Airport bird strike prevention method and system based on multi-sensor fusion

Abstract

The invention relates to the technical field of airport safety protection, and discloses an airport bird strike prevention method and system based on multi-sensor fusion. The method comprises the steps of collecting airport peripheral airspace radar detection, radio detection, laser monitoring, weather and other multi-mode sensor data, and performing time synchronization and space registration to generate a fusion data stream; features are extracted based on a dynamic weight distribution algorithm, target types are identified by using a multi-modal classification model, and birds and unmanned aerial vehicles are distinguished; constructing a dynamic risk assessment model based on a Markov chain to generate an intrusion risk level; and when the risk level exceeds a preset threshold value, triggering a multi-level alarm strategy and linking the expelling equipment to execute a self-adaptive expelling instruction. According to the method and system, through multi-sensor fusion, the monitoring precision is improved, the target is accurately recognized, the risk is evaluated in real time, efficient alarming and expelling are achieved, the airport bird strike prevention capacity is effectively improved, and aviation safety is guaranteed.

Description

Technical Field

[0001] The present invention relates to the technical field of airport security prevention, and specifically provides an airport bird strike prevention method and system based on multi-sensor fusion. Background Technique

[0002] With the booming development of the aviation industry, the problems of airport bird strikes and drone intrusions have become increasingly prominent, posing a serious threat to aviation safety. Bird activities are relatively frequent in the airspace around airports. Once an aircraft collides with a bird during takeoff and landing, it may lead to serious consequences such as engine failure and airframe damage. According to statistics, the annual global aviation economic losses caused by bird strikes reach hundreds of millions of dollars, and it may also trigger major accidents of plane crashes and deaths, seriously endangering the lives of passengers and crew.

[0003] At the same time, the popularization of drone technology has led to an increasing frequency of its appearance around airports. Some drones enter the airport's clear zone due to improper operation or illegal intrusion, interfering with normal aviation order. Drones are small in size, and traditional airport security equipment is difficult to effectively detect and identify them. When they fly close to an aircraft, they may collide with the aircraft or interfere with the aircraft's communication, navigation and other systems, posing a great potential risk to flight safety.

[0004] Currently, airports mainly use single-sensor monitoring means to prevent bird strikes and drone intrusions. For example, radar is used to monitor bird and aircraft targets, but the detection accuracy of radar for small birds and low-flying targets is limited, and false alarms and missed detections are likely to occur. Some airports also use visual monitoring systems, but they are greatly affected by environmental factors such as weather and light, and can hardly work properly under bad weather conditions. For the monitoring of drones, existing radio detection equipment can only obtain limited communication frequency band information, and it is difficult to comprehensively and accurately identify the drone model and flight intention.

[0005] In addition, the existing prevention systems have deficiencies in data processing and analysis. There is no effective fusion mechanism for different types of sensor data, and the complementary advantages of multi-source data cannot be fully utilized, resulting in inaccurate identification and risk assessment of targets. After detecting an intrusion target, the alarm and repelling measures are not timely and effective enough, and it is difficult to meet the strict requirements of airports for safety and efficiency. Summary of the Invention

[0006] The purpose of the present invention is to provide an airport bird strike prevention method and system based on multi-sensor fusion to solve the problems raised in the above background technique.

[0007] To achieve the above purpose, the present invention provides the following technical solution: An airport bird strike prevention method based on multi-sensor fusion, the method includes: Collect multimodal sensor data of the airspace around the airport, wherein the multimodal sensor data includes radar detection data, radio detection data, laser monitoring data and meteorological data; Performing time synchronization and spatial registration on the multimodal sensor data to generate a fused data stream; Extracting features from the fused data stream based on a dynamic weight allocation algorithm to generate target trajectory information and target attribute information; Using a multimodal classification model to identify the target type based on the target trajectory information and target attribute information, and distinguish between bird targets and drone targets; According to the target type recognition result and the target motion parameters, a dynamic risk assessment model based on a Markov chain is constructed to generate an intrusion risk level; When the intrusion risk level exceeds a preset threshold, a multi-level alarm strategy is triggered, and the expulsion device is linked to execute an adaptive expulsion instruction.

[0008] Preferably, the multimodal sensor data collected from the airspace surrounding the airport includes: Use three-coordinate low-altitude detection radar to collect target azimuth, altitude and speed information; Capture the electromagnetic signal characteristics of the drone communication frequency band through radio detection equipment; Use laser monitoring equipment to obtain the target's three-dimensional profile and motion trajectory; Integrate real-time weather data from long-range weather radar, including wind speed, air pressure and precipitation information.

[0009] Preferably, the time synchronization and space registration include: Perform Kalman filtering correction on the timestamp of multimodal sensor data to eliminate clock deviation between devices; Based on the geographic information system coordinates, the detection areas of radar, laser and radio equipment are spatially overlapped and calibrated to establish a unified coordinate system.

[0010] Preferably, the dynamic weight allocation algorithm satisfies the following relationship:

[0011] in, For the The weight of each sensor, For the Sensor confidence for each sensor, For the The data integrity factor for each sensor is and is the dynamic adjustment coefficient, is the total number of sensors.

[0012] Preferably, the multimodal classification model includes: Using a convolutional neural network to extract the target contour features from the laser monitoring data; Using a long short-term memory network to analyze the temporal features of radar and radio data; Through transfer learning, fusing the bird morphology database and the UAV model database to generate the target classification result.

[0013] Preferably, the construction of the dynamic risk assessment model includes: Defining the state space as the flight altitude, speed of the intrusion target, and the relative distance from the core area; Training the state transition probability matrix through historical intrusion event data; Calculating the risk entropy value based on the real-time target motion parameters to determine the risk level.

[0014] Preferably, the risk entropy value satisfies the following relationship:

[0015] Wherein, is the risk entropy value, is the probability of the th state, is the total number of states.

[0016] Preferably, the multi-level alarm strategy includes: Dividing the warning levels according to the risk level, including first-level alert, second-level alert, and emergency response; Triggering the sound and light alarm device, the air traffic control communication system, and the mobile terminal to push warning information.

[0017] Preferably, the generation of the adaptive expulsion instruction includes: Based on the target type, matching the priority of the expulsion device, giving priority to starting the acoustic wave expulsion for bird targets and starting the electromagnetic interference for UAV targets; Dynamically adjusting the coverage range and power parameters of the expulsion device according to the target position.

[0018] Preferably, the present invention further includes an airport bird strike prevention system based on multi-sensor fusion, and the system includes: Data acquisition module: used to acquire multi-modal sensor data of the airspace around the airport, and the multi-modal sensor data includes radar detection data, radio detection data, laser monitoring data, and meteorological data; Data preprocessing module: performing time synchronization and spatial registration on the multi-modal sensor data to generate a fused data stream; Feature extraction module: extracting features from the fused data stream based on the dynamic weight allocation algorithm to generate target trajectory information and target attribute information; Target recognition module: using a multimodal classification model to identify the target type based on the target trajectory information and target attribute information, and distinguish between bird targets and drone targets; Risk assessment module: constructs a dynamic risk assessment model based on a Markov chain according to the target type recognition result and target motion parameters, and generates an intrusion risk level; Alarm and expulsion module: When the intrusion risk level exceeds the preset threshold, a multi-level alarm strategy is triggered, and the expulsion device is linked to execute an adaptive expulsion instruction.

[0019] Compared with the prior art, the present invention has the following beneficial effects: The present invention generates a fused data stream by collecting multimodal sensor data such as radar detection data, radio detection data, laser monitoring data and meteorological data, and performing time synchronization and spatial registration. This multi-sensor fusion method makes up for the shortcomings of a single sensor in terms of detection range, accuracy and environmental influence. For example, radar has advantages in long-distance detection, and laser monitoring equipment is more accurate in monitoring the contour and close-range motion trajectory of the target. The combination of the two can obtain target information more comprehensively and accurately, greatly improving the monitoring accuracy of targets in the airspace around the airport and reducing missed reports and false alarms.

[0020] By using a multimodal classification model, a convolutional neural network is used to extract target contour features from laser monitoring data, a long short-term memory network is used to analyze the temporal features of radar and radio data, and transfer learning is used to fuse the bird morphology database and the drone model database to accurately distinguish between bird targets and drone targets. This precise target recognition capability provides strong support for subsequent targeted preventive measures, avoiding resource waste and safety hazards caused by misjudgment of targets.

[0021] According to the target type recognition results and target motion parameters, a dynamic risk assessment model based on Markov chain is constructed to generate the intrusion risk level. The model takes into account factors such as the flight altitude, speed and relative distance of the intrusion target from the core area, and trains the state transition probability matrix through historical intrusion event data, and calculates the risk entropy value based on the real-time target motion parameters to determine the risk level. It can evaluate the threat level of the intrusion target to the airport in real time and dynamically, so that airport staff can understand the severity of potential risks in advance, and provide a scientific basis for formulating reasonable prevention strategies.

[0022] When the intrusion risk level exceeds the preset threshold, a multi-level alarm strategy is triggered, and the warning levels are divided according to the risk level, including first-level alert, second-level alert, and emergency response. Different levels of detailed warning information are respectively pushed through the acoustic and optical alarm devices, the air traffic control communication system, and the mobile terminal, ensuring that the staff in different positions at the airport can timely and accurately understand the risk situation of the intrusion target, so as to take corresponding countermeasures, improving the timeliness and effectiveness of the emergency response.

[0023] Based on the target type, the priorities of the repelling devices are matched. For bird targets, the acoustic wave repelling is preferentially activated, and for UAV targets, the electromagnetic interference is preferentially activated. The coverage range and power parameters of the repelling devices are dynamically adjusted according to the target position. This adaptive repelling method can quickly and effectively execute the repelling operation according to different target types and position characteristics, minimizing the threat of intrusion targets to the airport and ensuring the safe operation of the airport.

[0024] The airport bird strike prevention method and system based on multi-sensor fusion of the present invention form a complete and efficient security prevention system from data collection, processing, target recognition, risk assessment to alarm and repelling. Through the collaborative work of each link, the prevention ability of the airport against security threats such as bird strikes and UAV intrusions is significantly improved, reducing the probability of aviation accidents, ensuring the safety of passengers and crew members, and at the same time reducing the aviation economic losses caused by bird strikes and UAV intrusions, which has important economic and social significance. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 is the working principle diagram of the airport bird strike prevention method described in the present invention; Figure 2 is the schematic flow chart of multi-modal sensor data collection around the airport; Figure 3 is the schematic flow chart of feature extraction based on the dynamic weight assignment algorithm; Figure 4 is the schematic flow chart of constructing a dynamic risk assessment model based on the Markov chain. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0026] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0027] Please refer to Figures 1-4 , the present invention provides a technical solution: an airport bird strike prevention method based on multi-sensor fusion, the method includes: Collect multimodal sensor data in the airspace around the airport: Use multiple sensors to collect rich information, including radar detection data, radio detection data, laser monitoring data and meteorological data. These data monitor the targets and environmental conditions in the airspace around the airport from different dimensions, providing comprehensive basic information for subsequent analysis.

[0028] Perform time synchronization and spatial registration on multimodal sensor data to generate a fused data stream: Due to the differences in clocks and detection areas of different sensors, time synchronization and spatial registration operations are used to eliminate clock deviations between devices, calibrate the spatial overlap of each sensor's detection area, establish a unified coordinate system, and integrate multimodal sensor data into a consistent and correlated fused data stream for subsequent processing.

[0029] Based on the dynamic weight allocation algorithm, the fused data stream is featured extracted to generate target trajectory information and target attribute information: With the help of the dynamic weight allocation algorithm, the fused data stream is analyzed by comprehensively considering factors such as sensor confidence and data integrity, and the target trajectory information, such as the target's motion path, speed changes, etc., as well as the target's attribute information, such as size, shape, etc., are extracted, laying the foundation for identifying the target type.

[0030] Use multimodal classification models to identify target types based on target trajectory information and target attribute information, and distinguish between bird targets and drone targets: Use multimodal classification models, through convolutional neural networks, long short-term memory networks, and transfer learning and other technical means, to conduct in-depth analysis of the target's trajectory and attribute information, and accurately distinguish whether the target entering the airspace around the airport is a bird or a drone, so as to take targeted countermeasures.

[0031] According to the target type recognition results and target motion parameters, a dynamic risk assessment model based on Markov chain is constructed to generate the intrusion risk level: the state space is defined by combining the target type recognition results and the target motion parameters, such as flight altitude, speed, and relative distance to the core area. The state transition probability matrix is ​​trained through historical intrusion event data, and the risk entropy value is calculated based on the real-time target motion parameters to determine the risk level of the intrusion target to the airport.

[0032] When the intrusion risk level exceeds the preset threshold, the multi-level alarm strategy is triggered, and the expulsion device is linked to execute the adaptive expulsion command: Once the intrusion risk level reaches or exceeds the preset threshold, the system immediately activates the multi-level alarm strategy, sends alarm information to relevant personnel and systems, and links the expulsion device to dynamically adjust the priority, coverage and power parameters of the expulsion device according to the target type and location, execute adaptive expulsion commands, and eliminate potential threats.

[0033] The present invention will be further described below in conjunction with Embodiments 1 to 5:

[0034] Embodiment 1: This embodiment mainly elaborates on the acquisition method of multimodal sensor data and the specific implementation of time synchronization and spatial registration. Its function is to obtain accurate and unified basic data, providing a reliable basis for subsequent analysis and processing.

[0035] In the data acquisition link, a three-coordinate low-altitude detection radar is used to collect target azimuth, altitude, and speed information. The three-coordinate low-altitude detection radar uses the principle of electromagnetic wave reflection to scan the low-altitude area around the airport. When the electromagnetic wave it emits encounters a target, it will be reflected back. The radar calculates the azimuth angle, elevation angle, and altitude information of the target accurately by receiving the reflected wave and based on information such as the propagation time and frequency change of the electromagnetic wave. At the same time, the radial velocity of the target is measured through the Doppler effect to obtain the target's motion speed information. This radar can effectively detect flying targets at a long distance, providing early warning information for airport bird strike prevention.

[0036] The electromagnetic signal characteristics in the communication frequency band of the UAV are captured by a radio detection device. The UAV communicates through specific frequency bands during flight, and the radio detection device can monitor these frequency bands in real time. It can identify characteristic information such as the frequency, modulation method, and signal strength of the UAV communication signal. These characteristic information are like the "fingerprints" of the UAV. By comparing with the known UAV model database, it helps to more accurately identify UAV targets, distinguish different types of UAVs, and provide support for subsequent countermeasures.

[0037] A laser surveillance device is used to obtain the three-dimensional contour and motion trajectory of the target. The laser surveillance device emits laser beams, and when the laser beams encounter a target, they are reflected back. The device calculates the distance between the target and the device by measuring the time difference between the emission and reception of the laser. By measuring the distances in multiple different directions and combining the scanning angle information of the device, the three-dimensional contour of the target can be constructed. At the same time, as the target moves, the laser surveillance device continuously tracks and measures, recording the position information of the target at different times, so as to obtain the motion trajectory of the target. This device can provide high-precision target shape and motion information, which is of great significance for identifying targets such as birds and small UAVs.

[0038] Integrate the real-time meteorological data of the remote meteorological radar, including wind speed, air pressure, and precipitation information. The remote meteorological radar detects meteorological targets in the atmosphere, such as clouds, raindrops, etc., by emitting electromagnetic waves, and inversely calculates meteorological parameters according to the characteristics of the reflected waves. Wind speed information can reflect the flow state of the atmosphere and affect the flight trajectories of birds and drones; changes in air pressure may cause changes in the flight altitude of birds; precipitation information may affect the detection performance of sensors and the flight environment of birds and drones. Integrating these meteorological data into the system helps to more comprehensively analyze the motion state of the target and improve the accuracy and reliability of bird strike prevention.

[0039] In terms of time synchronization, perform Kalman filter correction on the timestamps of multi-modal sensor data to eliminate the clock deviation between devices. Due to the differences in the clock accuracies of different sensors, problems of time inconsistency will occur during the data acquisition process. Kalman filter is an optimal linear recursive filtering algorithm that uses the estimated value of the previous moment and the measured value of the current moment, and through two steps of prediction and update, continuously optimizes the estimation of the timestamp. In practical applications, first establish a Kalman filter model according to the clock characteristics and historical data of the sensors. Then, input the data timestamps collected by the sensors as measured values into the model. The model gradually eliminates the clock deviation through continuous iterative calculations, enabling the data of different sensors to be synchronized in time and ensuring the accuracy and consistency of data processing.

[0040] In terms of spatial registration, perform spatial overlap calibration on the detection areas of radar, laser, and radio devices based on the coordinates of the geographic information system to establish a unified coordinate system. The geographic information system (GIS) has accurate geographic coordinate information. Based on its coordinates, calibrate the detection areas of each sensor. For radar, laser, and radio devices, determine their positions and detection ranges in the geographic coordinate system respectively. Through coordinate transformation algorithms, convert the data collected by each sensor into a unified geographic coordinate system, so that the data of different sensors are consistent in space. In this way, in the subsequent data fusion and analysis process, the target can be accurately located and tracked, avoiding misjudgment and missed judgment problems caused by spatial inconsistency.

[0041] Embodiment 2: This embodiment introduces the application and implementation of the dynamic weight allocation algorithm. The role of this algorithm is to reasonably allocate the weights of each sensor during the multi-sensor data fusion process, highlight the data role of the dominant sensors, and improve the accuracy of feature extraction.

[0042] The dynamic weight allocation algorithm satisfies the following relationship:

[0043] Among them, is the weight of the th sensor, is the sensor confidence of the th sensor, is the data integrity factor of the th sensor, and is the dynamic adjustment coefficient, is the total number of sensors.

[0044] In practical applications, the sensor confidence reflects the reliability of the sensor measurement data. For example, for a three-coordinate low-altitude detection radar, if its equipment status is good, the signal strength is stable, and the environmental interference is small, then its sensor confidence is relatively high; on the contrary, if the radar is affected by factors such as electromagnetic interference and equipment aging, and the accuracy of the measurement data decreases, its sensor confidence will decrease. The data integrity factor is used to measure the integrity of the data collected by the sensor. For example, for a laser surveillance device, if part of the measurement data is missing at a certain moment due to reasons such as the target being blocked, then its data integrity factor will decrease accordingly; when the device is working properly and can collect target information completely, the data integrity factor is relatively high.

[0045] The dynamic adjustment coefficient and are adjusted according to the actual application scenario and requirements. In areas with complex airport perimeters, frequent bird and drone activities, if more attention is paid to the reliability of the sensor, and the data of sensors with high reliability is preferentially used for feature extraction, the value of can be appropriately increased; if more attention is paid to the integrity of the data, and all the data collected by all sensors is fully utilized, even if the reliability of some data is slightly low, the value of can also be appropriately increased.

[0046] During the operation of the system, first, according to factors such as the performance parameters of each sensor, historical measurement data, and the current working environment, determine the sensor confidence and the data integrity factor of each sensor. Then, according to the preset dynamic adjustment coefficient and , as well as the total number of sensors , substitute them into the above formula to calculate the weight of each sensor. When extracting features from the fused data stream, according to the calculated weight , the data collected by each sensor is weighted. The data of sensors with higher weights account for a larger proportion in the feature extraction process and have a greater impact on the finally generated target trajectory information and target attribute information, thus highlighting the role of the data of the dominant sensors and improving the accuracy and reliability of feature extraction.

[0047] Embodiment 2: This embodiment details the construction and working principle of the multi-modal classification model, whose function is to accurately distinguish bird targets from drone targets through the analysis of multi-modal data, providing accurate target type information for subsequent risk assessment and countermeasures.

[0048] The multi-modal classification model includes using a convolutional neural network to extract the target contour features in the laser monitoring data. The convolutional neural network (CNN) is a deep learning model specifically designed for processing image data, which has characteristics such as local perception and weight sharing, and can automatically extract features in images. The target three-dimensional contour information obtained by the laser monitoring device can be converted into image-form data and input into the convolutional neural network. The convolutional neural network extracts the features of the target contour through a series of operations of convolutional layers, pooling layers, and fully connected layers. For example, the convolutional kernels in the convolutional layer can detect low-level features such as edges and textures in the image, and the pooling layer compresses and reduces the dimensions of the features, retaining key information. After multiple layers of operations, the feature vectors of the target contour are finally output in the fully connected layer, and these feature vectors contain important information such as the shape and size of the target, which helps to distinguish birds from drones.

[0049] Use a long short-term memory network to analyze the temporal features of radar and radio data. Radar and radio data have temporality, that is, the data shows certain patterns as it changes over time. The long short-term memory network (LSTM) is a special recurrent neural network that can effectively process time series data and solve the problems of gradient vanishing and gradient explosion in traditional recurrent neural networks. Input the data of the target position, speed, etc. detected by the radar that change over time, and the data of the changes in the characteristics of the drone communication signals detected by the radio over time into the long short-term memory network. The LSTM can selectively remember and forget information through the gating mechanism, and can learn the long-term dependencies and short-term change trends in the data. For example, by analyzing the temporal features of the radar data, it can be judged whether the flight trajectory of the target conforms to the typical flight patterns of birds or drones; by analyzing the temporal features of the radio data, the change rules of the drone communication signals can be identified, and the model and status of the drone can be further determined.

[0050] Fusing the bird morphology database and the UAV model database through transfer learning to generate the target classification result. Transfer learning is a technique that transfers the knowledge learned in one task to another related task. In the present invention, first, a bird morphology database and a UAV model database are respectively established, and these databases contain a large amount of characteristic information of known birds and UAVs. Then, a deep learning model trained in other related image classification or target recognition tasks, such as a convolutional neural network model trained on a large-scale image dataset, is transferred to this system. By fine-tuning some parameters of the model, it is adapted to the classification tasks of birds and UAVs. The target contour features of the extracted laser monitoring data and the temporal features of radar and radio data are input into the model after transfer learning. The model combines the information in the bird morphology database and the UAV model database, conducts comprehensive analysis and judgment, and finally generates the target classification result to accurately distinguish bird targets from UAV targets.

[0051] Embodiment 4:

[0052] This embodiment elaborates in detail the construction of a dynamic risk assessment model based on Markov chain and the calculation of risk entropy value. Its function is to accurately evaluate the risk level of an intrusion target to the airport according to the motion state and historical data of the target, providing a basis for formulating reasonable preventive measures.

[0053] When constructing the dynamic risk assessment model, the state space is defined as the flight altitude, speed of the intrusion target, and the relative distance from the core area. Flight altitude is one of the important factors for risk assessment, and different altitudes pose different threats to the airport. For example, an intrusion target approaching the aircraft takeoff and landing altitude has a significantly higher risk than a target flying in a higher or lower airspace. Speed can also reflect the threat level of the target, and a target flying at high speed may pose a greater threat to the airport in a short time. The relative distance from the core area is directly related to the potential impact of the target on the key facilities of the airport, and the closer the distance, the higher the risk. By taking these three parameters as the dimensions of the state space, the state of the intrusion target can be described comprehensively and accurately.

[0054] The state transition probability matrix is trained through historical intrusion event data. The Markov chain assumes that the state of the system at a future moment depends only on the current state and is independent of the past historical states. In this system, a large amount of historical intrusion event data is used to count the transition frequencies between different states. For example, count the number of times of transitioning to other states after a certain period of time in a state of a certain flight altitude, speed, and relative distance from the core area. According to these statistical data, the state transition probability is calculated, and the state transition probability matrix is constructed. This matrix reflects the possibility of the intrusion target transitioning between different states and provides a basis for subsequent calculation of the risk entropy value.

[0055] Calculate the risk entropy value based on the real-time target motion parameters to determine the risk level. The risk entropy value satisfies the following relationship:

[0056] where is the risk entropy value, is the probability of the th state, is the total number of states.

[0057] In actual calculation, first determine the current state of the target according to the real-time obtained target motion parameters. Then, combined with the state transition probability matrix, calculate the probabilities of the target being in different states . Substitute these probability values into the risk entropy value formula for calculation. The larger the risk entropy value, the higher the uncertainty of the target state and the greater the risk to the airport. According to the pre-set risk entropy value threshold, the risk level is divided into different levels, such as low risk, medium risk, high risk, etc. For example, when the risk entropy value is less than a certain threshold, it is determined as low risk; when the risk entropy value is within a certain range, it is determined as medium risk; when the risk entropy value is greater than another threshold, it is determined as high risk. In this way, the risk level of the intrusion target to the airport can be dynamically and accurately evaluated according to the real-time motion state of the target.

[0058] Example 5: The multi-level alarm strategy includes dividing the warning levels according to the risk level, including first-level alert, second-level alert and emergency response. When the intrusion risk level reaches the first-level alert, it indicates that the threat of the target to the airport is small, but still needs attention. At this time, the system triggers the audible and visual alarm device to issue low-frequency and low-brightness audible and visual alarms in relevant areas of the airport to remind the staff to pay attention to observing the target dynamics. At the same time, the air traffic control communication system sends a brief warning message to the controller, such as the approximate location and type of the target, but does not affect the normal air traffic control process. The mobile terminal also pushes a prompt message to the relevant security personnel to inform that there is a potential intrusion target and they need to be vigilant.

[0059] When the risk level is upgraded to the second-level alert, the threat of the target to the airport increases. The audible and visual alarm device issues louder and brighter audible and visual alarms to attract more people's attention. The air traffic control communication system details the motion parameters, risk level and other information of the target to the controller, and the controller adjusts the flight takeoff and landing arrangements according to the situation to ensure flight safety. The warning information pushed by the mobile terminal is more detailed, including the real-time location of the target, the expected motion trajectory, etc. After receiving the notice, the security personnel start to prepare the corresponding repelling equipment and stand by at any time.

[0060] When the risk level reaches the emergency response level, it indicates that the target poses a great threat to the airport and may soon have a serious impact on flight safety. At this time, the audible and visual alarm devices emit strong audible and visual alarms, covering all relevant areas of the airport to ensure that all personnel can be informed of the danger in a timely manner. The air traffic control communication system urgently notifies all flights to suspend takeoff and landing operations to avoid conflicts with the intruding target. The mobile terminal pushes emergency warning messages, requiring security personnel to take immediate action to drive away the intruding target.

[0061] The generation of adaptive drive-away instructions includes matching the priority of drive-away devices based on the target type. For bird targets, the acoustic drive-away is started preferentially, and for drone targets, electromagnetic interference is started preferentially. For bird targets, acoustic drive-away is a relatively effective method. According to the auditory characteristics of different birds, appropriate frequencies and intensities of sound waves are selected for emission. For example, for some common birds, sound waves with frequencies between 2000 Hz and 5000 Hz and intensities between 80 dB and 100 dB can be used to startle and drive them away. The acoustic drive-away device dynamically adjusts the emission direction and coverage range according to the position of the target to ensure that the area where the bird target is located can be covered.

[0062] For drone targets, electromagnetic interference is a commonly used drive-away means. By emitting electromagnetic signals with specific frequencies and powers, the communication and navigation systems of the drones are interfered, causing them to lose control or make an emergency landing. According to the model and communication frequency band of the drone, appropriate electromagnetic interference frequencies are selected. For example, for drones using the 2.4 GHz frequency band for communication, electromagnetic interference signals with a center frequency near 2.4 GHz and a certain bandwidth are emitted. At the same time, according to the distance and flight attitude of the drone, the power parameters of the electromagnetic interference device are dynamically adjusted to ensure that the interference signal can effectively act on the drone and prevent it from flying normally.

[0063] During the entire alarm and drive-away process, the system closely combines the risk level and the target type, conveys threat information in a timely manner through a multi-level alarm strategy, and accurately and effectively executes the drive-away operation through adaptive drive-away instructions, minimizing the occurrence probability of bird strikes and drone intrusion incidents at the airport and ensuring the safe operation of the airport.

[0064] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or device.

[0065] Although embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An airport bird strike prevention method based on multi-sensor fusion, characterized in that, The following steps are involved: Collect multimodal sensor data of the airspace around the airport, wherein the multimodal sensor data includes radar detection data, radio detection data, laser monitoring data and meteorological data; Performing time synchronization and spatial registration on the multimodal sensor data to generate a fused data stream; Extracting features from the fused data stream based on a dynamic weight allocation algorithm to generate target trajectory information and target attribute information; Using a multimodal classification model to identify the target type based on the target trajectory information and target attribute information, and distinguish between bird targets and drone targets; According to the target type recognition result and the target motion parameters, a dynamic risk assessment model based on a Markov chain is constructed to generate an intrusion risk level; When the intrusion risk level exceeds a preset threshold, a multi-level alarm strategy is triggered, and the expulsion device is linked to execute an adaptive expulsion instruction.

2. The airport bird strike prevention method based on multi-sensor fusion according to claim 1, wherein The multimodal sensor data collected in the airspace around the airport includes: Use three-coordinate low-altitude detection radar to collect target azimuth, altitude and speed information; Capture the electromagnetic signal characteristics of the drone communication frequency band through radio detection equipment; Use laser monitoring equipment to obtain the target's three-dimensional profile and motion trajectory; Integrate real-time weather data from long-range weather radar, including wind speed, air pressure and precipitation information.

3. The airport bird strike prevention method based on multi-sensor fusion according to claim 1, wherein, The time synchronization and space registration include: Perform Kalman filtering correction on the timestamp of multimodal sensor data to eliminate clock deviation between devices; Based on the geographic information system coordinates, the detection areas of radar, laser and radio equipment are spatially overlapped and calibrated to establish a unified coordinate system.

4. The airport bird strike prevention method based on multi-sensor fusion according to claim 1, wherein, The dynamic weight allocation algorithm satisfies the following relationship: ; Among them, is the weight of the th sensor, is the sensor confidence of the th sensor, is the data integrity factor of the th sensor, and are dynamic adjustment coefficients, is the total number of sensors.

5. The airport bird strike prevention method based on multi-sensor fusion according to claim 1, wherein The multimodal classification model includes: Convolutional neural network is used to extract target contour features from laser monitoring data; Analyze the temporal characteristics of radar and radio data using long short-term memory networks; The bird morphology database and the drone model database are integrated through transfer learning to generate target classification results.

6. The method for preventing airport bird strikes based on multi-sensor fusion according to claim 1, characterized in that, The construction of the dynamic risk assessment model includes: The state space is defined as the flight altitude, speed and relative distance of the invading target from the core area; Train the state transition probability matrix through historical intrusion event data; The risk entropy value is calculated based on the real-time target motion parameters to determine the risk level.

7. The method for preventing airport bird strikes based on multi-sensor fusion according to claim 6, wherein The risk entropy value satisfies the following relationship: ; Among them, is the risk entropy value, is the probability of the -th state, is the total number of states.

8. The method for preventing airport bird strikes based on multi-sensor fusion according to claim 1, characterized in that The multi-level alarm strategy includes: The warning levels are divided into level one, level two and emergency response according to the risk level; Trigger the sound and light alarm equipment, air traffic control communication system and mobile terminal to push warning information.

9. The method for preventing airport bird strikes based on multi-sensor fusion according to claim 1, wherein The generation of the adaptive drive-away instruction includes: The priority of the repelling device is matched based on the target type. For bird targets, the acoustic repelling is initiated first, while for drone targets, the electromagnetic interference is initiated first. Dynamically adjust the coverage and power parameters of the drive-away device based on the target location.

10. An airport bird strike prevention system based on multi-sensor fusion, characterized in that, include: Data acquisition module: used to collect multimodal sensor data in the airspace around the airport, including radar detection data, radio detection data, laser monitoring data and meteorological data; Data preprocessing module: performs time synchronization and spatial registration on the multimodal sensor data to generate a fused data stream; Feature extraction module: extracts features from the fused data stream based on a dynamic weight allocation algorithm to generate target trajectory information and target attribute information; Target recognition module: using a multimodal classification model to identify the target type based on the target trajectory information and target attribute information, and distinguish between bird targets and drone targets; Risk assessment module: constructs a dynamic risk assessment model based on a Markov chain according to the target type recognition result and target motion parameters, and generates an intrusion risk level; Alarm and expulsion module: When the intrusion risk level exceeds the preset threshold, a multi-level alarm strategy is triggered, and the expulsion device is linked to execute an adaptive expulsion instruction.

Images