Airport bird situation automatic detection and intelligent bird repelling device and method

Abstract

The invention discloses an airport bird condition automatic detection and intelligent bird repelling device and method, and the method comprises the steps: collecting airport airspace environment data in real time through a multi-source sensor array, carrying out the preprocessing of the environment data, and generating a fusion data set; a target detection model is constructed based on a deep learning algorithm to perform bird target recognition on the fusion data set, and target feature information is extracted and comprises bird positions, movement tracks, the number and types; calculating a threat level of bird activities to flights through a risk assessment model established based on fuzzy logic and a Bayesian network in combination with airport flight take-off and landing time sequence data and target feature information; triggering a grading early warning signal according to the threat grade, and linking the bird repelling equipment to execute directional bird repelling operation. According to the invention, an airport bird situation automatic detection and emergency prevention and control system is formed, the bird strike risk detection precision, the early warning accuracy and the bird repelling efficiency are improved, and the aviation operation safety is guaranteed.

Description

Technical Field

[0001] The present invention relates to the technical field of automatic detection of bird conditions and risk prevention and control at airports, and in particular to an automatic detection of bird conditions and intelligent bird-repelling device and method at airports. Background Art

[0002] In recent years, with the rapid development of the global air transport industry, airport flight traffic has continued to increase, and flight safety accidents caused by bird strikes have occurred frequently. Bird monitoring and prevention have gradually become one of the core issues of airport safety management; traditional airport bird monitoring methods generally use manual patrols, ground telescope observations, single radar or single camera detection methods, which have problems such as limited detection range, weak target recognition ability, insufficient real-time performance, and low monitoring accuracy, which are difficult to meet the high standards of safety for modern aviation operations.

[0003] In order to cope with the increasingly complex and changeable patterns of bird activities, existing technologies have gradually developed towards multi-sensor fusion detection. Some bird monitoring technologies that combine radar and optical equipment or are based on machine vision algorithms have emerged. Although they have improved the ability to detect bird targets to a certain extent, the existing technical solutions still have much room for improvement in information fusion accuracy, complex scene recognition capabilities and risk assessment accuracy.

[0004] Specifically, the existing airport bird monitoring technology generally has the following shortcomings: First, the fusion of multi-source sensor data is not deep enough, and the multi-modal data is not fully coordinated, resulting in low efficiency in the comprehensive utilization of environmental data and affecting the accuracy of target detection; second, in the target identification process, there is a lack of accurate extraction of comprehensive characteristic information such as bird flight trajectories, species, and numbers, which makes it impossible to effectively predict bird behavior and accurately divide risk levels; third, in terms of bird activity risk assessment, most methods only use simple rule models, fail to effectively integrate airport flight takeoff and landing data for real-time threat assessment, and cannot accurately and quickly form effective early warnings for flight safety.

[0005] Therefore, how to effectively improve the accuracy of automated bird detection and bird strike risk prediction at airports, and further realize efficient early warning and bird-repelling linkage, is an important technical issue that needs to be solved urgently.

[0006] In summary, the existing airport bird detection technology has the problems of insufficient detection accuracy, low data fusion efficiency and untimely risk warning. The present invention improves the accuracy of airport bird detection and risk assessment by utilizing multi-source sensor fusion, deep learning target recognition and fuzzy logic combined with the risk assessment model of Bayesian network, and realizes timely and effective graded warning and bird repellent linkage. Summary of the invention

[0007] The purpose of this section is to summarize some aspects of the embodiments of the present invention and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the abstract of the specification and the title of the invention of this application to avoid blurring the purpose of this section, the abstract of the specification and the title of the invention, and such simplifications or omissions cannot be used to limit the scope of the present invention.

[0008] In view of the above existing problems, the present invention is proposed.

[0009] In order to solve the above technical problems, the present invention provides the following technical solutions: an automated detection and intelligent bird-repelling method for airport birds, comprising: collecting airport airspace environmental data in real time through a multi-source sensor array, and pre-processing the environmental data to generate a fused data set; Building a target detection model based on a deep learning algorithm to identify bird targets in the fused data set and extract target feature information, wherein the target feature information includes the location, movement trajectory, number and type of birds; Combining the airport flight take-off and landing time series data with the target feature information, the threat level of bird activities to flights is calculated through a risk assessment model based on fuzzy logic and Bayesian network; A graded warning signal is triggered according to the threat level, and the bird-repelling equipment is linked to perform a directional bird-repelling operation.

[0010] As a preferred solution of the automatic detection method of bird situation at an airport described in the present invention, the multi-source sensor array includes a radar device, a visible light camera, an infrared thermal imager and a sonar detector, and the airport airspace environment data collected in real time by the multi-source sensor array includes radar data, visible light image data, infrared thermal imaging data, sonar data and meteorological information around the airport runway; Among them, the radar data includes distance, radial velocity and azimuth, the visible light image data includes frame sequence images, target pixel coordinates and color texture features, the infrared thermal imaging data includes target thermal radiation intensity and thermal imaging coordinates, the sonar data includes echo intensity, delay time and target azimuth, and the meteorological information around the airport runway includes wind direction, wind speed and visibility.

[0011] As a preferred solution of the automatic detection method for bird sightings at airports described in the present invention, the environmental data is preprocessed to generate a fused data set, including: Performing unified time calibration on the environmental data; Perform bandpass filtering and background noise suppression on radar, infrared and sonar data, and perform background modeling and moving target segmentation on visible light images; According to the coordinate systems of different sensors, the target detection results are converted into a unified spatial coordinate system, and the Kalman filter algorithm is used to realize the association and redundancy removal of multi-source tracking results; Conduct secondary screening on targets with lower confidence levels and eliminate duplicate targets; Annotate the bounding boxes of birds in visible light and infrared thermal images; Visible light and infrared data are enhanced by image flipping, rotation, brightness adjustment and noise addition methods; The processed data and the annotation information are matched according to the timestamp to form a unified serial data set, which is the fused data set.

[0012] As a preferred solution of the automatic detection method for bird situation at an airport described in the present invention, a target detection model is constructed based on a deep learning algorithm to identify bird targets in the fused data set and extract target feature information, including: A bidirectional long short-term memory network is used to perform multi-level fusion of radar, infrared and visible light features; For the small-scale and easily overlapping features of bird targets, an attention mechanism is set to enhance the key area features; Use the labeled fusion data set for supervised training, and the training loss function adopts a multi-task weighted form: in, is the classification loss, is the bounding box regression loss, is the multimodal consistency loss; , , They are the weighted coefficients for classification loss, bounding box regression loss, and multimodal consistency loss respectively; Perform multimodal matching on the pre-processed video frames, infrared thermal imaging frames and radar / sonar data, and the network outputs the predicted position, type and confidence of the bird target; If the confidence level is higher than the threshold , it is determined to be a valid bird target, and its flight trajectory, number and appearance characteristics are further extracted; The identified target information is indexed by the target ID to generate a target feature information set including the bird's location, movement trajectory, number and type.

[0013] As a preferred solution of the automatic detection method for bird sightings at airports described in the present invention, the identification and extraction of the target detection model includes: Track the trajectory of the same target detected by continuous video frames or radar, and use the Kalman filter algorithm to obtain the precise motion trajectory; The position of each moment in the trajectory and speed Discretized storage is performed to form a trajectory sequence; Conduct statistical analysis on flight speed, altitude and direction to determine whether birds have behavioral patterns of flying quickly across the runway or forming flocks; Combining deep learning classification networks with biological prior knowledge, the body shape, feather characteristics and thermal distribution information of birds in visible light and infrared images are classified and identified; Based on the identification results, the bird species are divided into different categories such as small finches, raptors, gulls, wading birds, etc., and are marked with clustering behavior labels or nocturnal activity frequency labels.

[0014] As a preferred solution of the automatic detection method for bird activity at an airport described in the present invention, the threat level of bird activity to flights is calculated by combining the airport flight take-off and landing time sequence data with the target feature information through a risk assessment model based on fuzzy logic and Bayesian network, including: Set the number of birds n and population density , the distance d from the center of the runway and the flight altitude h are the key indicators, and a corresponding fuzzy set is established for each indicator; Using membership function Fuzzy processing of each indicator: Among them, a and b are interval thresholds; Input the fuzzified indicators into the Bayesian network nodes to form event conditions; Combined with the aircraft take-off and landing frequency and take-off and landing time information in the flight take-off and landing time series data, the posterior probability of bird strike risk is calculated; The evaluation values ​​of each indicator output by fuzzy logic and the posterior probability of Bayesian network are combined to determine the threat level of bird activities to flights according to the weighted rule; If the comprehensive score exceeds the preset upper threshold, it is judged as high risk or extremely high risk; If the comprehensive score is in the middle range, it is judged as medium risk; If the comprehensive score is below the lower threshold, it is judged as low risk.

[0015] As a preferred solution of the automatic detection method for bird situation at an airport according to the present invention, the threat levels include: When the comprehensive score is in the range of [0, 0.3), it is judged as low threat level L1, which means that the number of birds is very small and far away from the critical areas of the runway, and basically does not pose a safety impact on aircraft takeoff and landing; When the comprehensive score is in [0.3, 0.5), it is judged as medium threat level L2, then the number of birds is moderate or occasionally enters the runway safety buffer zone, and monitoring and regular warning are required; When the comprehensive score is between [0.5, 0.7), it is judged as a high threat level L3, which means that the bird activities are frequent and have obviously approached the flight path, posing a potential risk to flights that are taking off or about to take off or land. When the comprehensive score is ≥0.7, it is judged as an extremely high threat level L4, then large flocks of birds repeatedly appear on the runway and its surroundings, posing a serious threat to flight safety.

[0016] As a preferred solution of the automatic detection method for bird situation at an airport of the present invention, a graded warning signal is triggered according to the threat level, and a bird-repelling device is linked to perform a directional bird-repelling operation, including: When the threat level is L1, only text / voice reminders are sent to tower personnel to remind them to stay focused; When the threat level is L2, an audio signal or light flashing warning is triggered to remind ground and flight crews to pay close attention to the bird situation at the airport; When the threat level is L3, a high-frequency harsh sound wave or strong light warning is triggered, and real-time risk warnings are issued to flights about to take off and land; When the threat level is L4, the highest level emergency alarm is triggered, and the control department is advised to delay flight takeoff and landing, and immediately activate a linkage mechanism of multiple bird-repelling measures.

[0017] As a preferred solution of the automatic detection method for bird situation at an airport according to the present invention, the linked bird-repelling device performs a directional bird-repelling operation, including: When the threat level is ≥ L2, the sonic repelling device will be automatically activated to send out adjustable frequency high-decibel sonic interference; When the threat level is ≥ L3, in addition to sound wave interference, laser or strong beam bird repellent devices are turned on, and bionic sounds such as the calls of birds of prey can be played to repel them; When the threat level is L4, multiple devices are activated simultaneously to implement a comprehensive bird repellent solution including laser, strong light beam, bionic natural enemy and aerosol.

[0018] The present invention also discloses an intelligent bird-repelling device based on the aforementioned automatic detection method for bird conditions at an airport, comprising a monitoring module arranged on a runway, the monitoring module comprising a radar device, a visible light camera, an infrared thermal imager and a sonar detector, and the monitoring modules are all connected to a control host; It also includes an early warning module, which includes a tower display screen, a voice player, and warning lights arranged on both sides of the runway; The device also includes a bird-repelling execution module, which includes a frequency-adjustable sound wave transmitter, a strong light, a bionic sound player and an aerosol generator. Beneficial effects of the present invention: 1. The radar equipment, visible light camera, infrared thermal imager and sonar detector work together to realize multi-dimensional and multi-modal comprehensive perception of airport bird information, which effectively makes up for the problem of limited detection range of a single sensor or missed target detection and false alarm. The pre-processed data fusion process further eliminates the data inconsistency between sensors, provides higher quality and more reliable environmental data, and thus improves the accuracy and stability of subsequent bird target detection; 2. Through the precise identification and dynamic tracking of bird targets under complex backgrounds and multi-target interference conditions, the accuracy and reliability of bird activity identification are effectively improved. At the same time, the bird activity tracks, quantity and bird species characteristics are accurately extracted, which further provides a sufficient and accurate data basis for subsequent risk assessment, helps to accurately judge bird behavior patterns and clarify the potential risk types posed to airport flight safety; 3. By incorporating flight take-off and landing time data into the risk assessment process, the intersection of aircraft operation status and bird activity in space and time is effectively considered. The uncertainty of qualitative and quantitative indicators is handled through fuzzy logic, and the risk probability is quantitatively inferred using Bayesian networks. This comprehensively realizes the refined and real-time analysis of the impact of bird activities on airport flight safety, accurately divides the risk level, and greatly improves the timeliness of bird strike incident warning judgment; 4. Through the direct linkage between risk assessment results and the actual prevention and control measures of the airport, different levels of warning signals are automatically triggered in real time according to the risk level and corresponding bird-repelling measures are implemented, thereby achieving the optimal scheduling and precise implementation of the airport's bird-repelling resources, avoiding the defects of slow response and waste of resources of traditional manual patrols or single warning methods; further enhancing the airport's rapid response and efficient handling capabilities to sudden bird strike risks, and significantly improving the airport's flight safety prevention and control level. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative labor. Among them: Figure 1 It is a flow chart of the automatic detection method of bird situation at an airport shown in the present invention. DETAILED DESCRIPTION

[0020] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, but not all of the embodiments.

[0021] Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without making any creative work should fall within the scope of protection of the present invention.

[0022] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0023] According to an embodiment of the present invention, Figure 1 The flowchart shown in the figure shows an automatic detection method of bird situation at an airport and an intelligent method for driving away birds, comprising: S1. Collect airport airspace environment data in real time through a multi-source sensor array, and pre-process the environment data to generate a fused data set. The following points need to be explained in this step: The multi-source sensor array includes radar equipment, visible light cameras, infrared thermal imagers and sonar detectors. The airport airspace environment data collected in real time by the multi-source sensor array includes radar data, visible light image data, infrared thermal imaging data, sonar data and meteorological information around the airport runway; Among them, radar data includes distance, radial velocity and azimuth, visible light imaging data includes frame sequence images, target pixel coordinates and color texture features, infrared thermal imaging data includes target thermal radiation intensity and thermal imaging coordinates, sonar data includes echo intensity, delay time and target azimuth, and meteorological information around the airport runway includes wind direction, wind speed and visibility; As examples, radar equipment includes multi-beam scanning radar and pulse Doppler radar for detecting the flight speed, distance and direction of bird targets within a range of 0.5 km to 10 km; As an example, the visible light camera includes a high-resolution dynamic camera module for acquiring the morphological characteristics and activity videos of birds within a visual range of 0 to 2 kilometers; As an example, the infrared thermal imager includes medium-wave infrared and long-wave infrared imaging modules to detect the heat signature characteristics of birds within a range of 0 to 1.5 kilometers in night / low-light conditions; As an example, a sonar detector includes sonar transmitting and receiving units distributed in horizontal and vertical directions, which are used to identify and locate the acoustic echo characteristics of birds within a range of 0 to 1 km.

[0024] In an optional implementation, the fused data set is generated by preprocessing the environmental data, including: Perform unified time calibration on environmental data; Perform bandpass filtering and background noise suppression on radar, infrared and sonar data, and perform background modeling and moving target segmentation on visible light images; According to the coordinate systems of different sensors, the target detection results are converted into a unified spatial coordinate system, and the Kalman filter algorithm is used to realize the association and redundancy removal of multi-source tracking results; Conduct secondary screening on targets with lower confidence levels and eliminate duplicate targets; Annotate the bounding boxes of birds in visible light and infrared thermal images; Visible light and infrared data are enhanced by image flipping, rotation, brightness adjustment and noise addition methods; The processed data and the annotation information are matched according to the timestamp to form a unified serial data set, which is the fused data set.

[0025] S2. Build a target detection model based on deep learning algorithm to identify birds in the fused data set and extract target feature information, which includes bird location, movement trajectory, quantity and species. The following points need to be explained in this step: A bidirectional long short-term memory network is used to perform multi-level fusion of radar, infrared and visible light features; For the small-scale and easily overlapping features of bird targets, an attention mechanism is set to enhance the key area features; Use the labeled fusion data set for supervised training, and the training loss function adopts a multi-task weighted form: in, is the classification loss, is the bounding box regression loss, is the multimodal consistency loss; , , They are the weighted coefficients for classification loss, bounding box regression loss, and multimodal consistency loss respectively; Perform multimodal matching on the pre-processed video frames, infrared thermal imaging frames and radar / sonar data, and the network outputs the predicted position, type and confidence of the bird target; If the confidence level is higher than the threshold , it is determined to be a valid bird target, and its flight trajectory, number and appearance characteristics are further extracted; The identified target information is indexed by the target ID to generate a target feature information set including the bird's location, movement trajectory, number and type.

[0026] Furthermore, the identification and extraction of target detection models include: Track the trajectory of the same target detected by continuous video frames or radar, and use the Kalman filter algorithm to obtain the precise motion trajectory; The position of each moment in the trajectory and speed Discretized storage is performed to form a trajectory sequence; Conduct statistical analysis on flight speed, altitude and direction to determine whether birds have behavioral patterns of flying quickly across the runway or forming flocks; Combining deep learning classification networks with biological prior knowledge, the body shape, feather characteristics and thermal distribution information of birds in visible light and infrared images are classified and identified; Based on the identification results, bird species are divided into different categories such as small finches, raptors, gulls, wading birds, etc., and are marked with clustering behavior labels or nocturnal activity frequency labels.

[0027] Specifically, each detected bird target is assigned a unique ID; The same target detected in different frames or sensors is matched and merged into different time records under the same ID; For each target ID, key information such as location, movement trajectory, quantity and type is recorded and output as a unified data structure, for example: in, Indicates the current position of the target in the spatial coordinate system. It represents the motion trajectory sequence from the first detection to the current moment. Indicates the number of birds in the same group or frame. Indicates specific bird species, such as finches, raptors, gulls, and wading birds. This includes additional tags for whether the birds are flying in groups or are nocturnal.

[0028] As an example, the Kalman filter algorithm is used to track the same seagull target in consecutive frames to obtain the position at each moment within 5 seconds. and speed , if the average speed is 12 m / s and the direction is stable, it is judged to be normal flight and has not changed significantly; If more than 10 seagulls are detected in succession in the same area, flying in a similar direction and at a similar altitude, they will be marked as "seagull group activity" to remind subsequent risk assessments to pay attention to the possibility of group passing near the runway safety zone; Use a deep learning classification network to perform secondary recognition on high-resolution snapshot images to confirm the bird's body shape and feather color, and match them with infrared thermal distribution. If the features match those of a raptor, update the label to Hawk. If a large number of infrared thermal targets are detected moving during the night time period (e.g. 23:00~05:00), the "nighttime activity frequency" tag will be added; Exemplarily, the generated data record is: in, Indicates that the category is seagull, Indicates the frequency of nighttime activities; The above information indicates that the target with ID = 0027 is a single seagull, whose position has changed sequentially over the past 2 seconds, and there is no flock flying behavior.

[0029] S3. Combine the airport flight take-off and landing time sequence data with the target feature information, and calculate the threat level of bird activities to flights through a risk assessment model based on fuzzy logic and Bayesian network. Among them, what needs to be explained in this step is: Set the number of birds n and population density , the distance d from the center of the runway and the flight altitude h are the key indicators, and a corresponding fuzzy set is established for each indicator; For example, taking the number of birds n as an example, it is set as: : The number of birds is very small; : The number of birds is moderate; : The bird population is huge; Using membership function Fuzzy processing of each indicator: Among them, a and b are interval thresholds; It should be noted that the four key indicators Calculate the fuzzy membership values ​​separately , and transform each indicator into a corresponding fuzzy language variable to form fuzzy bird activity information ; For example, if And preset The interval threshold is ,but , meaning that the bird population is large; make Indicated by The bird activity event conditions are obtained by synthesizing the fuzzy reasoning results; make Represents the airport flight take-off and landing time series data, including the take-off and landing frequency and time of flights in the current or future time; make Indicates the bird strike threat event level, taking ; In the embodiment of the present invention, the posterior probability of a specific threat level occurring under given bird activity conditions and flight take-off and landing timing information is calculated through a risk assessment model, and the following mathematical expression formula is given as an example: in, represents the prior probability of observing a specific bird activity under given threat level and flight schedule, represents the prior probability of a certain threat level under a given flight takeoff and landing frequency, is a normalized term, corresponding to the overall probability of a certain bird activity occurring under given flight conditions; The evaluation values ​​of each indicator output by the comprehensive fuzzy logic and the posterior probability of the Bayesian network are combined to determine the threat level of bird activities to flights according to the weighted rule; that is, on the basis of fuzzy membership processing and the posterior probability of the Bayesian network, the results are further weighted and combined to obtain the final comprehensive score S: in, For The weighted average value obtained by comprehensive calculation of fuzzy indicators, , is a weight factor used to find a balance between fuzzy logic scoring and probability inference results; When S is high, it means that bird activities overlap significantly with flight operations, and the risk increases accordingly; Combined with the final score S, it is mapped to the interval [0,1]. Further, the threat levels include: When the comprehensive score is in the range of [0, 0.3), it is judged as low threat level L1, which means that the number of birds is very small and far away from the critical areas of the runway, and basically does not pose a safety impact on aircraft takeoff and landing; When the comprehensive score is in [0.3, 0.5), it is judged as medium threat level L2, then the number of birds is moderate or occasionally enters the runway safety buffer zone, and monitoring and regular warning are required; When the comprehensive score is between [0.5, 0.7), it is judged as a high threat level L3, which means that the bird activities are frequent and have obviously approached the flight path, posing a potential risk to flights that are taking off or about to take off or land. When the comprehensive score is ≥0.7, it is judged as an extremely high threat level L4, then large flocks of birds repeatedly appear on the runway and its surroundings, posing a serious threat to flight safety.

[0030] S4. Trigger a graded warning signal according to the threat level, and link the bird-repelling device to perform a directional bird-repelling operation. The following points need to be explained in this step: When the threat level is L1, only text / voice reminders are sent to tower personnel to remind them to stay focused; When the threat level is L2, an audio signal or light flashing warning is triggered to remind ground and flight crews to pay close attention to the bird situation at the airport; When the threat level is L3, a high-frequency harsh sound wave or strong light warning is triggered, and real-time risk warnings are issued to flights about to take off and land; When the threat level is L4, the highest level emergency alarm is triggered, and the control department is advised to delay flight takeoff and landing, and immediately activate a linkage mechanism of multiple bird-repelling measures.

[0031] In an optional embodiment, the bird-repelling device linkage includes: When the threat level is ≥ L2, the sonic repelling device will be automatically activated to send out adjustable frequency high-decibel sonic interference; When the threat level is ≥ L3, in addition to sound wave interference, laser or strong beam bird repellent devices are turned on, and bionic sounds such as the calls of birds of prey can be played to repel them; When the threat level is L4, multiple devices are activated simultaneously to implement a comprehensive bird repellent solution including laser, strong light beam, bionic natural enemy, and aerosol.

[0032] In an optional embodiment, the directional bird-repelling operation includes: Automatically adjust the scanning direction and angle of the sound wave or light beam based on the real-time detected bird target location information; When large-scale flocks of birds appear, multiple bird-repelling devices are coordinated to operate synchronously according to multiple target locations to maximize the dispersal effect; Dynamically monitor the effectiveness of the dispersal operation. If the birds still pose an obvious threat, maintain high-intensity joint operations. If the birds have left the critical airspace, automatically downgrade to low intensity and stop dispersing to avoid unnecessary resource consumption.

[0033] In an optional implementation, the linkage with flight takeoff and landing includes: When the threat level is ≥ L3 and the bird-repelling effect is poor, a high-risk alert will be automatically sent to the aviation control department, suggesting delaying or changing the take-off and landing time; After the bird-scaring is completed and the threat level is reduced to L1, the warning status will be automatically lifted and the tower will be notified to resume normal flight scheduling.

[0034] The aforementioned preprocessing method and feature extraction method for collecting environmental data can be performed using methods and means in the prior art and will not be described in detail in this example.

[0035] Furthermore, the present embodiment discloses an intelligent bird-repelling device based on the aforementioned automatic detection method for bird conditions at an airport, including a monitoring module, wherein the monitoring module includes a radar device, a visible light camera, an infrared thermal imager and a sonar detector, and the monitoring modules are all connected to a control host. Specifically, the radar device may be a multi-beam scanning radar and a pulse Doppler radar, which are deployed at commanding heights around the airport, covering a range of 0.5-10 kilometers, and detecting the distance, speed and direction of birds. The visible light camera may be a high-resolution dynamic camera module, which is installed on lamp poles on both sides of the runway, covering a range of 0-2 kilometers, and collecting bird morphological characteristics and activity videos. The infrared thermal imager can be a medium / long wave infrared imaging module, deployed on a low-light area bracket, covering 0-1.5 kilometers, and detecting bird heat signals at night; the sonar detector may include horizontal and vertical sonar arrays, distributed on the ground outside the runway, covering 0-1 kilometers, and locating the position of birds through echoes; it also includes meteorological sensors installed around the runway to monitor meteorological data such as wind direction, wind speed, and visibility in real time.

[0036] It also includes a central processing unit, including a control host and a data storage server. The control host is deployed in the airport control room and integrates a data fusion module, a deep learning target detection model (such as BiLSTM+attention mechanism), and a risk assessment model (fuzzy logic and Bayesian network). It is responsible for processing sensor data, generating threat levels and outputting control instructions; the data storage server is used to store fused data sets, flight takeoff and landing timing data and historical threat records.

[0037] It also includes early warning devices, A tower display screen used to display bird locations, threat levels, and bird-repelling status in real time, a voice broadcast module that sends voice alerts (such as L1-L4 level prompts) to tower personnel, and warning lights distributed on both sides of the runway that trigger flashing strong lights (L2 and above).

[0038] It also includes a bird-repellent execution module, including a frequency-adjustable sound wave transmitter installed on the periphery of the runway to emit high-decibel sound waves in a directionally controlled manner, a strong light deployed in key areas of the runway to emit laser beams or strong light interference, a bionic sound player that plays bird-repellent audio such as the calls of birds of prey, and an aerosol generator located around the runway to release irritating aerosols.

[0039] Furthermore, the placement of each module is as follows: radar equipment is set on the tower or high point around the airport to achieve 360° airspace coverage; visible light / infrared cameras are set on lamp poles on both sides of the runway, staggered at intervals of 200 meters; sonar detectors are set on the ground outside the runway, in a circular array layout; meteorological sensors are set in the middle and both ends of the runway to monitor local meteorological changes; sonic transmitters are set at the four corners of the runway to cover the near-field area (0-1 km); strong light lamps are set on both sides of the runway centerline to directionally scan the take-off and landing routes; aerosol generators are set outside the runway safety buffer zone to avoid affecting the flight line of sight; the central processing unit and data server are located in the airport control room, directly connected to the tower operating console.

[0040] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. An automatic detection and intelligent bird-repelling method for airport birds, characterized in that: include: Collect airport airspace environment data in real time through a multi-source sensor array, and pre-process the environment data to generate a fused data set; Building a target detection model based on a deep learning algorithm to identify bird targets in the fused data set and extract target feature information, wherein the target feature information includes the location, movement trajectory, number and type of birds; Combining the airport flight take-off and landing time series data with the target feature information, the threat level of bird activities to flights is calculated through a risk assessment model based on fuzzy logic and Bayesian network; A graded warning signal is triggered according to the threat level, and the bird-repelling equipment is linked to perform a directional bird-repelling operation.

2. The method for automatic detection of bird conditions and intelligent bird repelling at an airport according to claim 1, characterized in that: The multi-source sensor array includes radar equipment, visible light cameras, infrared thermal imagers and sonar detectors. The airport airspace environment data collected in real time by the multi-source sensor array includes radar data, visible light image data, infrared thermal imaging data, sonar data and meteorological information around the airport runway; Among them, the radar data includes distance, radial velocity and azimuth, the visible light image data includes frame sequence images, target pixel coordinates and color texture features, the infrared thermal imaging data includes target thermal radiation intensity and thermal imaging coordinates, the sonar data includes echo intensity, delay time and target azimuth, and the meteorological information around the airport runway includes wind direction, wind speed and visibility.

3. The method for automatic detection of bird conditions and intelligent bird repelling at an airport according to claim 1 or 2, characterized in that: Preprocessing the environmental data to generate a fused data set includes: Performing unified time calibration on the environmental data; Perform bandpass filtering and background noise suppression on radar, infrared and sonar data, and perform background modeling and moving target segmentation on visible light images; According to the coordinate systems of different sensors, the target detection results are converted into a unified spatial coordinate system, and the Kalman filter algorithm is used to realize the association and redundancy removal of multi-source tracking results; Conduct secondary screening on targets with low confidence and eliminate duplicate targets; Annotate the bounding boxes of birds in visible light and infrared thermal images; Visible light and infrared data are enhanced by image flipping, rotation, brightness adjustment and noise addition methods; The processed data and the annotation information are matched according to the timestamp to form a unified serial data set, which is the fused data set.

4. The method for automatic detection of bird conditions and intelligent bird repelling at an airport according to claim 3, characterized in that: A target detection model is constructed based on a deep learning algorithm to identify bird targets in the fused data set and extract target feature information, including: A bidirectional long short-term memory network is used to perform multi-level fusion of radar, infrared and visible light features; For the small-scale and easily overlapping features of bird targets, an attention mechanism is set to enhance the key area features; Use the labeled fusion data set for supervised training, and the training loss function adopts a multi-task weighted form: ； in, is the classification loss, is the bounding box regression loss, is the multimodal consistency loss; , , , are the weighted coefficients of classification loss, bounding box regression loss, and multimodal consistency loss respectively; Perform multimodal matching on the pre-processed video frames, infrared thermal imaging frames and radar / sonar data, and the network outputs the predicted position, type and confidence of the bird target; If the confidence level is higher than the threshold , it is determined to be a valid bird target, and its flight trajectory, number and appearance characteristics are further extracted; The identified target information is indexed by the target ID to generate a target feature information set including the bird's location, movement trajectory, number and type.

5. The method for automatic detection of bird conditions and intelligent bird repelling at an airport according to claim 4, characterized in that: The identification and extraction of the target detection model includes: Track the trajectory of the same target detected by continuous video frames or radar, and use the Kalman filter algorithm to obtain the precise motion trajectory; The position of each moment in the trajectory and speed Discretized storage is performed to form a trajectory sequence; statistical analysis is performed on the flight speed, altitude and direction to determine whether birds have a behavior pattern of quickly flying over the runway or forming a flock flight; Combining deep learning classification networks with biological prior knowledge, the body shape, feather characteristics and thermal distribution information of birds in visible light and infrared images are classified and identified; According to the identification results, bird species are divided into different categories and labeled with clustering behavior labels or nocturnal activity frequency labels.

6. The method for automatic detection of bird conditions and intelligent bird repelling at an airport according to claim 4, characterized in that: Combining the airport flight take-off and landing time sequence data with the target feature information, the threat level of bird activities to flights is calculated through a risk assessment model based on fuzzy logic and Bayesian network, including: Set the number of birds n and population density , the distance d from the center of the runway and the flight altitude h are the key indicators, and a corresponding fuzzy set is established for each indicator; Using membership function Fuzzy processing of each indicator: ； Among them, a and b are interval thresholds; Input the fuzzified indicators into the Bayesian network nodes to form event conditions; Combined with the aircraft take-off and landing frequency and take-off and landing time information in the flight take-off and landing time series data, the posterior probability of bird strike risk is calculated; The evaluation values ​​of each indicator output by fuzzy logic and the posterior probability of Bayesian network are combined to determine the threat level of bird activities to flights according to the weighted rule; If the comprehensive score exceeds the preset upper threshold, it is judged as high risk or extremely high risk; If the comprehensive score is in the middle range, it is judged as medium risk; If the comprehensive score is below the lower threshold, it is judged as low risk.

7. The method for automatic detection of bird conditions and intelligent bird repelling at an airport according to claim 6, characterized in that: The threat levels include: When the comprehensive score is in the range of [0, 0.3), it is judged as low threat level L1, which means that the number of birds is very small and far away from the critical areas of the runway, and basically does not pose a safety impact on aircraft takeoff and landing; When the comprehensive score is in [0.3, 0.5), it is judged as medium threat level L2, then the number of birds is moderate or occasionally enters the runway safety buffer zone, and monitoring and regular warning are required; When the comprehensive score is between [0.5, 0.7), it is judged as a high threat level L3, which means that the bird activities are frequent and have obviously approached the flight path, posing a potential risk to flights that are taking off or about to take off or land. When the comprehensive score is ≥0.7, it is judged as an extremely high threat level L4, then large flocks of birds repeatedly appear on the runway and its surroundings, posing a serious threat to flight safety.

8. The method for automatic detection of bird conditions and intelligent bird repelling at an airport according to claim 6, characterized in that: According to the threat level, a graded warning signal is triggered, and the bird-repelling device is linked to perform a directional bird-repelling operation, including: When the threat level is L1, only text / voice reminders are sent to tower personnel to remind them to stay focused; When the threat level is L2, an audio signal or light flashing warning is triggered to remind ground and flight crews to pay close attention to the bird situation at the airport; When the threat level is L3, a high-frequency harsh sound wave or strong light warning is triggered, and real-time risk warnings are issued to flights about to take off and land; When the threat level is L4, the highest level emergency alarm is triggered, and the control department is advised to delay flight takeoff and landing, and immediately activate a linkage mechanism of multiple bird-repelling measures.

9. The method for automatic detection of bird conditions and intelligent bird repelling at an airport according to claim 8, characterized in that: The linked bird-repelling device performs a directional bird-repelling operation, including: When the threat level is ≥ L2, the sonic repelling device will be automatically activated to send out adjustable frequency high-decibel sonic interference; When the threat level is ≥ L3, in addition to sound wave interference, turn on the laser or strong beam bird repellent device and play bionic sound to repel the birds; When the threat level is L4, multiple devices are activated simultaneously to implement a comprehensive bird repellent solution including laser, strong light beam, bionic natural enemy and aerosol.

10. An intelligent bird-repelling device based on the automatic detection and intelligent bird-repelling method for airport birds as described in any one of claims 1 to 9, characterized in that: It includes a monitoring module arranged on the runway, the monitoring module includes a radar device, a visible light camera, an infrared thermal imager and a sonar detector, and the monitoring modules are all connected to the control host; It also includes an early warning module, which includes a tower display screen, a voice player, and warning lights arranged on both sides of the runway; It also includes a bird-repelling execution module, which includes a frequency-adjustable sound wave transmitter, a strong light, a bionic sound player and an aerosol generator.

Images