion behaviors and improving the comprehensiveness of intrusion behavior identification.

[0011] In one optional implementation, after the central server sends the warning information to the client, the airport perimeter monitoring and warning method further includes: the client responding to the user's video playback request by sending a video information acquisition request to the P2P network; the P2P network responding to the video information acquisition request by acquiring target node information and sending the target node information to the client, wherein the target node information is the index address of the storage node including the target video.

[0012] In one optional implementation, obtaining target node information includes: obtaining the load of the nearest storage node, where the nearest storage node is the storage node closest to the client; if the load of the nearest storage node is less than a load threshold, obtaining the index address of the nearest storage node as the target node information; if the load of the nearest storage node is greater than or equal to the load threshold, obtaining the index address of a non-nearest storage node whose load is less than the load threshold as the target node information, where the non-nearest storage node is the storage node not closest to the client. In this way, the P2P network can first obtain the storage node closest to the edge node where the intrusion behavior is detected, and send the target video stored in the nearest storage node to the client to improve the transmission speed of the target video; if the load of the storage node closest to the edge node where the intrusion behavior is detected is too high and the target video cannot be transmitted immediately, the P2P network can also send the target video from other storage nodes with lower loads to the client, avoiding the client having to wait for the target video to be transmitted and improving the speed at which the user obtains the target video.

[0013] In one optional implementation, after sending the target node information to the client, the airport perimeter monitoring and early warning method further includes: sending a video retrieval request from the client to the storage node pointed to by the index address in the target node information; the storage node pointed to by the index address in the target node information responds to the video retrieval request and sends the target video to the client. In this way, the client can find a suitable storage node based on the index address to retrieve the target video.

[0014] In one optional implementation, the airport perimeter monitoring and early warning system further includes a feedback unit, and the airport perimeter monitoring and early warning method further includes: receiving and analyzing user-provided feedback data through the feedback unit, the feedback data including early warning judgment data and video viewing data; sending the feedback data to a central server; optimizing the feature library in the edge nodes based on the early warning judgment data through the central server; and optimizing the video generation strategy based on the video viewing data. In this way, the feature library of intrusion behavior can be updated based on the user-provided early warning judgment data; and since video viewing data can reflect the viewing duration that users are more interested in, edge nodes can select target videos with more suitable durations, improving the accuracy of intrusion behavior prediction and the user's viewing experience of the target video.

[0015] Secondly, this application also provides an airport perimeter monitoring and early warning system, comprising: a central server configured to: send prediction instructions to edge nodes, generate early warning information based on behavior prediction results, and send the early warning information to clients; edge nodes configured to: acquire monitoring video of the airport perimeter in response to the prediction instructions; perform behavior prediction on target objects in image frames of the monitoring video to obtain behavior prediction results of the target objects, the behavior prediction results including whether the target objects exhibit intrusion behavior or not; if the behavior prediction result indicates that the target objects exhibit intrusion behavior, generate a video generation strategy for the target objects; send the video generation strategy to a P2P network, and send the behavior prediction results to the central server; the P2P network, comprising multiple storage nodes and multiple clients, configured to: acquire target video containing image frames of the target objects from the edge nodes based on the video generation strategy; cache the target video in a target storage node, and send the target video to a non-target storage node, wherein the target storage node is the storage node closest to the edge node among the multiple storage nodes, and the non-target storage node is the storage node not closest to the edge node among the multiple storage nodes.

[0016] It is understandable that the beneficial effects of the technical solution of the second aspect provided above can be referred to the beneficial effects of the first aspect and any of its optional implementation methods, and will not be repeated here. Attached Figure Description

[0017] To more clearly illustrate the technical solution of this application, the drawings used in the embodiments will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A schematic diagram of an intrusion monitoring system provided in an embodiment of this application;

[0019] Figure 2 A schematic diagram of an airport perimeter monitoring and early warning system provided in this application embodiment;

[0020] Figure 3 A flowchart illustrating an airport perimeter monitoring and early warning method provided in this application embodiment;

[0021] Figure 4 A schematic diagram illustrating the process of predicting the behavior of an edge node on a target object, provided in an embodiment of this application;

[0022] Figure 5 A schematic diagram illustrating a video generation strategy based on a surveillance video timeline, provided as an embodiment of this application;

[0023] Figure 6 A flowchart illustrating a method for obtaining target node information provided in this application embodiment;

[0024] Figure 7 A timing diagram of an airport perimeter monitoring and early warning method provided in an embodiment of this application.

[0025] Illustration:

[0026] Wherein: 101-Sensor; 102-Server; 103-Camera; 104-Client. Detailed Implementation

[0027] The embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following examples do not represent all embodiments consistent with this application.

[0028] It should be noted that the brief descriptions of terms in this application are only for the convenience of understanding the embodiments described below, and are not intended to limit the embodiments of this application. Unless otherwise stated, these terms should be understood in their ordinary and common meaning.

[0029] The terms "first," "second," "third," etc., used in the specification and accompanying drawings of this application are used to distinguish similar or related objects or entities, and do not necessarily imply a specific order or sequence, unless otherwise specified. It should be understood that such terms can be used interchangeably where appropriate.

[0030] To facilitate understanding of the plan, the following explanations are provided for relevant terms:

[0031] Edge nodes: These are devices or nodes located at the edge of a network that have certain data processing capabilities and can preprocess the collected data, such as filtering, aggregation, and calculation.

[0032] P2P network, or peer-to-peer network, is a distributed network architecture. In a P2P network, all nodes (also called peers) have equal status, and there is no centralized control by a central server. Each node can act as a client to obtain resources from other nodes, or as a server to provide resources to other nodes.

[0033] Installing an intrusion monitoring system within the airport perimeter can monitor intrusion activities and issue alarms. Upon triggering an alarm, the monitoring video will be transmitted to the client, allowing airport users to view the alarm information and make timely decisions, thereby ensuring the security of various facilities and overall operations within the airport area.

[0034] Figure 1 This is a schematic diagram of an intrusion monitoring system provided in an embodiment of this application. Figure 1 As shown, the intrusion monitoring system includes a sensor 101, a server 102, a camera 103, and a client 104. The workflow of the intrusion monitoring system includes: sensors 101 (including infrared sensors, fiber optic sensors, etc.) installed in the airport boundary area detect people or objects approaching the airport boundary; when sensor 101 detects an intrusion into the airport perimeter, it generates an electrical signal and sends it to the server 102; after receiving the electrical signal from sensor 101, server 102 sends a monitoring video acquisition command to camera 103, controlling camera 103 to capture monitoring video of the area where sensor 101 emitted the electrical signal, and transmits the monitoring video to client 104; simultaneously, server 102 generates alarm information and sends it to client 104, thus realizing the function of alarming and transmitting monitoring video of intrusion into the airport perimeter.

[0035] However, the aforementioned intrusion monitoring system has certain limitations. Signals are only transmitted to the central server and alarm information is issued when an intruder is detected by the sensors, reducing the real-time nature of intrusion detection. Furthermore, the client begins loading monitoring video after receiving the alarm, which can lead to video buffering due to network issues, making it difficult to quickly and continuously obtain complete intrusion information and reducing the user's efficiency in handling intrusion incidents.

[0036] In view of this, the embodiments of this application provide an airport perimeter monitoring and early warning method, which can solve the problems of untimely alarms for intrusion behavior and video transmission lag after alarms are generated in airport perimeter monitoring systems, resulting in low efficiency in handling intrusion behavior.

[0037] The airport perimeter monitoring and early warning method is applied to the airport perimeter monitoring and early warning system to enable early warning of intrusion behavior within the airport perimeter area and to load monitoring videos containing intrusion behavior in advance. Figure 2 This is a schematic diagram of an airport perimeter monitoring and early warning system provided as an embodiment of this application. Figure 2 As shown, the airport perimeter monitoring and early warning system includes a central server, edge nodes, and a P2P network. The P2P network comprises multiple storage nodes and multiple clients. In this system, the central server controls the system to predict intrusion attempts; edge nodes can be surveillance cameras with data processing capabilities, used to capture and process images from the surveillance videos; the P2P network interacts with the edge nodes, storing the surveillance videos in the storage nodes and transmitting them to the target clients.

[0038] The following section introduces the specific implementation method of airport perimeter monitoring and early warning. Figure 3 This is a flowchart illustrating an airport perimeter monitoring and early warning method provided in an embodiment of this application. Figure 3 As shown, the airport perimeter monitoring and early warning method includes the following steps:

[0039] S100: Sends prediction instructions to edge nodes through the central server.

[0040] In some embodiments, to activate the airport perimeter monitoring intrusion system, a central server can generate a prediction command and send the prediction command to edge nodes with computing capabilities.

[0041] S200: Edge nodes respond to prediction commands and acquire surveillance video of the airport perimeter.

[0042] To monitor airport perimeter intrusion, it is first necessary to acquire surveillance video of the airport perimeter. In some embodiments, after receiving a prediction instruction from the central server, the edge nodes begin acquiring surveillance video of the airport perimeter. Specifically, the edge nodes are surveillance cameras with computing capabilities, capable of collecting surveillance video of the airport perimeter area in real time.

[0043] In some embodiments, due to the large area of ​​an airport, multiple edge nodes can be set up within the airport perimeter to capture surveillance videos of different areas, thereby improving the comprehensiveness of intrusion monitoring.

[0044] S300: Edge nodes perform behavior prediction on target objects in image frames of surveillance video and obtain the behavior prediction results of the target objects.

[0045] In some embodiments, intrusion behavior is typically caused by a target object such as a person or animal. Therefore, to improve the efficiency of monitoring and early warning, edge nodes can perform preliminary detection on each frame of the acquired surveillance video to filter out image frames containing the target object. Behavior prediction is then performed on the target object to obtain the behavior prediction result. This behavior prediction result may indicate that the target object has engaged in intrusion behavior or that it has not.

[0046] For example, if a target object in a certain image frame is preparing to cross the airport perimeter fence, the edge node predicts that the target object is engaging in intrusion behavior; if a target object in a certain image frame is walking normally along the edge of the airport perimeter fence, the edge node predicts that the target object is not engaging in intrusion behavior.

[0047] In some embodiments, to improve the accuracy of intrusion behavior prediction, edge nodes can perform behavior prediction on intrusion target objects in two stages. Figure 4 This is a schematic diagram illustrating a process for predicting the behavior of an edge node on a target object, as provided in an embodiment of this application. Figure 4 As shown, the behavior prediction performed by edge nodes on target objects can include:

[0048] S401: Obtain the target object in the image frame of the surveillance video. First, obtain the target object for which behavior prediction needs to be performed, such as a person or animal.

[0049] S402: Determine if the target object exhibits abnormal behavior based on behavioral indicators. After acquiring the target object, proceed to the first stage to preliminarily determine if the target object exhibits abnormal behavior. If the target object exhibits abnormal behavior, proceed to the second stage (S403); if the target object does not exhibit abnormal behavior, perform behavior prediction on the next image frame containing the target object. Behavioral indicators may include holding an inappropriate tool, climbing, or impacting. For example, if the target object is holding a tool such as an iron rod, it is determined that the target object exhibits abnormal behavior.

[0050] It should be noted that abnormal behavior also includes other behaviors that may affect the security of the airport perimeter, which are not specifically limited in this application embodiment.

[0051] S403: Perform preprocessing and feature extraction on the image frame containing the target object to obtain abnormal behavior features. For target objects initially determined to have abnormal behavior, preprocessing, such as noise reduction, needs to be performed on the image frame containing the target object, followed by feature extraction to obtain abnormal behavior features.

[0052] S404: Input the abnormal behavior features into the behavior prediction model to obtain the behavior prediction results. Finally, the behavior prediction model performs prediction on the abnormal behavior features. If the abnormal behavior is determined by the behavior prediction model to be an intrusion behavior, the behavior prediction result is that the target object has intrusion behavior; if the abnormal behavior is determined by the behavior prediction model to be a non-intrusion behavior, the behavior prediction result is that the target object has no intrusion behavior.

[0053] In some embodiments, the behavior prediction model described above is a trained neural network model, which is trained by a central server and then sent to the edge nodes after training. The training process of the behavior prediction model by the central server includes:

[0054] First, behavioral sample data is acquired, which includes input images and result labels. The input images are divided into two categories: images where the target object exhibits intrusive behavior and images where the target object does not exhibit intrusive behavior. The corresponding result labels are "target object exhibits intrusive behavior" and "target object does not exhibit intrusive behavior."

[0055] Secondly, feature extraction is performed on the input image to obtain the behavioral features of the target object in the input image.

[0056] The behavioral features are then input into the behavior prediction model to be trained to obtain the prediction results output by the behavior prediction model.

[0057] Finally, the training loss is calculated based on the prediction results and the result labels. If the training loss is greater than or equal to the preset loss threshold, the model parameters of the behavior prediction model to be trained are adjusted according to the training loss. If the training loss is less than the preset loss threshold, the model parameters of the behavior prediction model to be trained are output to obtain the behavior prediction model.

[0058] In some embodiments, the edge node includes a feature library containing behavioral features that have been previously identified as intrusion behaviors. When the behavior prediction model performs behavior prediction, the feature library can be used to determine whether the target object has intrusion behaviors, thereby improving the accuracy of behavior prediction.

[0059] In some embodiments, as can be seen from step S403, the edge node can also obtain behavioral features by performing behavior prediction. If the behavior prediction result obtained after step S404 indicates that the target object has intrusion behavior, the edge node generates and sends the abnormal behavior features obtained in step S403 to the central server, and the central server updates the feature library in the edge node according to the abnormal behavior features.

[0060] Specifically, edge nodes send abnormal behavior features to the central server, which then sends these features to all edge nodes within the airport perimeter. If an edge node's feature library does not already contain the abnormal behavior feature, it will be added to the feature library to optimize the feature library of each edge node and improve the accuracy of edge nodes in predicting intrusion behavior.

[0061] S400: If the behavior prediction result indicates that the target object has intrusive behavior, the edge node generates a video generation strategy for the target object.

[0062] In some embodiments, if the behavior prediction result indicates that the target object is engaging in intrusion behavior, it means that there may be an intrusion within the airport perimeter. Therefore, users need to review the surveillance video to determine whether this situation requires intervention. To avoid video lag caused by network issues when acquiring real-time surveillance video, edge nodes can immediately generate a video generation strategy for the target object after predicting intrusion behavior. This strategy guides the airport perimeter monitoring and early warning system to store surveillance video containing image frames containing the target object. Once the surveillance video is stored in the system, users will not experience lag when retrieving it.

[0063] In some embodiments, the edge node generates a video generation strategy for the target object by: obtaining the time point of the image frame containing the target object in the monitoring video; and marking the video segment of the monitoring video that starts from a preset time before the time point and ends at a preset time after the time point as the target video to obtain the video generation strategy.

[0064] For example, Figure 5 This is a schematic diagram illustrating a video generation strategy based on a surveillance video timeline, provided as an embodiment of this application. For example... Figure 5 As shown, the surveillance video is 60 seconds long. The edge node performs behavior prediction on the target object in the image frame at the 30th second of the surveillance video. The behavior prediction result obtained is that the target object exhibits intrusion behavior. Figure 5 It can be seen that the image frame containing the target object appears in the monitoring video at 30 seconds. If the preset time is set to 10 seconds, then the video segments within the time frame of 20 to 40 seconds, which is 10 seconds before and 10 seconds after 30 seconds, will be marked as the target video.

[0065] It should be noted that the preset duration can be based on actual user needs, and no specific limitation is made in this embodiment.

[0066] S500: Edge nodes send video generation strategies to the P2P network and behavior prediction results to the central server.

[0067] In some embodiments, in order to coordinate and allocate the tasks that each node in the P2P network needs to complete, the P2P network also includes a task allocation node, which is used to allocate the caching task of the target video to each storage node in the P2P network according to the video generation strategy.

[0068] In some embodiments, since the airport perimeter monitoring and early warning system may include multiple edge nodes and multiple clients, the speed at which different clients acquire video from the edge nodes is affected by the distance between them. Furthermore, if multiple clients simultaneously request video from an edge node, that edge node cannot send the video to multiple clients at the same time, reducing the speed at which users acquire the target video. Therefore, to improve the speed at which users acquire the target video, the target video can be stored in a storage node of a P2P network. When a user needs to acquire the target video, the task allocation node in the P2P network can select a suitable storage node and send the target video to the client. Based on this, the edge node needs to send the video generation strategy to the task allocation node in the P2P network, which then allocates storage tasks for the target video to various storage nodes.

[0069] For example, if an edge node predicts that an intrusion is about to cross the airport perimeter fence in the 30th frame of the surveillance video, it generates a video generation strategy that "the 20th to 40th frames of the surveillance video are the target video" and sends the video generation strategy to the task allocation node in the P2P network.

[0070] In some embodiments, in order to notify users of intrusion behavior in a timely manner, when an edge node predicts that an intrusion behavior has occurred on a target object, it will also send the behavior prediction result to the central server for processing.

[0071] For example, if an edge node predicts that a target object in a certain frame of a surveillance video is exhibiting intrusion behavior, it will send the prediction result to the central server for processing.

[0072] S600: The P2P network uses a video generation strategy to obtain target videos containing image frames of the target object from edge nodes.

[0073] In some embodiments, in order to store the target video in a timely manner, after receiving the video generation strategy sent by the edge node, the task allocation node in the P2P network will obtain the target video containing the image frame where the target object is located from the edge node based on the video generation strategy.

[0074] For example, the edge node generates a video generation strategy that "the 20s-40s of the monitored video is the target video" and sends it to the task allocation node in the P2P network. Based on this video generation strategy, the task allocation node obtains the video segment from the edge node for the 20s-40s.

[0075] S700: The P2P network caches the target video to the target storage node and sends the target video to a non-target storage node.

[0076] In some embodiments, the closer the storage node is to the edge node, the faster the target video can be retrieved from the edge node. Therefore, the task allocation node in the P2P network is responsible for selecting the storage node closest to the edge node as the target storage node based on the distance between the edge node and the storage node, and caching the target video to the target storage node.

[0077] For example, the P2P network includes storage node 1, storage node 2, and storage node 3. Among them, storage node 1 is closest to the edge node that generates the video generation strategy, so the task allocation node selects storage node 1 as the target storage node and caches the target video in storage node 1.

[0078] In some embodiments, since the airport perimeter monitoring and early warning system includes multiple clients, when multiple clients simultaneously retrieve target video from the same storage node, the storage node may become overloaded, preventing it from sending the target video to multiple clients at the same time and reducing the speed at which some users can retrieve the target video. Therefore, the task allocation node in the P2P network is also responsible for sending the target video to other non-target storage nodes that are not closest to the edge node. When a user needs to retrieve the target video, the task allocation node coordinates and selects a suitable storage node to send the target video to the client.

[0079] For example, the P2P network includes storage node 1, storage node 2, and storage node 3. Storage node 1 is closest to the edge node that generates the video generation strategy. Storage nodes 2 and 3 are non-target storage nodes, and the task allocation node sends the target video cached in storage node 1 to storage nodes 2 and 3 respectively.

[0080] S800: The central server generates early warning information based on the behavior prediction results and sends the early warning information to the client.

[0081] In some embodiments, in order to promptly notify users of intrusion events, the central server, after receiving the behavior prediction results sent by the edge node, generates early warning information based on the behavior prediction results and sends the early warning information to the client.

[0082] For example, the central server receives a behavior prediction result from an edge node indicating that a target object is preparing to cross the fence, generates an early warning message "Intrusion has occurred at the airport perimeter" and sends it to the client. The client receives the early warning message and displays an alert.

[0083] In some embodiments, after the central server sends the warning information to the client, a warning notification appears on the client. When the user receives the warning notification and needs to view the target video, they will send a video playback request. At this time, the client can retrieve the target video that has been cached in the storage nodes of the P2P network according to the user's video playback request, without needing to load the video in real time, thus improving the smoothness of the user's viewing of the target video.

[0084] In some embodiments, the task allocation node in the P2P network can also select the storage node to send the target video to the client based on the distance relationship between the client and the storage node.

[0085] Specifically, in response to a user's video playback request, the client sends a video information retrieval request to the task allocation node in the P2P network; the task allocation node, in response to the video information retrieval request, obtains the target node information and sends the target node information to the client. The target node information includes the index address of the storage node containing the target video.

[0086] In some embodiments, the task allocation node in a P2P network needs to determine the target node information based on two conditions: the distance between the client and the storage node, and the load of the storage node. Figure 6 This is a flowchart illustrating a method for obtaining target node information provided in an embodiment of this application. Figure 6 As shown, the methods for obtaining target node information include: S601: Obtaining the nearest storage node to the client. Since the storage node closest to the client transmits the target video to the client the fastest, this node is obtained first.

[0087] If the load of the nearest storage node is less than the load threshold, then step S602 is executed: obtain the index address of the nearest storage node as the target node information. When the load of the nearest storage node does not exceed the load threshold, it indicates that the nearest storage node can still handle client requests. Therefore, the index address of the nearest storage node is used as the target node information, enabling the client to find the corresponding storage node based on the index address to obtain the target video.

[0088] If the load of the nearest storage node is greater than or equal to the load threshold, then step S603 is executed: obtain the index address of a non-nearest storage node whose load is less than the load threshold as the target node information. When the load of the nearest storage node exceeds the load threshold, it indicates that the nearest storage node can no longer immediately process the client's request. Therefore, to ensure the speed at which the client obtains the target video, the task allocation node in the P2P network will assign the index address of another storage node whose load does not exceed the load threshold and is relatively close to the client as the target node information, so that the client can find the corresponding storage node based on the index address to obtain the target video.

[0089] In some embodiments, once the client obtains the target node information, it can retrieve the target video based on that information: the client sends a video retrieval request to the storage node pointed to by the index address in the target node information, and the storage node responded to the request by sending the target video back to the client. Thus, the airport perimeter monitoring and early warning system achieves the function of quickly and smoothly sending the target video after the early warning to the user for viewing.

[0090] For example, a user sends a video playback request to client 1. Client 1 then sends a video information retrieval request to a task allocation node in the P2P network. The task allocation node determines the nearest storage node as storage node 1 based on the distance between client 1 and various storage nodes in the P2P network, and obtains the load of storage node 1. Finding that the load of storage node 1 is less than the load threshold, the task allocation node sends the index address of storage node 1 to client 1. After receiving the index address of storage node 1, client 1 locates storage node 1 using that address and sends a video retrieval request to it. Storage node 1 then sends the target video it stores to client 1.

[0091] For another example, a user sends a video playback request to client 2. Client 2 sends a video information retrieval request to the task allocation node in the P2P network. The task allocation node determines the nearest storage node as storage node 3 based on the distance between client 2 and various storage nodes in the P2P network, and obtains the load of storage node 3. It finds that the load of storage node 3 is greater than the load threshold, so it cannot immediately send the video to the client through storage node 3. Therefore, it continues to search for the second closest storage node 2 to the client. It finds that the load of storage node 2 is less than the load threshold, so it sends the index address of storage node 2 to client 2. After receiving the index address of storage node 2, client 2 locates storage node 2 through that index address and sends a video retrieval request to storage node 2. Storage node 2 can then send the target video it stores to client 2.

[0092] In some embodiments, the airport perimeter monitoring and early warning system further includes a feedback unit that can receive and analyze user-provided feedback data. The airport perimeter monitoring and early warning method can also optimize the feature library of the behavior prediction model and the duration of the target video based on user feedback data.

[0093] In some embodiments, the feedback data includes warning judgment data. Warning judgment data is data from the user's assessment of warning information sent to the client. For example, the warning information may indicate a genuine and dangerous intrusion, or it may be a false alarm indicating a non-dangerous intrusion. After receiving the warning judgment data input by the user, the feedback unit sends the warning judgment data to the central server. The central server can optimize the feature library in the edge nodes based on the warning judgment data. For example, it can remove behavioral features that indicate a false alarm indicating a non-dangerous intrusion from the edge nodes to reduce the false alarm rate.

[0094] In some embodiments, the feedback data also includes video viewing data. Video viewing data refers to the time period during which the user watches the target video. For example, the target video is 20 seconds long, and the user can start watching from the 5th second by dragging the progress bar, or exit watching at the 15th second. After receiving the user-inputted video viewing data, the feedback unit sends the data to the central server. The central server can adjust the duration of the target video captured by the edge nodes when generating the video generation strategy based on the video viewing data. For example, if the user starts watching the target video from the 5th second by dragging the progress bar and exits watching at the 15th second, the preset duration can be adjusted to 5 seconds. The video frame containing the target object, within 5 seconds before and 5 seconds after a specific time point in the monitoring video, is marked as the target video, making the target video length more consistent with the user's viewing habits and improving the user experience when watching the target video.

[0095] In summary, the airport perimeter monitoring and early warning method provided in this application includes five stages: model training, edge computing and decision-making, video caching and distribution, user request and response, and feedback and optimization. Figure 7 A timing diagram of an airport perimeter monitoring and early warning method provided in an embodiment of this application.

[0096] like Figure 7 As shown, during the model training phase, the central server obtains behavioral sample data from the data source, performs data cleaning and feature extraction on the behavioral sample data, then trains the behavioral prediction model, and finally sends the trained behavioral prediction model to the client.

[0097] During the edge computing and decision-making phase, after receiving the behavior prediction instruction from the central server, the edge nodes begin to predict intrusion behavior of target objects in the image frames of the acquired surveillance video. Based on the behavior prediction results, they generate video generation strategies for target objects exhibiting intrusion behavior. Simultaneously, they feed back the predicted intrusion features to the central server, which then updates the feature database across all edge nodes. In this phase, the central server also generates early warning notifications based on the feedback intrusion features and behavior prediction results, and sends these notifications to clients to provide alerts.

[0098] During the video caching and distribution phase, edge nodes send the video generation strategy to the task allocation nodes in the P2P network. The task allocation nodes cache the target video and distribute it to various storage nodes according to the video generation strategy.

[0099] During the user request and response phase, after receiving the warning notification, the user sends a video viewing request to the client. The client then sends a video information retrieval request to the task allocation node in the P2P network. In response to the video information retrieval request, the task allocation node sends the target node information to the client. The client can then retrieve the target video from the corresponding storage node based on the target node information. The storage node sends the target video back to the client, and the client then plays the target video for the user to watch.

[0100] During the feedback and optimization phase, the feedback unit receives and analyzes the feedback data provided by the user, and then sends the analysis results to the central server. The central server optimizes the feature library and video generation strategy based on the analysis results of the user feedback data.

[0101] This application also provides an airport perimeter monitoring and early warning system, comprising: a central server configured to: send prediction instructions to edge nodes, generate early warning information based on behavior prediction results, and send the early warning information to clients; edge nodes configured to: acquire monitoring video of the airport perimeter in response to the prediction instructions; perform behavior prediction on target objects in the image frames of the monitoring video to obtain behavior prediction results of the target objects, the behavior prediction results including whether the target objects exhibit intrusion behavior or not; if the behavior prediction result indicates that the target objects exhibit intrusion behavior, generate a video generation strategy for the target objects; send the video generation strategy to a P2P network, and send the behavior prediction results to the central server; the P2P network, comprising multiple storage nodes and multiple clients, configured to: acquire target video containing image frames of the target objects from the edge nodes based on the video generation strategy; cache the target video in a target storage node, and send the target video to a non-target storage node, the target storage node being the storage node closest to the edge nodes among the multiple storage nodes, and the non-target storage node being the storage node not closest to the edge nodes among the multiple storage nodes.

[0102] As can be seen from the above technical solutions, the airport perimeter monitoring and early warning method provided in this application can collect monitoring videos of the airport perimeter through edge nodes and perform behavior prediction on target objects in the image frames of the monitoring videos. When intrusion behavior of the target object is predicted, the behavior prediction result is sent to the central server, which generates early warning information and sends it to the client for early warning, thus playing a role in early warning before the intrusion behavior occurs. In addition, the edge nodes can also generate video generation strategies based on the behavior prediction results. The P2P network stores the monitoring images of the image frames containing the target object containing intrusion behavior according to the video generation strategy. Users can obtain the cached video when they need to view the monitoring video without waiting for the video to load, improving the smoothness of users viewing the monitoring video and thus improving the efficiency of handling intrusion behavior.

[0103] Similar parts between the embodiments provided in this application can be referred to mutually. The specific implementation methods provided above are only a few examples under the overall concept of this application and do not constitute a limitation on the scope of protection of this application. For those skilled in the art, any other implementation methods extended from the solution of this application without creative effort shall fall within the scope of protection of this application.

Claims

1. A method for monitoring and early warning of airport perimeter fencing, characterized in that, An application is made in an airport perimeter monitoring and early warning system. The system includes a central server, edge nodes, and a P2P network. The P2P network includes multiple storage nodes and multiple clients. The method includes: The central server sends prediction instructions to the edge nodes. The edge node responds to the prediction command and captures surveillance video of the airport perimeter. The edge nodes perform two-stage behavior prediction, including: The behavior indicators are used to determine whether the target object in the image frame of the surveillance video exhibits abnormal behavior. If the target object exhibits the abnormal behavior, then preprocessing and feature extraction are performed on the image frame containing the target object to obtain the abnormal behavior features; The abnormal behavior features are input into the behavior prediction model to obtain the behavior prediction result of the target object; the behavior prediction result includes whether the target object has intrusion behavior or whether the target object has no intrusion behavior. If the behavior prediction result indicates that the target object exhibits intrusion behavior, the edge node generates a video generation strategy for the target object, including: Obtain the time point of the image frame containing the target object in the surveillance video; The video segments of the surveillance video that start from a preset time before the time point and end at a preset time after the time point are marked as target videos; The edge node sends the video generation strategy to the P2P network and sends the behavior prediction result to the central server; The P2P network, based on the video generation strategy, obtains a target video containing the image frame where the target object is located from the edge nodes. The P2P network caches the target video to a target storage node and sends the target video to a non-target storage node, wherein the target storage node is the storage node that is closest to the edge node among the plurality of storage nodes, and the non-target storage node is the storage node that is not closest to the edge node among the plurality of storage nodes; The central server generates early warning information based on the behavior prediction results and sends the early warning information to the client. In response to the user's video playback request, the client sends a video information retrieval request to the P2P network; In response to the video information acquisition request, the P2P network acquires the load of the nearest storage node; if the load of the nearest storage node is less than a load threshold, it acquires the index address of the nearest storage node as target node information; if the load of the nearest storage node is greater than or equal to the load threshold, it acquires the index address of a non-nearest storage node whose load is less than the load threshold as target node information; and sends the target node information to the client; wherein, the nearest storage node is the storage node closest to the client, and the non-nearest storage node is the storage node not closest to the client; The client sends a video retrieval request to the storage node pointed to by the index address in the target node information; The storage node pointed to by the index address in the target node information responds to the video acquisition request and sends the target video to the client.

2. The airport perimeter monitoring and early warning method according to claim 1, characterized in that, The method further includes: Behavioral sample data is obtained through the central server, and the behavioral sample data includes input images and result labels; Feature extraction is performed on the input image to obtain the behavioral features of the target object in the input image; The behavioral features are input into the behavior prediction model to be trained to obtain the prediction results output by the behavior prediction model to be trained. Calculate the training loss based on the prediction results and the result labels; If the training loss is greater than or equal to a preset loss threshold, the model parameters of the behavior prediction model to be trained are adjusted according to the training loss. If the training loss is less than the preset loss threshold, the model parameters of the behavior prediction model to be trained are output to obtain the behavior prediction model; The behavior prediction model is sent to the edge node.

3. The airport perimeter monitoring and early warning method according to claim 1, characterized in that, The method further includes: If the behavior prediction result indicates that the target object has intrusion behavior, the edge node generates behavior features corresponding to the intrusion behavior of the target object; The edge node sends the behavioral characteristics to the central server; The central server updates the feature library in the edge nodes based on the behavioral characteristics.

4. The airport perimeter monitoring and early warning method according to claim 3, characterized in that, The system further includes a feedback unit, and the method further includes: The feedback unit receives and analyzes user-provided feedback data, including early warning judgment data and video viewing data. The feedback data is sent to the central server; The central server optimizes the feature library in the edge nodes based on the early warning judgment data, and optimizes the video generation strategy based on the video viewing data.

5. An airport perimeter monitoring and early warning system, characterized in that, include: The central server is configured to: send prediction instructions to edge nodes, generate early warning information based on behavior prediction results, and send the early warning information to the client; Edge nodes are configured to capture surveillance video of the airport perimeter in response to the prediction command; Perform two-stage behavior prediction, including: The behavior indicators are used to determine whether the target object in the image frame of the surveillance video exhibits abnormal behavior. If the target object exhibits the abnormal behavior, then preprocessing and feature extraction are performed on the image frame containing the target object to obtain the abnormal behavior features; The abnormal behavior features are input into the behavior prediction model to obtain the behavior prediction result of the target object; the behavior prediction result includes whether the target object has intrusion behavior or whether the target object has no intrusion behavior. If the behavior prediction result indicates that the target object exhibits intrusive behavior, a video generation strategy for the target object is generated, including: Obtain the time point of the image frame containing the target object in the surveillance video; The video segments of the surveillance video that start from a preset time before the time point and end at a preset time after the time point are marked as target videos; The video generation strategy is sent to the P2P network, and the behavior prediction result is sent to the central server; The P2P network, including multiple storage nodes and multiple clients, is configured to: based on the video generation strategy, obtain a target video containing the image frame where the target object is located from the edge nodes; The target video is cached to the target storage node, and the target video is sent to a non-target storage node, wherein the target storage node is the storage node that is closest to the edge node among the plurality of storage nodes, and the non-target storage node is the storage node that is not closest to the edge node among the plurality of storage nodes; The client is configured to: in response to a user's video playback request, send a video information retrieval request to the P2P network; The P2P network is further configured to: in response to the video information acquisition request, acquire the load of the nearest storage node; if the load of the nearest storage node is less than a load threshold, acquire the index address of the nearest storage node as target node information; if the load of the nearest storage node is greater than or equal to the load threshold, acquire the index address of a non-nearest storage node whose load is less than the load threshold as target node information; and send the target node information to the client; wherein, the nearest storage node is the storage node closest to the client, and the non-nearest storage node is the storage node not closest to the client; The client is also configured to send a video retrieval request to the storage node pointed to by the index address in the target node information; The storage node pointed to by the index address in the target node information is configured to send the target video to the client in response to the video acquisition request.

Images