Low-altitude flight real-time situation awareness method, system and device and storage medium

Abstract

The invention relates to the field of aviation safety and unmanned aerial vehicle management, and discloses a low-altitude flight real-time situation awareness method, system and device and a storage medium, and the method comprises the steps: deploying a radar group composed of a millimeter wave radar and an optical camera in an airport area, the deployment density of the radar group meeting the detection precision requirement of a small unmanned aerial vehicle, each device communicates with the real-time data processing platform through a distributed network; target detection data of the radar group, image data of the AI image recognition module, GPS positioning data and ADS-B flight data are received in real time; and processing the multi-source data through a preset data fusion algorithm, including time-space calibration of the unified timestamp and the space coordinate reference. A radar group composed of millimeter wave radars and optical cameras is deployed in an airport area, so that the problems that most traditional low-altitude monitoring systems only depend on a single radar data source, so that the omission ratio of small unmanned aerial vehicles is high, and the misjudgment rate of the small unmanned aerial vehicles and targets such as birds is high are solved.

Description

Technical Field

[0001] The present invention relates to the fields of aviation safety and drone management, and specifically to a method, system, device and storage medium for real-time situation awareness of low-altitude flight. Background Art

[0002] With the increasing popularity of drone technology, the safety management of low-altitude flights at airports has become increasingly important. Existing technologies have made significant progress in real-time monitoring, data analysis, and processing capabilities, providing a fundamental guarantee for airport safety.

[0003] Most traditional low-altitude monitoring systems rely solely on a single radar data source. Due to their insufficient ability to capture reflected signals from small drones and the lack of multi-dimensional feature verification, they result in a high rate of missed detection of small drones and a high rate of misjudgment of targets such as birds. Summary of the Invention

[0004] In response to the shortcomings of the existing technology, the present invention provides a real-time situational awareness method, system, device and storage medium for low-altitude flight, which solves the problem that most traditional low-altitude monitoring systems rely only on a single radar data source, have insufficient ability to capture the reflected signals of small drones and lack multi-dimensional feature verification, resulting in a high missed detection rate of small drones and a high misjudgment rate of targets such as birds.

[0005] In a first aspect, the present invention provides the following technical solutions: a method for real-time situation awareness of low-altitude flight, applied to a real-time data processing platform, comprising: Deploy a radar system consisting of millimeter-wave radars and optical cameras in the airport area. The deployment density of the radar system meets the accuracy requirements for small drone detection, and each device communicates with the real-time data processing platform through a distributed network. Receive target detection data from the radar group, image data from the AI ​​image recognition module, GPS positioning data, and ADS-B flight data in real time; Process multi-source data through preset data fusion algorithms, including: time-space alignment to unify timestamps and spatial coordinate benchmarks; matching features using machine learning models trained on historical data to identify and verify target types; and extracting position, velocity, radar reflection characteristics, and image texture features based on a deep learning framework for real-time correlation analysis. Generate flight situation information including location, speed, type, motion trajectory and threat level, and send it to the terminal device at a frequency of seconds; When a target is detected entering a no-fly zone or following an abnormal trajectory, the early warning mechanism is triggered, sending early warning information containing real-time position, speed, threat level, and historical trajectory to the air traffic control system; The real-time data processing platform adopts a distributed computing architecture, including edge computing nodes and cloud-based elastic computing resources. The edge nodes preprocess the raw data, and the cloud takes over the in-depth analysis tasks during peak load.

[0006] Preferably, the millimeter-wave radar and the optical camera in the radar group work together: the millimeter-wave radar detects the target position and motion parameters around the clock, the optical camera collects target image data, and the AI ​​image recognition module extracts visual features such as texture and shape based on the image.

[0007] Preferably, the machine learning model is trained by using historical radar echo data, image data, and flight data of drones, flying birds, and light aircraft, and supports online updates to adapt to new target features.

[0008] Preferably, the threat level is calculated based on target type, flight speed, distance from the airport runway and historical trajectory, and is divided into three levels: low, medium and high, corresponding to different early warning response strategies.

[0009] Preferably, the edge computing node preprocessing includes: target detection based on lightweight convolutional neural network, preliminary classification based on support vector machine, and trajectory prediction based on Kalman filter.

[0010] In a second aspect, the present invention provides the following technical solutions: a low-altitude flight real-time situation awareness system, comprising: The radar group, consisting of a millimeter-wave radar and an optical camera, is used to output target detection data and image data; The AI ​​image recognition module works in conjunction with the radar group to extract and classify target image features; Signal receiving unit, used to receive GPS positioning data and ADS-B flight data; The real-time data processing platform integrates and processes multi-source data through a distributed network, generates real-time flight situation information and implements real-time situation awareness methods for low-altitude flight.

[0011] In a third aspect, the present invention provides the following technical solutions: a low-altitude flight real-time situation awareness device, applied to a real-time data processing platform, comprising: Multi-source data receiving module, used to obtain data from radar group, AI image recognition module, GPS and ADS-B; Data fusion processing module, with built-in deep learning data fusion algorithm, performs spatiotemporal calibration, feature extraction, target classification and threat level assessment; Situation generation and transmission module, used to generate flight situation information containing multi-dimensional target information and push it in seconds; An early warning trigger module is used to generate early warning information containing real-time location, threat level and historical trajectory when an anomaly is detected; the mathematical expression of the data fusion algorithm is F = Activation (W2 · Activation (W1 · X + b1) + b2); W1, W2: weight matrix of the neural network layer, used for feature transformation and fusion; b1, b2: bias vectors of the neural network layer; F: The fused target feature vector, which contains flight situation information such as position, speed, type, motion trajectory, and threat level; Activation: Activation function, such as ReLU, Sigmoid, or Swish; X is the concatenation vector of the radar feature vector and the image feature vector, with the dimension D_radar+D_image.

[0012] Preferably, the data fusion processing module includes an edge computing pre-processing unit and a cloud collaborative processing unit: The edge computing pre-processing unit is deployed near the radar group and integrates a lightweight detection model and filtering algorithm to complete raw data noise reduction, target initial screening and preliminary positioning; The cloud-based collaborative processing unit is linked with the edge node through a high-speed encrypted channel to achieve real-time modeling of large-scale data based on the cloud-based GPU cluster.

[0013] In a fourth aspect, the present invention provides the following technical solution: a computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned method for real-time situational awareness of low-altitude flight when executing the computer program.

[0014] In a fifth aspect, the present invention provides the following technical solution: a readable storage medium having a computer program stored thereon, and the computer program, when executed by a processor, implements the above-mentioned method for real-time situational awareness of low-altitude flight.

[0015] The present invention provides a method, system, device, and storage medium for real-time situation awareness during low-altitude flight. This method has the following beneficial effects: 1. The present invention deploys a radar group consisting of millimeter-wave radar and optical cameras in the airport area, and uses an AI image recognition module to extract visual features such as target image texture and shape, and then fuses and analyzes the radar physical features with the visual semantic features, thereby improving the traditional low-altitude monitoring system, which mostly relies on a single radar data source. Due to the insufficient ability to capture the reflected signals of small drones and the lack of multi-dimensional feature verification, it results in a high missed detection rate of small drones and a high misjudgment rate of targets such as flying birds.

[0016] 2. The present invention adopts a distributed architecture that collaborates edge computing nodes and cloud-based elastic computing resources in a real-time data processing platform. The edge nodes perform lightweight CNN target detection, SVM preliminary classification, and Kalman filter trajectory prediction, thereby achieving near-end noise reduction and preprocessing of the original data. The cloud-based GPU cluster takes over the deep analysis task through a high-speed encrypted channel, thereby improving the traditional data processing platform that mostly adopts a centralized computing architecture. Due to high data transmission latency and limited local computing power, the update of flight situation information lags and cannot cope with high-density data processing needs.

[0017] 3. The present invention uses a data fusion algorithm to perform spatiotemporal calibration, multi-source feature extraction, and machine learning model classification, and comprehensively calculates the threat level (low / medium / high) based on target type, flight speed, distance from the runway, and historical trajectory, thereby generating structured warning information containing the threat level. This improves the traditional warning mechanism, which mostly adopts a "one-size-fits-all" strategy. Due to the lack of quantitative assessment of target risks and dynamic trajectory analysis, it causes unreasonable allocation of air traffic control resources and delayed response to high-threat targets.

[0018] 4. The present invention utilizes historical multi-source data of drones, flying birds, and light aircraft for training through machine learning models and supports online updates, thereby continuously optimizing target feature matching capabilities. This improves the problem that traditional detection models are mostly based on fixed parameter training and cannot adapt to the iteration of new drone features or changes in complex airspace environments, resulting in long-term system detection performance degradation and insufficient ability to identify new threats.

[0019] 5. The present invention uses an early warning trigger module to generate early warning information containing real-time location, threat level, and historical trajectory when an anomaly is detected. It also interacts with the air traffic control system in real time, thereby achieving a response within seconds to risks such as no-fly zone intrusions and abnormal trajectories. This improves the problem that traditional early warning information mostly lacks historical trajectory correlation and risk classification. Since air traffic controllers need to manually integrate data and judge threat levels, it results in poor decision-making timeliness and lengthy emergency response processes. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 This is a flow chart of the low-altitude flight real-time situation awareness method proposed by the present invention; Figure 2 This is a system architecture diagram of the low-altitude flight real-time situation awareness system proposed by the present invention; Figure 3 This is a device module diagram of the low-altitude flight real-time situation awareness device proposed by the present invention. DETAILED DESCRIPTION

[0021] The following will clearly and completely describe the technical solution of the present invention in conjunction with the accompanying drawings. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0022] Please see the attached Figure 1 In a first embodiment of the present invention, a method for real-time situation awareness of low-altitude flight is provided, which is applied to a real-time data processing platform and includes: A radar system consisting of millimeter-wave radars and optical cameras is deployed in the airport area. The deployment density of the radar system meets the accuracy requirements for small drone detection, and each device communicates with the real-time data processing platform through a distributed network. Receive target detection data from the radar group, image data from the AI ​​image recognition module, GPS positioning data, and ADS-B flight data in real time; Process multi-source data through preset data fusion algorithms, including: Time and space calibration unifies the timestamp and spatial coordinate reference; Use machine learning models trained with historical data to match features, identify, and verify target types; Extract position, speed, radar reflection features, and image texture features based on a deep learning framework and conduct real-time correlation analysis; Generate flight situation information including location, speed, type, motion trajectory and threat level, and send it to the terminal device at a frequency of seconds; When a target is detected entering a no-fly zone or following an abnormal trajectory, the early warning mechanism is triggered, sending early warning information containing real-time position, speed, threat level, and historical trajectory to the air traffic control system; The real-time data processing platform adopts a distributed computing architecture, including edge computing nodes and elastic computing resources in the cloud. The edge nodes preprocess the raw data, and the cloud takes over the in-depth analysis tasks during peak load.

[0023] Specifically, through the collaboration of a radar group consisting of millimeter-wave radar and an optical camera and an AI image recognition module, combined with a multi-source data fusion algorithm, multi-dimensional information such as target position, speed, radar reflection characteristics, and image texture characteristics is obtained to solve the problem of insufficient detection accuracy of small drones by traditional systems and reduce the missed detection rate; the distance between adjacent radar nodes is ≤500 meters, ensuring that the detection probability of targets with a radar reflection area of ​​≤0.01㎡ is ≥95%; edge computing nodes are used to pre-process the raw data (target detection, classification, and preliminary positioning) to reduce data transmission volume and delay; elastic computing resources in the cloud are used to meet large-scale data processing needs and break through local computing bottlenecks. Ensure that flight situation information is updated at a frequency of seconds to improve data processing efficiency and real-time performance; based on the fused target information, monitor the target position and trajectory in real time. When the target is detected entering a no-fly zone or an abnormal flight trajectory occurs, the early warning mechanism is triggered, and accurate early warning information containing real-time position, speed, threat level and historical trajectory is sent to the air traffic control system, providing timely basis for air traffic control decision-making and enhancing airport security protection capabilities; distributed computing architecture (edge ​​computing + cloud collaboration) supports independent data processing by edge nodes when the network is interrupted to ensure system stability; high-speed encrypted channels ensure data transmission security and adapt to high-density, fast-paced low-altitude monitoring scenarios in airports.

[0024] Please see the attached Figure 1 The millimeter-wave radar and optical camera in the radar group work together: the millimeter-wave radar detects the target position and motion parameters all day long, the optical camera collects target image data, and the AI ​​image recognition module extracts visual features such as texture and shape based on the image.

[0025] Specifically, millimeter-wave radar is not affected by weather, and provides all-weather basic detection data such as target distance, speed, and direction, solving the problem of blind spots in monitoring of traditional optical means in harsh environments; the optical camera collects target images, and the AI ​​module extracts visual features such as texture and shape from them, which complement the position and motion parameters of the millimeter-wave radar, and realizes the "radar physical characteristics + visual semantic features" multimodal description of small UAVs, improving target recognition accuracy and anti-interference ability; the machine learning model combined with multi-source feature training can accurately distinguish low-altitude targets such as drones and birds, reducing the misjudgment caused by similar echoes of traditional single radars, especially for small UAVs with small radar reflection areas, and effective identification can be achieved with the assistance of visual features. The radar group composed of millimeter-wave radar and optical camera uses the all-weather detection characteristics of millimeter-wave radar to continuously monitor low-altitude flying targets regardless of weather conditions, and the optical camera assists in identifying the characteristics of the target. The two complement each other and greatly enhance the detection capability of small UAVs. The deployment density of the radar group meets the detection accuracy requirements for small drones, ensuring that any activities of small drones in the airport area can be effectively monitored, reducing the risk of missed detection of small drones due to insufficient detection accuracy, thereby improving the safety of low-altitude flight in the airport; each radar node, AI image recognition module, GPS signal receiving device and ADS-B signal receiving device communicate with the real-time data processing platform through a distributed network, so that all kinds of data collected by these devices can be aggregated to the processing platform in real time. The radar node provides target detection data, the AI ​​image recognition module provides target feature data, the GPS signal receiving device provides precise positioning data, and the ADS-B signal receiving device provides flight data. Multiple data are processed on the real-time data processing platform through a preset data fusion algorithm to achieve efficient fusion of multi-source data, providing a comprehensive and accurate data basis for subsequent precise analysis of target status; the distributed network architecture makes data transmission more efficient and reduces data transmission delay.

[0026] Please see the attached Figure 1 The machine learning model is trained through historical radar echo data, image data and flight data including drones, flying birds and light aircraft, and supports online updates to adapt to new target features.

[0027] Specifically, the model is trained using historical multi-source data (radar echoes, images, flight data) from drones, birds, light aircraft, etc., to learn the characteristic differences of different targets (such as radar reflection intensity, image contours, and flight trajectory patterns), thereby solving the problem of traditional systems' insufficient ability to distinguish between small drones and birds and other targets, and reducing the false alarm rate; it supports online model updates, can incorporate new target data (such as new drone models) in real time, continuously optimize feature recognition capabilities, avoid detection failures caused by target type iterations, and ensure that the system maintains stable performance in complex and changeable low-altitude environments; based on real-time data, it continuously adjusts model parameters to adapt to the target distribution characteristics of different time periods and different airspaces, improves the recognition efficiency of emerging threats (such as modified drones), and enhances the long-term reliability and security of the system.

[0028] Please see the attached Figure 1 The threat level is calculated based on the target type, flight speed, distance from the airport runway and historical trajectory, and is divided into three levels: low, medium and high, corresponding to different early warning response strategies.

[0029] Specifically, the risk level is quantitatively assessed based on the target type (e.g., drones pose a higher threat than birds), flight speed (high-speed targets pose a higher threat), distance from the runway (close-range targets have a higher risk) and historical trajectory (abnormal trajectories indicate potential threats); differentiated response strategies are matched to low-, medium- and high-level threats (e.g., low-level threats only require monitoring, while high-level threats trigger interception plans), and air traffic control resources are concentrated to prioritize high-risk targets, thereby improving emergency response efficiency; the standardized threat level system provides controllers with a clear basis for decision-making, reduces human error, and supports the system to automatically trigger corresponding handling processes (e.g., notifying the military of high-threat targets), shortening the time link from early warning to action.

[0030] Please see the attached Figure 1 ,The edge computing node preprocessing includes: target detection based on lightweight ,convolutional neural network, preliminary classification based on support ,vector machine, and trajectory prediction based on Kalman filter.

[0031] Specifically, lightweight CNN quickly detects targets and filters redundant information, and SVM performs preliminary classification to compress data dimensions, reduce the amount of original data transmission, and alleviate network bandwidth pressure, making it particularly suitable for real-time monitoring scenarios in high-density airspace. Edge nodes deployed close to the radar group use Kalman filtering to predict target trajectories in real time, and can independently complete basic processing (such as target screening and trajectory tracking) even when the network is delayed or interrupted, ensuring uninterrupted data processing and meeting the needs of second-level situation updates. Pre-processed key feature data (such as target category and preliminary position) is transmitted to the cloud to avoid computational congestion caused by direct transmission of raw data, allowing the cloud to focus on in-depth analysis (such as threat level assessment and long-term trajectory modeling), thereby improving the overall system processing efficiency.

[0032] Example 2: Please see the attached Figure 2 In a second embodiment of the present invention, the present invention provides a low-altitude flight real-time situation awareness system including a real-time data processing platform, including: The radar group, consisting of a millimeter-wave radar and an optical camera, is used to output target detection data and image data; The AI ​​image recognition module works in conjunction with the radar group to extract and classify target image features; Signal receiving unit, used to receive GPS positioning data and ADS-B flight data; The real-time data processing platform integrates and processes multi-source data through a distributed network, generates real-time flight situation information and implements real-time situation awareness methods for low-altitude flight.

[0033] Specifically, a radar system consisting of a millimeter-wave radar and an optical camera enables all-weather target detection and image acquisition. The signal receiving unit obtains precise positioning and flight status data, and the AI ​​image recognition module supplements visual features to construct a complete dataset covering the target's physical attributes, visual semantics, and dynamic trajectory, addressing the insufficient detection capabilities of traditional single data sources. The real-time data processing platform integrates multi-source data through a distributed network and utilizes a fusion algorithm based on a deep learning framework to perform spatiotemporal calibration, feature association, and target recognition. It transforms heterogeneous information such as radar echoes, image pixels, and positioning coordinates into unified situational information (position, speed, type, trajectory, etc.), improving the accuracy and timeliness of data processing. Based on flight situation information updated at the sub-second level, air traffic controllers can obtain real-time information on the distribution and movement trends of low-altitude targets. Combined with a threat level response mechanism, this system enables rapid early warning and response to risks such as no-fly zone incursions and abnormal trajectories, enhancing the proactive and precise management of airport safety. The distributed network and edge-cloud collaborative architecture support localized data preprocessing and dynamic migration of computing tasks, reducing network transmission latency while overcoming local computing resource limitations. This adapts to the high-volume data processing needs of high-density flight scenarios and ensures system stability and scalability.

[0034] Example 3: Please see the attached Figure 3 In a third embodiment of the present invention, the present invention provides a low-altitude flight real-time situation awareness device, which is applied to a real-time data processing platform, comprising: Multi-source data receiving module, used to obtain data from radar group, AI image recognition module, GPS and ADS-B; Data fusion processing module, with built-in deep learning data fusion algorithm, performs spatiotemporal calibration, feature extraction, target classification and threat level assessment; Situation generation and transmission module, used to generate flight situation information containing multi-dimensional target information and push it in seconds; The early warning trigger module is used to generate early warning information containing real-time location, threat level and historical trajectory when an anomaly is detected; The mathematical expression of the data fusion algorithm is F = Activation (W2 · Activation (W1 · X + b1) + b2); W1, W2: weight matrix of the neural network layer, used for feature transformation and fusion; b1, b2: bias vectors of the neural network layer; F: The fused target feature vector, which contains flight situation information such as position, speed, type, motion trajectory, and threat level; Activation: Activation function, such as ReLU, Sigmoid, or Swish; X is the concatenation vector of the radar feature vector and the image feature vector, with the dimension D_radar+D_image.

[0035] Specifically, the multi-source data receiving module can integrate the physical detection data (distance, speed, direction) of the radar group, the visual feature data (texture, shape) of the AI ​​image recognition module, GPS positioning data and ADS-B flight data (heading, altitude) to form a full-factor data set of the target's position, motion state, appearance characteristics and identity attributes, solving the detection blind spot problem caused by the one-sided information of the traditional single data source; millimeter wave radar data guarantees all-weather detection capability, optical image data enhances the target visual recognition accuracy, positioning and flight data provide time and space benchmarks, multi-source data verify and complement each other, reducing the risk of misjudgment or missed detection caused by environmental interference (such as weather, electromagnetic noise) of a single sensor; standardized data reception The receiving interface provides a unified input format for the data fusion algorithm, ensuring the efficient execution of subsequent processing processes such as spatiotemporal calibration and feature matching, which is the basis for achieving accurate target identification, trajectory prediction and threat assessment; the data fusion processing module eliminates the timestamp differences and spatial coordinate deviations of multi-source data through spatiotemporal calibration, unifies the "relative position" of radar detection, the "absolute coordinates" of GPS, and the "viewing angle range" of image acquisition to the same benchmark, solves the problem of data heterogeneity, and provides a reliable foundation for subsequent analysis; uses a deep learning framework to automatically extract physical features such as frequency and phase from radar echoes, extracts semantic features such as texture and contour from image data, and combines dynamic features such as speed and heading in flight data to form a multi-dimensional target image. The feature vector of the attribute is used to improve the recognition of small, slow and small targets such as small drones; the machine learning model based on historical training data is used to classify the fused features and distinguish target types such as drones, birds, and compliant aircraft; at the same time, combined with parameters such as target flight speed, distance from the runway, and historical trajectory, the threat level (low / medium / high) is calculated through an algorithm to achieve quantitative assessment and hierarchical management of risks; through feature extraction and threat assessment, structured information including target type, location, and risk level is quickly generated, providing a direct basis for real-time situation generation, warning triggering, and air traffic control disposal, shortening the time link from data collection to decision response, and enhancing the system's emergency handling capabilities; through the integration of situation generation and sending modules Multi-dimensional information such as target location, speed, type, motion trajectory, and threat level generates real-time flight status at a frequency of seconds, enabling air traffic controllers to dynamically grasp the distribution and changing trends of low-altitude targets, solving the problem of delayed information updates in traditional systems. Standardized situation information formats (such as dynamic data containing historical trajectories) provide controllers with intuitive and comprehensive decision-making basis, supporting them to quickly judge the threat level of targets and formulate response strategies (such as adjusting the take-off and landing sequence of flights and initiating drone countermeasures). Highly timely information push ensures that abnormal situations (such as no-fly zone intrusions) are discovered in a timely manner. Combined with the early warning trigger module, it can shorten the time difference from monitoring to disposal, improve the airport's response efficiency to sudden security threats, and reduce accident risks.The situation information updated in seconds can be synchronized with other air traffic control subsystems (such as radar navigation and communication systems) in real time to enhance the coordination of various links, realize the closed-loop management of the entire process of low-altitude flight, and improve the overall safety management efficiency; the early warning trigger module monitors the boundaries of the no-fly zone and flight trajectory rules in real time based on the fused target information. When it is detected that the target enters the no-fly zone illegally or makes abnormal maneuvers (such as emergency stops and turns), it immediately generates early warning information containing real-time position, threat level and historical trajectory, solving the problem of traditional monitoring lagging behind in identifying abnormal behavior; the early warning information integrates the target dynamic data (such as speed, heading) and historical trajectory to assist air traffic controllers in quickly Identify target movement trends and potential risk areas. For example, historical trajectory analysis can distinguish between accidental and intentional intrusions, providing a basis for differentiated response. Binding threat levels to warning information (e.g., low-level warnings only prompt monitoring, while high-level warnings trigger interception plans) can guide the air traffic control system to prioritize resource allocation to high-risk targets, optimize emergency response processes, and avoid resource waste. Structured warning information (including location, level, and trajectory) is directly connected to the air traffic control command system, reducing manual information integration time and supporting automated response processes (e.g., notifying the military of high-threat target coordinates). This improves the efficiency of the entire chain from warning to action, ensuring the safety of the airspace around airports.

[0036] Please see the attached Figure 3 ,The data fusion processing module includes an edge computing pre-processing unit and a cloud collaborative ,processing unit: the edge computing pre-processing unit is deployed near the radar group, ,integrates a lightweight detection model and filtering algorithm to complete raw data ,noise reduction, target initial screening and preliminary positioning; The cloud-based collaborative processing unit interacts with edge nodes through high-speed encrypted channels and realizes real-time modeling of large-scale data based on cloud-based GPU clusters.

[0037] Specifically, the edge computing preprocessing unit uses a lightweight detection model (such as lightweight CNN) to perform noise reduction on the radar's original echo and image pixel data, filter out invalid clutter and background interference, and eliminate irrelevant signals (such as ground vehicles and birds) through initial target screening, reducing the amount of redundant data that needs to be transmitted to the cloud and reducing network bandwidth pressure; the physical location close to the radar group deployment shortens the data processing link, and combined with real-time algorithms such as Kalman filtering, completes the initial positioning and trajectory prediction of the target on the edge side, and can independently output basic situation information (such as rough target coordinates and movement direction) even in the event of network delays or interruptions, ensuring the system is accurate in seconds. Continuity of updating capability; key data after pre-processing (such as target category confidence, preliminary position coordinates) is transmitted to the real-time data processing platform, so that the cloud can focus on deep tasks (such as threat level assessment, multi-target correlation analysis), avoid computing congestion caused by direct transmission of raw data, and improve the overall system's concurrent processing capability for high-density targets; for the complex electromagnetic environment or weather conditions of the airport, real-time noise reduction and dynamic filtering on the edge side can improve the quality of front-end data, reduce false detections caused by signal interference, enhance the robustness of the system in harsh environments, and ensure the stability of low-altitude monitoring; the cloud-based collaborative processing unit uses the parallel processing of the cloud-based GPU cluster Computing advantages: Real-time modeling of large-scale data (such as multi-target trajectory data in high-density airspace) pre-processed by edge nodes, solving the problem that the edge side cannot handle sudden large-volume data due to computing power limitations, and ensuring the stability of the system during peak hours; based on deep learning models (such as long short-term memory network LSTM, graph neural network GNN), historical trajectories, multi-target interaction relationships, etc. are analyzed to achieve target flight trend prediction and potential risk modeling (such as group cluster target warning), and enhance the system's forward-looking judgment ability for complex scenarios; automatically adjust cloud resource allocation according to real-time data traffic (such as increasing the number of GPU nodes ), avoid resource waste or insufficient computing power caused by fixed hardware configuration, reduce system deployment and operation and maintenance costs, and adapt to the fluctuating characteristics of airport airspace traffic; high-speed encrypted channels ensure confidentiality during data transmission, and cloud security mechanisms (such as access control and data desensitization) meet the compliance requirements of the aviation regulatory field, while supporting multi-tenant data isolation to protect the data privacy of users at different airports; the cloud centrally stores all historical data, supports model training and parameter updates based on the latest samples (such as algorithm upgrades for new drone features), and realizes rapid iteration of detection strategies through the linkage between edge nodes and the cloud, continuously improving system performance.

[0038] Example 4: The fourth embodiment of the present invention is based on the same inventive concept. The present invention proposes a computer device, including: a processor and a memory; the processor and the memory communicate with each other; the memory is used to store instructions; the processor is used to execute the instructions in the memory to execute the low-altitude flight real-time situation awareness method of the above embodiment.

[0039] Embodiment 5: The fifth embodiment of the present invention, based on the same inventive concept, proposes a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the low-altitude flight real-time situation awareness method of the above embodiment.

[0040] It should be understood that various parts of the present invention can be implemented using hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, an application-specific integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0041] Finally, it should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A low-altitude flight real-time situation awareness method, applied to a real-time data processing platform, characterized in that: include: Deploy a radar system consisting of millimeter-wave radars and optical cameras in the airport area. The deployment density of the radar system meets the accuracy requirements for small drone detection, and each device communicates with the real-time data processing platform through a distributed network. Receive target detection data from the radar group, image data from the AI ​​image recognition module, GPS positioning data, and ADS-B flight data in real time; Process multi-source data through preset data fusion algorithms, including: time-space alignment to unify timestamps and spatial coordinate benchmarks; matching features using machine learning models trained on historical data to identify and verify target types; and extracting position, velocity, radar reflection characteristics, and image texture features based on a deep learning framework for real-time correlation analysis. Generate flight situation information including location, speed, type, motion trajectory and threat level, and send it to the terminal device at a frequency of seconds; When a target is detected entering a no-fly zone or following an abnormal trajectory, the early warning mechanism is triggered, sending early warning information containing real-time position, speed, threat level, and historical trajectory to the air traffic control system; The real-time data processing platform adopts a distributed computing architecture, including edge computing nodes and cloud-based elastic computing resources. The edge nodes preprocess the raw data, and the cloud takes over the in-depth analysis tasks during peak load.

2. The low-altitude flight real-time situation awareness method according to claim 1, characterized in that: The millimeter-wave radar and optical camera in the radar group work together: the millimeter-wave radar detects the target position and motion parameters around the clock, the optical camera collects target image data, and the AI ​​image recognition module extracts visual features such as texture and shape based on the image.

3. The low-altitude flight real-time situation awareness method according to claim 1, characterized in that: The machine learning model is trained using historical radar echo data, image data, and flight data from drones, birds, and light aircraft, and supports online updates to adapt to new target features.

4. The low-altitude flight real-time situation awareness method according to claim 1, characterized in that: The threat level is calculated based on target type, flight speed, distance from the airport runway and historical trajectory, and is divided into three levels: low, medium and high, corresponding to different early warning response strategies.

5. The low-altitude flight real-time situation awareness method according to claim 1, characterized in that: The edge computing node preprocessing includes: target detection based on lightweight convolutional neural network, preliminary classification based on support vector machine, and trajectory prediction based on Kalman filter.

6. Low-altitude flight real-time situation awareness system, characterized by: include: The radar group consists of a millimeter-wave radar and an optical camera, which is used to output target detection data and image data; The AI ​​image recognition module works in conjunction with the radar group to extract and classify target image features; Signal receiving unit, used to receive GPS positioning data and ADS-B flight data; A real-time data processing platform integrates and processes multi-source data through a distributed network, generates real-time flight situation information and executes the method described in any one of claims 1 to 5.

7. Low-altitude flight real-time situation awareness device, applied to real-time data processing platform, characterized by: include: Multi-source data receiving module, used to obtain data from radar group, AI image recognition module, GPS and ADS-B; Data fusion processing module, with built-in deep learning data fusion algorithm, performs spatiotemporal calibration, feature extraction, target classification and threat level assessment; Situation generation and transmission module, used to generate flight situation information containing multi-dimensional target information and push it in seconds; The early warning trigger module is used to generate early warning information containing real-time location, threat level and historical trajectory when an anomaly is detected; The mathematical expression of the data fusion algorithm is F=Activation(W2·Activation(W1·X+b1)+b2); W1, W2: weight matrix of the neural network layer, used for feature transformation and fusion; b1, b2: bias vectors of the neural network layer; F: The fused target feature vector, which contains flight situation information such as position, speed, type, motion trajectory, and threat level; Activation: Activation function, such as ReLU, Sigmoid, or Swish; X is the concatenation vector of the radar feature vector and the image feature vector, with the dimension D_radar+D_image.

8. The low-altitude flight real-time situation awareness device according to claim 7, characterized in that: The data fusion processing module includes an edge computing pre-processing unit and a cloud collaborative processing unit: The edge computing pre-processing unit is deployed near the radar group and integrates a lightweight detection model and filtering algorithm to complete raw data noise reduction, target initial screening and preliminary positioning; The cloud-based collaborative processing unit is linked with the edge node through a high-speed encrypted channel to achieve real-time modeling of large-scale data based on the cloud-based GPU cluster.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the low-altitude flight real-time situation awareness method as described in any one of claims 1 to 5 is implemented.

10. A readable storage medium, characterized in that: The storage medium stores a computer program, which, when executed by a processor, implements the low-altitude flight real-time situation awareness method according to any one of claims 1 to 5.

Images