Safety monitoring and early warning system for low-altitude aircraft and ground equipment

Abstract

The invention relates to the technical field of low-altitude aircraft safety monitoring, and discloses a low-altitude aircraft and ground equipment safety monitoring and early warning system, and the system comprises an aircraft monitoring module which is used for collecting the dynamic position data of an aircraft and setting an airspace monitoring interval; the environment sensing module is used for dividing space nodes and analyzing environment data to generate security feature vectors; the risk positioning module is used for extracting key abnormal indexes and establishing risk grading rules; the dynamic prediction module is used for identifying a track correlation mode to calibrate an index and calculating a safety offset; the threshold value optimization module is used for deducing an optimal safety threshold value and generating a risk deviation sequence; and the early warning decision module analyzes the sequence and integrates the sequence into an early warning response decision scheme. The system realizes low-altitude global monitoring, risk analysis and dynamic early warning through multi-module cooperation, improves the accuracy and timeliness of low-altitude airspace safety management, and is suitable for safety guarantee of low-altitude aircrafts and ground equipment.

Description

Technical Field

[0001] The present invention relates to the technical field of low-altitude aircraft safety monitoring, and in particular to a safety monitoring and early warning system for low-altitude aircraft and ground equipment. Background Art

[0002] With the rapid development of the general aviation industry, the number and frequency of low-altitude aircraft have increased significantly, posing unprecedented challenges to the safety management of low-altitude airspace. Traditional low-altitude safety monitoring methods suffer from limited monitoring range, insufficient real-time data, and delayed risk warnings, making them unable to meet the safety requirements of complex airspace conditions.

[0003] From a technical perspective, existing low-altitude monitoring systems primarily rely on single sensors or isolated monitoring nodes, failing to form a comprehensive, three-dimensional monitoring network. This results in blind spots in the collection of dynamic aircraft position data. For example, in areas with complex terrain or environments with strong electromagnetic interference, traditional radar or radio monitoring equipment is prone to signal loss or false alarms, making it difficult to accurately obtain the aircraft's real-time position and status information. Furthermore, existing systems' analysis of environmental data remains limited to simple threshold judgments, lacking the ability to deeply analyze environmental signal characteristics and effectively identifying potential risk factors in complex environments. For example, changes in meteorological conditions (such as strong winds and thunderstorms) and the airspace's electromagnetic environment can interfere with an aircraft's navigation system, but traditional systems struggle to capture subtle changes in these environmental parameters in real time and establish models linking them to flight safety.

[0004] In terms of risk identification and early warning decision-making, existing technologies mostly use fixed-threshold risk assessment models, lack a dynamic calibration mechanism, and are unable to adapt to the changing risk characteristics of different airspaces and at different times. For example, the risk levels faced by aircraft on busy low-altitude routes and in remote, open low-altitude areas vary significantly, but traditional systems use a unified risk assessment standard, resulting in insufficient accuracy and pertinence in early warning results. In addition, the decision-making plans of existing early warning systems are often based on a single risk indicator and lack the integration and correlation analysis of multi-dimensional risk data, making it difficult to generate comprehensive and scientific early warning response strategies. For example, when an aircraft experiences both an abnormal trajectory and a sudden change in environmental parameters, traditional systems may not be able to identify the correlation between the two in a timely manner, resulting in delayed warnings or decision-making errors.

[0005] With the widespread use of new low-altitude aircraft, such as drones, low-altitude traffic scenarios have become more complex, placing higher demands on the intelligence and integration of safety monitoring systems. The limitations of existing technologies in multi-source data fusion, dynamic risk prediction, and threshold optimization are becoming increasingly prominent. There is an urgent need for a new safety monitoring and early warning system that can achieve full-area real-time monitoring, multi-dimensional risk analysis, and dynamic early warning decision-making. This will ensure the safe operation of low-altitude aircraft and ground equipment, and promote the healthy development of the low-altitude economy. Summary of the Invention

[0006] The purpose of the present invention is to provide a low-altitude aircraft and ground equipment safety monitoring and early warning system to solve the problems raised in the above background technology.

[0007] To achieve the above-mentioned object, the present invention provides the following technical solution: a low-altitude aircraft and ground equipment safety monitoring and early warning system, the system comprising: Aircraft monitoring module, used to collect dynamic position data of low-altitude aircraft in real time and set airspace monitoring intervals corresponding to safety levels; The environmental perception module is used to divide the airspace monitoring area into multiple spatial nodes, analyze the signal characteristics of the environmental data of each spatial node, and generate the security feature vector corresponding to the spatial node; The risk location module is used to extract key abnormal indicators from the security feature vector, establish risk classification rules associated with spatial nodes, and obtain signal correction parameters corresponding to the rules; Dynamic prediction module, used to identify trajectory correlation patterns in signal correction parameters, dynamically calibrate key anomaly indicators based on the correlation patterns, and calculate the safety offset of each spatial node under different calibration strategies; The threshold optimization module is used to derive the optimal safety threshold based on the safety offset and generate a risk deviation sequence by comparing the current environmental characteristic value with the optimal safety threshold; The early warning decision module is used to analyze the risk deviation sequence and integrate the risk deviation sequence into an early warning response decision plan based on the safety deviation trend of the spatial nodes.

[0008] Preferably, the implementation of the environment perception module includes: constructing an environment feature library corresponding to the spatial node, the environment feature library containing environment data and a security parameter vector of a signal feature mapping; Similar signal matching is performed on the security parameter vector, and the spatial clustering groups of the security parameter vector are divided according to the matching results; the signal focus points of the environmental data are extracted from the spatial clustering groups, and the focus points are set as spatial nodes.

[0009] Preferably, the spatial clustering groups for dividing the security parameter vectors further include: According to the spatial attributes and risk intensity in the safety parameter vector, the signal distribution, anomaly polarity and environmental density parameters are extracted, and the environmental feature labels are generated based on the above parameters; The environmental feature labels are associated with the security parameter vectors. By calculating the signal similarity between the feature labels, the parameter vectors with similarity higher than the preset security threshold are screened to form a spatial clustering group.

[0010] Preferably, the implementation method of generating the security feature vector corresponding to the spatial node includes: For each spatial node, according to the node's trajectory position in the airspace monitoring interval, the node's signal fluctuation data within the preset window is obtained, and the node's signal fluctuation coefficient is calculated; When the signal fluctuation coefficient exceeds the first safety threshold, the node is marked as a high-risk node, and its environmental data is extracted to form a safety feature vector; when the signal fluctuation coefficient is lower than the first safety threshold, the node is marked as a steady-state node, and the environmental data of the node's adjacent nodes are superimposed, and the superimposed data is reconstructed into a safety feature vector.

[0011] Preferably, the implementation of the risk location module includes: Separate the height anomaly parameters, speed anomaly parameters, and trajectory fluctuation parameters from the safety feature vector, and generate risk classification rules for spatial nodes based on the above parameters; If the number of spatial nodes covered by the current risk classification rule is less than the preset safety threshold, the security feature vectors of adjacent spatial nodes are traversed, and the signal indicators not included in the classification rules of the adjacent nodes are added to the current rule.

[0012] Preferably, the implementation of the dynamic prediction module includes: obtaining a timing factor of the signal switching frequency and a fluctuation factor of the trajectory jump amplitude in the trajectory association mode; A spatial weight matrix associated with the time series factor and the volatility factor is constructed, and the safety offset under different calibration strategies is determined according to the distribution probability of each element in the matrix.

[0013] Preferably, constructing the spatial weight matrix further includes: Identify the trajectory period characteristics of the timing factor. If the current period characteristics completely match the preset signal period, set the timing factor as the starting index of the spatial weight matrix. Calculate the trajectory matching degree of the time series factor and the volatility factor, and generate the intermediate index and terminal index of the spatial weight matrix in descending order of matching degree; The path of the termination index is backtracked, and when the matching degree of the termination index is lower than the preset matching threshold, it is output as the final distribution of the spatial weight matrix.

[0014] Preferably, the implementation of calculating the safety offset includes: Statistically calculate the mean and range of the time series factor of each terminal index in the spatial weight matrix, and calculate the global variance of all index factors; The timing offset coefficient is obtained by subtracting the mean of the timing factor of a single termination index from the mean of the timing factor of the adjacent index and dividing it by the global variance. At the same time, the ratio of the volatility factor range to the global variance is calculated, and the weighted sum of the two is used as the safety offset of the index.

[0015] Preferably, the implementation method of deriving the optimal safety threshold includes: Extract the risk pattern closest to the current safety offset from the historical data and calculate the cosine similarity between the two in trajectory distribution as the first dynamic reference value; Count the difference in the number of abnormal peaks between the current safety offset and the historical risk pattern, and use the difference as the second dynamic reference value; Based on the linear combination of the first dynamic reference value and the second dynamic reference value, an optimal safety threshold in a preset safety threshold table is matched.

[0016] Preferably, the implementation of the early warning decision module includes: dividing the safety deviation direction of each node in the risk deviation sequence into a positive deviation interval and a negative deviation interval; The convergence rate of risk deviation in the positive offset interval and the diffusion rate of risk deviation in the negative offset interval are extracted, and the two are conflict-fused according to the trajectory weights of spatial nodes to generate the adjustment parameters of the early warning response decision-making plan.

[0017] Compared with the prior art, the present invention has the following beneficial effects: At the data collection and monitoring level, the aircraft monitoring module can collect dynamic position data of low-altitude aircraft in real time and set differentiated airspace monitoring intervals according to safety levels, achieving precise monitoring of areas with different risk levels. The environmental perception module divides the airspace monitoring interval into spatial nodes, analyzes the signal characteristics of the environmental data of each node, and generates a safety feature vector. This breaks through the monitoring limitations of traditional single sensors and builds a three-dimensional monitoring network with full coverage, effectively improving the comprehensiveness and real-time nature of data collection. For example, by building an environmental feature library and matching similar signals, it can automatically identify key feature points in the environmental data and set them as spatial nodes, achieving precise positioning of risk hazards in complex environments.

[0018] In terms of risk analysis and positioning, the risk positioning module extracts key anomaly indicators from the security feature vector, establishes risk grading rules associated with spatial nodes, and dynamically adjusts the rules according to the number of nodes, ensuring the accuracy and flexibility of risk assessment. For example, when the number of nodes covered by the current risk grading rules is insufficient, the system automatically traverses the security feature vectors of adjacent nodes and supplements the signal indicators that are not included, avoiding missed risks. The dynamic prediction module dynamically calibrates key anomaly indicators by identifying trajectory correlation patterns in signal correction parameters, calculates safety offsets under different calibration strategies, and realizes forward-looking predictions of risk development trends. Constructing a spatial weight matrix and combining the analysis of time series factors and volatility factors can deeply explore the spatiotemporal correlation characteristics between data, making the prediction results more in line with the actual risk evolution law.

[0019] In terms of threshold optimization and early warning decision-making, the threshold optimization module derives the optimal safety threshold based on the safety offset, and generates a risk deviation sequence by comparing the current environmental characteristic value with the threshold, solving the problem that traditional fixed thresholds cannot adapt to dynamic risk changes. For example, by extracting similar risk patterns in historical data and calculating dynamic reference values, the threshold can be dynamically adjusted so that the system can maintain optimal monitoring accuracy in different airspace environments. The early warning decision module analyzes the risk deviation sequence and integrates the early warning response decision plan based on the safety offset trend of the spatial node. By dividing the positive and negative offset intervals and analyzing the risk diffusion and convergence rate, it generates targeted adjustment parameters, realizing intelligent processing of the entire chain from risk identification to decision response, significantly improving the scientific nature and timeliness of early warning decisions.

[0020] Furthermore, data interaction and algorithmic collaboration between the system's modules form a closed-loop safety monitoring and early warning mechanism. The fully automated process, from dynamic data collection to risk analysis, prediction, and decision-making, not only reduces manual intervention costs but also improves the system's response speed and processing efficiency. Furthermore, the in-depth integration and analysis of multi-dimensional data enhances the system's adaptability to complex airspace conditions, effectively addressing the safety management challenges posed by the growing number of low-altitude aircraft and the diversification of flight scenarios, providing a solid technical foundation for the development of the low-altitude economy. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 This is a working principle diagram of the low-altitude aircraft and ground equipment safety monitoring and early warning system of the present invention; Figure 2 Design graphs generated for spatial cluster groups; Figure 3 Design graphs for security feature vector generation; Figure 4 Generate design diagrams for risk ranking rules. DETAILED DESCRIPTION

[0022] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0023] See also Figures 1-4 The low-altitude aircraft and ground equipment safety monitoring and early warning system of the present invention is specifically implemented as follows: The system includes an aircraft monitoring module, an environmental perception module, a risk positioning module, a dynamic prediction module, a threshold optimization module and an early warning decision module. The modules work together to achieve low-altitude safety monitoring and early warning.

[0024] Aircraft monitoring module: Real-time collection of dynamic position data such as longitude, latitude, altitude, speed, and heading of aircraft through satellite navigation receivers, radar sensors, or ADS-B (Automatic Dependent Surveillance-Broadcast) devices deployed in low-altitude airspace. At the same time, according to the type of aircraft (such as unmanned aerial vehicles, light aircraft), flight tasks (such as cargo transport, plant protection), and airspace management requirements, set up airspace monitoring intervals corresponding to safety levels. For example, set a first-level monitoring interval of 5 kilometers in radius for densely populated areas, and a second-level monitoring interval of 10 kilometers in radius for open areas, to achieve hierarchical monitoring.

[0025] Environmental perception module: Divide multiple spatial nodes (such as 100m x 100m x 100m cubic grid nodes) within the airspace monitoring interval, and collect environmental data such as temperature, humidity, air pressure, wind speed, and electromagnetic signals through sensors (such as weather sensors, electromagnetic monitoring devices) deployed at each node. Perform signal feature analysis on the environmental data of each node, including frequency domain and time domain feature extraction such as Fourier transform and wavelet analysis, to generate a safety feature vector containing characteristic parameters (such as frequency components, amplitude fluctuations).

[0026] Risk positioning module: Extract key abnormal indicators (such as high abnormal values, sudden speed changes, and trajectory deviation) from the safety feature vector, and establish risk classification rules associated with spatial nodes (such as three levels of risk: low risk, medium risk, and high risk). At the same time, obtain signal correction parameters (such as sensor calibration coefficients, data transmission delay compensation values) through data interfaces for subsequent dynamic calibration.

[0027] Dynamic prediction module: Identify trajectory correlation patterns in signal correction parameters through machine learning algorithms (such as Hidden Markov Model, Recurrent Neural Network), such as periodic fluctuations and sudden turning patterns of aircraft trajectories. Perform dynamic calibration on key abnormal indicators based on correlation patterns, and use Kalman filtering or particle filtering algorithms to calculate safety offsets of each spatial node under different calibration strategies (such as real-time calibration and lagging calibration), quantifying the degree of risk deviation.

[0028] Threshold optimization module: Based on safety offsets, derive the optimal safety threshold through historical data comparison and statistical analysis. Specifically, calculate the difference between the current environmental characteristic value and the optimal threshold to generate a risk deviation sequence containing the risk deviation degree of each node, providing a data basis for early warning.

[0029] Early warning decision module: Analyze the risk deviation sequence using fuzzy logic or decision tree algorithms, and combine the safety offset trend (such as offset rate and offset direction) of the spatial node to integrate discrete risk information into a hierarchical early warning response decision scheme, such as triggering sound and light alarms or sending control instructions to the aircraft.

[0030] Example 1: The core function of the environmental perception module is to build a structured monitoring node system within the airspace monitoring range and generate safety feature vectors corresponding to the nodes through signal feature analysis, providing a data basis for subsequent risk assessment. The specific implementation is as follows: The environmental perception module must construct an environmental feature library corresponding to each spatial node. This library, built based on historical monitoring data, stores environmental data for each region within the low-altitude airspace, as well as safety parameter vectors generated through signal feature mapping. Environmental data encompasses multidimensional physical quantities such as temperature and humidity, air pressure, wind speed and direction, electromagnetic signal strength, and light intensity. These data are continuously collected and stored by various sensors deployed in the low-altitude airspace (e.g., meteorological sensors, electromagnetic monitoring equipment, and optical cameras). The safety parameter vector is a structured representation of the raw environmental data after feature extraction. Signal processing algorithms (e.g., Fourier transform and wavelet transform) analyze the data features in the time, frequency, and amplitude domains, generating a multidimensional vector containing parameters such as frequency components, amplitude fluctuations, and trend terms. For example, for electromagnetic environmental data for a specific area, a Fourier transform is used to extract the energy distribution of each frequency band. Combined with time-domain parameters such as peak and mean values, a safety parameter vector containing 10 dimensional features is generated to characterize the electromagnetic environment status of that area.

[0031] Based on the construction of the environmental feature library, similar signal matching processing is performed on the security parameter vectors to achieve the division of spatial cluster groups. Similar signal matching uses a clustering algorithm to calculate the similarity between security parameter vectors and classify vectors with similar environmental characteristics into the same cluster group. In specific implementation, a density clustering algorithm (such as DBSCAN) or a partitioning clustering algorithm (such as K-means) can be used to judge similarity based on the distance metric between vectors (such as Euclidean distance and cosine similarity). For example, for two security parameter vectors, their Euclidean distance is calculated. If the distance is less than a preset threshold, they are considered to be similar signals and should be included in the same cluster group. Through clustering processing, the security parameter vectors of the entire airspace monitoring interval are divided into several spatial cluster groups, and the vectors in each group correspond to areas with similar environmental characteristics.

[0032] When delineating spatial clusters, it's necessary to combine the airspace attributes and risk intensity in the safety parameter vector to further extract signal distribution, anomaly polarity, and environmental density parameters to generate environmental feature labels. Airspace attributes include the three-dimensional coordinates (longitude, latitude, and altitude) of the spatial node and the level of the monitoring interval to which it belongs (e.g., Level 1 or Level 2), reflecting the node's location and importance in the airspace. Risk intensity is obtained through historical data statistics, such as the frequency of low-altitude aircraft anomaly incidents in a region, which is used to measure the potential risk level of the area. Signal distribution parameters, such as standard deviation and interquartile range, describe the fluctuation range and central tendency of the data in each dimension of the feature vector. Anomaly polarity parameters identify the direction of data anomalies (e.g., positive anomalies when the temperature is above a threshold or negative anomalies when it is below a threshold). Environmental density parameters indicate the density of sensor deployment within a unit space and reflect the reliability of data collection. Through comprehensive analysis of these parameters, an environmental feature label is generated, containing information such as airspace location, risk level, signal characteristics, and anomaly type. Each safety parameter vector is associated with one or more feature labels for subsequent clustering and screening.

[0033] Environmental feature labels are associated with security parameter vectors. By calculating signal similarity between the feature labels, parameter vectors with similarity exceeding a preset security threshold are selected to form spatial cluster groups. Feature label similarity can be calculated using methods such as the Jaccard coefficient and edit distance. Matching rules are designed for categorical attributes (such as monitoring interval level and anomaly polarity) and numerical attributes (such as risk intensity and signal distribution parameters) within the labels. For example, for the categorical attribute "monitoring interval level," similarity is calculated as 1 only when two labels have the same level. For the numerical attribute "risk intensity," similarity is calculated using the normalized absolute value of the difference, with smaller differences indicating higher similarity. By setting a comprehensive similarity threshold (e.g., 0.8), security parameter vectors with label similarity exceeding this threshold are grouped together in the same spatial cluster group. This ensures high consistency in environmental characteristics among the vectors within the group, reducing redundant computations in subsequent monitoring and improving processing efficiency.

[0034] After completing the spatial cluster group division, it is necessary to extract the signal focus points of the environmental data from each cluster group and set them as spatial nodes. The extraction of signal focus points is based on the feature analysis of the safety parameter vectors within the cluster group. Through dimensionality reduction algorithms such as principal component analysis (PCA), the direction of the principal components in the feature space is determined, and key points that can represent the core characteristics of the cluster group are identified. For example, for a cluster group containing 100 safety parameter vectors, the first two principal components are obtained through PCA calculation. The center of mass or extreme point of the cluster group in the principal component space is used as the signal focus point, which can best reflect the overall environmental characteristics of the cluster group. After setting the signal focus point as a spatial node, a node network covering the entire airspace monitoring range is formed. Each node corresponds to a spatial area with typical environmental characteristics, realizing structured monitoring of low-altitude airspace.

[0035] In actual deployment, the density of spatial node divisions can be dynamically adjusted based on monitoring needs. For example, in areas surrounding airports with frequent aircraft activity, smaller spatial grids (e.g., 50m x 50m x 50m) can be used to improve monitoring accuracy. In open rural areas, larger spatial grids (e.g., 200m x 200m x 200m) can be used to reduce deployment costs. Furthermore, the environmental feature database needs to be regularly updated. Newly collected environmental data is incorporated through incremental learning algorithms, and cluster analysis and node adjustments are performed again to ensure that spatial nodes always align with current airspace environmental characteristics, maintaining the accuracy and timeliness of system monitoring.

[0036] Furthermore, the environmental perception module must handle noise and outliers in sensor data. Before analyzing signal features, the raw environmental data is preprocessed using algorithms such as sliding average filtering and median filtering to remove random noise. Outliers that clearly fall outside a reasonable range are corrected through data interpolation or neighborhood data filling to ensure that the generated safety parameter vector truly reflects the actual environmental state. During the clustering process, outlier vectors that cannot be assigned to any cluster group are individually marked as abnormal areas, and the monitoring frequency of these areas is increased to promptly detect sudden environmental changes or equipment failures.

[0037] The environmental perception module converts raw environmental data into structured spatial nodes, establishing a monitoring system that dynamically reflects airspace environmental characteristics. This system reduces data redundancy through cluster analysis and ensures node representativeness through signal focus extraction. This provides efficient and accurate input data for subsequent risk location and dynamic prediction modules, laying the technical foundation for the entire low-altitude aircraft and ground equipment safety monitoring and early warning system.

[0038] Example 2: The generation of security feature vectors is one of the core links of the environmental perception module. Its purpose is to generate feature vectors that can accurately represent the risk status of nodes based on the signal fluctuation data of spatial nodes through differentiated processing strategies, providing key data support for subsequent risk positioning and early warning. The specific implementation is as follows: For each spatial node within the airspace monitoring range, the system needs to obtain its signal fluctuation data within the preset window in real time. The length of the preset window can be flexibly set according to monitoring needs, for example, to 5 minutes, 10 minutes or 15 minutes, to capture environmental change trends within a short period of time. The signal fluctuation data covers the real-time measurement values ​​of various environmental parameters collected by the node, such as temperature and humidity, air pressure, wind speed, electromagnetic signal strength, etc., as well as dynamic data related to low-altitude aircraft, such as the altitude, speed, and heading of the aircraft when passing through the node. These data are collected in real time by sensors deployed at the node and transmitted to the system backend for processing via the wireless communication network.

[0039] The system needs to calculate the signal fluctuation coefficient of each spatial node to evaluate the environmental stability or risk level of the node. The signal fluctuation coefficient is an indicator that comprehensively measures the amplitude of data changes, and is obtained by statistical analysis of the signal data within a preset window. Specifically, the system first calculates the numerical fluctuation range of each environmental parameter within the preset window, such as the difference between the maximum and minimum temperature values, the root mean square value of the wind speed, etc., and then standardizes the fluctuation values ​​of these parameters to eliminate the influence of dimensional differences. Finally, a comprehensive signal fluctuation coefficient is generated by weighted summation. The weighting factor can be pre-set according to the importance of each parameter. For example, the weight of the electromagnetic signal strength can be set to 0.3, the weight of the aircraft speed can be set to 0.25, the weight of the temperature and humidity can be set to 0.15, etc., reflecting the different degrees of influence of different parameters on low-altitude safety.

[0040] Based on the calculated signal fluctuation coefficient, the system classifies spatial nodes into high-risk and stable nodes, and implements different safety feature vector generation strategies. When the signal fluctuation coefficient exceeds the first safety threshold, the node is identified as high-risk. The first safety threshold is a pre-set critical value based on historical data and airspace management requirements, used to distinguish normal from abnormal fluctuations. For example, for a certain monitoring area, historical data analysis determined that when the signal fluctuation coefficient exceeds 0.6, the probability of a low-altitude safety incident increases significantly. Therefore, the first safety threshold is set to 0.6. For high-risk nodes, the system directly extracts environmental data within a preset window, including the raw values, extreme fluctuations, and changing trends of various parameters collected in real time, to form a multi-dimensional safety feature vector. For example, the safety feature vector of a high-risk node may include 10 dimensions of features, such as current temperature, humidity, air pressure, wind speed, peak electromagnetic signal strength, abnormal aircraft altitude, sudden speed changes, trajectory deviation, signal fluctuation coefficient, and risk level indicator, comprehensively reflecting the node's abnormal status and risk characteristics.

[0041] When the signal fluctuation coefficient falls below the first safety threshold, the node is marked as a steady-state node, indicating that its current environmental state is relatively stable and no significant anomalies have been detected. For steady-state nodes, the system does not directly use its own environmental data to generate a safety feature vector. Instead, it processes the data of adjacent nodes through signal superposition to reduce data redundancy and improve computational efficiency. Specifically, the system first determines the range of adjacent nodes of the steady-state node. In three-dimensional space, it typically selects eight adjacent nodes centered on the node (for example, nodes in the east, south, west, north, and four diagonal directions horizontally, and nodes adjacent to each other vertically). The system then obtains the environmental data of these adjacent nodes within the same preset window. The data of adjacent nodes is then subjected to signal superposition processing, which involves summing or averaging parameters of the same type. For example, the temperature values ​​of nine nodes (including the steady-state node itself) are averaged, and the wind speed values ​​are vector-synthesized to form the superimposed comprehensive data.

[0042] After signal superposition, the system needs to reconstruct the superimposed data to generate a safety feature vector for the steady-state node. This reconstruction process uses signal processing algorithms to remove noise and redundant information from the superimposed data and extract key parameters that characterize the overall environmental characteristics of the area. For example, a sliding average algorithm is used to smooth the superimposed temperature data to remove random noise. Principal component analysis (PCA) is used to reduce the dimensionality of the multidimensional superimposed data, extracting the first few principal components as characteristic parameters to form a streamlined safety feature vector. The reconstructed safety feature vector not only reflects the overall environmental status of the steady-state node and its adjacent area, but also effectively reduces data dimensionality, reducing the computational complexity of subsequent data processing.

[0043] It should be noted that the signal superposition processing of a steady-state node and its adjacent nodes is not a simple data accumulation, but rather a weighted process based on the spatial distance and signal correlation between the nodes. For example, adjacent nodes closer to the steady-state node are assigned higher weights, while nodes farther away are assigned lower weights to reflect the impact of spatial proximity on environmental characteristics. The weight coefficients can be determined using a Gaussian function or inverse distance weighting, for example, where the weight is inversely proportional to the square of the distance, ensuring that closer adjacent nodes contribute more to the steady-state node's feature vector.

[0044] Furthermore, the system must regularly evaluate and adjust the first safety threshold to adapt to changes in different time periods or airspace environments. For example, during peak aircraft activity, when the normal fluctuation range of environmental parameters may increase, the first safety threshold can be appropriately raised to avoid misjudgments. At night or during low-traffic periods, when the environment is relatively stable, the first safety threshold can be appropriately lowered to improve anomaly detection sensitivity. Threshold adjustments can be based on statistical analysis of historical data or automatically optimized through machine learning algorithms to ensure the system maintains reasonable risk detection accuracy.

[0045] In practical applications, the dimensions and specific parameters of the safety feature vector can be flexibly configured based on monitoring requirements. For example, in areas with complex electromagnetic environments, additional feature dimensions related to electromagnetic signals can be added, such as interference intensity and harmonic components in different frequency bands. In areas with variable weather conditions, the fluctuation characteristics of meteorological parameters such as temperature, humidity, air pressure, and wind speed can be emphasized. Furthermore, the system must establish a feature vector update mechanism to ensure real-time reflection of the latest environmental conditions and aircraft dynamics, thereby avoiding risk assessment distortions caused by data lags.

[0046] Example 3: The core function of the risk location module is to extract key anomaly indicators from the security feature vector, establish risk classification rules associated with spatial nodes, and dynamically optimize the rule content based on the rule coverage to ensure accurate identification of risk nodes and their levels in low-altitude airspace. The specific implementation is as follows: The risk location module needs to separate the key abnormality indicators used for risk classification from the safety feature vector. These indicators are the core parameters that reflect the safety risks of low-altitude aircraft and ground equipment. Specifically, the system extracts altitude abnormality parameters, speed abnormality parameters and trajectory fluctuation parameters from the safety feature vector. The altitude abnormality parameter characterizes the degree of deviation between the actual flight altitude of the aircraft and the specified altitude of the airspace. For example, for a no-fly zone or a restricted altitude area, the difference between the aircraft altitude and the specified value is the altitude abnormality parameter; the speed abnormality parameter reflects the deviation ratio of the aircraft flight speed from the speed limit requirement of the airspace. For example, if the aircraft speed exceeds the speed limit by 10%, the speed abnormality parameter value is recorded as 10%; the trajectory fluctuation parameter is measured by the aircraft heading change rate or the distance the trajectory deviates from the preset route. For example, when the heading change angle per unit time exceeds a certain threshold or the trajectory deviates from the route by more than the set value, it is determined to be a trajectory fluctuation abnormality.

[0047] Based on the above three types of key abnormal parameters, the system establishes risk classification rules for spatial nodes. Risk classification adopts a multi-level classification system, for example, divided into three levels of low risk, medium risk, and high risk, each level corresponding to a different parameter threshold combination. Take the three-level risk as an example: Low-risk rules: height anomaly value is less than 5 meters, speed anomaly ratio is less than 10%, and trajectory fluctuation angle is less than 15° / second; Medium risk rule: height anomaly value is between 5 meters and 15 meters, speed anomaly ratio is between 10% and 20%, and trajectory fluctuation angle is between 15° / second and 30° / second; High-risk rules: The height anomaly value is greater than or equal to 15 meters, or the speed anomaly ratio is greater than or equal to 20%, or the trajectory fluctuation angle is greater than or equal to 30° / second.

[0048] The above thresholds can be adjusted according to airspace management regulations, aircraft types and historical risk data. For example, a lower speed anomaly threshold can be set for light drones, and a higher fault tolerance threshold can be set for manned aircraft.

[0049] After establishing the initial risk classification rules, the system needs to evaluate whether the coverage of the rules meets monitoring requirements. The evaluation metric is the ratio of the number of spatial nodes covered by the current risk classification rules to the total number of nodes. If this ratio is less than a preset safety threshold (e.g., 5%), it indicates that the existing rules are not effectively covering potential risk nodes and require rule optimization. The preset safety threshold can be set based on the airspace risk level. For example, a lower threshold (e.g., 3%) can be set in high-risk areas (such as those around airports) to ensure that the rules are more sensitive to risks; a higher threshold (e.g., 8%) can be set in low-risk areas to avoid excessive triggering of warnings.

[0050] When the rule optimization mechanism is triggered, the system traverses the security feature vectors of adjacent spatial nodes and extracts signal indicators not included in the current rules to expand the applicability of the rules. The range of adjacent nodes can be defined as the three-dimensional spatial neighboring nodes centered on the current node (such as one grid node each in the top, bottom, left, right, front, and back), ensuring coverage of spatially related areas. During the traversal process, the system uses feature engineering methods to identify parameters in the security feature vector that have not yet been included in the risk classification rules, such as wind speed influence coefficient, electromagnetic interference intensity, pressure gradient change rate in environmental data, or acceleration mutation value and communication signal strength attenuation rate in aircraft dynamic data.

[0051] For newly extracted signal indicators, the system uses the feature splitting criterion of the decision tree algorithm (such as information gain or Gini index) to evaluate their contribution to risk classification and selects indicators with higher contribution to add to the current rules. Taking information gain as an example, its calculation formula is:

[0052] in, Before partitioning the dataset The information entropy reflects the uncertainty of the data set; After the division The information entropy of the subsets; is the number of subsets, and The original dataset and the The larger the information gain, the stronger the classification ability of the indicator for risk classification, and the more it should be included in the rules.

[0053] When adding new indicators to the risk classification rules, the system adopts a hierarchical expansion strategy. First, for nodes not covered by existing rules, the main types of anomalies in their security feature vectors are analyzed. For example, the height anomaly value of a certain type of node does not exceed the threshold, but the electromagnetic interference intensity increases significantly, resulting in unstable aircraft communication signals. At this time, the system adds a threshold condition for electromagnetic interference intensity on the basis of existing rules, such as "electromagnetic interference intensity is greater than threshold A and lasts for more than T minutes" as a new medium-risk or high-risk rule condition. Secondly, for situations involving multiple new indicators at the same time, a combination of logical AND or logical OR is used to construct a composite rule, such as "wind speed is greater than threshold B and height anomaly value is greater than threshold C" as a joint trigger condition to ensure the accuracy and robustness of the rule.

[0054] The rule optimization process adheres to the principle of minimal intervention, gradually adding new indicators to existing rules to avoid completely overturning the original rules and causing system instability. Each time a new indicator is added, the system recalculates the number of nodes covered by the rule. If the preset safety threshold is still not reached, the system continues to traverse the feature vectors of other adjacent nodes, repeating the feature extraction and rule expansion process until the rule coverage meets the requirements or no new indicators can be extracted.

[0055] Furthermore, risk grading rules must have a dynamic update mechanism to adapt to changes in the airspace environment and adjustments to monitoring requirements. For example, if the electromagnetic environment changes due to the addition of new ground equipment (such as a radar station) in a certain area, the system automatically identifies changes in electromagnetic interference-related indicators in the area's security signature vector, triggering a rule optimization process to incorporate electromagnetic interference parameters into the risk grading rules. During rule updates, the system retains historical rule versions for traceability and rollback, ensuring system auditability and stability.

[0056] In practical applications, the risk location module also needs to address the dimensional differences between different types of indicators. For numerical indicators such as altitude anomalies and speed anomalies, standardization methods (such as Z-score standardization) are used to convert them into dimensionless standardized values, making them easier to compare and combine with other indicators. For physical indicators such as electromagnetic interference intensity and pressure gradient, different weighting coefficients are assigned based on their impact on low-altitude safety. For example, electromagnetic interference intensity is weighted 0.4, altitude anomalies are weighted 0.3, speed anomalies are weighted 0.2, and trajectory fluctuations are weighted 0.1. A weighted summation is used to generate a comprehensive risk index to assist in risk level determination.

[0057] Risk grading rules are validated through historical data backtesting. The system selects security feature vector data from the past period, classifies it using the current risk grading rules, and then counts the number of nodes at each risk level and the subsequent occurrence of actual security incidents to evaluate the accuracy and recall of the rules. If a rule is found to contain misjudgments or omissions, the rule threshold or indicator combination is adjusted manually or automatically until the predetermined monitoring accuracy requirements are met.

[0058] Example 4: The core goal of the dynamic prediction module is to dynamically calibrate key anomaly indicators by analyzing trajectory correlation patterns in signal correction parameters and quantify safety offsets under different calibration strategies, providing data support for threshold optimization and early warning decision-making. The following describes the implementation details with specific examples: 1. Trajectory Correlation Pattern and Signal Factor Extraction The dynamic prediction module first identifies the trajectory correlation pattern from the signal correction parameters, which reflects the time correlation of the movement of low-altitude aircraft in the airspace and the characteristics of the environment. For example, in urban airspace, unmanned aerial vehicle delivery routes may exhibit periodic characteristics of high-frequency return at fixed time periods (such as lunch and dinner peaks); in the vicinity of an airport, manned aircraft takeoff and landing trajectories may exhibit repetitive dive or climb patterns along fixed routes. The system trains a machine learning model (such as a recurrent neural network RNN or a long short-term memory network LSTM) through historical trajectory data to capture the temporal dependence in the trajectory and extract two key factors: The time factor of signal switching frequency: represents the frequency of changes in the state of the aircraft trajectory per unit time, such as the number of heading changes per minute, the number of height layer switches, etc. Taking a certain unmanned aerial vehicle inspection route as an example, if it makes 8 heading adjustments in 10 minutes, its time factor can be quantified as 0.8 times per minute.

[0059] The fluctuation factor of trajectory jump amplitude: measures the spatial distance between adjacent time points of the trajectory, reflecting the stability of the aircraft's movement. For example, a certain aircraft has coordinates (116.3°E, 39.9°N, 100m) at time t and coordinates (116.35°E, 39.92°N, 120m) at time t+1. The Euclidean distance between the two points is 500 meters, so the fluctuation factor is 500 meters.

[0060] II. Spatial weight matrix construction process The spatial weight matrix is used to characterize the spatial distribution characteristics and correlation of the time factor and the fluctuation factor, and its construction process includes the following steps: Periodic feature matching of time factor: the system first analyzes the periodic characteristics of the time factor, such as determining whether it exhibits a daily, weekly, or random fluctuation pattern. Taking a scenic tour helicopter as an example, its flight trajectory performs a circular route every hour during the period 9:00-17:00 each day, exhibiting a clear hourly periodicity. If the periodicity of the extracted time factor matches the pre-set signal periodicity (such as the average period of 1 hour statistically), the time factor is set as the starting index of the spatial weight matrix, ensuring that the time logic of the matrix is consistent with the actual trajectory pattern. The starting index corresponds to the first row or column of the matrix, anchoring the time dimension feature distribution.

[0061] Factor matching degree calculation and index sorting: calculate the trajectory matching degree of the time factor and the fluctuation factor to measure their synergy in spatial movement. The matching degree can be calculated by the dynamic time warping (DTW) algorithm or the cosine similarity algorithm, for example: For the time factor sequence and the fluctuation factor sequence , calculate the cosine similarity after normalization, and the larger the value, the higher the correlation between the two.

[0062] The spatial weight matrix is ​​sorted from highest to lowest by degree of fit, generating the intermediate and final indices. The intermediate indices correspond to the middle levels of the matrix, reflecting moderately strong associations between factors; the final indices correspond to the bottom levels of the matrix, representing combinations of factors with lower but still statistically significant fits. For example, if one set of factors has a fit of 0.7 (high), another 0.4 (medium), and a third 0.2 (low), the index order would be first group, then second group, then third group.

[0063] Path backtracking and matrix optimization: Paths are backtracked for the ending indexes to check whether their matching scores fall below a preset matching threshold (e.g., 0.3). If a given ending index's matching score is 0.25, which is below the threshold, it is removed from the matrix or marked as low-confidence data. Only the matching indexes and their weights are retained as the final matrix distribution. For example, when constructing a 3×3 spatial weight matrix, the index with insufficient matching scores in the lower right corner is backtracked to form a matrix with a valid index distribution of 8 nodes.

[0064] 3. Safety offset calculation example The safety offset is used to quantify the degree of deviation of risk indicators under different calibration strategies. Its calculation is combined with the index characteristics of the spatial weight matrix. The specific steps are as follows: Factor statistics and global variance calculation: Taking a spatial node as an example, its spatial weight matrix contains 5 terminal indexes, each of which corresponds to a set of time series factor means and volatility factor ranges: Index 1: The mean of the time series factor is 0.5 times / minute, and the range of the fluctuation factor is 300 meters; Index 2: The mean of the time series factor is 0.8 times / minute, and the range of the fluctuation factor is 500 meters; Index 3: The mean of the time series factor is 0.6 times / minute, and the range of the fluctuation factor is 400 meters; Index 4: The mean of the time series factor is 0.7 times / minute, and the range of the fluctuation factor is 600 meters; Index 5: The mean of the time series factor is 0.4 times / minute, and the range of the fluctuation factor is 200 meters.

[0065] Calculate the global variance of all index factors to reflect the degree of dispersion of factor distribution. For example, the global variance of the mean of the time series factor is 0.02 (times / minute) 2 , the global variance of the fluctuation factor range is 20,000 meters 2 .

[0066] Calculation of the ratio of the timing offset coefficient to the fluctuation factor: For index 1, its adjacent index is index 5 (the average timing factor is 0.4 times / minute), and the timing offset coefficient is calculated as follows:

[0067] At the same time, calculate the ratio of the volatility factor range to the global variance:

[0068] Weighted summation of safety offsets: Default timing offset coefficient weight is 0.6, fluctuation factor ratio weight is 0.4, calculate the safety offset of index 1:

[0069] This value indicates the degree of deviation of the risk indicator of the spatial node relative to the baseline state under the current calibration strategy. The larger the value, the more significant the risk deviation.

[0070] 4. Calibration Strategy and Matrix Application The spatial weights matrix supports a variety of calibration strategies, such as: Real-time calibration strategy: Calculates the safety offset based on the index with the highest matching degree (e.g., the index with a matching degree of 0.7). This strategy is suitable for scenarios where the aircraft's trajectory changes suddenly, such as sharp turns when avoiding obstacles.

[0071] Lag calibration strategy: Integrates factor data from multiple intermediate indexes (such as indexes with matching degrees of 0.4-0.6) to smooth short-term fluctuations, suitable for risk assessment during the stable flight phase.

[0072] Prediction calibration strategy: Utilizes the time series period characteristics (such as daily cycles) in the matrix to predict the safety offset of the next time period in advance. It is suitable for regular flight scenarios (such as the repetitive routes of agricultural plant protection drones).

[0073] Taking the real-time calibration strategy as an example, when the system detects that a drone has a rapid switching of heading in urban airspace (the timing factor suddenly rises to 1.2 times / minute), it quickly locates the index with the highest matching degree through the spatial weight matrix (assuming the matching degree is 0.8, the corresponding fluctuation factor range is 800 meters), calculates the safety offset to 2.5, triggers a medium-risk warning, and prompts ground control personnel to pay attention to the flight status of the drone.

[0074] 5. Dynamic Adjustment and Application Scenarios The system needs to update the spatial weight matrix regularly to adapt to changes in the airspace environment. For example: When seasonal meteorological conditions change (e.g., strong winds in winter cause increased fluctuations in aircraft trajectories), the distribution range of the fluctuation factor expands, and the global variance and index weights need to be recalculated; The addition of new aircraft types (such as heavy cargo drones) causes changes in the characteristics of the timing factors, and it is necessary to update the starting index of the matrix and the matching calculation logic through incremental learning.

[0075] In practical applications, the dynamic prediction module can be linked to the aircraft control system. For example, when the safety offset exceeds a preset threshold, the system automatically sends a calibration command to the aircraft, adjusting the flight speed or route to reduce the risk of deviation. Simultaneously, the calibrated trajectory data is fed back to the module to optimize subsequent matrix construction, forming a closed-loop control process.

[0076] The dynamic prediction module achieves a dynamic, quantitative assessment of the risk status of low-altitude aircraft through trajectory correlation pattern analysis, spatial weight matrix construction, and safety offset calculation. This module, combined with specific flight scenarios and environmental characteristics, provides a multi-strategy risk calibration basis. This provides a data foundation for the adaptive threshold adjustment of the threshold optimization module and the precise response of the early warning decision module, enhancing the predictive capabilities and real-time performance of the entire monitoring and early warning system.

[0077] Example 5: The threshold optimization module dynamically derives the optimal safety threshold, enabling the system to adapt to changes in the airspace environment; the early warning decision module generates a graded response plan based on the risk deviation sequence to achieve precise control of low-altitude risks. The following describes the implementation method in detail based on specific scenarios: 1. Derivation Process of Optimal Safety Threshold The core task of the threshold optimization module is to dynamically determine the optimal safety threshold based on the safety offset, ensuring that the threshold can accurately reflect the risk characteristics of the current airspace. Taking a low-altitude airspace as an example, the system monitors the safety offset of a spatial node in real time as 0.8. The corresponding optimal threshold needs to be derived through the following steps: Historical risk pattern matching: The system retrieves the risk pattern closest to the current safety offset from the historical database. Historical risk patterns store environmental characteristics and risk event data corresponding to different safety offsets over a period of time (e.g., the past three months). Each pattern includes key attributes such as trajectory distribution and the number of abnormal peaks. For example, the current safety offset of 0.8 may match two historical patterns: Mode A: Occurs in the spring afternoon, with severe convective weather in the airspace, a safety offset of 0.78, scattered fluctuations in trajectory distribution, and three abnormal peaks; Mode B: Occurs during large-scale events. Dense drone flights cause electromagnetic interference, with a safety offset of 0.82. The trajectory distribution is concentrated around the event venue, with 5 abnormal peaks.

[0078] The first dynamic reference value is determined by calculating the similarity between the current safety offset and historical patterns in trajectory distribution (e.g., using Euclidean distance based on spatial coordinates or a dynamic time warping algorithm based on trajectory shape). Assuming the Euclidean distance between the current trajectory distribution and pattern A is 150 meters, and the Euclidean distance with pattern B is 80 meters, pattern B is determined to be the most similar historical pattern, and the first dynamic reference value is the similarity coefficient between the two (e.g., 0.9).

[0079] Abnormal Peak Count Difference Analysis: This tool calculates the difference in the number of abnormal peaks between the real-time data corresponding to the current safety offset and the historical risk pattern. Abnormal peaks refer to the number of times an environmental parameter or aircraft dynamic parameter exceeds a temporary threshold, such as the number of times the electromagnetic signal strength exceeds 50dBμV or the number of times the aircraft speed exceeds the speed limit. The current data has four abnormal peaks, while the historical number of abnormal peaks for pattern B is five. The difference is -1 (one less than the historical number), which serves as the second dynamic reference value.

[0080] Optimal threshold matching: Based on the linear combination of the first dynamic reference value and the second dynamic reference value (for example, the first reference value accounts for 0.7 and the second reference value accounts for 0.3), the comprehensive reference value is calculated:

[0081] The system has a pre-set safety threshold table that stores the optimal thresholds for different comprehensive reference values. For example, when the comprehensive reference value is between 0.3 and 0.4, the optimal safety thresholds are an abnormal height of 8 meters, a speed abnormality of 15%, and an electromagnetic interference intensity of 45dBμV. By looking up the table and matching, the optimal thresholds for the current scenario are determined, enabling dynamic adaptive adjustment of the thresholds.

[0082] 2. Response Logic of the Early Warning Decision Module The early warning decision module generates an early warning response plan based on the risk deviation sequence. The specific steps are as follows: Deviation interval division: Taking 10 spatial nodes within a certain airspace monitoring interval as an example, the risk deviation sequence is: [+0.5, +0.3, -0.1, +0.7, -0.2, +0.4, -0.3, +0.6, -0.4, +0.2], where positive values ​​indicate positive deviations (increased risk) and negative values ​​indicate negative deviations (decreased risk). The system divides consecutive deviations of the same sign into intervals: Positive offset range: nodes 1-2 (+0.5, +0.3), node 4 (+0.7), node 6 (+0.4), node 8 (+0.6), node 10 (+0.2); Negative offset range: node 3 (-0.1), node 5 (-0.2), node 7 (-0.3), node 9 (-0.4).

[0083] Rate extraction and weight fusion: For the positive deviation interval, calculate the convergence rate of the risk deviation (i.e. the magnitude of the deviation value decrease per unit time). For example, the deviation value of node 1-2 decreases from +0.5 to +0.3, with a time interval of 5 minutes, the convergence rate is (0.5-0.3) / 5=0.04 / minute. For the negative deviation interval, calculate the diffusion rate of the risk deviation (i.e. the magnitude of the deviation value increase per unit time), such as the deviation value of node 9 increases from -0.4 to -0.3 (absolute value decreases), the diffusion rate is (-0.3-(-0.4)) / 5=0.02 / minute (here the actual is convergence, but if the negative deviation indicates risk reduction, the diffusion rate can be understood as the risk rebound speed).

[0084] The trajectory weight of each spatial node is set according to its position importance, for example, the weight of the node near the airport runway is 0.9, and the weight of the node far from the airport is 0.3. The convergence rate and the diffusion rate are fused according to the node weight, for example, the convergence rate of node 1 (weight 0.9) in the positive interval is 0.04, and the convergence rate of node 2 (weight 0.7) is 0.03, and the weighted average is obtained as the comprehensive convergence rate of the positive interval:

[0085] The diffusion rate of node 9 (weight 0.8) in the negative interval is 0.02, and the diffusion rate of node 7 (weight 0.5) is 0.01, and the weighted average is obtained as the comprehensive diffusion rate of the negative interval:

[0086] Early warning response scheme generation: According to the comprehensive convergence rate and diffusion rate, generate adjustment parameters to drive different early warning measures: If the comprehensive convergence rate of the positive interval is lower than the preset threshold (such as 0.02 / minute), it indicates that the risk continues to rise and is not controlled, triggering a high-risk warning, starting the ground radar tracking and aircraft communication blocking plan; If the diffusion rate of the negative interval is higher than the preset threshold (such as 0.03 / minute), it indicates that the risk reduction trend is reversed, and there may be a secondary risk, triggering a medium-risk warning, increasing the sensor scanning frequency of this area; If both are within the safe range, maintain the current monitoring level, and only record the risk deviation data.

[0087] Taking the comprehensive convergence rate of the positive interval 0.036 / minute as an example, the value is higher than the preset threshold 0.02 / minute, indicating that the risk increase trend is alleviated, the system generates a yellow warning (medium risk), notifies the airspace administrator through a short message, and highlights the relevant nodes on the monitoring interface, prompting to strengthen manual patrol.

[0088] Three, dynamic scene application example Scenario 1: Sudden Meteorological Risk: A summer thunderstorm occurred in a low-altitude airspace. The system detected a sudden increase in the safety offset of multiple nodes to 1.2, matching the historical thunderstorm pattern (first dynamic reference value 0.85). The current number of abnormal peaks was 6 (the historical pattern was 4, with a difference of +2). The combined reference value was calculated as 0.7 × 0.85 + 0.3 × 2 = 1.195, and the optimal matching thresholds were an altitude abnormality of 10 meters and a wind speed abnormality of 25%. The risk deviation sequence showed multiple nodes with positive deviations and a convergence rate of 0.01 / minute (below the threshold), triggering a red alert (high risk). The system automatically sent a return command to all aircraft in flight and broadcast a warning to ground personnel to stop launching drones.

[0089] Scenario 2: Large-Scale Event Support: While performing a low-altitude security mission around a sports stadium, the system detected a safety deviation of 0.6, matching the historical event support pattern (first dynamic reference value 0.92). The number of abnormal peaks was consistent with the historical pattern (difference 0). The optimal thresholds were set at an electromagnetic interference intensity of 40dBμV and an aircraft density of 5 aircraft / square kilometer. Some nodes in the risk deviation sequence showed negative deviations (reduced electromagnetic interference), but the diffusion rate was 0.04 / minute (higher than the threshold of 0.03 / minute), indicating a potential deterioration in the electromagnetic environment. The system issued an orange alert (medium risk), activated backup electromagnetic monitoring equipment, and coordinated with the drone control team to strengthen patrols around the venue.

[0090] 4. System Collaboration and Data Closed Loop The threshold optimization and warning decision-making modules must interact with other modules in real time: the threshold optimization module's optimal threshold value is fed back to the environmental perception module to adjust the baseline parameters for signal feature analysis; the warning decision-making module's response instructions are synchronized with the aircraft monitoring module to update the safety level of the monitored interval (for example, from level 2 to level 1); and historical warning data is stored in the environmental feature library to optimize future risk pattern matching and threshold derivation logic. Through this closed-loop mechanism, the system continuously learns from changes in airspace characteristics, improving the accuracy of threshold setting and the relevance of warning responses, achieving intelligent management of the entire process from risk monitoring and threshold adaptation to decision execution.

[0091] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "includes," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0092] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A low-altitude aircraft and ground equipment safety monitoring and early warning system, characterized in that: include: Aircraft monitoring module, used to collect dynamic position data of low-altitude aircraft in real time and set airspace monitoring intervals corresponding to safety levels; The environmental perception module is used to divide the airspace monitoring area into multiple spatial nodes, analyze the signal characteristics of the environmental data of each spatial node, and generate the security feature vector corresponding to the spatial node; The risk location module is used to extract key abnormal indicators from the security feature vector, establish risk classification rules associated with spatial nodes, and obtain signal correction parameters corresponding to the rules; Dynamic prediction module, used to identify trajectory correlation patterns in signal correction parameters, dynamically calibrate key anomaly indicators based on the correlation patterns, and calculate the safety offset of each spatial node under different calibration strategies; The threshold optimization module is used to derive the optimal safety threshold based on the safety offset and generate a risk deviation sequence by comparing the current environmental characteristic value with the optimal safety threshold; The early warning decision module is used to analyze the risk deviation sequence and integrate the risk deviation sequence into an early warning response decision plan based on the safety deviation trend of the spatial nodes.

2. The low-altitude aircraft and ground equipment safety monitoring and early warning system according to claim 1 is characterized in that: The implementation of the environment perception module includes: building an environment feature library corresponding to the spatial node, the environment feature library contains the environment data and the security parameter vector of the signal feature mapping; Similar signal matching is performed on the security parameter vector, and the spatial clustering groups of the security parameter vector are divided according to the matching results; the signal focus points of the environmental data are extracted from the spatial clustering groups, and the focus points are set as spatial nodes.

3. The low-altitude aircraft and ground equipment safety monitoring and early warning system according to claim 2 is characterized in that: The spatial clustering groups that divide the security parameter vectors also include: According to the spatial attributes and risk intensity in the safety parameter vector, the signal distribution, anomaly polarity and environmental density parameters are extracted, and the environmental feature labels are generated based on the above parameters; The environmental feature labels are associated with the security parameter vectors. By calculating the signal similarity between the feature labels, the parameter vectors with similarity higher than the preset security threshold are screened to form a spatial clustering group.

4. The low-altitude aircraft and ground equipment safety monitoring and early warning system according to claim 1 is characterized in that: The implementation methods of generating the security feature vector corresponding to the spatial node include: For each spatial node, according to the node's trajectory position in the airspace monitoring interval, the node's signal fluctuation data within the preset window is obtained, and the node's signal fluctuation coefficient is calculated; When the signal fluctuation coefficient exceeds the first safety threshold, the node is marked as a high-risk node, and its environmental data is extracted to form a safety feature vector; when the signal fluctuation coefficient is lower than the first safety threshold, the node is marked as a steady-state node, and the environmental data of the node's adjacent nodes are superimposed, and the superimposed data is reconstructed into a safety feature vector.

5. The low-altitude aircraft and ground equipment safety monitoring and early warning system according to claim 1 is characterized in that: The implementation methods of the risk positioning module include: Separate the height anomaly parameters, speed anomaly parameters, and trajectory fluctuation parameters from the safety feature vector, and generate risk classification rules for spatial nodes based on the above parameters; If the number of spatial nodes covered by the current risk classification rule is less than the preset safety threshold, the security feature vectors of adjacent spatial nodes are traversed, and the signal indicators not included in the classification rules of the adjacent nodes are added to the current rule.

6. The low-altitude aircraft and ground equipment safety monitoring and early warning system according to claim 1 is characterized in that: The implementation of the dynamic prediction module includes: obtaining the timing factor of the signal switching frequency and the fluctuation factor of the trajectory jump amplitude in the trajectory association mode; A spatial weight matrix associated with the time series factor and the volatility factor is constructed, and the safety offset under different calibration strategies is determined according to the distribution probability of each element in the matrix.

7. The low-altitude aircraft and ground equipment safety monitoring and early warning system according to claim 6 is characterized in that: Constructing a spatial weight matrix also includes: Identify the trajectory period characteristics of the timing factor. If the current period characteristics completely match the preset signal period, set the timing factor as the starting index of the spatial weight matrix. Calculate the trajectory matching degree of the time series factor and the volatility factor, and generate the intermediate index and terminal index of the spatial weight matrix in descending order of matching degree; The path of the termination index is backtracked, and when the matching degree of the termination index is lower than the preset matching threshold, it is output as the final distribution of the spatial weight matrix.

8. The low-altitude aircraft and ground equipment safety monitoring and early warning system according to claim 7 is characterized in that: The implementation methods for calculating the safety offset include: Statistically calculate the mean and range of the time series factor of each terminal index in the spatial weight matrix, and calculate the global variance of all index factors; The timing offset coefficient is obtained by subtracting the mean of the timing factor of a single termination index from the mean of the timing factor of the adjacent index and dividing it by the global variance. At the same time, the ratio of the volatility factor range to the global variance is calculated, and the weighted sum of the two is used as the safety offset of the index.

9. The low-altitude aircraft and ground equipment safety monitoring and early warning system according to claim 1 is characterized in that: The implementation methods for deriving the optimal safety threshold include: Extract the risk pattern closest to the current safety offset from the historical data and calculate the cosine similarity between the two in trajectory distribution as the first dynamic reference value; Count the difference in the number of abnormal peaks between the current safety offset and the historical risk pattern, and use the difference as the second dynamic reference value; Based on the linear combination of the first dynamic reference value and the second dynamic reference value, an optimal safety threshold in a preset safety threshold table is matched.

10. The low-altitude aircraft and ground equipment safety monitoring and early warning system according to claim 1 is characterized in that: The implementation of the early warning decision module includes: dividing the positive deviation interval and the negative deviation interval according to the safety deviation direction of each node in the risk deviation sequence; The convergence rate of risk deviation in the positive offset interval and the diffusion rate of risk deviation in the negative offset interval are extracted, and the two are conflict-fused according to the trajectory weights of spatial nodes to generate the adjustment parameters of the early warning response decision-making plan.

Images