Low-altitude aircraft state monitoring method and system based on 5g base station tower

Abstract

The application provides a low-altitude aircraft state monitoring method and system based on a 5G base station tower, which comprises the following steps: receiving low-altitude multi-modal observation data strings collected by distributed sensing nodes deployed on the 5G base station tower; performing multi-modal feature trajectory weaving processing on the low-altitude multi-modal observation data strings to obtain a dynamic feature trajectory set; constructing a trajectory correlation evolution network based on the dynamic feature trajectory set to generate trajectory correlation evolution parameters, which are used to represent the evolution law data of the correlation strength between different feature trajectories over time; performing state trend bifurcation deduction based on the trajectory correlation evolution parameters to obtain a state trend bifurcation result; and outputting a low-altitude aircraft state comprehensive monitoring result according to the state trend bifurcation result. The application can improve the accuracy and comprehensiveness of low-altitude aircraft state monitoring.

Description

Technical Field

[0001] This invention relates to the field of air traffic monitoring, and more specifically, to a method and system for monitoring the status of low-altitude aircraft based on 5G base station towers. Background Technology

[0002] With the expansion of low-altitude aircraft application scenarios, analyzing real-time status data can accurately determine their operational status to ensure low-altitude airspace safety. Currently, observational data such as the aircraft's airspace position, signal characteristics, or motion parameters are typically collected by sensing devices deployed in fixed locations. This data is then processed and used for prediction to output status monitoring results. However, existing technologies struggle to effectively construct cross-correlation relationships between characteristic trajectories of different modes, resulting in feature dimensions being limited to simple modes and lacking correlative support. Furthermore, insufficient analysis of feature correlations makes it difficult to capture the dynamic evolution of correlation strength over time, leading to a less in-depth characterization of the intrinsic mechanisms of aircraft state changes. Consequently, monitoring results cannot comprehensively cover potential changes in aircraft state, affecting the accuracy and comprehensiveness of low-altitude aircraft status monitoring. Summary of the Invention

[0003] This invention provides a method and system for monitoring the status of low-altitude aircraft based on 5G base station towers.

[0004] In a first aspect, embodiments of the present invention provide a method for monitoring the status of low-altitude aircraft based on 5G base station towers, comprising: receiving low-altitude multimodal observation data strings collected by distributed sensing nodes deployed on 5G base station towers; performing multimodal feature trajectory weaving processing on the low-altitude multimodal observation data strings to obtain a dynamic feature trajectory set; constructing a trajectory association evolution network based on the dynamic feature trajectory set to generate trajectory association evolution parameters, wherein the trajectory association evolution parameters are used to characterize the evolution law data of the association strength between different feature trajectories over time; performing state trend bifurcation inference based on the trajectory association evolution parameters to obtain state trend bifurcation results; and outputting a comprehensive monitoring result of the low-altitude aircraft status based on the state trend bifurcation results.

[0005] Secondly, embodiments of the present invention provide a computer system, including: a memory storing a computer program; and a processor for loading the computer program to implement the low-altitude aircraft status monitoring method based on 5G base station towers as described above.

[0006] The present invention provides a method for monitoring the status of low-altitude aircraft based on 5G base station towers. It receives low-altitude multimodal observation data strings within the coverage area of ​​the 5G base station towers. By performing multimodal feature trajectory weaving processing on these data strings, a dynamic feature trajectory set is obtained, comprising airspace feature trajectories, signal feature trajectories, motion feature trajectories, and cross-correlation trajectory data between these three types of feature trajectories. This ensures that the acquired features not only cover trajectory information of a single mode but also include the correlation relationships between feature trajectories of different modes, thereby enriching the feature dimensions and enhancing the correlation between features. Based on this dynamic feature trajectory set, a trajectory correlation evolution network is constructed to generate trajectories representing the evolution of the correlation strength between different feature trajectories over time. Correlation evolution parameters can capture the dynamic changes in characteristic correlations, breaking through the limitations of static correlation analysis and improving the ability to characterize the intrinsic mechanisms of aircraft state changes. Based on these trajectory correlation evolution parameters, state trend bifurcation inference is performed to obtain state trend bifurcation results that include multiple possible state trend data generated based on different correlation evolution paths. This can systematically cover the multiple path possibilities of aircraft state evolution and avoid ignoring uncertainties in single trend prediction. Finally, based on these state trend bifurcation results, comprehensive monitoring results of low-altitude aircraft state are output, which can more comprehensively and accurately indicate the dominant state trend category and corresponding confidence level data of the aircraft during the monitoring period, improving the accuracy and comprehensiveness of low-altitude aircraft state monitoring. Attached Figure Description

[0007] Figure 1 This is a flowchart of a low-altitude aircraft status monitoring method based on a 5G base station tower, provided by an embodiment of the present invention.

[0008] Figure 2 This is a schematic diagram of the composition of a computer system provided in an embodiment of the present invention. Detailed Implementation

[0009] Please see Figure 1 , Figure 1 A flowchart illustrating a low-altitude aircraft status monitoring method based on a 5G base station tower, provided as an embodiment of the present invention, is shown. This method can be executed by a computer system and may include the following steps:

[0010] Step S100: Receive the low-altitude multimodal observation data string collected by the distributed sensing nodes deployed on the 5G base station tower.

[0011] The low-altitude multimodal observation data string is collected by distributed sensing nodes deployed on 5G base station towers. It contains various types of observation information, such as airspace observation data, signal observation data, and motion observation data arranged in time series. These data strings cover many aspects of information related to low-altitude aircraft. The distributed sensing nodes are multiple data acquisition devices distributed on 5G base station towers. They can observe the low-altitude area from different angles and positions to obtain comprehensive and accurate aircraft-related data.

[0012] In practical applications, these distributed sensing nodes can be various types of sensors. For example, optical sensors can be used to collect image information of the aircraft, and by analyzing the images, information such as the aircraft's appearance and flight attitude can be obtained; radar sensors can measure parameters such as the aircraft's distance and speed; and signal monitoring sensors can capture the signal spectrum data emitted by the aircraft, which can reflect information such as the aircraft's communication status and operating mode. These different types of sensors integrate the collected data to form a low-altitude multimodal observation data string.

[0013] Step S200: Perform multimodal feature trajectory weaving processing on the low-altitude multimodal observation data string to obtain a dynamic feature trajectory set.

[0014] Multimodal feature trajectory weaving is used to integrate and analyze different types of observation data from low-altitude multimodal observation datasets, uncovering the correlations and features between the data to form a trajectory set that reflects the dynamic characteristics of low-altitude aircraft. The dynamic feature trajectory set is a set of trajectory data containing various dynamic feature information of the aircraft, such as airspace feature trajectories, signal feature trajectories, and motion feature trajectories, as well as cross-correlation trajectory data between these three types of feature trajectories. This trajectory data can describe the flight state and behavior of the aircraft from different perspectives.

[0015] In one implementation, step S200 may specifically include the following steps S210 to S270:

[0016] Step S210: Separate the airspace observation substring, signal observation substring, and motion observation substring from the low-altitude multimodal observation data string. The airspace observation substring contains time-series arranged three-dimensional coordinate data of the aircraft, the signal observation substring contains time-series arranged signal spectrum data of the aircraft, and the motion observation substring contains time-series arranged acceleration vector data of the aircraft.

[0017] The airspace observation substring is extracted from the low-altitude multimodal observation data string and is used to describe the aircraft's position information in space. The aircraft's three-dimensional coordinate data is arranged in a time series, clearly reflecting the changes in the aircraft's spatial position at different times. The signal observation substring contains the signal spectrum data emitted by the aircraft, which is also arranged in a time series. Analysis of the signal spectrum data reveals changes in the aircraft's communication status, operating mode, and other information over time. The motion observation substring contains the aircraft's acceleration vector data, arranged in a time series, reflecting changes in the aircraft's motion state at different times, such as acceleration, deceleration, and turning.

[0018] In practice, these three substrings can be separated from the low-altitude multimodal observation data string through data parsing and classification methods. For example, based on the data format and characteristics, data containing three-dimensional coordinate information can be classified into the spatial domain observation substring, data containing signal spectrum information into the signal observation substring, and data containing acceleration vector information into the motion observation substring.

[0019] For example, a low-altitude multimodal observation data string is a large file containing various data types, encoded according to a specific format. A data parsing program can be written to read the data line by line, determining which substring the data belongs to based on its beginning identifier or specific format characteristics. For a spatial observation substring, the program can identify lines containing 3D coordinate data and extract them, then organize them chronologically; for a signal observation substring, the program can identify lines containing signal spectrum data and perform the same processing; for a motion observation substring, the program can identify lines containing acceleration vector data and organize them.

[0020] Step S220: Perform trajectory point density clustering on the spatial observation substring to obtain a spatial trajectory cluster, which contains a set of spatial observation data points with spatial location correlation.

[0021] Trajectory point density clustering calculates the distances and densities between data points in a substring of airspace observations, grouping spatially correlated data points together to form different clusters. An airspace trajectory cluster is a set of spatially correlated airspace observation data points that are close to each other in space and may represent the flight trajectory of an aircraft over a period of time. When performing trajectory point density clustering, the spatial location and temporal order of the data points need to be considered. For example, an aircraft may make brief stops or circle during flight, and these data points may be relatively dense in space. Density clustering can group these dense data points together to form an airspace trajectory cluster.

[0022] In one implementation, step S220 may specifically include the following steps S221 to S227:

[0023] Step S221: Extract all spatial observation data points from the spatial observation substring. Each spatial observation data point contains three-dimensional coordinate values ​​and corresponding timestamp data.

[0024] Airspace observation data points are the basic elements constituting an airspace observation substring. Each data point contains the three-dimensional coordinates of the aircraft at a specific moment and its corresponding timestamp. The three-dimensional coordinates accurately represent the aircraft's position in space, while the timestamp records the specific time corresponding to that position. When extracting airspace observation data points, data parsing methods can be used to identify the three-dimensional coordinates and timestamp of each data point from the airspace observation substring. For example, the information of each data point can be extracted and stored in a corresponding data structure based on the data format and delimiters.

[0025] Step S222: Calculate the spatial Euclidean distance between any two spatial observation data points and generate a distance matrix containing the spatial distance values ​​of all data point pairs.

[0026] Spatial Euclidean distance is the straight-line distance between two points in three-dimensional space. By calculating the spatial Euclidean distance between any two spatial observation data points, their relative spatial positions can be obtained. The distance matrix is ​​a two-dimensional matrix, where each element represents the spatial Euclidean distance between two spatial observation data points.

[0027] When calculating spatial Euclidean distance, the Euclidean distance formula can be used: For two three-dimensional coordinate points (x1, y1, z1) and (x2, y2, z2), the spatial Euclidean distance d between them can be calculated using the following formula: When generating the distance matrix, you can iterate through all pairs of data points, calculate the spatial Euclidean distance between them using the formula above, and store the results in the distance matrix.

[0028] For example, given n spatial observation data points, the distance matrix D is an n×n matrix. All pairs of data points (i,j) can be iterated using a nested loop, where i and j represent the indices of the data points. For each pair of data points, the distance between them is calculated using the Euclidean distance formula, and the result is stored in the i-th row and j-th column and the j-th row and i-th column of matrix D (because the distance is symmetric).

[0029] Step S223: Set a density threshold and a distance threshold. The density threshold is the minimum number of data points required to form a cluster, and the distance threshold is the maximum allowed distance between data points.

[0030] Density threshold and distance threshold are two parameters in trajectory point density clustering algorithms. The density threshold determines how many other data points are needed around a given data point to form a cluster. If the number of data points surrounding a data point is less than the density threshold, that data point may be considered noise. The distance threshold limits the maximum allowable distance between data points. If the distance between two data points is greater than the distance threshold, they are unlikely to belong to the same cluster. In practical applications, the density and distance thresholds need to be adjusted based on the specific application scenario and data characteristics. For example, if the aircraft's flight trajectory is dense and the distance between data points is relatively small, the distance threshold can be appropriately reduced; if the data points are sparse, the distance threshold can be appropriately increased.

[0031] Step S224: Based on the distance matrix and density threshold, identify core data points. Core data points are spatial observation data points that contain at least a density threshold of other data points within the distance threshold range.

[0032] Core data points are spatial observation data points that contain at least a density threshold of other data points within a distance threshold range. Core data points are typically located at the center of clusters, surrounded by denser data points, representing important locations in the aircraft's flight path.

[0033] When identifying core data points, the distance matrix can be traversed, and for each data point, the number of data points within a distance threshold range can be counted. If this number is not less than the density threshold, then the data point is marked as a core data point.

[0034] Step S225: Group the data points whose distance to the core data points is less than the distance threshold into the same cluster to obtain the initial spatial cluster.

[0035] After identifying the core data points, data points that are less than a distance threshold from the core data points can be grouped into the same cluster, forming an initial airspace cluster. These data points are spatially close to the core data points and may represent the flight trajectory of the aircraft within the same time period.

[0036] When classifying data points, all data points can be traversed. For each data point, check if its distance to the core data point is less than a distance threshold. If so, the data point is assigned to the cluster corresponding to that core data point.

[0037] Step S226: Calculate the average timestamp of all data points in the initial spatial cluster, and sort the initial spatial cluster by time according to the average timestamp to obtain a time-ordered spatial cluster sequence.

[0038] The timestamp mean is the arithmetic mean of the timestamps of all data points in the initial airspace cluster. By calculating the timestamp mean and sorting the initial airspace clusters by time, these clusters can be arranged in chronological order to form a time-ordered sequence of airspace clusters. This more clearly reflects the temporal order of the aircraft's flight trajectory.

[0039] To calculate the timestamp mean, the timestamps of all data points in the initial spatial cluster are summed, and then divided by the number of data points to obtain the timestamp mean. When sorting by time, sorting algorithms such as quicksort or bubble sort can be used to sort the initial spatial cluster in ascending order of timestamp mean.

[0040] In one implementation, step S226 may specifically include the following steps S2261 to S2266:

[0041] Step S2261: Traverse each initial spatial cluster, collect the timestamp data of all data points within the cluster, and obtain the cluster timestamp set.

[0042] A cluster timestamp set is a collection of timestamp data from all data points within an initial spatial cluster. By traversing each initial spatial cluster and collecting the timestamp data from its data points, the cluster timestamp set for that cluster can be obtained.

[0043] In practical implementation, a loop can be used to traverse all initial spatial clusters. For each cluster, the timestamp data of its data points is stored in a list or array, forming a cluster timestamp set. For example, given an initial spatial cluster E containing 5 data points E1, E2, E3, E4, and E5, with timestamps t1, t2, t3, t4, and t5 respectively, by traversing the cluster and collecting these 5 timestamps, a cluster timestamp set {t1, t2, t3, t4, t5} can be formed.

[0044] Step S2262: Calculate the arithmetic mean of all timestamps in the cluster timestamp set to obtain the average timestamp of the initial spatial cluster.

[0045] The arithmetic mean is the sum of a set of data divided by the number of data points. The average timestamps of an initial spatial cluster can be obtained by calculating the arithmetic mean of all timestamps in the cluster timestamp set. To calculate the arithmetic mean, all timestamps in the cluster timestamp set are added together and then divided by the number of elements in the set.

[0046] Step S2263: Sort all initial spatial clusters in ascending order of timestamp mean to obtain a sorted list of spatial clusters.

[0047] The sorted list of spatial clusters is obtained by arranging all the initial spatial clusters in ascending order of their timestamp mean. Sorting algorithms, such as quicksort and bubble sort, can be used for this process. These algorithms compare and swap the initial spatial clusters based on their timestamp mean to obtain the final sorted list.

[0048] Step S2264: Check the difference in the average timestamps of adjacent clusters in the sorted spatial cluster list. When the difference is less than the preset time interval threshold, they are determined to be consecutive clusters.

[0049] The preset time interval threshold is a pre-defined time difference value used to determine whether adjacent initial airspace clusters are consecutive clusters. If the average difference of the timestamps of adjacent clusters is less than the threshold, the two clusters are considered to be consecutive in time, possibly representing the flight trajectory of the aircraft within a continuous time period.

[0050] When checking the timestamp mean difference, you can iterate through the sorted list of spatial clusters, calculate the timestamp mean difference between two adjacent clusters, and compare it with a preset time interval threshold.

[0051] Step S2265: Perform timestamp mean fusion on the continuous clusters, and update the timestamp mean of the continuous clusters to the arithmetic mean of the timestamp means of all continuous clusters.

[0052] After identifying consecutive clusters, to more accurately represent the temporal continuity of these clusters, their timestamp means can be merged. For example, the timestamp mean of consecutive clusters can be updated to the arithmetic mean of the timestamp means of all consecutive clusters. When merging timestamp means, the timestamp means of consecutive clusters can be added together first, and then divided by the number of consecutive clusters to obtain the merged timestamp mean.

[0053] Step S2266: Combine the merged continuous clusters with the non-continuous clusters to form a time-ordered spatial cluster sequence.

[0054] After the timestamps of consecutive clusters are fused, the fused consecutive clusters and non-consecutive clusters are recombine in chronological order to form a time-ordered spatial cluster sequence. This sequence can more accurately reflect the temporal order and continuity of the aircraft's flight trajectory.

[0055] For example, the sorted list of spatial clusters contains three contiguous clusters I1, I2, and I3, and two non-contiguous clusters J1 and J2. After timestamp mean fusion, the timestamp mean of the contiguous clusters is updated to... The merged continuous clusters [I1,I2,I3] and non-continuous clusters [J1,J2] are recombined in chronological order to obtain the temporally ordered spatial cluster sequence [J1,I1,I2,I3,J2].

[0056] Step S227: Calculate the spatial overlap of adjacent clusters in the temporally ordered spatial cluster sequence, merge adjacent clusters with a spatial overlap greater than a preset overlap threshold, and obtain the merged spatial cluster as the spatial trajectory cluster.

[0057] Spatial overlap is the degree to which adjacent temporally ordered spatial clusters overlap in space. By calculating spatial overlap, it can be determined whether adjacent clusters represent the flight trajectories of aircraft within the same region. The preset overlap threshold is a pre-defined overlap ratio used to determine whether to merge adjacent clusters.

[0058] When calculating spatial overlap, geometric methods can be used, such as calculating the ratio of the intersection area to the union area of ​​the spatial regions covered by two clusters. If this ratio is greater than a preset overlap threshold, the two adjacent clusters are merged into one cluster, resulting in a merged spatial cluster, which is the final spatial trajectory cluster.

[0059] Step S230: Perform spectral feature drift tracking on the observed signal substring to obtain the signal drift trajectory, which contains the drift path data of the signal spectral features over time.

[0060] Spectral feature drift tracking is a method for analyzing signal spectral data in a signal observation substring. By tracking the changes in signal spectral features over time, the signal drift trajectory is obtained. The signal drift trajectory records the position and change path of the signal spectral features at different times, reflecting the dynamic changes of the aircraft signal.

[0061] When tracking spectral drift, spectral analysis algorithms such as Fourier transform and wavelet transform can be used. These algorithms can transform a signal from the time domain to the frequency domain, extracting its spectral features. Then, by comparing and analyzing the spectral features at different times, its drift path over time can be tracked. For example, for a signal emitted by an aircraft, a Fourier transform is used to convert its time-domain signal at different times into a frequency-domain signal, extracting its spectral features. Then, the spectral features at adjacent times are compared, and the changes in frequency and amplitude are recorded. By continuously performing such comparisons and recordings, the drift path of the signal's spectral features over time can be obtained, forming the signal drift trajectory.

[0062] Step S240: Perform acceleration direction coherence analysis on the motion observation substring to obtain the motion direction trajectory, which contains the path data of the coherent change of the acceleration vector direction over time.

[0063] Acceleration direction coherence analysis is a method for analyzing acceleration vector data in a motion observation substring. By analyzing the coherent changes in the acceleration vector direction over time, the motion direction trajectory is obtained. The motion direction trajectory records the path of change in the acceleration vector direction of the aircraft at different times, reflecting the aircraft's motion state and directional changes.

[0064] When performing coherence analysis of acceleration direction, vector analysis methods can be used, such as calculating the angle between acceleration vectors at adjacent moments. If the angle is small, it indicates that the acceleration vector direction is consistent over time; if the angle is large, it indicates that the acceleration vector direction has changed significantly. By analyzing the acceleration vectors at a series of adjacent moments, the coherent change path of the acceleration vector direction over time can be obtained, forming the trajectory of motion.

[0065] For example, consider a motion observation substring of an aircraft containing acceleration vector data at different times. For two adjacent times t1 and t2, the acceleration vectors are a1 and a2, respectively. The angle θ between these two vectors is calculated. If θ is less than a preset angle threshold, the acceleration vector direction is considered continuous between these two times. By continuously calculating the angle between adjacent times and recording the changes in the acceleration vector direction, the motion trajectory is obtained.

[0066] Step S250: Calculate the cross-correlation degree between the spatial trajectory cluster and the signal drift trajectory, and generate the first cross-correlation trajectory, which contains the correlation path data between the spatial position change and the signal spectrum drift.

[0067] Cross-correlation is an indicator that measures the degree of correlation between airspace trajectory clusters and signal drift trajectories. By calculating this correlation, the relationship between changes in the airspace position of an aircraft and signal spectral drift can be understood. The first cross-correlation trajectory records the correlation path data between changes in airspace position and signal spectral drift, providing more comprehensive information for aircraft status monitoring. When calculating cross-correlation, the characteristics and changes of both airspace trajectory clusters and signal drift trajectories need to be considered comprehensively. This can be achieved by extracting key features from both, such as the center coordinates of the airspace trajectory clusters and the spectral characteristics of the signal drift trajectory, and then analyzing their correlation.

[0068] As one implementation method, step S250, calculating the cross-correlation degree between the spatial trajectory cluster and the signal drift trajectory to generate the first cross-correlation trajectory, may specifically include the following steps S251~S258:

[0069] Step S251: Extract the cluster center trajectory from the spatial trajectory cluster. The cluster center trajectory contains the path data of the change of the center coordinates of each spatial trajectory cluster over time.

[0070] The cluster center trajectory is extracted from the spatial trajectory clusters, and it records the path of the center coordinates of each spatial trajectory cluster over time. The cluster center coordinates can be obtained by calculating the average coordinates of all data points in the spatial trajectory cluster.

[0071] When extracting the cluster center trajectory, each spatial trajectory cluster can be traversed, its center coordinates can be calculated, and the changes of the center coordinates over time can be recorded.

[0072] For example, for a spatial trajectory cluster L, which contains multiple spatial observation data points (x1, y1, z1), (x2, y2, z2), ..., calculate the average coordinates of these data points: , where n is the number of data points. By continuously calculating the center coordinates of the spatial trajectory cluster at different times, the cluster center trajectory is obtained. Step S252: Extract the spectral feature main trajectory from the signal drift trajectory. The spectral feature main trajectory contains the drift path data of the strongest frequency component in the signal spectrum over time.

[0073] The master spectral feature trajectory is extracted from the signal drift trajectory and records the drift path of the most energetic frequency component in the signal spectrum over time. In the signal spectrum, the most energetic frequency component is usually the most representative feature, and its changes can reflect the main trend of the signal. When extracting the master spectral feature trajectory, the spectral data in the signal drift trajectory can be analyzed to identify the most energetic frequency component at each moment and record its drift path over time.

[0074] For example, a signal drift trajectory contains spectral data at different times. For the spectral data at each time point, spectral analysis methods are used to identify the frequency components with the strongest energy. By continuously performing such analysis and recording, the main spectral characteristic trajectory is obtained.

[0075] Step S253: Unify the time axis of the cluster center trajectory and the main trajectory of the spectral features to obtain a time-synchronized trajectory pair.

[0076] Time-synchronized trajectory pairs are obtained by unifying the time axes of the cluster center trajectory and the main spectral feature trajectory. Since the cluster center trajectory and the main spectral feature trajectory may be recorded at different time scales, it is necessary to unify their time axes for subsequent analysis and calculation. Interpolation or sampling methods can be used to unify the time axes. For example, a common time axis can be selected, and the data of the cluster center trajectory and the main spectral feature trajectory can be interpolated or sampled according to this common time axis, so that they have corresponding trajectory data at the same time points.

[0077] For example, the time interval of the cluster center trajectory is 1 second, and the time interval of the main spectral feature trajectory is 2 seconds. A common time axis with a time interval of 1 second is selected. For the main spectral feature trajectory, linear interpolation is used to calculate the corresponding spectral feature value at each 1-second point in time. In this way, time-synchronized cluster center trajectories and main spectral feature trajectories are obtained, forming a time-synchronized trajectory pair.

[0078] Step S254: Calculate the feature vector of the time synchronization trajectory pair at each timestamp. The feature vector contains the cluster center coordinate vector and the main frequency vector of the spectrum.

[0079] An eigenvector is a vector obtained by combining key features of a time-synchronized trajectory at each timestamp. At each timestamp, the cluster center coordinate vector represents the center coordinates of the spatial trajectory cluster at that moment, and the dominant frequency vector represents the most energetic frequency component in the signal spectrum at that moment. When calculating the eigenvector, the cluster center coordinate vector and the dominant frequency vector can be combined in a specific order. For example, the three components of the cluster center coordinate vector and one component of the dominant frequency vector can be arranged sequentially to form a four-dimensional eigenvector.

[0080] Step S255: Construct a joint feature matrix based on the feature vectors of all timestamps. The joint feature matrix contains feature vector data arranged in chronological order.

[0081] The joint feature matrix is ​​a matrix obtained by arranging the feature vectors of all timestamps in chronological order. This matrix contains key feature information of time-synchronized trajectory pairs at different timestamps, providing a data foundation for subsequent analysis and calculations.

[0082] When constructing the joint feature matrix, the feature vector of each timestamp can be treated as a row of the matrix, arranged sequentially in chronological order. For example, if there are n timestamps and each feature vector is an m-dimensional vector, then the joint feature matrix is ​​an n×m matrix.

[0083] Step S256: Perform singular value decomposition on the joint characteristic matrix to obtain the left singular matrix, the singular value matrix, and the right singular matrix.

[0084] Singular Value Decomposition (SVD) decomposes a matrix into the product of three matrices: the left singular matrix, the singular value matrix, and the right singular matrix. By performing SVD on the joint characteristic matrix, key features and structural information of the matrix can be extracted.

[0085] The formula for singular value decomposition can be found in existing techniques and will not be elaborated here. When performing singular value decomposition, matrix factorization algorithms, such as Singular Value Decomposition (SVD), can be used. These algorithms can decompose the joint characteristic matrix into a left singular matrix, a singular value matrix, and a right singular matrix. For example, for the joint characteristic matrix A mentioned above, using the SVD algorithm yields a left singular matrix U, a singular value matrix ∑, and a right singular matrix V.

[0086] Step S257: Extract the left and right singular vectors corresponding to the largest singular value in the singular value matrix, calculate their inner product, and obtain the cross-correlation coefficient.

[0087] The cross-correlation coefficient is an indicator that measures the degree of cross-correlation between spatial trajectory clusters and signal drift trajectories. It is obtained by extracting the left and right singular vectors corresponding to the maximum singular value in the singular value matrix and calculating their inner product. The left and right singular vectors corresponding to the maximum singular value are the most important vectors in singular value decomposition, as they contain the main information of the joint feature matrix. Calculating the inner product of these two vectors yields a numerical value, which is the cross-correlation coefficient.

[0088] Step S258: Arrange the cross-correlation coefficients by timestamp to generate the first cross-correlation trajectory.

[0089] The first cross-correlation trajectory is the trajectory obtained by arranging the cross-correlation coefficients according to timestamps. By arranging the cross-correlation coefficients by timestamps, the change in the degree of cross-correlation between the spatial trajectory cluster and the signal drift trajectory over time can be clearly seen.

[0090] When generating the first cross-correlation trajectory, the cross-correlation coefficient for each timestamp can be used as the trajectory value for that timestamp, arranged sequentially in chronological order. For example, the cross-correlation coefficients can be stored in a list or array, where the list index corresponds to the timestamp and the elements in the list correspond to the cross-correlation coefficients.

[0091] Step S260: Calculate the cross-correlation degree between the signal drift trajectory and the motion direction trajectory, and generate a second cross-correlation trajectory. The second cross-correlation trajectory contains the correlation path data between the signal spectrum drift and the change in motion direction.

[0092] Cross-correlation is an indicator that measures the degree of correlation between the signal drift trajectory and the motion direction trajectory. By calculating this correlation, the relationship between the aircraft's signal spectrum drift and changes in motion direction can be understood. The second cross-correlation trajectory records the correlation path data between signal spectrum drift and changes in motion direction, providing more comprehensive information for aircraft status monitoring. When calculating cross-correlation, the characteristics and changes of both the signal drift trajectory and the motion direction trajectory need to be considered comprehensively. This can be achieved by extracting key features from both, such as signal spectrum characteristics and acceleration vector direction, and then analyzing their correlation.

[0093] In one implementation, step S260 may specifically include the following steps S261 to S267:

[0094] Step S261: Extract the spectral drift rate trajectory from the signal drift trajectory. The spectral drift rate trajectory contains the path data of the change of the drift speed of the signal spectral features over time. The drift speed is calculated by the ratio of the difference in spectral features between adjacent timestamps to the time interval.

[0095] The spectral drift rate trajectory is extracted from the signal drift trajectory and records the path of change of the signal's spectral characteristics over time. The drift rate is obtained by calculating the ratio of the difference in spectral characteristics between adjacent timestamps to the time interval, reflecting how quickly the signal's spectral characteristics change over time.

[0096] When extracting the spectral drift rate trajectory, the signal drift trajectory can be traversed. For two adjacent timestamps, the difference in their spectral characteristics is calculated, and then divided by the time interval to obtain the drift rate within that time interval. By continuously performing such calculations and recording, the path of the spectral drift rate change over time can be obtained, forming the spectral drift rate trajectory.

[0097] For example, for a signal drift trajectory, its spectral characteristics at timestamps t1 and t2 are f1 and f2, respectively, with a time interval Δt = t2 - t1. The drift velocity v = (f2 - f1) / Δt is calculated. By continuously calculating the drift velocity at adjacent timestamps, the spectral drift rate trajectory is obtained.

[0098] Step S262: Extract the direction change acceleration trajectory from the motion direction trajectory. The direction change acceleration trajectory contains the path data of the change rate of acceleration in the motion direction over time. The rate of change of acceleration is calculated by the ratio of the difference in acceleration vector between adjacent timestamps to the time interval.

[0099] The directional acceleration trajectory is extracted from the motion direction trajectory and records the path of the rate of change of acceleration in the motion direction over time. The rate of change of acceleration is obtained by calculating the ratio of the difference in acceleration vectors between adjacent time stamps to the time interval, reflecting how quickly the aircraft's motion direction changes. When extracting the directional acceleration trajectory, the motion direction trajectory can be traversed. For two adjacent time stamps, the difference in their acceleration vectors is calculated, and then divided by the time interval to obtain the rate of change of acceleration within that time period. By continuously performing this calculation and recording, the path of the rate of change of acceleration in the motion direction over time is obtained, forming the directional acceleration trajectory. Step S263: Unify the time axes of the spectral drift rate trajectory and the directional acceleration trajectory to obtain a synchronization rate-acceleration trajectory pair. The synchronization rate-acceleration trajectory pair contains the spectral drift rate value and the directional acceleration value at the same time stamp.

[0100] The synchronization rate-acceleration trajectory pair is obtained by unifying the time axes of the spectral drift rate trajectory and the directional change acceleration trajectory. Since the spectral drift rate trajectory and the directional change acceleration trajectory may be recorded at different time scales, it is necessary to unify their time axes for subsequent analysis and calculation.

[0101] When unifying the time axis, interpolation or sampling methods can be used. For example, a common time axis can be selected, and the data of the spectral drift rate trajectory and the directional acceleration trajectory can be interpolated or sampled according to this common time axis, so that they have corresponding trajectory data at the same time points. For example, the time interval of the spectral drift rate trajectory is 1 second, and the time interval of the directional acceleration trajectory is 2 seconds. A common time axis with a time interval of 1 second is selected. For the directional acceleration trajectory, linear interpolation is used to calculate the corresponding acceleration rate value at each 1-second time point. In this way, time-synchronized spectral drift rate trajectory and directional acceleration trajectory are obtained, forming a synchronized rate-acceleration trajectory pair.

[0102] Step S264: Calculate the dynamic cross-correlation coefficient of the synchronization rate-acceleration trajectory pair at each time point. The dynamic cross-correlation coefficient is calculated by dividing the sum of the products of the rate value and acceleration value at the current time point and the preset number of time points before and after by the product of the root mean square of the rate and the root mean square of the acceleration for the corresponding time period.

[0103] The dynamic cross-correlation coefficient is a measure of the correlation between the synchronization rate and acceleration trajectory at each time point. By calculating this coefficient, we can understand the degree of correlation between the signal spectrum drift rate and the change in acceleration in the direction of motion.

[0104] When calculating the dynamic cross-correlation coefficient, it is necessary to consider the velocity and acceleration values ​​of the current timestamp and a preset number of timestamps before and after it. The specific calculation method is to add the products of the velocity and acceleration values ​​of these timestamps, and then divide by the product of the root mean square of the velocity and the root mean square of the acceleration for the corresponding time period.

[0105] For example, with a preset quantity of 2, for timestamp t, consider timestamps. rate value and acceleration value Calculate the sum of the products of the velocity and acceleration values ​​for these timestamps. Calculate the root mean square rate for the corresponding time period. and root mean square of acceleration Then the dynamic cross-relation coefficient .

[0106] Step S265: Extract the peak and valley points from the dynamic cross-correlation coefficients to generate a characteristic extreme value sequence. The characteristic extreme value sequence contains the peak and valley values ​​of the dynamic cross-correlation coefficients arranged in chronological order and their corresponding timestamp data.

[0107] The eigenvalue sequence is obtained by extracting the peak and valley points from the dynamic cross-correlation coefficients. These peak and valley points reflect the extreme values ​​of the correlation between the signal's spectral drift rate and the change in acceleration in the direction of motion. When extracting peak and valley points, the dynamic cross-correlation coefficient sequence can be traversed, comparing adjacent coefficient values. If a coefficient value is greater than its preceding and following coefficient values, it is considered a peak point; if a coefficient value is less than its preceding and following coefficient values, it is considered a valley point. The values ​​of these peak and valley points and their corresponding timestamps are recorded and arranged in chronological order to form the eigenvalue sequence.

[0108] Step S266: Fit a trend line to the characteristic extreme value sequence to obtain the extreme value trend line, which contains the overall trend data of the characteristic extreme values ​​over time.

[0109] Trendline fitting is a method for analyzing characteristic extreme value sequences. By fitting the extreme value trend line, we can understand the overall trend of characteristic extreme values ​​over time.

[0110] When fitting trend lines, methods such as polynomial fitting or linear fitting can be used. For example, when using polynomial fitting, a suitable polynomial order can be selected, and the eigenvalue sequence can be fitted using the least squares method to obtain the fitting polynomial. Substituting the timestamps as independent variables and the eigenvalues ​​as dependent variables into the fitting polynomial yields the extreme value trend line.

[0111] For example, for the eigenvalue sequence [(0.8,t3),(0.1,t6),(0.9,t9)], a quadratic polynomial is used for fitting, let the fitting polynomial be y=ax 2 +bx+c. Solving for the coefficients a, b, and c using the least squares method yields a fitted polynomial. Substituting different timestamps into the fitted polynomial produces the corresponding predicted eigenvalues, forming an extreme value trend line.

[0112] Step S267: Combine the dynamic cross-correlation coefficient and the extreme value trend line to generate the second cross-correlation trajectory.

[0113] The second cross-correlation trajectory is obtained by combining the dynamic cross-correlation coefficient and the extreme value trend line. By combining them, the correlation between signal spectrum drift and changes in motion direction can be reflected more comprehensively.

[0114] When combining, the dynamic cross-correlation coefficient sequence and the extreme value trend line sequence can be merged in chronological order. For example, elements in the dynamic cross-correlation coefficient sequence and the extreme value trend line sequence can be matched according to their timestamps to form a new sequence, which is the second cross-correlation trajectory.

[0115] Step S270: Fuse the spatial trajectory cluster, signal drift trajectory, motion direction trajectory, first cross-correlation trajectory and second cross-correlation trajectory to obtain a dynamic feature trajectory set.

[0116] Fusion is the process of integrating airspace trajectory clusters, signal drift trajectories, motion direction trajectories, first cross-correlation trajectories, and second cross-correlation trajectories. Through fusion, a set of dynamic feature information of the aircraft can be obtained, namely, a set of dynamic feature trajectories.

[0117] When performing fusion, data association and feature combination methods can be used. For example, these trajectories can be aligned along a timeline, and then the key features of each timestamp can be combined to form a comprehensive feature vector. By collecting and organizing the feature vectors of all timestamps, a dynamic feature trajectory set is obtained.

[0118] For example, for spatial trajectory clusters, signal drift trajectories, motion direction trajectories, first cross-correlation trajectories, and second cross-correlation trajectories, at each timestamp, their key features, such as cluster center coordinates, dominant frequency of the spectrum, acceleration vector direction, and cross-correlation coefficient, are combined to form a multi-dimensional feature vector. The feature vectors of all timestamps are arranged in chronological order to obtain a dynamic feature trajectory set. Step S300: Based on the dynamic feature trajectory set, a trajectory association evolution network is constructed to generate trajectory association evolution parameters. These parameters characterize the evolution of the association strength between different feature trajectories over time.

[0119] The trajectory association evolution network is a network model used to analyze the evolution of the correlation strength between different characteristic trajectories over time. By constructing this network, we can gain a deeper understanding of the interrelationships and changes among different characteristic trajectories of an aircraft. The trajectory association evolution parameters are generated from the trajectory association evolution network; they quantitatively describe the evolution of the correlation strength between different characteristic trajectories over time. When constructing the trajectory association evolution network, it is necessary to consider the spatiotemporal correlation and mutual influence between different characteristic trajectories. The network model can employ structures such as deep learning networks or graph neural networks, obtaining the trajectory association evolution parameters by learning and analyzing a dynamic set of characteristic trajectories.

[0120] In one implementation, step S300 may specifically include the following steps S310 to S370:

[0121] Step S310: Initialize the trajectory association evolution network structure. The trajectory association evolution network structure includes a trajectory alignment layer, an association strength evolution layer, and a parameter feedback layer. The trajectory alignment layer is used to receive the dynamic feature trajectory set and perform time axis alignment processing. The association strength evolution layer is used to calculate the evolution data of the association strength between different feature trajectories over time. The parameter feedback layer is used to adjust the network connection weights based on the evolution data and output the evolution parameters.

[0122] The trajectory association evolution network structure is a multi-layered network structure consisting of a trajectory alignment layer, an association strength evolution layer, and a parameter feedback layer. The main function of the trajectory alignment layer is to unify the time axis of the dynamic feature trajectory set, ensuring that different feature trajectories have corresponding trajectory data at the same time point. The association strength evolution layer is responsible for calculating the evolution of the association strength between different feature trajectories over time, obtaining the changing patterns of association strength through analysis and processing of the trajectory data. The parameter feedback layer adjusts the network's connection weights based on the evolution data of association strength, enabling the network to better learn and adapt to the association relationships between different feature trajectories, ultimately outputting trajectory association evolution parameters.

[0123] When initializing the trajectory association evolution network structure, the specific structure and function of each layer can be defined. For example, the trajectory alignment layer can use interpolation or sampling methods to align the time axis; the association strength evolution layer can use neural networks or graph neural networks to calculate the association strength; and the parameter feedback layer can use optimization algorithms such as gradient descent to adjust the network connection weights.

[0124] For example, the trajectory alignment layer can use linear interpolation to calculate the corresponding values ​​for missing data points on the common time axis for different feature trajectories. The association strength evolution layer can use a three-layer fully connected neural network, taking the time-aligned feature trajectory data as input and outputting the association strength between different feature trajectories. The parameter feedback layer can use the stochastic gradient descent algorithm to adjust the connection weights of the neural network based on the evolving association strength data.

[0125] Step S320: Input the dynamic feature trajectory set into the trajectory alignment layer, perform timestamp deviation compensation processing, and obtain a time-aligned trajectory set. The time-aligned trajectory set contains the trajectory data of each feature trajectory on a unified time axis.

[0126] Timestamp deviation compensation is a process of unifying the time axis of a dynamic feature trajectory set. By compensating for timestamp deviations, different feature trajectories have corresponding trajectory data at the same time point, forming a time-aligned trajectory set.

[0127] When performing timestamp offset compensation, it is necessary to consider the potential deviations in timestamp sequences for different characteristic trajectories. This can be achieved by determining a baseline timestamp sequence, calculating the time deviation between other timestamp sequences and the baseline timestamp sequence, and then correcting the characteristic trajectories based on the deviation.

[0128] In one implementation, step S320 may specifically include the following steps S321 to S326:

[0129] Step S321: Extract the timestamp sequence of each feature trajectory from the dynamic feature trajectory set to obtain the spatial timestamp sequence, signal timestamp sequence, motion timestamp sequence, first cross timestamp sequence, and second cross timestamp sequence.

[0130] The spatial timestamp sequence, signal timestamp sequence, motion timestamp sequence, first cross-interference timestamp sequence, and second cross-interference timestamp sequence are timestamp sequences extracted from the spatial trajectory cluster, signal drift trajectory, motion direction trajectory, first cross-interference associated trajectory, and second cross-interference associated trajectory in the dynamic feature trajectory set, respectively. These timestamp sequences record the time information of each feature trajectory at different times. When extracting the timestamp sequences, the dynamic feature trajectory set can be traversed, and for each feature trajectory, its contained timestamp data can be extracted and arranged in chronological order to form the corresponding timestamp sequence.

[0131] Step S322: Determine the reference timestamp sequence, which is the timestamp sequence with the highest sampling frequency among all timestamp sequences.

[0132] A reference timestamp sequence is a sequence used to unify other timestamp sequences. Selecting the timestamp sequence with the highest sampling frequency as the reference timestamp sequence can ensure higher accuracy when unifying the timeline.

[0133] When determining the baseline timestamp sequence, the sampling frequencies of all timestamp sequences can be compared, and the sequence with the highest sampling frequency can be selected as the baseline timestamp sequence. The sampling frequency can be obtained by calculating the reciprocal of the time interval between adjacent timestamps.

[0134] Step S323: For each non-reference timestamp sequence, calculate its time deviation from the reference timestamp sequence. The time deviation includes the difference between each timestamp in the non-reference timestamp sequence and the most recent timestamp in the reference timestamp sequence.

[0135] Time deviation is an indicator that measures the time difference between a non-reference timestamp sequence and a reference timestamp sequence. By calculating the time deviation, we can understand the time offset of each non-reference timestamp sequence relative to the reference timestamp sequence.

[0136] When calculating the time skew, the non-baseline timestamp sequence can be traversed. For each timestamp, the nearest timestamp in the baseline timestamp sequence is found, and the difference between them is calculated. These differences are then arranged in chronological order to form the time skew sequence.

[0137] For example, for non-reference timestamp sequences and reference timestamp sequence For timestamps Find the most recent timestamp in the baseline timestamp sequence. Calculate the difference Similarly, calculation and The time deviation sequence is obtained by comparing the time deviation with the most recent timestamp in the reference timestamp sequence. .

[0138] Step S324: Construct a deviation compensation function based on the time deviation amount. The deviation compensation function contains the mapping relationship data between the time deviation amount and the feature trajectory correction amount.

[0139] The deviation compensation function is used to correct the feature trajectory corresponding to a non-reference timestamp sequence. It establishes a mapping relationship between the time deviation and the correction amount of the feature trajectory. Through this function, the feature trajectory can be corrected according to the time deviation, so that the non-reference timestamp sequence is aligned with the reference timestamp sequence in time.

[0140] When constructing the deviation compensation function, either a linear function or a nonlinear function can be used. For example, when using a linear function, the deviation compensation function can be y = kx + b, where x is the time deviation, y is the characteristic trajectory correction, and k and b are coefficients to be determined. These coefficients can be determined using methods such as the least squares method, so that the deviation compensation function can fit the relationship between the time deviation and the characteristic trajectory correction well.

[0141] Step S325: Correct the feature trajectory corresponding to the non-reference timestamp sequence based on the deviation compensation function to obtain the corrected feature trajectory, which includes trajectory data after compensating for time deviation.

[0142] The corrected feature trajectories are obtained by correcting the feature trajectories corresponding to non-reference timestamp sequences according to a deviation compensation function. This correction aligns these feature trajectories temporally with the reference timestamp sequence, facilitating subsequent analysis and calculations.

[0143] When correcting feature trajectories, the feature trajectories corresponding to non-reference timestamp sequences can be traversed. For each data point, the correction amount is calculated using a deviation compensation function based on its corresponding time deviation. This correction amount is then applied to the data point to obtain the corrected data point. All corrected data points are arranged in chronological order to form the corrected feature trajectory.

[0144] Step S326: Combine the feature trajectory corresponding to the base timestamp sequence with all the corrected feature trajectories to obtain a time-aligned trajectory set.

[0145] A time-aligned trajectory set is a collection obtained by combining the feature trajectories corresponding to the baseline timestamp sequence with all corrected feature trajectories. This combination ensures that all feature trajectories have corresponding trajectory data on a unified time axis, facilitating subsequent correlation strength calculation and analysis. During the combination, the feature trajectories corresponding to the baseline timestamp sequence and the corrected feature trajectories can be arranged in chronological order to form a unified trajectory set.

[0146] Step S330: Input the time-aligned trajectory set into the association strength evolution layer and calculate the initial association strength matrix. The initial association strength matrix contains the association strength values ​​of feature points corresponding to different feature trajectories at the same time stamp.

[0147] The initial association strength matrix describes the association strength between corresponding feature points of different feature trajectories at the same time stamp. Calculating this matrix reveals the initial association relationships between different feature trajectories, providing a foundation for subsequent analysis of the evolution of association strength. When calculating the initial association strength matrix, the association strength evolution layer analyzes the data in the time-aligned trajectory set. Various methods can be used to calculate association strength, such as the correlation coefficient method and the mutual information method. Taking the correlation coefficient method as an example, for feature points of different feature trajectories at the same time stamp, the Pearson correlation coefficient can be calculated between them. The closer the value of this coefficient is to 1 or -1, the stronger the linear correlation between the two feature points, and the greater the association strength; the closer the value is to 0, the weaker the correlation. For example, in the time-aligned trajectory set, at time stamp t, there is feature point a of feature trajectory A. t and feature point b of feature trajectory B t First, calculate the mean values ​​a' and b' of feature trajectories A and B at multiple timestamps. Then, calculate the difference between each feature point and the mean value, i.e., a'. t -a' and b t Next, the sum of the products of these differences is calculated and then divided by the product of the number of feature points and their respective standard deviations. This yields the Pearson correlation coefficient between the feature points of feature trajectories A and B at time stamp t. This coefficient is used as the correlation strength between them at that time stamp. Performing this calculation for all combinations of time stamps and different feature trajectories yields the initial correlation strength matrix.

[0148] Step S340: Perform intensity evolution modeling based on the initial correlation strength matrix to generate intensity evolution curves, which contain the curve data of the change of each correlation strength value over time.

[0149] Intensity evolution modeling involves analyzing and processing the data in the initial correlation intensity matrix to reveal the changing patterns of correlation intensity between different characteristic trajectories over time. By generating intensity evolution curves, the dynamic changes in correlation intensity can be observed intuitively, providing important data for monitoring and predicting aircraft status.

[0150] In one implementation, step S340 may specifically include the following steps S341 to S346:

[0151] Step S341: Extract the timestamp corresponding to each association strength value from the initial association strength matrix to obtain a set of strength-time data pairs. The set of strength-time data pairs contains multiple data pairs consisting of timestamps and corresponding association strength values.

[0152] The intensity-time data pair set is obtained by pairing the correlation intensity values ​​in the initial correlation intensity matrix with their corresponding timestamps. This set clearly shows at which timestamp each correlation intensity value was calculated, laying the foundation for subsequent time series analysis.

[0153] When extracting intensity-time data pairs, it is necessary to iterate through each element of the initial association strength matrix, recording the association strength value of that element and its corresponding timestamp. For example, for element C in the i-th row and j-th column of the initial association strength matrix... ij Its corresponding timestamp is t ij This forms an intensity-time data pair (t ij C ij By collecting all such data pairs, we obtain the intensity-time data pair set.

[0154] Step S342: Perform time series segmentation on the intensity-time data pair set, dividing the continuous timestamps into multiple evolution periods, each evolution period containing a preset number of intensity-time data pairs.

[0155] Time series segmentation is used to better analyze the evolution of association strength by dividing consecutive timestamps into multiple relatively independent evolution periods. Each evolution period contains a predetermined number of intensity-time data pairs, allowing for local analysis of changes in association strength within each period.

[0156] The preset quantity needs to be adjusted based on the specific application scenario and data characteristics. If the preset quantity is too small, it may result in insufficient data information for each time period, failing to accurately reflect the changing patterns of correlation strength; if the preset quantity is too large, it may mask some local variation characteristics.

[0157] Step S343: For each evolution period, calculate the mean, variance, and maximum value of the correlation strength values ​​within that evolution period to obtain the period characteristic values.

[0158] Time-period eigenvalues ​​are a set of numerical values ​​describing the variation characteristics of association strength within each evolutionary period, including the mean, variance, and maximum value. The mean reflects the average level of association strength within that period; the variance reflects the dispersion of association strength within that period, with a larger variance indicating greater fluctuations in association strength; and the maximum value represents the highest possible association strength within that period, reflecting the peak value that association strength may reach within that period.

[0159] To calculate the mean, sum all association strength values ​​within the evolution period and then divide by the number of data pairs. To calculate the variance, first square the difference between each association strength value and the mean, then average these squared values. The maximum value is obtained by comparing all association strength values ​​within the period.

[0160] Step S344: Construct a feature sequence based on the time period feature values ​​of all evolution periods. The feature sequence contains time period feature value data arranged in chronological order.

[0161] The feature sequence is a sequence obtained by arranging the time-series feature values ​​of all evolution periods in chronological order. This sequence can comprehensively show the characteristic changes in association strength across different evolution periods, facilitating subsequent trend analysis and modeling. When constructing the feature sequence, the mean, variance, and maximum value of each time period are arranged sequentially according to the chronological order of the evolution periods.

[0162] Step S345: Perform polynomial fitting on the feature sequence to obtain the fitted polynomial. The independent variable of the fitted polynomial is the evolution period number, and the dependent variable is the period feature value.

[0163] Polynomial fitting approximates the data variation patterns in a feature sequence by finding a suitable polynomial. The independent variable of the fitting polynomial is the evolution period number, and the dependent variable is the period feature value. This allows a mathematical function to describe the variation of the period feature value with the evolution period. Different polynomial orders can be chosen when performing polynomial fitting. A higher order results in a more accurate fit to the data points, but it may also lead to overfitting, causing a decrease in the model's generalization ability on new data. Appropriate polynomial orders are typically selected using methods such as cross-validation.

[0164] For example, choose the quadratic polynomial y=ax 2 The polynomial curve is fitted using the formula +bx+c, where x is the evolution period number and y is the period characteristic value (such as mean, variance, or maximum). The coefficients a, b, and c are solved using the least squares method to minimize the sum of squared errors between the polynomial curve and the data points in the characteristic sequence.

[0165] Step S346: Generate a continuous intensity evolution curve based on the fitted polynomial. The intensity evolution curve contains the continuous change path data of the associated intensity values ​​over the evolution period.

[0166] The intensity evolution curve is generated based on a fitted polynomial, showing the continuous change path of the correlation intensity value over the evolution period. Through this curve, the changing trend of the correlation intensity throughout the monitoring process can be observed intuitively, such as whether it is increasing, decreasing, or fluctuating.

[0167] When generating intensity evolution curves, different evolution period numbers are substituted into the fitting polynomial to calculate the corresponding period characteristic values. These values ​​are then connected in order of evolution period numbers to obtain continuous intensity evolution curves.

[0168] Step S350: Extract the evolutionary characteristic parameters of the intensity evolution curve. The evolutionary characteristic parameters include the slope change rate of the curve, the peak occurrence time, and the decay period data.

[0169] Evolutionary characteristic parameters are parameters that describe the characteristics of the intensity evolution curve, and can help understand the evolution law of the correlation intensity and the changes in the aircraft state. The rate of change of the slope of the curve reflects the speed of the correlation intensity change; the moment of peak occurrence indicates the time when the correlation intensity reaches its maximum value, which may correspond to a key change in the aircraft state; the decay period data reflects the periodic law of the correlation intensity decaying after reaching the peak.

[0170] The rate of change of the slope of the curve can be obtained by differentiating the intensity evolution curve. For discrete data points, numerical differentiation methods can be used, such as calculating the slope change between adjacent data points. The peak occurrence time can be found by traversing the data points of the intensity evolution curve to find the timestamp corresponding to the maximum value. The extraction of decay period data can be achieved using methods such as spectral analysis to analyze the change pattern of the correlation intensity over time after reaching the peak, and to identify the periodic decay characteristics.

[0171] For example, for the data point sequence [y1, y2, ..., y] of the intensity evolution curve n ] Calculate the slope k of adjacent data points i =(y i+1 -y i ) / (x i+1 -x i (where x) i (This is the evolution period number), and then the rate of change of adjacent slopes is calculated to obtain the rate of change of the curve's slope. By comparing the magnitudes of the data points, the maximum value y is found. max Corresponding evolutionary period number x peak This is the moment when the peak occurs. Spectral analysis methods such as Fourier transform are used to analyze the data after the peak is reached to identify the decay period.

[0172] Step S360: Input the evolutionary feature parameters into the parameter feedback layer, adjust the node connection weights of the association strength evolution layer, and obtain the adjusted association strength matrix.

[0173] The role of the parameter feedback layer is to adjust the node connection weights of the correlation strength evolution layer according to the evolutionary characteristic parameters, so that the network can better adapt to the evolution law of correlation strength and improve the accuracy of aircraft status monitoring.

[0174] When adjusting node connection weights, optimization algorithms such as gradient descent can be used. Gradient descent calculates the gradient of the loss function with respect to the node connection weights, and then updates the weights in the opposite direction of the gradient, gradually decreasing the value of the loss function. The loss function can be defined based on evolutionary feature parameters, such as the difference between the slope change rate, peak occurrence time, and decay period and the expected values. After multiple iterative updates, the adjusted association strength matrix is ​​obtained.

[0175] Step S370: Regenerate the intensity evolution curve based on the adjusted correlation intensity matrix, and repeat the steps of extracting evolution feature parameters and adjusting node connection weights until the stability index of the intensity evolution curve meets the preset stability conditions. Use the evolution feature parameters at this time as the trajectory correlation evolution parameters.

[0176] Repeating the above steps is to continuously optimize the model of the correlation strength evolution layer, making the strength evolution curve more stable and able to accurately reflect the evolution law of the correlation strength between different feature trajectories. The stability index is the standard for measuring whether the strength evolution curve is stable, and the preset stability condition is a pre-set threshold for the stability index.

[0177] In one implementation, step S370 may specifically include the following steps S371 to S377:

[0178] Step S371: Extract the stability index of the current intensity evolution curve. The stability index includes the curve fluctuation entropy and the trend consistency coefficient. The curve fluctuation entropy is calculated by the probability distribution entropy value of the curve within a preset time window. The trend consistency coefficient is calculated by the degree of agreement between the actual change trend of the curve and the preset benchmark trend.

[0179] Curve fluctuation entropy is an indicator that measures the degree of fluctuation of the intensity evolution curve within a preset time window, reflecting the uncertainty and randomness of the curve. It is calculated by taking the probability distribution entropy value of the curve within the preset time window; the larger the entropy value, the more complex the fluctuation of the curve and the higher the uncertainty. The trend consistency coefficient, on the other hand, assesses the degree of matching between the actual trend of the curve and a preset benchmark trend. The preset benchmark trend can be an ideal trend of correlation strength determined based on historical data or prior knowledge. The trend consistency coefficient is obtained by comparing the actual trend of the curve (e.g., rising, falling, or stable) with the preset benchmark trend and calculating the degree of agreement between the two.

[0180] For example, for intensity evolution curve data points within a preset time window, the values ​​of the data points are divided into several intervals, the frequency of data points appearing in each interval is counted, and then the curve fluctuation entropy is calculated according to the general formula for calculating information entropy. For the trend consistency coefficient, the actual trend of the curve can be represented by a series of symbols (such as "+" for rising, "-" for falling, and "0" for stable). The preset baseline trend is also represented by the same symbols. Then the proportion of the same symbols in both is calculated as the trend consistency coefficient.

[0181] Step S372: Set the fluctuation entropy threshold and the consistency coefficient threshold. The fluctuation entropy threshold is the maximum allowable curve fluctuation value, and the consistency coefficient threshold is the minimum allowable trend consistency value.

[0182] The fluctuation entropy threshold and the consistency coefficient threshold are specific quantitative indicators of the preset stability conditions. The fluctuation entropy threshold specifies the maximum allowable degree of curve fluctuation. If the curve fluctuation entropy exceeds this threshold, it indicates that the curve fluctuation is too drastic, and the model needs further adjustment. The consistency coefficient threshold specifies the minimum degree of agreement between the actual trend of the curve and the preset benchmark trend. If the trend consistency coefficient is lower than this threshold, it indicates that the trend of the curve differs significantly from the expectation, and the model also needs adjustment.

[0183] The threshold values ​​for fluctuation entropy and consistency coefficient need to be determined based on the specific application scenario and data characteristics. A suitable threshold range can be found through analysis and experimentation with a large amount of historical data, allowing the model to accurately reflect the evolution of correlation strength while ensuring stability.

[0184] Step S373: Determine whether the curve fluctuation entropy is less than the fluctuation entropy threshold and whether the trend consistency coefficient is greater than the consistency coefficient threshold.

[0185] This step assesses the stability of the intensity evolution curve by comparing the relationship between the curve fluctuation entropy and the fluctuation entropy threshold, and the trend consistency coefficient and the consistency coefficient threshold, to determine whether the curve meets the preset stability conditions.

[0186] If the curve fluctuation entropy is less than the fluctuation entropy threshold and the trend consistency coefficient is greater than the consistency coefficient threshold, it indicates that the fluctuation of the curve is within the allowable range, and the actual change trend is highly consistent with the preset benchmark trend, indicating that the curve has good stability.

[0187] Step S374: When the curve fluctuation entropy is less than the fluctuation entropy threshold and the trend consistency coefficient is greater than the consistency coefficient threshold, the stability index of the intensity evolution curve is determined to meet the preset stability condition.

[0188] When the judgment condition in step S373 is met, the stability index of the intensity evolution curve is considered to have met the preset requirements. At this point, the adjustment of the connection weights of the nodes in the correlation intensity evolution layer can be stopped, and the current evolution characteristic parameters can be used as the trajectory correlation evolution parameters. This means that the model can adapt well to the evolution law of correlation intensity and accurately reflect the correlation between different characteristic trajectories, providing reliable parameters for the inference of the aircraft's state trend.

[0189] Step S375: When the curve fluctuation entropy is greater than or equal to the fluctuation entropy threshold, adjust the time window length of the correlation intensity evolution layer, increase the curve smoothness, and then regenerate the intensity evolution curve.

[0190] If the curve's fluctuation entropy is greater than or equal to the fluctuation entropy threshold, it indicates that the curve's fluctuations are too drastic, requiring adjustment of the correlation strength evolution layer. Adjusting the time window length is an effective method; by increasing the time window length, data can be averaged over a larger time range, thereby increasing the curve's smoothness and reducing fluctuations.

[0191] When adjusting the time window length, the window size can be increased appropriately, for example, increasing the original time window length from 10 timestamps to 15 timestamps. Then, based on the adjusted time window, the intensity-time data pairs are re-segmented, and steps S340-S370 are repeated to regenerate the intensity evolution curve.

[0192] Step S376: When the trend consistency coefficient is less than or equal to the consistency coefficient threshold, adjust the node activation function parameters of the correlation strength evolution layer to enhance the trend capture capability and then regenerate the strength evolution curve.

[0193] If the trend consistency coefficient is less than or equal to the consistency coefficient threshold, it indicates that the actual trend of the curve does not match the preset baseline trend well. It is necessary to adjust the node activation function parameters of the correlation strength evolution layer to enhance the network's ability to capture the correlation strength evolution trend.

[0194] Node activation functions are functions used in neural networks to introduce non-linear factors. Common activation functions include the Sigmoid function and the ReLU function. By adjusting the parameters of the activation function, such as adjusting the slope of the ReLU function, the output characteristics of the nodes can be changed, enabling the network to better learn and capture the changing trends in the strength of associations.

[0195] Step S377: Repeat the steps of generating the intensity evolution curve and judging the stability index until the curve fluctuation entropy is less than the fluctuation entropy threshold and the trend consistency coefficient is greater than the consistency coefficient threshold.

[0196] This step is an iterative process. By continuously adjusting the parameters of the associated intensity evolution layer, the intensity evolution curve is regenerated, and the stability index is judged until the intensity evolution curve meets the preset stability conditions.

[0197] In each iteration, based on the curve fluctuation entropy and trend consistency coefficient, an appropriate adjustment method is selected (such as adjusting the time window length or node activation function parameters). Then, the process of steps S340-S370 is repeated to continuously optimize the model, so that the model can accurately reflect the evolution law of the correlation strength between different feature trajectories.

[0198] For example, after the first adjustment, if the curve fluctuation entropy is still greater than the fluctuation entropy threshold and the trend consistency coefficient is less than the consistency coefficient threshold, the time window length of the correlation intensity evolution layer and the node activation function parameters are adjusted again to regenerate the intensity evolution curve. The stability index is then evaluated again until the stability condition is met.

[0199] Step S400: Perform state trend bifurcation deduction based on trajectory association evolution parameters to obtain state trend bifurcation results. The state trend bifurcation results include multiple possible state trend data generated based on different association evolution paths.

[0200] State trend bifurcation extrapolation uses trajectory correlation evolution parameters to predict and analyze the future state trend of an aircraft. Since the state of an aircraft is influenced by various factors, different correlation evolution paths may lead to different state development trends. Therefore, bifurcation extrapolation can obtain multiple possible state trend data, providing more comprehensive information for aircraft state monitoring and decision-making.

[0201] In one implementation, step S400 may specifically include the following steps S410-S480:

[0202] Step S410: Determine the starting time point and duration of the simulation. The starting time point is the current monitoring time point, and the simulation duration is the length of the continuous observation period starting from the starting time point.

[0203] The starting time point of the simulation is the starting point for the state trend bifurcation simulation. For example, selecting the current monitoring time point allows the simulation to be based on the latest aircraft state information. The simulation duration specifies the time range of the simulation, which is the length of the continuous observation period starting from the starting time point.

[0204] The simulation duration needs to be adjusted based on the specific application scenario and requirements. If the simulation duration is too short, it may not be able to fully reflect the changing trends of the aircraft's state; if the simulation duration is too long, the accuracy of the simulation results may be affected due to increased future uncertainties.

[0205] Step S420: Construct a trend bifurcation decision tree based on trajectory association evolution parameters. The trend bifurcation decision tree contains a multi-path trend prediction structure with the association strength evolution law as the branching condition.

[0206] Trend bifurcation decision trees are models used for state trend bifurcation inference. They construct a multi-path trend prediction structure based on the evolution of association strength as the branching condition. In the decision tree, each node represents a state, and each branch represents a possible association strength evolution path. By inferring along different branches, different state trends can be obtained.

[0207] When constructing a trend bifurcation decision tree, branching conditions are determined based on information such as the slope change rate, peak occurrence time, and decay period in the trajectory association evolution parameters. For example, if the slope change rate is greater than a certain threshold, one branch may be selected for further analysis; if the peak occurrence time is within a specific time range, another branch may be selected.

[0208] Step S430: Use the dynamic feature trajectory data of the current monitoring time point as the initial input, input it into the trend bifurcation decision tree, and generate multiple initial trend paths. Each initial trend path contains one possible state evolution direction data.

[0209] Using the dynamic characteristic trajectory data of the current monitoring time point as the initial input allows the trend bifurcation decision tree to extrapolate from the current aircraft state. By inputting this data, the decision tree will calculate along different branches based on the evolution of correlation strength and branching conditions, generating multiple initial trend paths.

[0210] Each initial trend path represents a possible direction of state evolution, reflecting the possible development trend of the aircraft state under different correlation strength evolution paths.

[0211] For example, the dynamic trajectory data at the current monitoring time point includes information such as airspace location, signal spectrum, and direction of motion. This data is input into a trend bifurcation decision tree. Based on the evolution of correlation strength, the decision tree generates three initial trend paths, representing different possible flight directions of the aircraft, different possible changes in the signal spectrum, etc.

[0212] Step S440: Calculate the path confidence for each initial trend path. The path confidence includes the probability of the path occurring, determined based on the trajectory association evolution parameters.

[0213] Path confidence is an indicator that measures the probability of each initial trend path occurring, and it is determined based on trajectory association evolution parameters. Information such as the rate of change of slope, the time of peak occurrence, and the decay period in trajectory association evolution parameters can reflect the stability and reliability of evolution paths with different association strengths, thus affecting the probability of each path occurring.

[0214] When calculating path confidence, probabilistic models, such as Bayesian probabilistic models, can be used. By analyzing historical data and learning trajectory correlation evolution parameters, the probability of each initial trend path occurring under different conditions can be estimated.

[0215] For example, a Bayesian probability model can be established based on historical data and trajectory correlation evolution parameters. For a certain initial trend path, considering factors such as the rate of change of the slope of the correlation strength and the time of peak occurrence, the probability of the path occurring is calculated to be 0.6, i.e., the path confidence level is 0.6.

[0216] Step S450: Select initial trend paths with a path confidence greater than a preset confidence threshold as candidate trend paths. The candidate trend paths contain data on the direction of state evolution with a high probability.

[0217] The preset confidence threshold is a pre-defined probability value used to filter out high-probability initial trend paths. Initial trend paths with a confidence level greater than the preset confidence threshold are considered candidate trend paths, as these paths are more likely to represent the actual future development trend of the aircraft. The preset confidence threshold is determined based on the specific application scenario and the accuracy requirements.

[0218] Step S460: Evaluate the path stability of each candidate trend path to obtain the path stability, which includes the fluctuation data of the trend path during the simulation period.

[0219] Path stability assessment aims to further filter out more reliable trend paths. By evaluating the degree of fluctuation of each candidate trend path over the simulation period, path stability is obtained. The higher the path stability, the more stable the trend path is over the simulation period, and the more likely it is to represent the actual development trend of the aircraft.

[0220] In one implementation, step S460 may specifically include the following steps S461-S466:

[0221] Step S461: Extract the state feature values ​​of all timestamps from the candidate trend path to obtain the path feature sequence, which contains state feature values ​​arranged in chronological order.

[0222] State feature values ​​are a set of numerical values ​​describing the state of an aircraft, such as its airspace position, signal spectrum intensity, and velocity. State feature values ​​for all timestamps are extracted from the candidate trend path and arranged chronologically to form a path feature sequence. For example, a candidate trend path may contain 10 timestamps within the simulation duration, each corresponding to a state feature value, such as the x-coordinate of its airspace position. Arranging these 10 state feature values ​​chronologically yields the path feature sequence [x1, x2, ..., x...]. 10 ].

[0223] Step S462: Calculate the first-order difference sequence of the path feature sequence, which contains the difference data of adjacent timestamp state feature values.

[0224] The first-order difference sequence is obtained by processing the path feature sequence, reflecting the changes in state feature values ​​between adjacent time stamps. By calculating the first-order difference sequence, the degree of fluctuation of state feature values ​​between adjacent time points can be understood. For the path feature sequence [x1, x2, ..., x...], ... n ], first-order difference sequence Δx i =x i+1 -x i (i=1,2,……,n-1).

[0225] Step S463: Calculate the standard deviation of the first-order difference sequence to obtain the volatility index, which contains the overall volatility data of the path characteristic sequence.

[0226] Standard deviation is a statistic that measures the dispersion of data. By calculating the standard deviation of a first-difference sequence, we can obtain the overall volatility of the path characteristic sequence, i.e., the volatility index. The larger the volatility index, the more drastic the fluctuation of the path characteristic sequence, and the worse the stability of the path.

[0227] Step S464: Perform sliding window smoothing on the path feature sequence to obtain a smoothed feature sequence, which contains state feature values ​​after eliminating short-term fluctuations.

[0228] Sliding window smoothing works by sliding a fixed-size window across the path feature sequence and averaging the data within the window, thereby eliminating short-term fluctuations and obtaining a smooth feature sequence. The choice of window size needs to be adjusted according to the specific application scenario and data characteristics. A larger window results in a better smoothing effect, but may mask some local variations; a smaller window results in a weaker smoothing effect, but retains more detailed information.

[0229] Step S465: Calculate the rate of change of curvature of the smooth feature sequence to obtain the curvature index, which contains data on the change in the degree of curvature of the smooth feature sequence.

[0230] The rate of change of curvature reflects the degree of curvature variation in a smooth feature sequence. By calculating the curvature index, we can further understand the stability of the path. The larger the rate of change of curvature, the more drastic the change in the curvature of the path, and the worse the path stability may be.

[0231] When calculating the rate of change of curvature, the second derivative of the smoothed feature sequence can be taken to obtain the change in curvature. For discrete data points, numerical differentiation can be used for approximate calculation.

[0232] Step S466: Input the volatility index and curvature index into the preset stability evaluation function to obtain the path stability. The preset stability evaluation function contains the weighted combination relationship data of the volatility index and curvature index.

[0233] The preset stability evaluation function is a function used to comprehensively evaluate path stability. It combines the volatility index and curvature index in a weighted manner to obtain a comprehensive path stability value.

[0234] The weights in the weighted combination need to be adjusted based on the specific application scenario and the degree of emphasis placed on volatility and curvature. For example, the preset stability evaluation function can be expressed as S. p =w1S f +w2S c S p For path stability, S f S is a volatility indicator. c The curvature index is represented by w1 and w2, which are weights, and w1 + w2 = 1.

[0235] Step S470: Sort the candidate trend paths based on path stability, and select the top N candidate trend paths with the highest stability, where N is an integer greater than 1.

[0236] Ranking candidate trend paths based on path stability allows for the selection of more stable and reliable paths. The top N most stable candidate trend paths are chosen, as these are more likely to represent the actual future development trend of the aircraft, providing more valuable information for subsequent analysis and decision-making. The value of N needs to be determined based on the specific application scenario and the requirements for comprehensiveness and accuracy of information. If N is too small, some important trend information may be missed; if N is too large, it will increase the complexity of subsequent analysis.

[0237] As one implementation method, after selecting the top N candidate trend paths with the highest stability in step S470, the following steps S471-S477 may also be included:

[0238] Step S471: Calculate the trend similarity between any two paths in the first N candidate trend paths. The trend similarity is calculated by the mean of the cosine similarity of the state feature values ​​of the two paths at the same time stamp.

[0239] Trend similarity is an indicator that measures the degree of similarity between two candidate trend paths. It is obtained by calculating the mean cosine similarity of the state feature values ​​of the two paths at the same time stamp. Cosine similarity is a commonly used vector similarity measure. It measures the similarity between two vectors by calculating the cosine of the angle between them. The closer the value is to 1, the more similar the two vectors are.

[0240] For any two paths A and B among the first N candidate trend paths, at each same timestamp, their state feature values ​​are combined into vectors Va and Vb, and the cosine similarity is calculated. Then, the average of the cosine similarities over all timestamps is taken to obtain the trend similarity between the two paths.

[0241] Step S472: Cluster the top N candidate trend paths based on trend similarity to obtain at least one trend cluster. Each trend cluster contains candidate trend paths with a trend similarity greater than a preset similarity threshold.

[0242] Clustering is the process of grouping similar candidate trend paths into the same cluster. By clustering based on trend similarity, the top N candidate trend paths can be divided into several trend clusters. A preset similarity threshold is used to determine whether two paths belong to the same cluster. When the trend similarity of two paths is greater than this threshold, they are grouped into the same trend cluster. The preset similarity threshold needs to be adjusted according to the specific application scenario and the required clustering accuracy. If the threshold is set too high, it may result in too few paths in each cluster, leading to insignificant clustering results; if the threshold is set too low, it may group some highly different paths into the same cluster.

[0243] By using a clustering algorithm (such as hierarchical clustering), paths with a trend similarity greater than 0.8 are grouped into the same cluster, resulting in two trend clusters: Cluster 1 contains paths 1, 2, and 4; Cluster 2 contains paths 3 and 5.

[0244] Step S473: Select the candidate trend path with the highest path confidence from each trend cluster as the cluster representative path to obtain the cluster representative path set.

[0245] The cluster representative path is a representative path selected from each trend cluster. Selecting the candidate trend path with the highest path confidence as the cluster representative path can ensure that the path in each cluster can best represent the trend characteristics of that cluster.

[0246] Within each trend cluster, the path confidence of each path is compared, and the path with the highest confidence is selected as the cluster representative path. All cluster representative paths are combined to obtain the cluster representative path set.

[0247] Step S474: Perform conflict detection on the paths in the cluster representative path set, and identify conflict time segments where the difference in state feature values ​​is greater than a preset difference threshold. The conflict time segments contain time interval data where the state feature values ​​of different cluster representative paths do not match at the same timestamp.

[0248] Conflict detection aims to identify conflicts between different paths within a set of cluster-representing paths. A conflict is considered to exist when the difference in state characteristic values ​​between different cluster-representing paths at the same timestamp exceeds a preset difference threshold. Conflict time segments record the time intervals in which these conflicts occur, providing a basis for subsequent processing. The preset difference threshold is adjusted based on the specific application scenario and the tolerance for conflicts.

[0249] For example, for paths P1 and P2 in the cluster-representing path set, the state feature values ​​at timestamps t1, t2, and t3 are respectively (x 11 ,x 12 ,x 13 ) and (y 11 ,y 12 ,y 13 With a preset difference threshold of 2, the difference in state feature values ​​at each timestamp is calculated. If at timestamp t2, If the value is greater than 2, then a conflict is considered to exist at timestamp t2, and the conflict time segment is recorded as [t2, t2].

[0250] Step S475: For each conflict time segment, calculate the weighted state feature value based on the path stability and path confidence of the cluster representative path. The weighted state feature value is calculated by dividing the sum of the products of the state feature value of each path and the corresponding path stability and path confidence by the sum of the products of the path stability and path confidence.

[0251] Weighted state feature values ​​are used to address the mismatch in state feature values ​​between different clusters representing paths in conflict time segments. They are calculated by weighting the state feature values ​​of each path in conjunction with path stability and path confidence. This approach comprehensively considers both path stability and confidence, resulting in a more reasonable set of state feature values.

[0252] For each timestamp within a conflict time segment, there are m clusters representing paths, and the state feature value of the i-th path is x. i The path stability is s i The path confidence is c. i Then the weighted state eigenvalues .

[0253] Step S476: Replace the original state feature values ​​within the conflict time segment with weighted state feature values ​​to obtain the fusion trend path.

[0254] The fused trend path is obtained by replacing the original state feature values ​​within conflicting time segments with weighted state feature values. This method can resolve conflicts and merge the differences between paths representing different clusters, resulting in a more consistent trend path. At each timestamp within a conflicting time segment, the original state feature values ​​are replaced with weighted state feature values, while the state feature values ​​within non-conflicting time segments remain unchanged, thus obtaining the fused trend path.

[0255] For example, for the conflict time segment [t2, t2], the original state characteristic value is (x 12 ,x 22 The calculated weighted state characteristic value is x. w , will x 12 and x 22 Replace with x w This yields the state feature value of the fusion trend path at that timestamp.

[0256] Step S477: Combine the fused trend path with the cluster representative path of the non-conflicting time segment to update the state trend bifurcation result.

[0257] By combining the fused trend path with the cluster representative path of non-conflicting time segments to form a complete trend path, the state trend bifurcation result is updated. This yields a more accurate and consistent aircraft state trend prediction result, providing a more reliable basis for aircraft state monitoring and decision-making.

[0258] For example, the state feature values ​​of the cluster representative path of the non-conflicting time segment at timestamps t1 and t3 are respectively (x 11 ,x 21 ) and (x 13 ,x 23 The state characteristic value of the fusion trend path in the conflict time segment t2 is x. w Combining them together yields the updated state trend bifurcation result [(x 11 ,x 21 ),x w ,(x 13 ,x 23 )).

[0259] Step S480: Merge the top N candidate trend paths and their corresponding path confidence to obtain the state trend bifurcation result.

[0260] The fusion of the top N candidate trend paths and their corresponding path confidence scores aims to comprehensively consider all possible state trends. By fusing these paths and confidence scores, a more comprehensive and accurate state trend bifurcation result can be obtained. The fusion can be achieved using methods such as weighted averaging, where the state feature values ​​of each path are weighted according to their path confidence scores to obtain the final state trend bifurcation result.

[0261] Step S500: Output the comprehensive monitoring results of the low-altitude aircraft status based on the status trend bifurcation results.

[0262] The comprehensive monitoring results of low-altitude aircraft status indicate the dominant status trend category and corresponding confidence level data of the aircraft during the monitoring period. The status trend bifurcation results encompass multiple possible aircraft status trends. By analyzing and processing these results, the comprehensive monitoring results of low-altitude aircraft status can be output. The comprehensive monitoring results clearly define the dominant status trend category of the aircraft during the monitoring period, such as normal flight, abnormal flight, and imminent landing, and also provide the corresponding confidence level data for the status trend, providing important reference for aircraft supervision and decision-making.

[0263] When determining the dominant state trend category, the path confidence and characteristics of each trend in the state trend bifurcation results can be used for judgment. For example, if the path confidence of a certain state trend is significantly higher than that of other trends, and its characteristics meet a specific state category definition, then this state category is determined as the dominant state trend category. The corresponding confidence data can be directly adopted from the path confidence of this state trend.

[0264] In an optional implementation, the present invention may further include a process of holographic situation tracking and command and dispatch, specifically, it may further include the following processes:

[0265] An airspace situation model is constructed based on the state trend bifurcation results and a preset airspace environmental parameter set. This preset airspace environmental parameter set includes parameters related to geographical obstacles, electromagnetic interference, and meteorological impacts within the monitoring area. This airspace situation model integrates aircraft state trend data with airspace environmental impact data. The aircraft state trend data includes multiple possible state trend data from the state trend bifurcation results, while the airspace environmental impact data includes data on the limitations of geographical obstacles on flight paths, the attenuation of signal transmission due to electromagnetic interference, and the impact of meteorological conditions on flight performance.

[0266] A real-time update process is performed on the airspace situation model to obtain a dynamically updated airspace situation model. This real-time update process includes adjusting the aircraft state trend data in the model based on newly received low-altitude multimodal observation data strings, and updating the airspace environmental impact data in the model based on external environmental data sources, including real-time meteorological data sent by the meteorological monitoring system and terrain update data sent by the geographic information system. A set of scheduling instructions is generated based on the dynamically updated airspace situation model and a preset set of scheduling rules. The preset set of scheduling rules includes aircraft priority rules, conflict avoidance rules, and path optimization rules. Aircraft priority rules are used to determine the air traffic priority of different types of aircraft; conflict avoidance rules are used to identify potential flight path intersections and generate avoidance strategies; and path optimization rules are used to plan the shortest flight path while meeting safety constraints.

[0267] The scheduling instruction set includes heading adjustment instructions, speed adjustment instructions, and communication frequency adjustment instructions for aircraft with different state trend categories. These instructions are used to schedule and control the aircraft.

[0268] Specifically, the above embodiments involve the process of holographic situation tracking and command and dispatch, including the following steps:

[0269] A spatial situation model is constructed based on the state trend bifurcation results and a preset set of spatial environment parameters:

[0270] The preset airspace environment parameter set is a dataset containing various important environmental information within the monitoring area. Among them, the geographical obstacle parameters cover information such as the location, height, and shape of geographical objects that may obstruct the flight of aircraft, such as mountains, tall buildings, and communication towers. The electromagnetic interference parameters record data such as the location, intensity, and frequency range of various electromagnetic interference sources in the area, which may come from communication base stations, industrial equipment, etc. The meteorological impact parameters include meteorological conditions such as wind speed, wind direction, temperature, air pressure, and precipitation within the monitoring area.

[0271] The airspace situational awareness model is a comprehensive model that integrates aircraft state trend data with airspace environmental impact data. Aircraft state trend data originates from state trend bifurcation results, containing multiple possible state trend data generated based on different correlation evolution paths. This data reflects the aircraft's possible flight state over a future period, such as flight direction, speed changes, and the possibility of anomalies. Airspace environmental impact data consists of the effects of factors such as geographical obstacles, electromagnetic interference, and meteorological conditions on aircraft flight. For example, data on the limitations of geographical obstacles on flight paths indicates which areas the aircraft needs to avoid during flight; data on the attenuation of electromagnetic interference on signal transmission shows the degree to which electromagnetic interference weakens the signal strength between the aircraft and the ground control center at different locations and frequencies; and data on the impact of meteorological conditions on flight performance illustrates how different meteorological conditions, such as strong winds and heavy rain, affect the aircraft's flight speed, stability, and energy consumption.

[0272] When constructing an airspace situation model, the first step is to preprocess the state trend bifurcation results and the preset airspace environmental parameter set. For the state trend bifurcation results, the various possible state trend data are organized according to time sequence and state category to give them a clear structure. For the preset airspace environmental parameter set, a unified coordinate system transformation and data format standardization are performed on geographical obstacle parameters, electromagnetic interference parameters, and meteorological impact parameters to ensure that they can be fused in the same model. Then, data fusion algorithms, such as Kalman filtering and Bayesian fusion algorithms, are used to fuse the aircraft state trend data and airspace environmental impact data. Taking the Kalman filtering algorithm as an example, it can continuously update the state estimate based on the current state estimate and new observation data, thereby achieving accurate modeling of the aircraft state and airspace environmental impact. In this way, the aircraft state trend and airspace environmental factors are organically combined to construct an airspace situation model that can comprehensively reflect the low-altitude flight situation.

[0273] For example, within a city's low-altitude monitoring area, a pre-set set of airspace environmental parameters indicates the presence of several tall buildings (geographic obstacle parameters), a nearby communication base station generating strong electromagnetic interference (electromagnetic interference parameters), and ongoing heavy rain (meteorological impact parameters). State trend bifurcation results suggest the aircraft may have several different flight paths and state trends. By constructing an airspace situation model and integrating this information, it becomes clear how the aircraft is affected by geographical obstacles, electromagnetic interference, and meteorological conditions under different state trends.

[0274] Perform real-time update processing on the airspace situation model to obtain a dynamically updated airspace situation model:

[0275] Real-time updates are a crucial step in ensuring that the airspace situation model accurately reflects the current reality. This process mainly includes two aspects: first, adjusting the aircraft status trend data in the model based on newly received low-altitude multimodal observation data strings; and second, updating the airspace environmental impact data in the model based on external environmental data sources.

[0276] The newly received low-altitude multimodal observation data strings are continuously collected by distributed sensing nodes located on 5G base station towers. These data strings contain information on the aircraft's latest position, speed, attitude, signal characteristics, and other aspects. By analyzing and processing this new data, changes in the aircraft's state can be detected in a timely manner, and corresponding adjustments can be made to the aircraft's state trend data in the airspace situation model. For example, if new data shows a sudden increase in the aircraft's speed, then the speed trend data of the aircraft in the model needs to be updated to reflect this change.

[0277] External environmental data sources are crucial for obtaining the latest airspace environmental information. Meteorological monitoring systems provide real-time meteorological data within the monitored area, such as changes in wind speed, wind direction, temperature, and precipitation. Geographic information systems (GIS) can send updated terrain data, such as information on land use changes and new building construction. Based on this real-time data, the airspace environmental impact data in the airspace situation model is updated. For example, when real-time meteorological data from the meteorological monitoring system shows a sudden increase in wind speed, the model's data on the impact of meteorological conditions on flight performance needs to be updated to reflect the impact of strong winds on aircraft flight stability and energy consumption.

[0278] When performing real-time updates, an efficient data reception and processing mechanism is required. For newly received low-altitude multimodal observation data strings, the data parsing module must parse them into a format suitable for model updates. Then, state prediction algorithms, such as particle filtering and neural network prediction algorithms, are used to re-predict and adjust the aircraft's state trend based on the new data. For data sent from external environmental data sources, the data interface module must integrate it into the airspace situation model, and the corresponding airspace environmental impact data must be updated promptly according to the data type and update frequency. Simultaneously, to ensure the model's real-time performance and accuracy, the update process needs to be monitored and evaluated to ensure that the model accurately reflects the current airspace situation after each update.

[0279] For example, at a certain moment, newly received low-altitude multimodal observation data shows that the aircraft's heading has changed, while real-time meteorological data sent by the meteorological monitoring system indicates new changes in wind speed and direction. At this time, the aircraft status trend data and the impact data of meteorological conditions on flight performance in the airspace situation model are immediately updated to obtain a dynamically updated airspace situation model, which can accurately reflect the current flight status of the aircraft and the airspace environment.

[0280] A set of scheduling instructions is generated based on the dynamically updated airspace situation model and a set of preset scheduling rules.

[0281] The pre-defined scheduling rule set is a system of rules used to regulate aircraft scheduling behavior. Among these, the aircraft priority rules determine the air traffic priority of different aircraft based on factors such as aircraft type and mission importance. For example, aircraft performing emergency rescue missions typically have higher priority, and other aircraft must give way to them during flight. The main function of the conflict avoidance rules is to identify potential intersections between the flight paths of different aircraft in the dynamically updated airspace situation model and generate corresponding avoidance strategies based on these intersections. For example, if it is found that the flight paths of two aircraft may intersect at a certain moment, the conflict avoidance rules will calculate the specific values ​​by which one aircraft needs to change its course or speed based on the aircraft's speed, position, and other information to avoid a collision. The path optimization rules, on the other hand, plan the shortest flight path for aircraft while meeting safety constraints. Safety constraints include avoiding geographical obstacles, avoiding areas with strong electromagnetic interference, and considering the impact of weather conditions on flight. By comprehensively considering these factors and using path planning algorithms such as A* algorithm and Dijkstra's algorithm, a path can be found for the aircraft that ensures both safety and minimizes the flight distance.

[0282] The scheduling instruction set is a set of instructions generated based on the dynamically updated airspace situation model and the preset scheduling rule set. Among them, the heading adjustment instruction is used to instruct the aircraft to change its flight direction to avoid geographical obstacles, avoid conflicts with other aircraft, or follow an optimized flight path; the speed adjustment instruction can control the aircraft's flight speed, for example, to appropriately reduce the speed in strong winds to ensure flight safety, or to increase the speed to reach the destination on time; the communication frequency adjustment instruction is to cope with the impact of electromagnetic interference on signal transmission. When the aircraft enters a strong electromagnetic interference area, the communication frequency is adjusted to ensure the quality of communication with the ground control center.

[0283] When generating the set of scheduling instructions, the dynamically updated airspace situation model is first analyzed to identify potential conflicts and flight paths requiring optimization. Then, scheduling decisions are made for aircraft with different state trend categories based on rules in the preset scheduling rule set. For each aircraft, based on its current state and environmental conditions, it is determined whether a heading adjustment command, speed adjustment command, or communication frequency adjustment command needs to be issued. For example, if the model indicates that an aircraft is about to enter a geographical obstacle area, a heading adjustment command is generated based on conflict avoidance rules and path optimization rules, instructing the aircraft to change its flight direction to avoid the obstacle. Finally, the generated set of scheduling instructions is sent to the corresponding aircraft to achieve scheduling control of the aircraft.

[0284] For example, in a busy low-altitude airspace, a dynamically updated airspace situation model shows that the flight paths of two aircraft are about to intersect, and one of the aircraft is about to enter a region of strong electromagnetic interference. Based on a pre-set set of scheduling rules, a set of scheduling instructions is generated, including issuing a heading adjustment instruction to one of the aircraft to change its flight direction to avoid a collision; and issuing a communication frequency adjustment instruction to the aircraft entering the electromagnetic interference region to adjust its communication frequency to ensure communication quality. Through these scheduling instructions, effective scheduling and control of the aircraft is achieved, ensuring the safety and orderliness of low-altitude flight.

[0285] It is understood that the various algorithms involved in the above descriptions of the embodiments of the present invention can be obtained from relevant content in the prior art. To save space, they will not be elaborated on in the embodiments of the present invention. In addition, those skilled in the art can supplement the details based on common knowledge in the art when implementing the solutions of the present invention. For example, they can use normalization to eliminate dimensional conflicts before feature fusion, use interpolation to eliminate dimensional differences, reasonably set thresholds based on historical data, experience or business scenario requirements, train the model based on a general model training method, set the number of layers in the model structure based on actual needs, select activation functions, etc. The present invention will not provide redundant descriptions of overly detailed implementation processes here.

[0286] Please see Figure 2 , Figure 2This is a schematic diagram of a computer system provided in an embodiment of the present invention. The computer system includes at least a processor 101, a communication interface 102, and a memory 103. The processor 101, communication interface 102, and memory 103 can be connected via a bus or other means. The processor 101 (or Central Processing Unit, CPU) is the computing and control core of the computer system, capable of parsing various instructions and processing various data within the computer system. The communication interface 102 may optionally include a standard wired interface or a wireless interface (such as Wi-Fi, mobile communication interface, etc.), and can be used to send and receive data under the control of the processor 101; the communication interface 102 can also be used for data transmission and interaction within the computer system. The memory 103 is a storage device in the computer system used to store programs and data. It is understood that the memory 103 here can include the computer system's built-in memory, or it can include extended memory supported by the computer system. The memory 103 provides storage space, which stores the computer system's operating system; this invention does not limit this storage space.

[0287] In one embodiment, the processor 101 executes the low-altitude aircraft status monitoring method based on 5G base station towers provided in the above embodiments of the present invention by running a computer program in the memory 103.

Claims

1. A method for monitoring the status of low-altitude aircraft based on 5G base station towers, characterized in that, include: Receive low-altitude multimodal observation data strings collected by distributed sensing nodes deployed on 5G base station towers; The low-altitude multimodal observation data string is processed by multimodal feature trajectory weaving to obtain a dynamic feature trajectory set. Specifically, this includes: separating airspace observation substrings, signal observation substrings, and motion observation substrings from the low-altitude multimodal observation data string; the airspace observation substrings contain time-series arranged three-dimensional coordinate data of the aircraft; the signal observation substrings contain time-series arranged aircraft signal spectrum data; and the motion observation substrings contain time-series arranged aircraft acceleration vector data; performing trajectory point density clustering on the airspace observation substrings to obtain airspace trajectory clusters, which contain a set of airspace observation data points with spatial location correlations; and performing spectral feature drift tracking on the signal observation substrings to obtain signal drift trajectories, which include the drift of signal spectrum features over time. Path data; Acceleration direction coherence analysis is performed on the motion observation substring to obtain the motion direction trajectory, which includes path data showing the coherent change of the acceleration vector direction over time; the cross-correlation degree between the spatial trajectory cluster and the signal drift trajectory is calculated to generate a first cross-correlation trajectory, which includes path data showing the correlation between spatial position change and signal spectrum drift; the cross-correlation degree between the signal drift trajectory and the motion direction trajectory is calculated to generate a second cross-correlation trajectory, which includes path data showing the correlation between signal spectrum drift and motion direction change; the spatial trajectory cluster, the signal drift trajectory, the motion direction trajectory, the first cross-correlation trajectory, and the second cross-correlation trajectory are fused to obtain the dynamic feature trajectory set; Based on the dynamic feature trajectory set, a trajectory association evolution network is constructed to generate trajectory association evolution parameters. These parameters are used to characterize the evolutionary pattern of the association strength between different feature trajectories over time. Based on the trajectory association evolution parameters, state trend bifurcation is deduced to obtain state trend bifurcation results; The comprehensive monitoring results of the low-altitude aircraft status are output based on the state trend bifurcation results.

2. The method as described in claim 1, characterized in that, The step of constructing a trajectory association evolution network based on the dynamic feature trajectory set and generating trajectory association evolution parameters includes: An initial trajectory association evolution network structure is established, which includes a trajectory alignment layer, an association strength evolution layer, and a parameter feedback layer. The trajectory alignment layer is used to receive a dynamic feature trajectory set and perform time axis alignment processing. The association strength evolution layer is used to calculate the evolution data of the association strength between different feature trajectories over time. The parameter feedback layer is used to adjust the network connection weights based on the evolution data and output the evolution parameters. The dynamic feature trajectory set is input into the trajectory alignment layer and timestamp deviation compensation is performed to obtain a time-aligned trajectory set, which contains trajectory data of each feature trajectory on a unified time axis. The time-aligned trajectory set is input into the association strength evolution layer to calculate the initial association strength matrix, which contains the association strength values ​​of feature points corresponding to different feature trajectories at the same timestamp; Intensity evolution modeling is performed based on the initial correlation strength matrix to generate an intensity evolution curve, which includes the curve data of the change of each correlation strength value over time. The evolutionary characteristic parameters of the intensity evolution curve are extracted, including the slope change rate, peak occurrence time, and decay period data of the curve. The evolutionary feature parameters are input into the parameter feedback layer, and the node connection weights of the association strength evolution layer are adjusted to obtain the adjusted association strength matrix. Based on the adjusted correlation strength matrix, the intensity evolution curve is regenerated, and the steps of extracting evolution feature parameters and adjusting node connection weights are repeated until the stability index of the intensity evolution curve meets the preset stability condition. The evolution feature parameters at this time are then used as the trajectory correlation evolution parameters.

3. The method as described in claim 1, characterized in that, The process of performing state trend bifurcation deduction based on the trajectory association evolution parameters to obtain state trend bifurcation results includes: Determine the starting time point and duration of the simulation, wherein the starting time point is the current monitoring time point, and the duration of the simulation is the length of the continuous observation period starting from the starting time point; A trend bifurcation decision tree is constructed based on the trajectory association evolution parameters. The trend bifurcation decision tree includes a multi-path trend prediction structure with the association strength evolution law as the branching condition. The dynamic characteristic trajectory data at the current monitoring time point is used as the initial input and input into the trend bifurcation decision tree to generate multiple initial trend paths. Each initial trend path contains one possible state evolution direction data. Calculate the path confidence score for each initial trend path, whereby the path confidence score includes a numerical value indicating the probability of the path occurring, determined based on trajectory association evolution parameters. Initial trend paths with a confidence level greater than a preset confidence threshold are selected as candidate trend paths, and the candidate trend paths contain data on the direction of state evolution with a high probability. For each candidate trend path, a path stability assessment is performed to obtain the path stability, which includes the fluctuation data of the trend path within the simulation period. The candidate trend paths are sorted based on the path stability, and the top N candidate trend paths with the highest stability are selected, where N is an integer greater than 1. The state trend bifurcation result is obtained by fusing the first N candidate trend paths and their corresponding path confidence scores.

4. The method as described in claim 1, characterized in that, The step of performing trajectory point density clustering on the spatial observation substring to obtain spatial trajectory clusters includes: Extract all spatial observation data points from the spatial observation substring, each spatial observation data point containing three-dimensional coordinate values ​​and corresponding timestamp data; Calculate the spatial Euclidean distance between any two spatial observation data points and generate a distance matrix, which contains the spatial distance values ​​of all data point pairs. Set a density threshold and a distance threshold, where the density threshold is the minimum number of data points required to form a cluster, and the distance threshold is the maximum allowed distance between data points; Based on the distance matrix and density threshold, core data points are identified, which are spatial observation data points that contain at least a density threshold of other data points within the distance threshold range. Data points whose distance to the core data points is less than the distance threshold are grouped into the same cluster to obtain the initial spatial clusters; Calculate the average timestamp of all data points in the initial spatial cluster, and sort the initial spatial cluster by time according to the average timestamp to obtain a time-ordered spatial cluster sequence. The spatial overlap of adjacent clusters in the temporally ordered spatial cluster sequence is calculated, and adjacent clusters with a spatial overlap greater than a preset overlap threshold are merged to obtain a merged spatial cluster, which is used as the spatial trajectory cluster. The step of calculating the cross-correlation degree between the spatial trajectory cluster and the signal drift trajectory to generate the first cross-correlation trajectory includes: Extract the cluster center trajectory from the spatial trajectory clusters, the cluster center trajectory containing the path data of the change of the center coordinates of each spatial trajectory cluster over time; Extract the main spectral feature trajectory from the signal drift trajectory, the main spectral feature trajectory containing the drift path data of the frequency component with the strongest energy in the signal spectrum over time; By unifying the time axes of the cluster center trajectory and the main spectral feature trajectory, a time-synchronized trajectory pair is obtained; Calculate the feature vector of the time synchronization trajectory pair at each timestamp, the feature vector containing the cluster center coordinate vector and the main frequency vector of the spectrum; A joint feature matrix is ​​constructed based on the feature vectors of all timestamps, and the joint feature matrix contains feature vector data arranged in chronological order. Singular value decomposition is performed on the joint feature matrix to obtain the left singular matrix, the singular value matrix, and the right singular matrix; Extract the left and right singular vectors corresponding to the largest singular value in the singular value matrix, calculate their inner product, and obtain the cross-correlation coefficient. The cross-correlation coefficients are arranged by timestamp to generate the first cross-correlation trajectory.

5. The method as described in claim 2, characterized in that, The step of inputting the dynamic feature trajectory set into the trajectory alignment layer and performing timestamp deviation compensation processing to obtain a time-aligned trajectory set includes: The timestamp sequence of each feature trajectory is extracted from the set of dynamic feature trajectories to obtain the spatial timestamp sequence, the signal timestamp sequence, the motion timestamp sequence, the first cross timestamp sequence, and the second cross timestamp sequence; A baseline timestamp sequence is determined, which is the timestamp sequence with the highest sampling frequency among all timestamp sequences; For each non-reference timestamp sequence, calculate its time deviation from the reference timestamp sequence. The time deviation includes the difference between each timestamp in the non-reference timestamp sequence and the most recent timestamp in the reference timestamp sequence. A deviation compensation function is constructed based on the time deviation, and the deviation compensation function includes mapping relationship data between the time deviation and the feature trajectory correction amount; The feature trajectory corresponding to the non-reference timestamp sequence is corrected based on the deviation compensation function to obtain the corrected feature trajectory, which includes trajectory data after compensating for time deviation. The feature trajectory corresponding to the reference timestamp sequence is combined with all the corrected feature trajectories to obtain the time-aligned trajectory set.

6. The method as described in claim 2, characterized in that, The step of modeling the intensity evolution based on the initial correlation strength matrix to generate an intensity evolution curve includes: Extract the timestamp corresponding to each association strength value from the initial association strength matrix to obtain an intensity-time data pair set, which contains multiple data pairs consisting of timestamps and corresponding association strength values; The intensity-time data pair set is segmented into time series, dividing consecutive timestamps into multiple evolution periods, each evolution period containing a preset number of intensity-time data pairs; For each evolution period, the mean, variance, and maximum value of the correlation strength within that period are calculated to obtain the period characteristic value; A feature sequence is constructed based on the time-period feature values ​​of all evolution periods, and the feature sequence contains time-period feature value data arranged in chronological order; The feature sequence is fitted with a polynomial to obtain a fitted polynomial, wherein the independent variable of the fitted polynomial is the evolution period number and the dependent variable is the period feature value. A continuous intensity evolution curve is generated based on the fitted polynomial, and the intensity evolution curve contains the continuous change path data of the associated intensity value over the evolution period.

7. The method as described in claim 3, characterized in that, The process of evaluating the path stability of each candidate trend path to obtain path stability includes: Extract the state feature values ​​of all timestamps from the candidate trend path to obtain a path feature sequence, which contains state feature values ​​arranged in chronological order. Calculate the first-order difference sequence of the path feature sequence, wherein the first-order difference sequence contains the difference data of adjacent timestamp state feature values; Calculate the standard deviation of the first-order difference sequence to obtain a volatility index, which includes the overall volatility data of the path characteristic sequence; The path feature sequence is smoothed using a sliding window process to obtain a smoothed feature sequence. Calculate the rate of curvature change of the smoothed feature sequence to obtain a curvature index, which includes data on the degree of curvature change of the smoothed feature sequence; The fluctuation index and the curvature index are input into a preset stability evaluation function to obtain the path stability. The preset stability evaluation function contains weighted combination relationship data of the fluctuation index and the curvature index.

8. The method as described in claim 4, characterized in that, The calculation of the average timestamp of all data points in the initial spatial cluster, followed by sorting the initial spatial cluster by time according to the average timestamp, yields a time-ordered spatial cluster sequence, including: Traverse each initial spatial cluster, collect the timestamp data of all data points within the cluster, and obtain the cluster timestamp set; Calculate the arithmetic mean of all timestamps in the cluster timestamp set to obtain the average timestamp of the initial spatial cluster; Sort all initial spatial clusters in ascending order of their timestamp mean to obtain a sorted list of spatial clusters; Check the difference in the average timestamps of adjacent clusters in the sorted spatial cluster list. If the difference is less than a preset time interval threshold, they are determined to be consecutive clusters. Timestamp mean fusion is performed on continuous clusters, and the timestamp mean of the continuous clusters is updated to the arithmetic mean of the timestamp means of all continuous clusters; The merged continuous clusters and non-continuous clusters together form the time-ordered spatial cluster sequence.

9. The method as described in claim 1, characterized in that, After performing state trend bifurcation deduction based on the trajectory association evolution parameters to obtain the state trend bifurcation result, the method further includes: An airspace situation model is constructed based on the state trend bifurcation results and a preset airspace environment parameter set. The preset airspace environment parameter set includes geographical obstacle parameters, electromagnetic interference parameters, and meteorological impact parameters within the monitoring area. The airspace situation model is used to integrate aircraft state trend data and airspace environment impact data. The airspace situation model is updated in real time to obtain a dynamically updated airspace situation model. The real-time update process includes adjusting the aircraft state trend data in the model according to the newly received low-altitude multimodal observation data string, and updating the airspace environment impact data in the model according to the external environment data source. The external environment data source includes real-time meteorological data sent by the meteorological monitoring system and terrain update data sent by the geographic information system. Based on the dynamically updated airspace situation model and the preset scheduling rule set, a scheduling instruction set is generated. The preset scheduling rule set includes aircraft priority rules, conflict avoidance rules, and path optimization rules. The aircraft priority rules are used to determine the air traffic priority of different types of aircraft. The conflict avoidance rules are used to identify potential flight path intersections and generate avoidance strategies. The path optimization rules are used to plan the shortest flight path under the condition of meeting safety constraints. The scheduling instruction set includes heading adjustment instructions, speed adjustment instructions, and communication frequency adjustment instructions for aircraft with different state trend categories.

10. A computer system, characterized in that, include: A memory, wherein a computer program is stored; A processor is used to load the computer program to implement the low-altitude aircraft status monitoring method based on 5G base station towers as described in any one of claims 1-9.

Images