Intelligent bird strike prevention and control method and system based on multi-source data

Abstract

The invention discloses an intelligent bird strike prevention and control method and system based on multi-source data, and relates to the field of aviation safety, in particular to the intelligent bird strike prevention and control method based on the multi-source data. The intelligent bird strike prevention and control method based on the multi-source data comprises the steps that aircraft position data are collected; acquiring bird position data; preprocessing the data; constructing a hybrid space-time convolutional neural network, a long-short-term memory network and a physical constraint aircraft trajectory prediction model; constructing a bird trajectory prediction model of a mixed long and short-term memory network and a graph neural network; constructing a collision detection model combined with a mahalanobis distance; risk grade evaluation; and risk grading response. According to the invention, the problems that the data structure is single and different air defense measures are not taken for different risk levels in the prior art are solved. Therefore, the probability of bird strike accidents is effectively reduced, and the safety performance of air transportation is improved.

Description

Technical Field

[0001] This invention relates to the field of aviation safety, and more specifically, to a bird strike intelligent prevention and control method and system based on multi-source data. Background Technology

[0002] Traditional bird strike warning systems rely primarily on radar monitoring and human experience, resulting in low accuracy and timeliness. Current bird strike research largely focuses on airport bird control methods, with limited research on the quantitative calculation of bird strike risk. Airport bird control can only truly be effective and improved if based on bird strike risk calculation. Therefore, developing an intelligent system that can efficiently integrate multiple data sources, accurately predict bird strike risk, and automatically drive away nearby birds is of paramount importance.

[0003] Existing technologies typically rely on single bird monitoring or meteorological data for bird strike risk assessment. However, due to the lack of fusion analysis of multi-source heterogeneous data, it is difficult to comprehensively and accurately predict bird strike risk. Furthermore, existing technologies lack corresponding prevention and control measures for different risk levels, failing to effectively prevent bird strike accidents. To address the problems of limited data structure and lack of differentiated air defense measures for different risk levels in existing technologies, this invention proposes a multi-source data-based intelligent bird strike prevention and control method and system. This system acquires sensor data in real time, predicts the flight trajectories of aircraft and birds separately, determines the probability of bird strikes based on the trajectory prediction results, and implements graded responses to guide the execution of bird deterrence strategies, thereby avoiding collision events. This effectively reduces the probability of bird strike accidents and improves the safety performance of air transport. Summary of the Invention

[0004] The main objective of this invention is to provide a bird strike intelligent prevention and control method and system based on multi-source data, to at least solve the problems of the existing technology's single data structure and lack of different air defense measures for different risk levels. This will effectively reduce the probability of bird strike accidents and improve the safety performance of air transport.

[0005] To achieve the above objectives, a method and system for intelligent bird strike prevention and control based on multi-source data is provided.

[0006] In a first aspect, the present invention provides a bird strike intelligent prevention and control method based on multi-source data, the method comprising: The system collects aircraft location data, bird flight information, and flock flight information and stores them into aircraft data stream, bird data stream, and flock data stream, respectively. The data in the aircraft data stream is preprocessed and saved as a first data stream, and the data in the bird data stream and the bird flock data stream are preprocessed and saved as a second data stream. A first prediction model is established based on the first data stream, and the first flight trajectory is predicted and the first future trajectory at multiple time points is obtained through the first prediction model. A second prediction model is established based on the second data stream. The second prediction model is used to predict the second flight trajectory and obtain the second future trajectory at multiple time points. A collision risk field model is established using the spatiotemporal proximity of the first future trajectory and the second future trajectory, and the collision risk probability is output using the collision risk field model. The collision risk probability is matched with a preset risk level classification standard and a predicted risk level is output. The predicted risk level and prediction time are combined to drive the corresponding bird deterrence strategy.

[0007] Specifically, the preprocessing operation on the data in the aircraft data stream and saving it as a first data stream, and the preprocessing operation on the data in the bird data stream and the bird flock data stream and saving it as a second data stream, include: The aircraft data stream is deduplicated, its integrity is verified, and its format is converted before being saved as the first data stream. After fusing the bird data stream and the flock data stream, the data format is converted and stored in the fused data stream. Each bird in the fused data stream is assigned a unique ID number and saved as the second data stream.

[0008] Specifically, the step of establishing a first prediction model based on the first data stream, predicting the first flight trajectory using the first prediction model, and obtaining the first future trajectory at multiple time points includes: The first data stream is obtained as the input data for the first prediction model; The temporal and spatial features of the first data stream are extracted using a spatiotemporal convolutional layer. The first spatiotemporal feature is obtained by fusing the temporal and spatial features; The first spatiotemporal feature is obtained by performing temporal feature extraction processing on the first spatiotemporal feature using a long short-term memory network; The first future trajectory at multiple time points is obtained by applying physical constraints to the processed first spatiotemporal features using a physical constraint layer.

[0009] Specifically, the step of establishing a second prediction model based on the second data stream, predicting the second flight trajectory using the second prediction model, and obtaining the second future trajectory at multiple time points includes: The second data stream is obtained as input data for the second prediction model; The individual movement patterns of each bird in the bird data stream are encoded using a Kalman filter; A graph neural network is used to encode the spatial interaction relationships between each flock of birds in the bird flock data stream; The second future trajectory at multiple time points is obtained by decoding the individual's movement pattern and spatial interaction relationship using a long short-term memory network.

[0010] Specifically, the step of establishing a collision risk field model using the spatiotemporal proximity of the first future trajectory and the second future trajectory, and then outputting the collision risk probability using the collision risk field model, includes: The Mahalanobis distance is calculated using the first future trajectory and the second future trajectory to account for the uncertainty of the prediction; The Mahalanobis distance is incorporated into the spatial risk term and expressed by the formula: The collision risk probability = the spatial risk term × the temporal risk term Calculate the collision risk probability.

[0011] Specifically, the step of matching the collision risk probability with a preset risk level classification standard and outputting a predicted risk level includes: The collision risk probability is less than 0.4, and the output predicted risk level is low risk; If the collision risk probability is greater than or equal to 0.4 and less than 0.8, the output predicted risk level is medium risk. If the collision risk probability is greater than or equal to 0.8 and the expected collision time is less than or equal to 30 seconds, the output predicted risk level is high risk.

[0012] Furthermore, the estimated collision time includes: The predicted collision time is greater than 0 seconds and less than or equal to 30 seconds, which is a short-term prediction. The predicted collision time of 30 seconds or less than or equal to 60 seconds is considered a medium-term prediction. The predicted collision time is greater than 60 seconds and less than or equal to 120 seconds, which is a long-term prediction. The short-term forecast is used for collision warning, the medium-term forecast is used for strategy deployment, and the long-term warning is used for resource scheduling.

[0013] Secondly, the present invention provides a bird strike intelligent prevention and control system based on multi-source data, wherein the prevention and control system applies the prevention and control method described in the first aspect, and the prevention and control system includes: The data acquisition module is used to collect the location data of the aircraft, the flight information of the bird, and the flight information of the flock of birds, and store them into the aircraft data stream, the bird data stream, and the flock of birds data stream, respectively. A data preprocessing module is used to perform the preprocessing operation on the data in the aircraft data stream and save it as the first data stream, and to perform the preprocessing operation on the data in the bird data stream and the bird flock data stream and save it as the second data stream. The first trajectory prediction module is used to establish a first prediction model based on the first data stream, predict the first flight trajectory through the first prediction model, and obtain the first future trajectory at multiple time points. The second trajectory prediction module is used to establish a second prediction model based on the second data stream, predict the second flight trajectory through the second prediction model, and obtain the second future trajectory at multiple time points. A collision detection module is used to establish a collision risk field model using the spatiotemporal proximity of the first future trajectory and the second future trajectory, and to output the collision risk probability using the collision risk field model. A risk assessment module is used to match the collision risk probability with the preset risk level classification standard and output the predicted risk level. The graded response module, wherein the risk determination module is used to combine the predicted risk level and the predicted time to drive the corresponding bird deterrence strategy.

[0014] Specifically, the first trajectory prediction module includes: The first prediction model construction module is used to mix the spatiotemporal convolutional neural network and the long short-term memory network, and add a physical constraint layer to complete the construction of the first prediction model. The first future trajectory prediction module is used to predict the first flight trajectory using the first prediction model and obtain the first future trajectory at multiple time points.

[0015] Specifically, the second trajectory prediction module includes: The second prediction model building module is used to complete the construction of the first prediction model by combining the modeling ability of long short-term memory network for temporal motion and the modeling ability of graph neural network for group interaction. The second future trajectory prediction module is used to predict the second flight trajectory using the second prediction model and obtain the second future trajectory at multiple time points.

[0016] This application provides a bird strike intelligent prevention and control method and system based on multi-source data. The method collects aircraft location data, bird flight information, and flock flight information, storing them in aircraft data streams, bird data streams, and flock data streams, respectively. It preprocesses the data in the aircraft data stream and saves it as a first data stream, and preprocesses the data in the bird data stream and flock data stream and saves it as a second data stream. Based on the first data stream, a first prediction model is established to predict a first flight trajectory and obtain first future trajectories at multiple time points. Based on the second data stream, a second prediction model is established to predict a second flight trajectory and obtain second future trajectories at multiple time points. A collision risk field model is established using the spatiotemporal proximity of the first and second future trajectories, and the collision risk field model is used to output a collision risk probability. The collision risk probability is matched with a preset risk level classification standard to output a predicted risk level, and the predicted risk level is combined with the predicted time to drive corresponding bird deterrence strategies. Compared with traditional methods, the intelligent bird strike prevention method based on multi-source data provided in this application at least solves the problems of the existing technology's single data structure and lack of different air defense measures for different risk levels. This effectively reduces the probability of bird strike accidents and improves the safety performance of air transport. Attached Figure Description

[0017] The accompanying drawings, which form part of this application, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings: Figure 1 A flowchart illustrating the intelligent bird strike prevention and control method based on multi-source data provided in this application; Figure 2 A flowchart illustrating the intelligent bird strike prevention and control system based on multi-source data provided in this application. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0019] The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein.

[0020] In this invention, the terms "exemplary" or "for example" are used to indicate examples, illustrations, or descriptions. Any embodiment or design described as "exemplary" or "for example" in this application should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of terms such as "exemplary" or "for example" is intended to present the relevant concepts in a concrete manner.

[0021] This application provides a bird strike intelligent control method and system based on multi-source data. The method first collects aircraft position data, bird flight information, and flock flight information, storing them in aircraft data streams, bird data streams, and flock data streams, respectively. Then, it preprocesses the data in the aircraft data stream and saves it as a first data stream, and preprocesses the data in the bird data stream and flock data stream and saves it as a second data stream. Next, a first prediction model is established based on the first data stream to predict a first flight trajectory and obtain first future trajectories at multiple time points. Simultaneously, a second prediction model is established based on the second data stream to predict a second flight trajectory and obtain second future trajectories at multiple time points. A collision risk field model is established using the spatiotemporal proximity of the first and second future trajectories, and the collision risk probability is output using this model. The collision risk probability is then matched with a preset risk level classification standard to output a predicted risk level. Finally, the predicted risk level and predicted time are combined to drive a corresponding bird deterrence strategy.

[0022] The technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will be described below with reference to the accompanying drawings.

[0023] Figure 1 A flowchart illustrating the intelligent bird strike prevention and control method based on multi-source data provided in this application is shown below. Figure 1 As shown, this embodiment provides a bird strike intelligent control method based on multi-source data. The method includes: S101: Collects aircraft location data, bird flight information, and flock flight information and stores them into the aircraft data stream, bird data stream, and flock data stream, respectively.

[0024] The implementation of S101 specifically includes: 1. Deploy ADS-B ground stations around the airport to receive ADS-B signals transmitted by aircraft in real time and store them in the aircraft data stream. The ADS-B signals contain information such as the aircraft's unique identifier, position (latitude, longitude, and altitude), speed, and heading.

[0025] 2. High-precision cameras and other optoelectronic devices will be installed in key areas around the airport to capture bird images and location information, storing it in a bird data stream. These high-precision cameras and other optoelectronic devices will have automatic target detection and tracking capabilities, enabling them to identify and track bird flight in real time.

[0026] 3. Use radar systems to monitor a large area surrounding the airport, acquiring the location and movement information of bird flocks and storing it in a bird flock data stream. The radar data can provide parameters such as the distance, speed, and direction of the bird flocks.

[0027] S102: Perform preprocessing operations on the data in the aircraft data stream and save it as a first data stream, and perform preprocessing operations on the data in the bird data stream and the bird flock data stream and save it as a second data stream.

[0028] Specifically, the preprocessing operation on the data in the aircraft data stream and saving it as a first data stream, and the preprocessing operation on the data in the bird data stream and the bird flock data stream and saving it as a second data stream, include: The aircraft data stream is deduplicated, its integrity is verified, and its format is converted before being saved as the first data stream. After fusing the bird data stream and the flock data stream, the data format is converted and stored in the fused data stream. Each bird in the fused data stream is assigned a unique ID number and saved as the second data stream.

[0029] The implementation of S102 specifically includes: 1. Perform preprocessing operations on the data in the aircraft data stream. 1.1 Record the ADS-B data from the aircraft data stream as raw data into the database; 1.2 Perform deduplication on the data; 1.2.1 Establish a composite index based on timestamps and ICAO addresses (unique aircraft identifiers); 1.2.2 Use SQL window functions or in-memory hash tables to identify and delete duplicate records.

[0030] 1.3 Perform data integrity verification; 1.3.1 Check if any required fields are missing (ICAO, location, timestamp); 1.3.2 Verify whether the coordinate range is reasonable (latitude and longitude are within the valid range); 1.3.3 Check if the altitude value is within a reasonable range (typically -1000ft to 50000ft); 1.3.4 Verify the physical rationality of the velocity vector.

[0031] 1.4 Convert the data format; 1.4.1 Decode the raw hexadecimal message into structured data; 1.4.2 Unify the time format to UTC timestamp; 1.4.3 Standardize the unit of height to meters; 1.4.4 Unify the unit of speed to meters per second; 1.4.5 Change the coordinate system to the WGS84 coordinate system.

[0032] 1.5 Handle outliers in the data; 1.5.1 Identifying outliers using the sliding window statistical method; 1.5.2 Smooth out mutation data or mark it as suspicious data.

[0033] 2. Denoise the data in the bird data stream. 2.1 Static clutter suppression is applied to the bird data stream; 2.1.1 Use a Moving Target Indication (MTI) filter to eliminate echoes from stationary objects; 2.1.2 Apply the Constant False Alarm Rate (CFAR) detection algorithm to adaptively set the detection threshold.

[0034] 2.2 Perform dynamic clutter processing on the bird data stream; 2.2.1 Use Doppler filtering to separate targets with different velocities; 2.2.2 Applying spatiotemporal filtering techniques to eliminate meteorological clutter (rain, clouds, etc.).

[0035] 2.3 Perform point aggregation on the bird data stream; 2.3.1 Clustering is performed on multiple detection points based on adjacent distance gates and azimuth angles; 2.3.2 Calculate cluster centers as effective target locations.

[0036] 2.4 Format the bird data stream; 2.4.1 Convert polar coordinates (distance, azimuth, elevation) to Cartesian coordinates; 2.4.2 Add timestamps and synchronize with the system clock.

[0037] 3. Multi-sensor data association and fusion 3.1 Spatiotemporal alignment of multi-sensor data; 3.1.1 Add a PTP (Precise Time Protocol) timestamp to all data collected by all devices; 3.1.2 Use interpolation to align asynchronous data to a unified time base; 3.1.3 Convert all sensor data to the ENU (East-North-Sky) local coordinate system; 3.1.4 Consider the offset of the sensor installation position for coordinate correction.

[0038] 3.2 The bird data stream and the bird flock data stream are correlated and fused using the Hungarian algorithm; 3.2.1 Extract the following data from the bird flock data stream: position, velocity, and RCS (radar cross-section); 3.2.2 Extract the following features from the bird data stream: position, size, color, and shape. 3.2.3 Design distance metrics: Consider spatial distance (Euclidean distance), velocity difference, and feature similarity; 3.2.4 Weighted combination of multiple metrics:

[0039] in, Represents a weight of 1; This represents the difference in spatial distance between two observed targets; Represents weight two; A measure of the difference in velocity vectors; Represents a weight of three; Similarity measures representing multimodal features; 3.2.5 The Hungarian algorithm is used to solve for the optimal matching; 3.2.6 Setting global thresholds and relative thresholds for filtering cannot match; 3.2.7 Handling splitting and merging situations (one radar target corresponding to multiple visual targets or vice versa); 3.2.8 Assign a globally unique ID to the successfully associated target; 3.2.9 Maintain the target status table to record the lifecycle; 3.2.10 Handling situations where new targets emerge and old targets disappear.

[0040] 3.3 Data Fusion 3.3.1 Kalman filtering is used to determine the target position measured by radar and camera; 3.3.2 Dynamically adjust the fusion weights based on sensor accuracy; 3.3.3 Integrating multi-source velocity information yields a more accurate velocity vector; 3.3.4 Enhancing the reliability of target recognition by combining appearance features from camera data; 3.3.5 Calculate the fused covariance matrix; 3.3.6 Perform pattern-level conflict detection on conflict data.

[0041] 4. Integrate aircraft data stream, bird data stream, and flock data stream. 4.1 Data synchronization; 4.1.1 Time-align all target data (aircraft and birds) based on GPS time; 4.1.2 Motion model prediction is used to perform interpolation compensation for asynchronous data.

[0042] 4.2 Unify the spatial representation of all data; 4.2.1 Transform all target positions to the same reference coordinate system (such as the airport reference point coordinate system); 4.2.2 Consider the effect of Earth's curvature (for distant targets); 4.2.3 Convert barometric altitude to geometric altitude (considering local air pressure and temperature); 4.2.4 The same reference is used for bird altitude and aircraft altitude.

[0043] 4.3 Generate and output a uniform data frame containing the following fields: a target ID (aircraft ICAO or bird global ID); b. Timestamp; c is located at position (x, y, z); d velocity (v) x ,v y ,v z ); e. Target type (civil aircraft / general aviation / birds); f confidence score; g data source marker.

[0044] 5. Quality Control and Monitoring 5.1 Data quality indicators, including: a. Completeness rate (valid data / total data); b. Success rate of association; c. Time synchronization error statistics; d-coordinate transformation residual.

[0045] 5.2 Real-time monitoring; 5.2.1 Establish a health status panel for the data pipeline; 5.2.2 Set anomaly detection thresholds and alarm mechanisms; 5.2.3 Record the performance indicators of each stage of preprocessing.

[0046] 5.3 Feedback optimization; 5.3.1 Adjust the preprocessing parameters based on the subsequent trajectory prediction results; 5.3.2 Recalibrate sensor parameters periodically.

[0047] S103: Establish a first prediction model based on the first data stream, predict the first flight trajectory through the first prediction model, and obtain the first future trajectory at multiple time points.

[0048] Specifically, the step of establishing a first prediction model based on the first data stream, predicting the first flight trajectory using the first prediction model, and obtaining the first future trajectory at multiple time points includes: The first data stream is obtained as the input data for the first prediction model; The temporal and spatial features of the first data stream are extracted using a spatiotemporal convolutional layer. The first spatiotemporal feature is obtained by fusing the temporal and spatial features; The first spatiotemporal feature is obtained by performing temporal feature extraction processing on the first spatiotemporal feature using a long short-term memory network; The first future trajectory at multiple time points is obtained by applying physical constraints to the processed first spatiotemporal features using a physical constraint layer.

[0049] The implementation of S103 specifically includes: 1. Input the first data and process it. 1.1 Input tensor dimensions for data normalization.

[0050] 1.2 Data Preprocessing 1.2.1 Converting from geographic coordinate system to ENU coordinate system; 1.2.2 Heading angle vectorization decoding; 1.2.3 Dynamic feature normalization.

[0051] 2. Construct the ST-CNN module 2.1 Spatial Convolution Design 2.1.1 Use depthwise separable convolutions to reduce the number of parameters; 2.1.2 The GELU activation function provides smooth nonlinearity.

[0052] 2.2 Temporal Convolution Design 2.2.1 Introducing a channel attention mechanism to dynamically weight important features; 2.2.2 A 5-frame time window covers the typical maneuver duration (such as the turn start phase).

[0053] 3. Hierarchical LSTM timing modeling 3.1 An LSTM layer is constructed using a short-term memory layer (0.1-0 second scale), a medium-term memory layer (1-5 second scale), and a long-term memory layer (5-10 second scale). The short-term LSTM is first trained to predict 1-second trajectories, and the medium and long-term layers are gradually added after the short-term layer is frozen.

[0054] 3.2 Time dimension alignment of the output of multi-scale LSTM is performed using interpolation.

[0055] 4. Apply physical constraints (maximum acceleration and maximum turn rate) to the trajectory output from the upper layer to ensure that the prediction conforms to the aircraft dynamics. 4.1 Calculate the velocity and acceleration of adjacent prediction points:

[0056] in, Indicates the velocity at each predicted point. Indicates the output from the upper layer. This indicates the time interval between each prediction point. This indicates the maximum acceleration threshold.

[0057] 4.2 Limit the maximum acceleration threshold; 4.3 Compare the acceleration at each predicted point with the maximum acceleration threshold. If the threshold is exceeded, adjust the acceleration of the next predicted point according to the maximum acceleration. 4.4 Calculate the change in heading angle and the rate of change of constrained heading based on the predicted points:

[0058] in, This indicates the change in heading angle at each prediction point. Indicates the heading angle at each predicted point. Indicates the constrained rate of change of heading. This represents the maximum turning rate threshold.

[0059] 4.5 Set the maximum turning rate threshold; 4.6 Compare the rate of change of the constrained heading at each predicted point with the maximum turning rate threshold. If the rate of change exceeds the threshold, adjust the heading angle according to the maximum turning rate and recalculate the position.

[0060] 5. Output the aircraft coordinate data for future time series.

[0061] Specifically, the step of establishing the first prediction model and the second prediction model based on the first data stream and the second data stream respectively further includes: The second data stream is obtained as input data for the second prediction model.

[0062] The individual movement patterns of each bird were encoded using a Kalman filter.

[0063] The spatial interaction relationships among the bird flocks were encoded using a graph neural network.

[0064] The second future trajectory is decoded using a long short-term memory network based on the individual's movement pattern and the spatial interaction relationship.

[0065] The second coordinate position of the future time series is output based on the second future trajectory.

[0066] S104: Establish a second prediction model based on the second data stream, predict the second flight trajectory through the second prediction model, and obtain the second future trajectory at multiple time points.

[0067] Specifically, the step of establishing a second prediction model based on the second data stream, predicting the second flight trajectory using the second prediction model, and obtaining the second future trajectory at multiple time points includes: The second data stream is obtained as input data for the second prediction model; The individual movement patterns of each bird in the bird data stream are encoded using a Kalman filter; A graph neural network is used to encode the spatial interaction relationships between each flock of birds in the bird flock data stream; The second future trajectory at multiple time points is obtained by decoding the individual's movement pattern and spatial interaction relationship using a long short-term memory network.

[0068] The implementation of S104 specifically includes: 1. Input the second data stream 2. Extracting individual motion features using Kalman filtering 3. Construct a dynamic graph structure 3.1 Design node characteristics; 3.2 Constructing edge features.

[0069] 4. Construct a graph neural network encoder 4.1 Dynamically weight important neighbors using GATv2 convolutions (graph convolutions with attention mechanism); 4.2 Normalized stable training process; 4.3 Edge features participate in message passing calculation.

[0070] 5. Construct a spatiotemporal joint decoder 5.1 Decoding is performed using an LSTM decoder; 5.2 Use course learning and multi-task loss training strategies.

[0071] 6. Output the bird's coordinates for the future time series.

[0072] S105: Establish a collision risk field model using the spatiotemporal proximity of the first future trajectory and the second future trajectory, and output the collision risk probability using the collision risk field model.

[0073] Specifically, the step of establishing a collision risk field model using the spatiotemporal proximity of the first future trajectory and the second future trajectory, and then outputting the collision risk probability using the collision risk field model, includes: The Mahalanobis distance is calculated using the first future trajectory and the second future trajectory to account for the uncertainty of the prediction; The Mahalanobis distance is incorporated into the spatial risk term and expressed by the formula: The collision risk probability = the spatial risk term × the temporal risk term Calculate the collision risk probability.

[0074] The implementation of S105 specifically includes: 1. Calculate the aircraft trajectory:

[0075] in, Represents the coordinates of the aircraft's trajectory.

[0076] 2. Calculate the bird's trajectory:

[0077] in, Represents the flight path of a bird.

[0078] 3. Introduce Mahalanobis distance:

[0079] in, Represents Mahalanobis distance; The covariance matrix representing the bird's position prediction can be estimated using the covariance matrix obtained through Kalman filtering.

[0080] 4. Calculate the spatial risk item:

[0081] in, =1.5 represents the perceived uncertainty of the system.

[0082] 5. Calculate time-related risk items:

[0083] in, =0.2 indicates the time sensitivity coefficient; =30s indicates the critical warning time.

[0084] 6. Calculate the probability of collision risk: Collision risk probability = Spatial risk term × Temporal risk term S106: Match the collision risk probability with the preset risk level classification standard and output the predicted risk level, and combine the predicted risk level with the prediction time to drive the corresponding bird deterrence strategy.

[0085] Specifically, the step of matching the collision risk probability with a preset risk level classification standard and outputting a predicted risk level, and then combining the predicted risk level with the prediction time to drive a corresponding bird deterrence strategy, includes: The collision risk probability is less than 0.4, and the output predicted risk level is low risk; If the collision risk probability is greater than or equal to 0.4 and less than 0.8, the output predicted risk level is medium risk. If the collision risk probability is greater than or equal to 0.8 and the expected collision time is less than or equal to 30 seconds, the output predicted risk level is high risk.

[0086] Specifically, the estimated collision time includes: The predicted collision time is greater than 0 seconds and less than or equal to 30 seconds, which is a short-term prediction. The predicted collision time of 30 seconds or less than or equal to 60 seconds is considered a medium-term prediction. The predicted collision time is greater than 60 seconds and less than or equal to 120 seconds, which is a long-term prediction. The short-term forecast is used for collision warning, the medium-term forecast is used for strategy deployment, and the long-term forecast is used for resource scheduling.

[0087] The implementation of S106 specifically includes: 1. Match the collision risk probability with the preset risk level classification criteria; 2. Output the corresponding predicted risk level.

[0088] Figure 2 A flowchart illustrating the intelligent bird strike prevention and control system based on multi-source data provided in this application is shown below. Figure 2 As shown, this embodiment provides a bird strike intelligent control system based on multi-source data. This system applies... Figure 1 The bird strike intelligent prevention and control method based on multi-source data described in the embodiment includes a prevention and control system comprising: The data acquisition module is used to collect the location data of the aircraft, the flight information of the bird, and the flight information of the flock of birds, and store them into the aircraft data stream, the bird data stream, and the flock of birds data stream, respectively. A data preprocessing module is used to perform the preprocessing operation on the data in the aircraft data stream and save it as the first data stream, and to perform the preprocessing operation on the data in the bird data stream and the bird flock data stream and save it as the second data stream. The first trajectory prediction module is used to establish a first prediction model based on the first data stream, predict the first flight trajectory through the first prediction model, and obtain the first future trajectory at multiple time points. The second trajectory prediction module is used to establish a second prediction model based on the second data stream, predict the second flight trajectory through the second prediction model, and obtain the second future trajectory at multiple time points. A collision detection module is used to establish a collision risk field model using the spatiotemporal proximity of the first future trajectory and the second future trajectory, and to output the collision risk probability using the collision risk field model. A risk assessment module is used to match the collision risk probability with the preset risk level classification standard and output the predicted risk level. The graded response module, wherein the risk determination module is used to combine the predicted risk level and the predicted time to drive the corresponding bird deterrence strategy.

[0089] Specifically, the first trajectory prediction module includes: The first prediction model construction module is used to mix the spatiotemporal convolutional neural network and the long short-term memory network, and add a physical constraint layer to complete the construction of the first prediction model. The first future trajectory prediction module is used to predict the first flight trajectory using the first prediction model and obtain the first future trajectory at multiple time points.

[0090] Specifically, the second trajectory prediction module includes: The second prediction model construction module is used to construct the second prediction model by combining the modeling ability of long short-term memory networks for temporal motion and the modeling ability of graph neural networks for group interaction. The second future trajectory prediction module is used to predict the second flight trajectory using the second prediction model and obtain the second future trajectory at multiple time points.

[0091] The data acquisition module specifically includes: ADS-B signal acquisition of aircraft position data, acquisition of single bird position, wingspan, and flight speed information, and acquisition of information such as flock center coordinates, diffusion radius, density, and direction of movement of bird flocks; The collected information is stored in the aircraft data stream, the bird data stream, and the bird flock data stream, respectively.

[0092] The data preprocessing module specifically includes: The system performs deduplication, data integrity verification, data format conversion, and outlier handling on the aircraft data stream. Denoise the bird data stream; Merge bird flock data streams with bird data streams; Synchronize all processed and merged data.

[0093] The implementation of the first trajectory prediction module specifically includes: Data input; Spatial features are extracted using spatial convolution; Temporal features are extracted using temporal convolution; Perform feature dimension alignment in the LSTM layer; Adding physical constraints yields the first future trajectory.

[0094] The implementation of the second trajectory prediction module specifically includes: Data input; Kalman filtering is used to extract individual motion features; Construct a dynamic graph structure; Construct a graph neural network encoder; Construct a spatiotemporal joint decoder; Output the second future trajectory.

[0095] The implementation of the collision detection module specifically includes: Calculate the aircraft trajectory; Calculate bird trajectory; Introduce Mahalanobis distance; Calculate the collision probability.

[0096] The risk assessment module specifically includes: Compare collision probability with risk level classification standards; Output the risk level based on the comparison results.

[0097] The implementation of the tiered response module specifically includes: Response strategies are implemented based on risk levels.

Claims

1. A bird strike intelligent control method based on multi-source data, characterized in that, include: The system collects aircraft location data, bird flight information, and flock flight information and stores them into aircraft data stream, bird data stream, and flock data stream, respectively. The data in the aircraft data stream is preprocessed and saved as a first data stream, and the data in the bird data stream and the bird flock data stream are preprocessed and saved as a second data stream. A first prediction model is established based on the first data stream, and the first flight trajectory is predicted and the first future trajectory at multiple time points is obtained through the first prediction model. A second prediction model is established based on the second data stream. The second prediction model is used to predict the second flight trajectory and obtain the second future trajectory at multiple time points. A collision risk field model is established using the spatiotemporal proximity of the first future trajectory and the second future trajectory, and the collision risk probability is output using the collision risk field model. The collision risk probability is matched with a preset risk level classification standard and a predicted risk level is output. The predicted risk level and prediction time are combined to drive the corresponding bird deterrence strategy.

2. The intelligent bird strike prevention and control method based on multi-source data according to claim 1, characterized in that, The preprocessing operation on the data in the aircraft data stream and saving it as a first data stream, and the preprocessing operation on the data in the bird data stream and the bird flock data stream and saving it as a second data stream, include: The aircraft data stream is deduplicated, its integrity is verified, and its format is converted before being saved as the first data stream. After fusing the bird data stream and the flock data stream, the data format is converted and stored in the fused data stream. Each bird in the fused data stream is assigned a unique ID number and saved as the second data stream.

3. The intelligent bird strike prevention and control method based on multi-source data according to claim 1, characterized in that, The step of establishing a first prediction model based on the first data stream, predicting the first flight trajectory using the first prediction model, and obtaining the first future trajectory at multiple time points includes: The first data stream is obtained as the input data for the first prediction model; The temporal and spatial features of the first data stream are extracted using a spatiotemporal convolutional layer. The first spatiotemporal feature is obtained by fusing the temporal and spatial features; The first spatiotemporal feature is obtained by performing temporal feature extraction processing on the first spatiotemporal feature using a long short-term memory network; The first future trajectory at multiple time points is obtained by applying physical constraints to the processed first spatiotemporal features using a physical constraint layer.

4. The intelligent bird strike prevention and control method based on multi-source data according to claim 1, characterized in that, The step of establishing a second prediction model based on the second data stream, predicting the second flight trajectory using the second prediction model, and obtaining the second future trajectory at multiple time points includes: The second data stream is obtained as input data for the second prediction model; The individual movement patterns of each bird in the bird data stream are encoded using a Kalman filter; A graph neural network is used to encode the spatial interaction relationships between each flock of birds in the bird flock data stream; The second future trajectory at multiple time points is obtained by decoding the individual's movement pattern and spatial interaction relationship using a long short-term memory network.

5. The intelligent bird strike prevention and control method based on multi-source data according to claim 1, characterized in that, The step of establishing a collision risk field model using the spatiotemporal proximity of the first future trajectory and the second future trajectory, and then outputting the collision risk probability using the collision risk field model, includes: The Mahalanobis distance is calculated using the first future trajectory and the second future trajectory to account for the uncertainty of the prediction; The Mahalanobis distance is incorporated into the spatial risk term and expressed by the formula: The collision risk probability = the spatial risk term × the temporal risk term Calculate the collision risk probability.

6. The intelligent bird strike prevention and control method based on multi-source data according to claim 1, characterized in that, The step of matching the collision risk probability with a preset risk level classification standard and outputting a predicted risk level includes: The collision risk probability is less than 0.4, and the output predicted risk level is low risk; If the collision risk probability is greater than or equal to 0.4 and less than 0.8, the output predicted risk level is medium risk. If the collision risk probability is greater than or equal to 0.8 and the expected collision time is less than or equal to 30 seconds, the output predicted risk level is high risk.

7. The intelligent bird strike prevention and control method based on multi-source data according to claim 6, characterized in that, The estimated collision time includes: The predicted collision time is greater than 0 seconds and less than or equal to 30 seconds, which is a short-term prediction. The predicted collision time of 30 seconds or less than or equal to 60 seconds is considered a medium-term prediction. The predicted collision time is greater than 60 seconds and less than or equal to 120 seconds, which is a long-term prediction. The short-term forecast is used for collision warning, the medium-term forecast is used for strategy deployment, and the long-term forecast is used for resource scheduling.

8. A bird strike intelligent control system based on multi-source data, characterized in that, The prevention and control system applies the prevention and control method according to any one of claims 1 to 7, and the prevention and control system comprises: The data acquisition module is used to collect the location data of the aircraft, the flight information of the bird, and the flight information of the flock of birds, and store them into the aircraft data stream, the bird data stream, and the flock of birds data stream, respectively. A data preprocessing module is used to perform the preprocessing operation on the data in the aircraft data stream and save it as the first data stream, and to perform the preprocessing operation on the data in the bird data stream and the bird flock data stream and save it as the second data stream. The first trajectory prediction module is used to establish a first prediction model based on the first data stream, predict the first flight trajectory through the first prediction model, and obtain the first future trajectory at multiple time points. The second trajectory prediction module is used to establish a second prediction model based on the second data stream, predict the second flight trajectory through the second prediction model, and obtain the second future trajectory at multiple time points. A collision detection module is used to establish a collision risk field model using the spatiotemporal proximity of the first future trajectory and the second future trajectory, and to output the collision risk probability using the collision risk field model. A risk assessment module is used to match the collision risk probability with the preset risk level classification standard and output the predicted risk level. The graded response module, wherein the risk determination module is used to combine the predicted risk level and the predicted time to drive the corresponding bird deterrence strategy.

9. A bird strike intelligent control system based on multi-source data according to claim 8, characterized in that, The first trajectory prediction module includes: The first prediction model construction module is used to mix the spatiotemporal convolutional neural network and the long short-term memory network, and add a physical constraint layer to complete the construction of the first prediction model. The first future trajectory prediction module is used to predict the first flight trajectory using the first prediction model and obtain the first future trajectory at multiple time points.

10. A bird strike intelligent control system based on multi-source data according to claim 8, characterized in that, The second trajectory prediction module includes: The second prediction model construction module is used to construct the second prediction model by combining the modeling ability of long short-term memory networks for temporal motion and the modeling ability of graph neural networks for group interaction. The second future trajectory prediction module is used to predict the second flight trajectory using the second prediction model and obtain the second future trajectory at multiple time points.

Images