An airport abnormal situation-based real-time situation awareness early warning system and method

Abstract

This invention discloses a real-time situational awareness and early warning system and method based on airport anomalies, relating to the field of risk warning technology. It identifies the wingtip positions of birds, continuously records the distance between wingtip positions in adjacent frames as a relative displacement sequence, performs spectral analysis on the relative displacement sequence to generate a spectrum diagram containing the correspondence between frequency and amplitude, establishes a mapping relationship database between wind field parameters and wingtip displacement spectral characteristics based on historical wind field conditions during bird intrusions into the airport, including wind speed and wind direction angle, divides the data in the mapping relationship data, generates filtering ranges for the spectrum under different wind field conditions, periodically samples marked risk events, calculates the current risk value at each sampling time, continuously monitors the changing trend of the risk value within several sampling periods, and triggers an alarm signal when the risk value shows a monotonically increasing trend.

Description

Technical Field

[0001] This invention relates to the field of risk warning technology, specifically a real-time situational awareness and early warning system and method based on abnormal airport situations. Background Technology

[0002] As a critical infrastructure for air transport, the operational efficiency and safety level of airports have become core indicators. Within the airport airspace safety management system, the intrusion of birds and other moving objects remains a significant threat to aircraft takeoff and landing safety. Due to the high degree of uncertainty and suddenness of bird activity, constructing an air defense system capable of early warning and rapid response is of paramount practical importance for ensuring flight safety.

[0003] Existing technologies mainly rely on radar detection and photoelectric tracking systems. Their core working principle is to capture bird targets entering the monitoring area through electromagnetic waves or optical sensors, and continuously track their spatial position to depict their macroscopic flight trajectory. Furthermore, relevant mathematical models are used to predict the future flight path of the birds.

[0004] However, with the increasing demand for refined management of airport operational safety, especially in response to near-field and sudden bird strike risks, birds exhibit complex behavioral patterns under stress, making their behavior difficult to predict accurately. Therefore, the inherent response delays in dealing with sudden and stressful risk events, and the achievement of advanced perception and highly sensitive early warning of sudden intrusion threats, have become a technical challenge for those skilled in the art. Summary of the Invention

[0005] The purpose of this invention is to provide a real-time situational awareness and early warning system and method based on abnormal airport situations, so as to solve the problems raised in the prior art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a real-time situational awareness and early warning method based on abnormal airport situations;

[0007] Step S1: Identify the wingtip position of birds by continuously collecting bird flight video streams using camera equipment deployed around the airport runway, collecting the pixel coordinates of the bird's wingtip, continuously recording the distance between the wingtip positions of adjacent frames as a relative displacement sequence, performing spectral analysis on the relative displacement sequence, and generating a spectrum diagram containing the correspondence between frequency and amplitude.

[0008] Step S2: Based on the wind field conditions and corresponding wingtip displacement spectrum when birds invade the airport in historical records, the wind field conditions include wind speed and wind direction angle, and establish a mapping relationship database between wind field parameters and wingtip displacement spectrum characteristics.

[0009] Step S3: Divide the data in the mapping relationship data, establish the correspondence between the frequency distribution intervals of different wind field conditions and wingtip displacement, and generate the filtering range of the spectrum under different wind field conditions;

[0010] Step S4: Obtain the maximum value in the spectrum graph, compare the maximum value with the risk threshold, and mark the presence of a risk event at the current time when the maximum value exceeds the threshold;

[0011] Step S5: Periodically sample the marked risk events and calculate the current risk value at each sampling time. The risk value is defined as the ratio of the current maximum value of the spectrum to the risk threshold. Continuously monitor the changing trend of the risk value within several sampling periods. When the risk value shows a monotonically increasing trend, trigger an alarm signal.

[0012] Furthermore, step S1 includes:

[0013] Step S1-1: Acquire a bird image, sample the edge of the bird's left or right wing, identify the coordinates of the wingtip in the image, and set the coordinates as wingtip coordinates;

[0014] Step S1-2: From several consecutive frames of bird images, obtain the wingtip coordinates of each bird image, obtain the Euclidean distance between the wingtip coordinates of two adjacent frames, and collect all the Euclidean distances to form a relative displacement sequence.

[0015] Step S1-3: Perform a fast Fourier transform on each Euclidean distance in the relative displacement sequence and then collect the amplitude spectrum to form a spectrum diagram.

[0016] First, a target detection model is run to process each frame of the video stream to identify and locate bird targets.

[0017] After determining the bounding box of the bird target, a wingtip keypoint localization algorithm is then executed within the image area defined by the bounding box. The wingtip keypoint localization algorithm first uses an edge recognition algorithm to extract the edge lines of the bird's wing, and then performs a generalized Huffman transform algorithm to locate the position of the wingtip within the edge lines.

[0018] To ensure the consistency and stability of data processing, this invention focuses on tracking only one wingtip. It continuously processes N+1 frames of images from the video stream to obtain a time series containing N+1 wingtip coordinate points. Based on this coordinate series, the Euclidean distance between the wingtip coordinates of adjacent frames is calculated.

[0019] Finally, the Fast Fourier Transform (FFT) algorithm is applied to the relative displacement sequence D to transform it from the time domain to the frequency domain in order to analyze the frequency components of the wingtip flapping.

[0020] Construct a relative displacement sequence of length N Performing a Fast Fourier Transform (FFT) on the sequence yields its frequency domain representation:

[0021] , where dn represents the Euclidean distance between the wingtip positions in the nth frame and the (n-1)th frame;

[0022] The FFT calculation result is a complex sequence. This invention extracts its amplitude spectrum to generate a spectrum S(f) describing the relationship between the wingtip flapping frequency and the corresponding vibration amplitude. The horizontal axis of this spectrum is frequency f, and the vertical axis is amplitude A(f). This completes the extraction of frequency domain features from the original video.

[0023] Furthermore, step S2 includes:

[0024] Step S2-1: Obtain historical data, which includes historical wind field conditions when bird intrusion events occur at the airport. Simultaneously collect spectral data of bird wingtip displacement when bird intrusion events occur. Historical wind field conditions include historical wind speed and historical wind direction angle.

[0025] Step S2-2: Extract spectral features from historical wingtip displacement spectral data. The spectral features include: dominant frequency, peak amplitude of dominant frequency, energy centroid of the spectrum, and bandwidth of the spectrum.

[0026] Step S2-3: Establish the correlation between historical wind field conditions and corresponding spectral characteristics, and store the historical wind field conditions, spectral characteristics and correlations together to form a mapping database.

[0027] Step 2 aims to establish a historical knowledge base, the core of which is the correspondence between the environmental wind field conditions at the time of confirmed bird invasion risk events in history, including wind speed and direction, and key characteristic parameters of the bird wingtip displacement spectrum collected simultaneously, as the basis for the subsequent establishment of a dynamic risk assessment model.

[0028] First, all confirmed high-risk bird intrusion events are identified from the airport's historical flight data records, bird alert records, and manual observation logs. For each historical event, the system retrieves wind field data recorded by the airport's meteorological station within a specific time window before and after the event, for example, from 30 seconds before the event to 10 seconds after the event. Simultaneously, it retrieves video surveillance footage synchronized with the event and calculates the bird wingtip displacement spectrum data during the event using the method in step S1.

[0029] Next, a set of characteristic parameters that can quantitatively describe the spectral morphology are extracted from the wingtip displacement spectral data of each historical event.

[0030] The spectral characteristic parameters specifically include four dimensions: dominant frequency, peak amplitude of dominant frequency, spectral energy centroid, and spectral bandwidth.

[0031] The dominant frequency reflects the bird's primary wing-beating frequency and is the frequency point with the largest amplitude in the spectrum.

[0032] The peak amplitude of the dominant frequency corresponds to the dominant frequency and reflects the intensity of the flapping motion;

[0033] The center of gravity of spectral energy, also known as the spectral centroid, reflects the weighted average position of spectral energy on the frequency axis.

[0034] Spectral bandwidth represents the range of all frequency points where spectral energy exceeds half of its peak value, reflecting the degree of concentration of spectral energy.

[0035] Then, the extracted historical wind field conditions are associated and stored with their corresponding spectral feature sets, which can be stored in the database in the form of key-value pairs.

[0036] Furthermore, step S3 includes:

[0037] Step S3-1: Divide the airport into several sectors, classify the historical records according to the wind field conditions in the mapping relationship database, divide the wind speed into several intervals, divide the wind direction into several sectors, and form K wind field condition groups.

[0038] Step S3-2: For any wind field condition group, obtain the spectral characteristics of all bird invasion events in the wind field condition group. Based on the spectral characteristics of bird invasion events, statistically calculate the frequency distribution range of wingtip displacements corresponding to bird behavior.

[0039] Step S3-3: Based on the wind field conditions and the corresponding frequency distribution range, use regression analysis to establish a regression model that includes the mapping relationship between wind field conditions and the filtering range. Regression analysis methods include: multinomial regression, support vector regression, or neural network model.

[0040] Step S3 utilizes the database constructed in Step S2 to establish a regression model capable of dynamically predicting the wingtip flapping frequency range most likely to occur under current conditions based on real-time wind field input. This achieves the transformation from static historical data to dynamic prediction capabilities.

[0041] First, the data in the mapping relation database is preprocessed and partitioned. The wind direction angle θ is decomposed into two components: cos(θ) and sin(θ), and the input features of the wind field are established as a three-dimensional vector [V, cos(θ), sin(θ)].

[0042] Next, cluster analysis was performed on the wind field conditions in the database to identify typical wind field patterns with similar impacts. The K-means clustering algorithm was used to divide all historical wind field data points into K wind field condition groups. For each wind field condition group k (k=1,2,...,K), the distribution of the dominant frequencies of all risk events within that group was statistically analyzed. By constructing a histogram of the dominant frequencies and performing kernel density estimation, the 95% confidence interval of the dominant frequency of risk under that wind field condition was determined, denoted as the frequency distribution interval of the wingtip displacement.

[0043] Subsequently, a multivariate regression model was trained based on the processed data.

[0044] For example, a feedforward neural network model can be used as a regression model. The specific structure of this neural network includes the following:

[0045] Input layer: Contains 2 neurons, which receive wind speed V and wind direction angle θ respectively;

[0046] Hidden layer 1: Contains 16 neurons, activated by the ReLU function;

[0047] Hidden layer 2: Contains 8 neurons, activated by the ReLU function;

[0048] Output layer: Contains 2 neurons, using linear activation, to output the lower and upper boundaries of the predicted risk frequency interval, respectively; all wind field conditions in the database are used as training input, the corresponding risk frequency intervals are used as training labels, mean squared error is used as the loss function, and backpropagation training is performed through an optimizer until the model converges;

[0049] For example, by fitting the data using the least squares method and employing linear regression, a polynomial regression model can be constructed to determine the boundaries of the distribution interval and the polynomial coefficients, wind speed, and wind direction angle.

[0050] After training, the regression model mapping the wind field conditions to the filtering range has the ability to calculate the filtering range of the current high-risk wingtip frequency in real time based on any input real-time wind field conditions.

[0051] Furthermore, step S4 includes:

[0052] Step S4-1: At the current moment, obtain the real-time wind field conditions, which include real-time wind speed and real-time wind conditions. Based on the real-time wind field conditions, calculate the current filtering range under the current wind field conditions in the regression model obtained in step S3-3, denoted as [FLt, FUt], where FLt represents the lower boundary of the current filtering range and FUt represents the upper boundary of the current filtering range.

[0053] Step S4-2: Extract the spectral information within the current filtering range from the wingtip displacement spectrum at the current moment, and collect the amplitude values ​​corresponding to N frequencies respectively, where the j-th frequency is denoted as fj, fj∈[FLt, FUt]; calculate the weighted average amplitude AW. , where w(fj) represents the probability of a bird intrusion event occurring when frequency fi appears in the historical records, and A(fj) represents the amplitude value corresponding to frequency fj in the current filtering range;

[0054] Step S4-3: Multiply the weighted average amplitude AW by a preset safety factor C. safe Obtain the dynamic threshold T dy (t), that is C safe It is a constant greater than 1;

[0055] Step S4-4: Obtain the maximum amplitude of the frequency in the spectrum information of the current filtering range, denoted as Amax(t), and multiply the maximum amplitude by the dynamic threshold T. dy If Amax(t) > Tdy(t), then a risk event is marked as existing at the current time, and the occurrence time of this risk event, the corresponding wind field conditions, and the spectral characteristics are recorded in the risk event log.

[0056] Step S4 is the core execution step of the early warning method. By acquiring the wind field conditions of the current environment in real time, the risk frequency filtering range is dynamically calculated using the regression model trained in step S3. Then, the real-time collected bird wingtip spectrum is weighted and analyzed within this range to generate a dynamic risk threshold. The maximum amplitude in the spectrum is compared with this threshold to determine whether there is a potential risk event at the current moment.

[0057] Furthermore, step S5 includes:

[0058] Step S5-1: Periodically sample the times marked as risk events, with a sampling period of Δt, to obtain a series of risk event sampling times;

[0059] Step S5-2: At the i-th sampling time ti, calculate the risk value R(ti) at the i-th sampling time, R(ti) = Amax(ti) / TsD(ti), where TsD(ti) represents the dynamic threshold at time ti, and Amax(ti) represents the maximum amplitude of the spectrum at time ti;

[0060] Step S5-3: Calculate the risk value at each sampling time. When the risk value shows a monotonically increasing trend over M consecutive sampling periods, trigger an alarm signal, where M≥2.

[0061] Step S5 continuously monitors the potential risk events marked in Step S4. By calculating risk values ​​over a series of consecutive time points and analyzing their changing trends, the highest level alarm signal is triggered only when the risk shows a clear and continuous increasing trend. This trend-based judgment mechanism can effectively filter out instantaneous and accidental spectral jumps, further improving the accuracy and reliability of early warnings.

[0062] To better implement the above methods, a real-time situational awareness and early warning system for bird activity at airports is also proposed. The system includes: a spectrum management module, a mapping database management module, a filtering management module, a risk event management module, and an alarm module.

[0063] The spectrum management module is used to identify the position of bird wingtips, collect the pixel coordinates of bird wingtips, continuously record the Euclidean distance between wingtips between adjacent frames as a relative displacement sequence, apply fast Fourier transform to the relative displacement sequence for spectrum analysis, and generate a spectrum diagram containing the correspondence between frequency and amplitude.

[0064] The mapping database management module is used to establish a mapping relationship database between wind field parameters and wingtip displacement spectrum characteristics based on the wind field conditions and corresponding wingtip displacement spectrum when birds invaded the airport in historical records. The wind field conditions include wind speed and wind direction angle.

[0065] The filtering management module is used to divide the data in the mapping relationship data, establish the correspondence between the frequency distribution intervals of different wind field conditions and wingtip displacement, and generate the filtering range of the spectrum under different wind field conditions.

[0066] The risk event management module is used to obtain the maximum value in the spectrum graph, compare the maximum value with the risk threshold, and mark the existence of a risk event at the current time when the maximum value exceeds the threshold.

[0067] The alarm module is used to periodically sample marked risk events, calculate the current risk value at each sampling time, and trigger an alarm signal when the risk value shows a monotonically increasing trend.

[0068] Furthermore, the spectrum management module includes: an image acquisition unit, a displacement difference management unit, and a spectrum conversion unit. The image acquisition unit is used to acquire bird images, sample the edges of the bird's left or right wing, and identify the coordinates of the wingtip in the image. The displacement difference management unit is used to acquire the wingtip coordinates of each bird image, acquire the Euclidean distance between the wingtip coordinates in two adjacent frames, and collect all Euclidean distances to form a relative displacement sequence. The spectrum conversion unit is used to perform a fast Fourier transform on each Euclidean distance in the relative displacement sequence and then acquire the amplitude spectrum to form a spectrum diagram.

[0069] Furthermore, the mapping database management module includes: a historical data management unit, a spectrum feature management unit, and a mapping relationship management unit. The historical data management unit is used to acquire historical data and synchronously collect bird wingtip displacement spectrum data when bird invasion events occur. The spectrum feature management unit is used to extract spectrum features from historical wingtip displacement spectrum data. The mapping relationship management unit is used to establish the association between historical wind field conditions and corresponding spectrum features, and to associate and store historical wind field conditions, spectrum features, and association relationships to form a mapping relationship database.

[0070] Furthermore, the filtering management module includes: a wind field condition management unit, a frequency distribution management unit, and a condition mapping unit. The wind field condition management unit is used to classify historical records according to the wind field conditions in the mapping relationship database to form several wind field condition groups. The frequency distribution management unit is used to obtain the spectral characteristics of all bird invasion events in any wind field condition group, and to statistically calculate the frequency distribution interval of the wingtip displacement corresponding to the bird behavior based on the spectral characteristics of the bird invasion events. The condition mapping unit is used to establish a regression model that includes the mapping relationship between wind field conditions and filtering range based on the wind field conditions and the corresponding frequency distribution interval using regression analysis methods.

[0071] Furthermore, the risk event management module includes: a frequency filtering unit, a risk value calculation unit, and a risk judgment unit. The frequency filtering unit is used to acquire real-time wind field conditions at the current moment and calculate the current filtering range under the current wind field conditions based on the regression model. The risk value calculation unit is used to extract the spectrum information within the current filtering range from the wingtip displacement spectrum at the current moment and calculate the dynamic threshold. The risk judgment unit is used to compare the maximum amplitude with the dynamic threshold to determine whether a risk event exists.

[0072] Furthermore, the alarm module includes a periodic sampling unit and an information prompting unit. The periodic sampling unit is used to periodically sample the times marked as risk events and calculate the risk value at each sampling time. The information prompting unit is used to trigger an alarm signal based on the judgment conditions.

[0073] Compared with the prior art, the beneficial effects of the present invention are:

[0074] 1. A complete perception and early warning logic chain, from micro-behavioral characteristics to macro-security situation, has been constructed. This is achieved through in-depth analysis of the spectral characteristics of bird wingtip flapping and dynamic modeling combined with environmental factors, transforming the early warning mechanism from a passive response to lagging macro-trajectories.

[0075] 2. In response to the fact that birds are more likely to encounter emergencies due to external influences such as airport lights or noise under near-field conditions, the dynamic response mechanism provides early warning of their internal stress state through advanced insight. This enables the system to provide earlier and more reliable early warning information when dealing with sudden and high-risk bird strikes, thus providing strong technical support for the airport safety operation guarantee system. Attached Figure Description

[0076] Figure 1 This is a schematic diagram of the structure of a real-time situational awareness and early warning system based on abnormal airport situations according to the present invention.

[0077] Figure 2 This is a flowchart illustrating a real-time situational awareness and early warning method for abnormal airport situations according to the present invention. Detailed Implementation

[0078] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0079] Example: Figures 1-2 As shown, the present invention provides a technical solution: a real-time situational awareness and early warning method based on abnormal airport situations.

[0080] Step S1: Identify the wingtip position of birds by continuously collecting bird flight video streams using camera equipment deployed around the airport runway, collecting the pixel coordinates of the bird's wingtip, continuously recording the distance between the wingtip positions of adjacent frames as a relative displacement sequence, performing spectral analysis on the relative displacement sequence, and generating a spectrum diagram containing the correspondence between frequency and amplitude.

[0081] Step S1 includes:

[0082] Step S1-1: Acquire a bird image, sample the edge of the bird's left or right wing, identify the coordinates of the wingtip in the image, and set the coordinates as wingtip coordinates;

[0083] Step S1-2: From several consecutive frames of bird images, obtain the wingtip coordinates of each bird image, obtain the Euclidean distance between the wingtip coordinates of two adjacent frames, and collect all the Euclidean distances to form a relative displacement sequence.

[0084] Step S1-3: Perform a fast Fourier transform on each Euclidean distance in the relative displacement sequence and then collect the amplitude spectrum to form a spectrum diagram.

[0085] Step S2: Based on the wind field conditions and corresponding wingtip displacement spectrum when birds invade the airport in historical records, the wind field conditions include wind speed and wind direction angle, and establish a mapping relationship database between wind field parameters and wingtip displacement spectrum characteristics.

[0086] Step S2 includes:

[0087] Step S2-1: Obtain historical data, which includes historical wind field conditions when bird intrusion events occur at the airport. Simultaneously collect spectral data of bird wingtip displacement when bird intrusion events occur. Historical wind field conditions include historical wind speed and historical wind direction angle.

[0088] Step S2-2: Extract spectral features from historical wingtip displacement spectral data. The spectral features include: dominant frequency, peak amplitude of dominant frequency, energy centroid of the spectrum, and bandwidth of the spectrum.

[0089] Step S2-3: Establish the correlation between historical wind field conditions and corresponding spectral characteristics, and store the historical wind field conditions, spectral characteristics and correlations together to form a mapping database.

[0090] Step S3: Divide the data in the mapping relationship data, establish the correspondence between the frequency distribution intervals of different wind field conditions and wingtip displacement, and generate the filtering range of the spectrum under different wind field conditions;

[0091] Step S3 includes:

[0092] Step S3-1: Divide the airport into several sectors, classify the historical records according to the wind field conditions in the mapping relationship database, divide the wind speed into several intervals, divide the wind direction into several sectors, and form K wind field condition groups.

[0093] Step S3-2: For any wind field condition group, obtain the spectral characteristics of all bird invasion events in the wind field condition group. Based on the spectral characteristics of bird invasion events, statistically calculate the frequency distribution range of wingtip displacements corresponding to bird behavior.

[0094] Step S3-3: Based on the wind field conditions and the corresponding frequency distribution range, use regression analysis to establish a regression model that includes the mapping relationship between wind field conditions and the filtering range. Regression analysis methods include: multinomial regression, support vector regression, or neural network model.

[0095] Step S4: Obtain the maximum value in the spectrum graph, compare the maximum value with the risk threshold, and mark the presence of a risk event at the current time when the maximum value exceeds the threshold;

[0096] Step S4 includes:

[0097] Step S4-1: At the current moment, obtain the real-time wind field conditions, which include real-time wind speed and real-time wind conditions. Based on the real-time wind field conditions, calculate the current filtering range under the current wind field conditions in the regression model obtained in step S3-3, denoted as [FLt, FUt], where FLt represents the lower boundary of the current filtering range and FUt represents the upper boundary of the current filtering range.

[0098] Step S4-2: Extract the spectral information within the current filtering range from the wingtip displacement spectrum at the current moment, and collect the amplitude values ​​corresponding to N frequencies respectively, where the j-th frequency is denoted as fj, fj∈[FLt, FUt]; calculate the weighted average amplitude AW. , where w(fj) represents the probability of a bird intrusion event occurring when frequency fi appears in the historical records, and A(fj) represents the amplitude value corresponding to frequency fj in the current filtering range;

[0099] Step S4-3: Multiply the weighted average amplitude AW by a preset safety factor C. safe Obtain the dynamic threshold T dy (t), that is C safe It is a constant greater than 1;

[0100] Step S4-4: Obtain the maximum amplitude of the frequency in the spectrum information of the current filtering range, denoted as Amax(t), and multiply the maximum amplitude by the dynamic threshold T. dy If Amax(t) > Tdy(t), then a risk event is marked as existing at the current time, and the occurrence time of this risk event, the corresponding wind field conditions, and the spectral characteristics are recorded in the risk event log.

[0101] Step S5: Periodically sample the marked risk events, calculate the current risk value at each sampling time. The risk value is defined as the ratio of the current maximum value of the spectrum to the risk threshold. Continuously monitor the changing trend of the risk value within several sampling periods. When the risk value shows a monotonically increasing trend, trigger an alarm signal.

[0102] Step S5 includes:

[0103] Step S5-1: Periodically sample the times marked as risk events, with a sampling period of Δt, to obtain a series of risk event sampling times;

[0104] Step S5-2: At the i-th sampling time ti, calculate the risk value R(ti) at the i-th sampling time, R(ti) = Amax(ti) / TsD(ti), where TsD(ti) represents the dynamic threshold at time ti, and Amax(ti) represents the maximum amplitude of the spectrum at time ti;

[0105] Step S5-3: Calculate the risk value at each sampling time. When the risk value shows a monotonically increasing trend over M consecutive sampling periods, trigger an alarm signal, where M≥2.

[0106] To further illustrate the specific operation process of the method disclosed in this invention, an embodiment is provided below:

[0107] One day, a camera unit deployed at the end of the approach lighting system on an airport runway captured a bird flying at low altitude. The system then activated the early warning method of this invention to continuously monitor the bird.

[0108] Following the method in step S1, images were continuously acquired and processed to obtain the coordinates of the wingtip and calculate a sequence D consisting of relative displacement values. After performing an FFT transformation on sequence D, the wingtip displacement spectrum at the initial moment was obtained.

[0109] Proceed to step S4.

[0110] The real-time wind conditions obtained from the AWOS system were: wind speed Vt = 4.2 m / s, wind direction θt = 95°.

[0111] These two parameters are input into a pre-trained regression model that maps wind field conditions to the filtering range. The model outputs the predicted risk frequency filtering range as [FLt,FUt]=[6.8Hz,15.6Hz].

[0112] Next, the information of the spectrum S(f) in the interval [6.8Hz, 15.6Hz] was analyzed. Three relatively significant frequency peaks were found in this interval, and the correspondence between frequency and amplitude was established (7.5Hz, 8.7), (9Hz, 2.1), (10.5Hz, 1.3), (12Hz, 0.9).

[0113] The risk weights obtained from the historical database are as follows: w(7.5Hz) = 0.85, w(9Hz) = 0.3, w(10.5Hz) = 0.2, w(12Hz) = 0.1.

[0114] The system calculates the weighted average amplitude based on historical statistics to obtain the weights:

[0115] ,

[0116] Assuming a safety factor Csafe = 1.5, then:

[0117] Tdy = 1.5 × 5.78 = 8.67, and the current Amax = 8.7 > 8.67 is marked as a risk period.

[0118] Set the sampling period to 0.5s and calculate the risk value for 3 consecutive sampling periods.

[0119] R(t1)=1.003, R(t2)=1.071, R(t3)=1.143;

[0120] If the risk value sequence increases significantly, the system determines that the bird's risk status has a significant, continuous and highly certain increasing trend, and immediately triggers an alarm, sending a warning message to the control tower and the field management department.

[0121] The system includes: a spectrum management module, a mapping database management module, a filter management module, a risk event management module, and an alarm module;

[0122] The spectrum management module is used to identify the position of bird wingtips, collect the pixel coordinates of bird wingtips, continuously record the Euclidean distance between wingtips between adjacent frames as a relative displacement sequence, apply fast Fourier transform to the relative displacement sequence for spectrum analysis, and generate a spectrum diagram containing the correspondence between frequency and amplitude.

[0123] The spectrum management module includes an image acquisition unit, a displacement difference management unit, and a spectrum conversion unit. The image acquisition unit is used to acquire bird images, sample the edges of the bird's left or right wing, and identify the coordinates of the wingtip in the image. The displacement difference management unit is used to acquire the wingtip coordinates of each bird image, acquire the Euclidean distance between the wingtip coordinates of two adjacent frames, and collect all Euclidean distances to form a relative displacement sequence. The spectrum conversion unit is used to perform a fast Fourier transform on each Euclidean distance in the relative displacement sequence and then acquire the amplitude spectrum to form a spectrum diagram.

[0124] The mapping database management module is used to establish a mapping relationship database between wind field parameters and wingtip displacement spectrum characteristics based on the wind field conditions and corresponding wingtip displacement spectrum when birds invaded the airport in historical records. The wind field conditions include wind speed and wind direction angle.

[0125] The mapping database management module includes a historical data management unit, a spectrum feature management unit, and a mapping relationship management unit. The historical data management unit is used to acquire historical data and synchronously collect bird wingtip displacement spectrum data when bird invasion events occur. The spectrum feature management unit is used to extract spectrum features from historical wingtip displacement spectrum data. The mapping relationship management unit is used to establish the association between historical wind field conditions and corresponding spectrum features, and to store historical wind field conditions, spectrum features, and association relationships together to form a mapping relationship database.

[0126] The filtering management module is used to divide the data in the mapping relationship data, establish the correspondence between the frequency distribution intervals of different wind field conditions and wingtip displacement, and generate the filtering range of the spectrum under different wind field conditions.

[0127] The filtering management module includes a wind field condition management unit, a frequency distribution management unit, and a condition mapping unit. The wind field condition management unit is used to classify historical records according to the wind field conditions in the mapping relationship database, forming several wind field condition groups. The frequency distribution management unit is used to obtain the spectral characteristics of all bird invasion events in any wind field condition group, and to statistically calculate the frequency distribution interval of the wingtip displacement corresponding to the bird behavior based on the spectral characteristics of the bird invasion events. The condition mapping unit is used to establish a regression model that includes the mapping relationship between wind field conditions and filtering range based on wind field conditions and the corresponding frequency distribution interval using regression analysis methods.

[0128] The risk event management module is used to obtain the maximum value in the spectrum graph, compare the maximum value with the risk threshold, and mark the existence of a risk event at the current time when the maximum value exceeds the threshold.

[0129] The risk event management module includes a frequency filtering unit, a risk value calculation unit, and a risk judgment unit. The frequency filtering unit is used to obtain real-time wind field conditions at the current moment and calculate the current filtering range under the current wind field conditions based on the regression model. The risk value calculation unit is used to extract the spectrum information within the current filtering range from the wingtip displacement spectrum at the current moment and calculate the dynamic threshold. The risk judgment unit is used to compare the maximum amplitude with the dynamic threshold to determine whether a risk event exists.

[0130] The alarm module is used to periodically sample the marked risk events, calculate the current risk value at each sampling time, and trigger an alarm signal when the risk value shows a monotonically increasing trend.

[0131] The alarm module includes a periodic sampling unit and an information prompting unit. The periodic sampling unit is used to periodically sample the times marked as risk events and calculate the risk value at each sampling time. The information prompting unit is used to trigger an alarm signal based on the judgment conditions.

[0132] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. A real-time situational awareness and early warning method based on abnormal airport situations, characterized in that: The methods include: Step S1: Identify the wingtip position of the bird. Continuously collect bird flight video streams using camera equipment deployed around the airport runway. Collect the pixel coordinates of the bird's wingtip and continuously record the distance between the wingtip positions of adjacent frames as a relative displacement sequence. Perform spectral analysis on the relative displacement sequence to generate a spectrum diagram containing the correspondence between frequency and amplitude. Step S2: Based on the wind field conditions and corresponding wingtip displacement spectrum when birds invade the airport in historical records, the wind field conditions include wind speed and wind direction angle, and establish a mapping relationship database between wind field parameters and wingtip displacement spectrum characteristics. Step S3: Divide the data in the mapping relationship data, establish the correspondence between the frequency distribution intervals of different wind field conditions and wingtip displacement, and generate the filtering range of the spectrum under different wind field conditions; Step S4: Obtain the maximum value in the spectrum graph, compare the maximum value with the risk threshold, and mark the presence of a risk event at the current moment when the maximum value exceeds the threshold; Step S5: Periodically sample the marked risk events and calculate the current risk value at each sampling time. The risk value is defined as the ratio of the current maximum value of the spectrum to the risk threshold. Continuously monitor the changing trend of the risk value within several sampling periods. When the risk value shows a monotonically increasing trend, trigger an alarm signal.

2. The real-time situational awareness and early warning method based on abnormal airport situations according to claim 1, characterized in that: Step S1 includes: Step S1-1: Acquire a bird image, sample the edge of the bird's left or right wing, identify the coordinates of the wingtip in the image, and set the coordinates as wingtip coordinates; Step S1-2: From several consecutive frames of bird images, obtain the wingtip coordinates of each bird image, obtain the Euclidean distance between the wingtip coordinates of two adjacent frames, and collect all the Euclidean distances to form a relative displacement sequence. Step S1-3: Perform a fast Fourier transform on each Euclidean distance in the relative displacement sequence and then collect the amplitude spectrum to form a spectrum diagram.

3. The real-time situational awareness and early warning method based on abnormal airport situations according to claim 2, characterized in that: Step S2 includes: Step S2-1: Obtain historical data, which includes historical wind field conditions at the time of bird intrusion events at the airport. Simultaneously collect spectral data of bird wingtip displacement at the time of bird intrusion events. The historical wind field conditions include historical wind speed and historical wind direction angle. Step S2-2: Extract spectral features from the historical wingtip displacement spectral data. The spectral features include: dominant frequency, peak amplitude of dominant frequency, energy centroid of the spectrum, and bandwidth of the spectrum. Step S2-3: Establish the association between the historical wind field conditions and the corresponding spectral characteristics, and store the historical wind field conditions, spectral characteristics and association relationships together to form a mapping relationship database.

4. The real-time situational awareness and early warning method based on abnormal airport situations according to claim 3, characterized in that: Step S3 includes: Step S3-1: Divide the airport into several sectors, classify the historical records according to the wind field conditions in the mapping relationship database, divide the wind speed into several intervals, divide the wind direction into several sectors, and form K wind field condition groups. Step S3-2: For any wind field condition group, obtain the spectral characteristics of all bird invasion events in the wind field condition group. Based on the spectral characteristics of bird invasion events, statistically calculate the frequency distribution range of wingtip displacements corresponding to bird behavior. Step S3-3: Based on the wind field conditions and the corresponding frequency distribution range, a regression model including the mapping relationship between wind field conditions and the filtering range is established using regression analysis methods. The regression analysis methods include: multinomial regression, support vector regression, or neural network model.

5. The real-time situational awareness and early warning method based on abnormal airport situations according to claim 4, characterized in that: Step S4 includes: Step S4-1: At the current moment, obtain the real-time wind field conditions, which include real-time wind speed and real-time wind conditions. Based on the regression model obtained in step S3-3, calculate the current filtering range under the current wind field conditions in real time, denoted as [FLt, FUt], where FLt represents the lower boundary of the current filtering range and FUt represents the upper boundary of the current filtering range. Step S4-2: Extract the spectral information within the current filtering range from the wingtip displacement spectrum at the current moment, and collect the amplitude values ​​corresponding to N frequencies respectively, where the j-th frequency is denoted as fj, fj∈[FLt, FUt]; calculate the weighted average amplitude AW. , where w(fj) represents the probability of a bird intrusion event occurring when frequency fi appears in the historical records, and A(fj) represents the amplitude value corresponding to frequency fj in the current filtering range; Step S4-3: Multiply the weighted average amplitude AW by a preset safety factor C. safe Obtain the dynamic threshold T dy (t), that is C safe It is a constant greater than 1; Step S4-4: Obtain the maximum amplitude of the frequency in the spectrum information of the current filtering range, denoted as Amax(t), and compare the maximum amplitude with the dynamic threshold T. dy If Amax(t) > Tdy(t), then a risk event is marked as existing at the current time, and the occurrence time of this risk event, the corresponding wind field conditions, and the spectral characteristics are recorded in the risk event log.

6. The real-time situational awareness and early warning method based on abnormal airport situations according to claim 5, characterized in that: Step S5 includes: Step S5-1: Periodically sample the times marked as risk events, with a sampling period of Δt, to obtain a series of risk event sampling times; Step S5-2: At the i-th sampling time ti, calculate the risk value R(ti) at the i-th sampling time, R(ti) = Amax(ti) / TsD(ti), where TsD(ti) represents the dynamic threshold at time ti, and Amax(ti) represents the maximum amplitude of the spectrum at time ti; Step S5-3: Calculate the risk value at each sampling time. When the risk value shows a monotonically increasing trend over M consecutive sampling periods, trigger an alarm signal, where M≥2.

7. A real-time situational awareness and early warning system based on airport anomalies, used to execute the real-time situational awareness and early warning method based on airport anomalies as described in any one of claims 1-6, characterized in that: The system includes: The system includes a spectrum management module, a mapping database management module, a filter management module, a risk event management module, and an alarm module. The spectrum management module is used to identify the position of bird wingtips, collect the pixel coordinates of bird wingtips, continuously record the Euclidean distance between wingtips between adjacent frames as a relative displacement sequence, apply fast Fourier transform to the relative displacement sequence for spectrum analysis, and generate a spectrum diagram containing the correspondence between frequency and amplitude. The mapping database management module is used to establish a mapping relationship database between wind field parameters and wingtip displacement spectrum characteristics based on the wind field conditions and corresponding wingtip displacement spectrum when birds invaded the airport in historical records. The wind field conditions include wind speed and wind direction angle. The filtering management module is used to divide the data in the mapping relationship data, establish the correspondence between the frequency distribution intervals of different wind field conditions and wingtip displacement, and generate the filtering range of the spectrum under different wind field conditions. The risk event management module is used to obtain the maximum value in the spectrum graph, compare the maximum value with the risk threshold, and mark the existence of a risk event at the current time when the maximum value exceeds the threshold. The alarm module is used to periodically sample marked risk events, calculate the current risk value at each sampling time, and trigger an alarm signal when the risk value shows a monotonically increasing trend.

8. A real-time situational awareness and early warning system based on abnormal airport situations according to claim 7, characterized in that: The spectrum management module includes an image acquisition unit, a displacement difference management unit, and a spectrum conversion unit. The image acquisition unit is used to acquire bird images, sample the edges of the bird's left or right wing, and identify the coordinates of the wingtip in the image. The displacement difference management unit is used to acquire the wingtip coordinates of each bird image, acquire the Euclidean distance between the wingtip coordinates of two adjacent frames, and collect all Euclidean distances to form a relative displacement sequence. The spectrum conversion unit is used to acquire the amplitude spectrum after performing a fast Fourier transform on each Euclidean distance in the relative displacement sequence to form a spectrum diagram. The mapping database management module includes a historical data management unit, a spectrum feature management unit, and a mapping relationship management unit. The historical data management unit is used to acquire historical data and synchronously collect spectral data of bird wingtip displacement during bird invasion events. The spectrum feature management unit is used to extract spectral features from historical wingtip displacement spectral data. The mapping relationship management unit is used to establish the association between the historical wind field conditions and the corresponding spectral features, and to store the historical wind field conditions, spectral features, and association relationships together to form a mapping relationship database.

9. A real-time situational awareness and early warning system based on abnormal airport situations according to claim 7, characterized in that: The filtering management module includes a wind field condition management unit, a frequency distribution management unit, and a condition mapping unit. The wind field condition management unit is used to classify historical records according to the wind field conditions in the mapping relationship database to form several wind field condition groups. The frequency distribution management unit is used to obtain the spectral characteristics of all bird intrusion events in any wind field condition group, and statistically calculate the frequency distribution interval of the wingtip displacement corresponding to the bird behavior based on the spectral characteristics of the bird intrusion events. The condition mapping unit is used to establish a regression model including the mapping relationship between wind field conditions and filtering range based on the wind field conditions and the corresponding frequency distribution interval using regression analysis methods.

10. A real-time situational awareness and early warning system based on abnormal airport situations according to claim 7, characterized in that: The risk event management module includes a frequency filtering unit, a risk value calculation unit, and a risk judgment unit. The frequency filtering unit is used to acquire real-time wind field conditions at the current moment and calculate the current filtering range under the current wind field conditions in real time according to the regression model. The risk value calculation unit is used to extract the spectrum information of the current filtering range from the wingtip displacement spectrum map at the current moment and calculate the dynamic threshold. The risk judgment unit is used to compare the maximum amplitude with the dynamic threshold to determine whether a risk event exists. The alarm module includes a periodic sampling unit and an information prompting unit. The periodic sampling unit is used to periodically sample the times marked as risk events and calculate the risk value at each sampling time. The information prompting unit is used to trigger an alarm signal based on the judgment conditions.

Images