A GNSS spoofing jamming detection and positioning method and device based on ADS-B open source data

Abstract

The application belongs to the technical field of satellite navigation, and discloses a GNSS deception jamming detection and positioning method and device based on ADS-B open source data, which comprises the following steps: acquiring ADS-B track data; performing preliminary screening on track points based on a speed threshold, and classifying and marking the track points as normal track points, controlled false track points, entering boundary points and leaving boundary points; performing re-screening on the entering boundary points and the leaving boundary points, removing invalid points through space-time correlation constraints, and obtaining effective edge nodes; clustering the effective edge nodes, and dividing the edge nodes belonging to the same deception jamming source into the same cluster; fitting the edge nodes in each cluster into an arc, taking the center position of the fitted arc as the positioning result of the corresponding deception jamming source; and performing positioning accuracy evaluation and cross-period space-time stability verification. The application can realize low-cost, high-reliability and GNSS deception jamming source positioning suitable for non-cooperative areas.

Description

Technical Field

[0001] This invention relates to the field of satellite navigation technology, and in particular to a method and apparatus for detecting and locating GNSS spoofing interference based on ADS-B open-source data. Background Technology

[0002] Existing GNSS deception interference source localization technologies can be mainly divided into two categories based on different implementation paths: ground-based monitoring methods based on dedicated receiver networks and aerial surveillance data-driven methods based on ADS-B open-source data.

[0003] (I) Ground-based monitoring method based on dedicated receiver network

[0004] This method constructs a ground monitoring network by deploying multiple GNSS monitoring receivers and uses quality indicators such as pseudorange residuals, carrier-to-noise ratio changes, and Doppler consistency to identify deception interference. At the positioning level, a Time Difference of Arrival (TDOA) system is commonly used. The basic implementation steps are as follows: First, the arrival time or pseudorange observations of the deception signal are extracted from each monitoring station; second, common error terms such as receiver clock bias are eliminated through single or double pseudorange differences between stations to obtain the distance difference observations from the deception source to each station; then, a hyperbolic equation is constructed with the stations as the focus, and the location of the interference source is determined through hyperbolic intersection or least squares fitting. Recent studies have further introduced improved strategies such as AOA / TDOA joint positioning, FDOA / TDOA fusion, and intelligent optimization search to enhance accuracy.

[0005] The advantage of this type of method lies in its ability to achieve high-precision interference source localization at the signal level. However, its main drawbacks are: it is highly dependent on the dense deployment of dedicated monitoring receivers, making it difficult to implement in non-cooperative areas, marine areas, or non-cooperative environments lacking ground infrastructure; it is also costly and has limited coverage. Furthermore, existing detection methods are mostly based on residual verification strategies under the receiver's autonomous integrity monitoring framework, failing to fully utilize the spatial consistency constraints caused by the co-source propagation of deception signals under multi-station collaborative observation conditions. There is still room for improvement in the balance between detection sensitivity and false alarm rate.

[0006] (II) Data-driven approach for airborne surveillance based on ADS-B open-source data

[0007] With the establishment of global open-source ADS-B sensor networks such as OpenSky and ADS-B Exchange, this type of method achieves the detection and localization of deception interference through the mining and analysis of massive flight track data. Currently, most domestic and international research on GNSS interference based on ADS-B data focuses on suppression interference, with few specific studies on large-scale deception in regional conflicts. In terms of interference detection, existing methods mainly extract navigation quality parameters such as NACp and NIC for statistical anomaly discrimination, or identify disturbed flights by analyzing the location of abnormal track interruptions and signal power attenuation characteristics. Moreover, most existing algorithms rely on navigation quality indicators in ADS-B data to complete the perception and localization of interference, which is difficult to adapt to the problem of good signal self-consistency and the failure of traditional detection methods in GNSS deception scenarios. In terms of localization, existing methods are mainly divided into two technical routes: one is based on the spatial distribution of the interfered track points to perform grid search or probability matching to determine the approximate area of ​​the interference source; the other is based on the spatial geometric constraints of deception interference, extracting track jump points and performing spatial clustering and circular arc fitting, using the center of the fitted circle as the estimated location of the interference source.

[0008] The core advantage of this type of method is that it can directly utilize open-source ADS-B data without the need to deploy dedicated monitoring equipment. It is low-cost, has wide coverage, and is particularly suitable for interference situational awareness in non-cooperative overseas areas. However, existing methods still have the following significant shortcomings: First, the outlier discrimination mechanism is relatively simple, mainly relying on speed thresholds or navigation quality indicators, lacking multi-level joint discrimination based on data continuity, position convergence characteristics, and statistical significance. It is easy to misjudge non-deceptive factors such as route turns, weather turbulence, and data interruptions as outliers, and it is difficult to effectively distinguish mixed point sets caused by multiple interference sources. Second, the accuracy of the grid search method on which positioning is based is limited by the grid size, and the single global clustering in the circular arc fitting method is difficult to effectively separate different interference source point sets that are close to the deceptive target area, resulting in insufficient cluster purity. Third, the circular arc fitting stage generally lacks a systematic robustness processing strategy, and there is no effective mechanism for identifying and removing gross errors in the data. The least squares fitting results are easily dominated by gross errors, resulting in insufficient robustness of the positioning results. Fourth, most studies only provide single positioning results and lack spatiotemporal consistency verification methods such as cross-time confidence ellipse analysis and two-way coincidence rate matrix, making it difficult to evaluate the stability and reliability of the interference source positioning results in the time dimension.

[0009] In summary, existing technologies still have significant shortcomings in terms of applicability in non-cooperative areas, reliable detection of anomalies, separation of multi-source mixed point sets, robustness, and verification of localization results. Summary of the Invention

[0010] To address the shortcomings of the existing technologies, this invention provides a GNSS spoofing interference detection and localization method and apparatus based on ADS-B open-source data. It aims to solve technical problems such as the difficulty in deployment in non-cooperative environments, the failure of navigation quality indicators in spoofing scenarios, the single anomaly point discrimination mechanism, the difficulty in effectively separating mixed point sets of multiple interference sources, and the lack of robustness processing in circular arc fitting. This invention achieves low-cost, high-reliability GNSS spoofing interference source localization applicable to non-cooperative areas.

[0011] To achieve the above objectives, the present invention adopts the following technical solution:

[0012] This invention proposes a GNSS spoofing interference detection and localization method based on ADS-B open-source data, comprising:

[0013] Acquire ADS-B track data, which includes at least flight identifier, timestamp, and location coordinates;

[0014] Initial screening of waypoints is performed based on speed thresholds, classifying them into normal waypoints, controlled false waypoints, entry boundary points, and exit boundary points. Controlled false waypoints are those with speeds below a first set threshold, while entry and exit boundary points are those with speeds above a second set threshold. Entry and exit boundary points are then further screened, and invalid points are removed using spatiotemporal correlation constraints to obtain valid edge nodes.

[0015] Cluster the effective edge nodes, grouping edge nodes belonging to the same deception interference source into the same cluster;

[0016] For each cluster, the edge nodes are fitted with circular arcs, and the center position of the fitted circle is used as the localization result of the corresponding deception interference source.

[0017] Furthermore, the spatiotemporal correlation constraints include at least one of the following: deception region constraints, recapture duration constraints, deception time period constraints, and cruise elevation constraints;

[0018] The deception region constraint is to assign a region to each boundary point and remove stray points based on the spatial density clustering results of the controlled false track points.

[0019] The recapture duration constraint is based on a set threshold for the recapture time interval after deception, which excludes timeout jump points.

[0020] The deception period constraint is based on the temporal distribution of controlled false track points to determine the effective deception period and remove boundary points outside the period.

[0021] The cruise elevation constraint is based on a set elevation threshold to exclude boundary points during low-altitude take-off and landing phases.

[0022] Furthermore, the clustering employs a two-stage clustering strategy:

[0023] In the first stage, spatial clustering is performed using the location of the controlled spurious track points corresponding to each edge node as the clustering object, and edge nodes with similar convergence locations are aggregated into the same convergence cluster.

[0024] In the second stage, for each edge node within a convergence cluster, spatial clustering is performed again using its position before entering the deception zone as the clustering object. This further subdivides the point sets from different spatial orientations within the same convergence cluster into subclusters corresponding to different deception sources.

[0025] Furthermore, after completing the two-stage clustering, the process also includes: calculating the average position of all edge nodes within each sub-cluster as the cluster centroid, and merging sub-clusters whose centroid distance is less than the merging threshold.

[0026] Furthermore, the circular arc fitting employs a three-step hybrid robustness strategy:

[0027] The first step is to solve for the robust initial center of the circle based on Tukey's two-weight M-estimation;

[0028] The second step is to eliminate extreme outliers relative to the initial center of the circle based on the interquartile range criterion.

[0029] The third step is to perform robust iterative fine fitting based on the IGGIII equivalent weight function, and use the center of the fitted circle after convergence as the localization result of the deception interference source.

[0030] Furthermore, the first set threshold is 100 m / s, and the second set threshold is 500 m / s.

[0031] Furthermore, it also includes: dividing the full-time data into multiple sub-time periods according to time, fitting 95% confidence ellipses to the edge nodes of each sub-time period, and judging the stability of the location of the interference source by comparing the degree of overlap and centroid clustering of the confidence ellipses of each time period; and / or, constructing a two-way overlap rate matrix, calculating the two-way overlap rate of edge nodes between each time period, and quantitatively evaluating the consistency of the spatial distribution of edge nodes in different time periods.

[0032] Furthermore, after acquiring the ADS-B track data, the method further includes: performing an integrity check on the data, removing invalid records with missing timestamps or abnormal coordinates, and performing trajectory splicing and repair on the segmented tracks to reconstruct a complete track sequence.

[0033] This invention also proposes a GNSS spoofing interference detection and positioning device based on ADS-B open-source data, comprising:

[0034] The data acquisition module is used to acquire ADS-B track data, which includes at least flight identifier, timestamp, and location coordinates;

[0035] A dual-layer extraction module is used to initially screen waypoints based on a velocity threshold, classifying and labeling them as normal waypoints, controlled false waypoints, entry boundary points, and exit boundary points. The controlled false waypoints are those with velocities below a first set threshold, while the entry and exit boundary points are those with velocities above a second set threshold. The entry and exit boundary points are then further screened, and invalid points are removed using spatiotemporal correlation constraints to obtain valid edge nodes.

[0036] The clustering module is used to cluster the effective edge nodes, grouping edge nodes belonging to the same deception interference source into the same cluster;

[0037] The localization module is used to perform circular arc fitting on the edge nodes within each cluster, and the center position of the fitted circle is used as the localization result of the corresponding deception interference source.

[0038] Furthermore, it also includes:

[0039] The spatiotemporal stability assessment module is used to divide the full-time data into multiple sub-time periods according to time, fit 95% confidence ellipses to the edge nodes of each sub-time period, and judge the stability of the location of the interference source by comparing the degree of overlap and centroid clustering of the confidence ellipses of each time period; and / or, construct a two-way overlap rate matrix, calculate the two-way overlap rate of edge nodes between each time period, and quantitatively assess the consistency of the spatial distribution of edge nodes in different time periods.

[0040] The present invention also proposes a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described in any of the preceding claims.

[0041] Compared with the prior art, the present invention has the following beneficial effects:

[0042] 1. Wide-area deployment without dedicated equipment. This invention directly utilizes open-source ADS-B data, eliminating the need for dedicated monitoring receivers and completely freeing traditional ground-based monitoring methods from dependence on ground infrastructure. It is particularly suitable for deception and interference situational awareness in non-cooperative environments.

[0043] 2. Reliable detection of abnormal transition points in deception scenarios. This invention does not rely on easily failing navigation quality parameters such as NACp and NIC. It adopts a two-layer extraction architecture of "initial screening + secondary screening": the initial screening stage achieves preliminary separation of track anomalies based on velocity thresholds; the secondary screening stage constructs four spatiotemporal correlation criteria: deception area constraint, recapture duration constraint, deception time period constraint, and cruise elevation constraint, to eliminate various non-edge pseudo-anomalies layer by layer, effectively solving the technical problem of easy misjudgment by single threshold discrimination.

[0044] 3. Effective Separation and Robust Localization of Multi-Source Hybrid Point Sets. This invention employs a two-stage DBSCAN clustering and centroid merging strategy to effectively separate source point sets in complex scenarios where multiple deception sources coexist and target areas are close together, overcoming the shortcomings of insufficient purity in traditional single-clustering. A three-step hybrid robust strategy of "Tukey dual-weight M-estimation + interquartile range coarse elimination + IGGIII iterative fine fitting" is constructed, significantly improving localization robustness and accuracy.

[0045] 4. Verification of the spatiotemporal stability of positioning results. This invention introduces two sets of cross-time period consistency verification methods: 95% confidence elliptic analysis and two-way coincidence rate matrix, effectively solving the problem that existing studies lack criteria for evaluating the spatiotemporal stability of positioning results. Attached Figure Description

[0046] Figure 1 The basic principle of the ADS-B system provided in the embodiments of the present invention;

[0047] Figure 2 This is a schematic diagram of the real flight track and ADS-B flight track of a single flight penetrating the deception range, provided in an embodiment of the present invention.

[0048] Figure 3 This is one of the flowcharts for a GNSS spoofing interference detection and localization method based on ADS-B open-source data provided in an embodiment of the present invention;

[0049] Figure 4 A second flowchart of a GNSS spoofing interference detection and localization method based on ADS-B open-source data provided in this embodiment of the invention;

[0050] Figure 5 This is one of the structural schematic diagrams of a GNSS spoofing interference detection and positioning device based on ADS-B open source data provided in an embodiment of the present invention;

[0051] Figure 6 This is the second schematic diagram of a GNSS spoofing interference detection and positioning device based on ADS-B open-source data, provided as an embodiment of the present invention. Detailed Implementation

[0052] The present invention will be further explained below with reference to the accompanying drawings and specific embodiments:

[0053] ADS-B (The basic principles of the ADS-B system are as follows) Figure 1(As shown in Table 1) is a traffic surveillance technology based on GNSS and air-to-ground data links: the aircraft acquires positioning data through an onboard GNSS receiver, integrating information from sensors such as barometric altitude and three-dimensional velocity. The core navigation parameters and characteristics of ADS-B are shown in Table 1. When GNSS is interfered with or spoofed, the onboard GNSS receiver, after being spoofed, will output a false track and simultaneously generate self-consistent navigation quality parameters such as NACp and NIC. Therefore, it is difficult to locate the spoofing through navigation quality performance parameters; it is necessary to rely on the ADS-B track to obtain the coordinates of the edge nodes of the spoofing source, and then fit the data to obtain the location information of the spoofing source.

[0054] Table 1. Core Navigation Parameters and Characteristics of ADS-B

[0055]

[0056] Navigation deception targeting specific high-value targets can cause collateral damage to nearby civil aviation due to excessive power. In this case, the civil aviation encounters asynchronous primary deception, which directly leads to the loss of tracking loop in the GNSS receiver, causing the flight track to instantly jump to the preset deception target area. The instantaneous ground speed calculated from adjacent track points will far exceed the maximum cruising speed of civil aircraft (approximately 300 m / s). The track points before and after these abnormal speed jumps are the edge nodes of the deception source's influence range and are key to determining the scope of the deception's effect.

[0057] For continuously active deception sources, a single civil aviation flight entering or leaving the deception range will create two points of abnormally high velocity jumps, namely the edge nodes entering and leaving the deception range. Because the power threshold difference between receiver loss of lock and acquisition is approximately 5dB, this causes significant asymmetry between the two types of edge nodes: edge nodes entering the deception area are closer to the center of the deception source, and their locations are concentrated on the same isopower circle; edge nodes leaving the deception area, due to the combined effects of receiver reacquisition delay and the difference in acquisition threshold, will have their recovered track points deviate from the true edge node's circumference by tens of kilometers after leaving the deception area. Figure 2 As shown. Therefore, when determining the deception range, it is appropriate to use only the track transition points entering the deception zone as observations of the edge node positions.

[0058] In real-world scenarios, there are two types of non-edge node track jump points that significantly affect the determination of the true coverage range of the deception source: The first type is jump points caused by the intermittent power-on of the deception source. High-power deception sources will start and stop intermittently according to mission requirements. At the moment of start and stop, all flights within the deception range will simultaneously experience track jumps. These types of track jump points cannot reflect the true deception range. The second type is jump points during low-altitude take-off and landing. At this time, the flight altitude is low, and the deception distance is limited by the line-of-sight conditions rather than the range of the deception source due to the curvature of the earth and terrain obstruction.

[0059] Based on the above analysis, such as Figure 3 As shown, this invention provides a GNSS spoofing interference detection and localization method based on ADS-B open-source data, including:

[0060] S101: ADS-B data acquisition and preprocessing.

[0061] This invention obtains raw ADS-B track data of global flights from open-source platforms such as ADS-B Exchange. The raw data includes fields such as flight number, UTC timestamp, latitude and longitude coordinates, barometric altitude, and ground speed. First, the data undergoes an integrity check, removing invalid records with missing timestamps or obviously abnormal coordinates. Then, track stitching and repair are performed, as segmented tracks are correlated and matched based on flight number and UTC timestamp to reconstruct a complete and continuous spatiotemporal track sequence for a single flight. Based on this, derived features such as time intervals, great circle distances, instantaneous ground speeds, and heading change rates between adjacent track points are calculated point-by-point, providing multi-dimensional information support for subsequent anomaly detection. The preprocessed multi-flight track data is then uniformly visualized, providing an intuitive reference for deception characteristic analysis.

[0062] S102: Dual-layer extraction of track jump anomalies.

[0063] This invention designs a two-layer progressive extraction architecture of "primary screening + secondary screening" to achieve high-quality extraction of deception source edge nodes.

[0064] The initial screening stage uses speed thresholds to preliminarily extract track anomalies. Excluding takeoff and landing phases, the normal cruising speed of civil aircraft is generally in the range of 200-300 m / s. To avoid the influence of data gross errors and abnormal jitter, this invention sets two speed thresholds: one is an abnormal steady-state threshold, where a cruising speed below 100 m / s is identified as a controlled steady-state false track point. These points correspond to flights locked to false coordinates and maintaining a stable motion state after being deceived by GNSS. The second is an abnormal abrupt change threshold, where a cruising speed exceeding 500 m / s is identified as a track jump point. These points correspond to track discontinuities generated when a flight switches between normal navigation and deceived states. Track point attributes are categorized into four types and marked: normal true track points, controlled false track points, the last normal track point before entering the deception range (entry boundary point), and the first normal track point after leaving the deception range (exit boundary point). After initial screening, controlled false trackpoints are used for subsequent identification of deceiving target areas, while entry and exit boundary points are selected as candidate edge nodes for secondary screening.

[0065] The secondary screening stage employs a four-tiered spatiotemporal constraint system for refined composite screening, eliminating invalid points layer by layer. The first tier is the deception region constraint: spatial density clustering is performed on all controlled false track points marked in the initial screening to determine the latitude and longitude range of each deception target area and number each target area sequentially. Based on this, track points are screened region by region, and track points falling within each deception target area, along with their corresponding entry and exit boundary points, are re-marked according to region numbering. This achieves accurate differentiation of multiple deception sources while eliminating stray false anomalies that do not fall within any deception target area. The second tier is the reacquisition time constraint: a threshold for the reacquisition time interval after deception is set, and jump points exceeding this threshold are eliminated. This is because jump points with excessively long reacquisition times are significantly affected by receiver reacquisition delay and acquisition threshold differences, causing their positions to deviate from the true edge node's circumference range, introducing significant gross errors. The third constraint is the deception period constraint: based on the intraday second sequence of controlled false trackpoints, continuous deception periods and non-deception periods are determined according to the minimum start-stop interval threshold of the deception source. Boundary points not within the effective deception period are eliminated, thereby eliminating non-edge transition points caused by intermittent start-stop of the deception source. The fourth constraint is the cruise elevation constraint: a cruise elevation threshold is set for edge nodes, retaining only entry boundary points above this threshold and their corresponding transition points, eliminating boundary points affected by the Earth's curvature and terrain obstruction during low-altitude take-off and landing phases, and where the deception distance is limited by line-of-sight conditions rather than the actual coverage range of the deception source. After the above four constraints are used for screening, all effective edge nodes are summarized and classified by region to generate a standardized partitioned edge node dataset.

[0066] It should be noted that, due to the approximately 5dB difference between the power thresholds of receiver loss of lock and acquisition, the edge nodes entering and leaving the deception zone exhibit significant asymmetry: entry boundary points are concentrated on an equal-power circle centered on the deception source, while exit boundary points deviate from the true circle by tens of kilometers due to the reacquisition delay. Therefore, in the subsequent positioning stage, this invention only uses the track jump points entering the deception zone (i.e., entry boundary points) as edge node observations.

[0067] S103: Separation of interference source clusters by two-stage DBSCAN clustering and centroid merging.

[0068] For the partition edge nodes extracted by S102, since there may be multiple deception sources working at the same time and edge nodes of different deception sources are mixed and distributed, this invention adopts a two-stage DBSCAN clustering strategy combined with a centroid merging method to achieve accurate division of interference source clusters.

[0069] The first stage is convergence location clustering, which uses the controlled spoofed flight path points (i.e., the spoofed locations that stably converge after being deceived) corresponding to each edge node as the clustering objects and employs the DBSCAN algorithm for spatial clustering. Since the same deception source deceives flights to the same target area, edge nodes with close convergence locations naturally aggregate into the same candidate cluster.

[0070] The second stage is spatial location clustering. For each edge node within a convergent cluster, its position before the jump to enter the deception zone is used as the clustering object, and the DBSCAN algorithm is used again for clustering. Since the effective coverage of the same deception source exhibits an arc distribution characteristic centered on the interference source, the positions before the jump from the same deception source are distributed within a certain angular range centered on that interference source. This method can further subdivide the set of points from different spatial orientations within the same convergent cluster into subclusters corresponding to different deception sources.

[0071] After completing the two-stage clustering, a centroid merging step is introduced: the average latitude and longitude of all edge nodes within each sub-cluster are calculated as the cluster centroid. If the distance between the centroids of two clusters is less than the merging threshold, a disjoint-set data structure algorithm is used to merge them. This step effectively solves the problem of the same deception source being split into multiple clusters due to signal interruption or other reasons, ensuring that each final cluster corresponds to an independent deception interference source.

[0072] S104: Three-step hybrid robust least squares circular arc fitting and interference source localization.

[0073] Within each cluster obtained via S103, all effective edge nodes (entry boundary points) are distributed on the same arc centered at the deceptive interference source. Based on this geometric constraint, this invention uses arc fitting to infer the location of the interference source. Since gross errors inevitably exist in actual data, this invention constructs a three-step hybrid robustness strategy for robust fitting.

[0074] The first step is to solve for the robust initial center of the circle based on Tukey's dual-weight M-estimation: using the median of the planar coordinates of all edge nodes within the cluster as the initial center, the center position is updated iteratively with weights. In each iteration, the distance from each point to the current center is calculated, the robust scale is estimated based on the median absolute deviation, the standardized residuals are calculated, the weights of each point are determined by the Tukey dual-weight function, and the center is updated with the weighted median until convergence. Tukey's dual-weight estimation has a collapse rate of 50%, exhibiting strong resistance to gross errors and providing a robust initial center.

[0075] The second step is outlier removal based on the interquartile range criterion: Referring to the robust initial circle center obtained in the first step, calculate the Euclidean distance from each edge node to the circle center. Take the first and third quartiles of the distance sequence and calculate the interquartile range. Set a removal threshold, and identify points whose distance exceeds the threshold as extreme outliers and remove them, thus purifying the dataset participating in subsequent fitting.

[0076] The third step is robust iterative fine fitting based on the IGGIII equivalent weight function: weighted least squares circle fitting is performed on the retained points. After fitting, the radius residuals of each point are calculated, and robust scaling is estimated using the median absolute deviation to obtain standardized residuals. The observation points are classified according to the IGGIII three-segment weight function: points with smaller residuals retain unit weights, points with medium residuals are smoothly weighted down, and points with extremely large residuals are completely rejected. The weighted fitting is then performed again with the updated weight factors, iterating until the change in weight factors is less than a set threshold or the maximum number of iterations is reached. Finally, a final fine fitting is performed with the converged weight factors to obtain the coordinates of the fitted circle's center and radius, where the center coordinates represent the estimated location of the deceptive interference source.

[0077] Furthermore, such as Figure 4 As shown, after S104, the following is also included:

[0078] S105: Positioning accuracy assessment and cross-time period spatiotemporal stability verification.

[0079] This invention comprehensively evaluates positioning results from two dimensions: internal consistency accuracy and spatiotemporal stability. Regarding internal consistency accuracy, the radial residuals from the effective edge nodes involved in the fitting to the center of the fitted circle are calculated, and then accuracy indicators such as residual mean square error and coordinate component mean square error are assessed. Furthermore, using the center of the fitted circle as a reference, the azimuth distribution of each effective point is calculated, and an angular coverage index is defined to evaluate the spatial coverage of the fitted arc, ensuring the reliability of the fitted geometry.

[0080] Regarding spatiotemporal stability, this invention introduces a segmented consistency analysis method. The full-time data is divided into multiple sub-time periods, and a 95% confidence ellipse is fitted to the edge nodes of each sub-time period. By comparing the overlap and centroid clustering of the core regions of the confidence ellipses in each time period, the stability of the interference source's location in the temporal dimension is determined. Simultaneously, a bidirectional overlap rate matrix is ​​constructed, and the bidirectional overlap rate of edge nodes between each time period is calculated using a set distance threshold, quantitatively assessing the consistency of the spatial distribution of edge nodes across different time periods. These verification methods effectively support the determination of the continuous operating state and location stability of deceptive interference sources within a long-term window.

[0081] In summary, this invention achieves reliable acquisition of edge nodes through a two-layer progressive extraction architecture, effectively separates mixed point sets of multiple interference sources through two-stage clustering and centroid merging, achieves robust interference source localization through three-step hybrid robust fitting, and verifies cross-time stability through confidence ellipse and coincidence rate matrix. Finally, it forms a complete GNSS spoofing interference detection and localization technology solution based on ADS-B open source data.

[0082] As a specific implementation method, the present invention also proposes another GNSS spoofing interference detection and localization method based on ADS-B open-source data, including:

[0083] S201: ADS-B Data Acquisition and Preprocessing.

[0084] This invention obtains raw ADS-B flight track data from the ADS-B Exchange open-source platform (https: / / globe.adsbexchange.com). The data format is KML (Keyhole Markup Language), but JSON or CSV formats can also be used as alternative data sources. The raw data includes at least the flight number (call sign), UTC timestamp, and latitude and longitude coordinates, and preferably also includes barometric altitude and ground speed information. In the KML file, time information is stored in... <when>In the tag, geographic coordinates are stored in <gx:coord>The two are in a one-to-one correspondence within the tags.

[0085] As one possible implementation, in this embodiment, the preprocessing operation is automatically executed on the computing server as a software program, including the following sub-steps:

[0086] (1) Trajectory splicing and repair: Read the original segmented track data, perform correlation matching based on flight number and UTC time stamp, and reconstruct a complete and continuous spatiotemporal track sequence for a single flight. The specific method is as follows: arrange track points with the same flight number in ascending order of UTC time. If the time interval between two adjacent track segments is less than a set threshold (preferably 300 seconds), they are determined to be data from the same flight and spliced; if the time interval exceeds the threshold, they are determined to be different segments and processed independently. This step avoids the misjudgment of abnormal points caused by the erroneous amplification of the time interval between adjacent track points due to missing data.

[0087] (2) Calculation of derived features: Based on the preprocessed continuous track sequence, the spatiotemporal distance and instantaneous ground speed of adjacent track points are calculated point by point. The instantaneous ground speed is calculated using the Haversine great circle distance formula, as shown in Equation (1), to eliminate the influence of the Earth's curvature on the distance calculation.

[0088]

[0089] (3) Track visualization: The latitude and longitude coordinates of the preprocessed track points are converted into local plane coordinates of the station center with the center of the target area as the origin. The conversion method adopts the Northeast-Eastern-Upper-Southern (ENU) coordinate system projection. The preprocessed multi-flight track data is uniformly visualized to provide an intuitive reference for determining the clustering range in the subsequent deception area constraint.

[0090] S202: Dual-layer extraction of track jump anomalies.

[0091] As one possible implementation method, in this embodiment, the present invention adopts a two-layer progressive extraction architecture of "initial screening + secondary screening", and all discrimination criteria are automatically executed by software program without manual intervention.

[0092] S202.1: Initial screening stage: Preliminary extraction of track anomalies based on speed threshold determination.

[0093] For the preprocessed continuous track sequence, the takeoff and landing phases are excluded first: only track points with an air pressure altitude greater than 3000m are retained for subsequent discrimination. This elevation threshold can be adjusted within the range of 2000~4000m according to the terrain features.

[0094] Instantaneous ground speed is calculated point by point for each retained track point, and a dual threshold determination is applied:

[0095] (1) Abnormal steady state determination: When the instantaneous ground speed of multiple consecutive track points is lower than 100 m / s, the track points in that segment are marked as controlled steady-state false track points (attribute marked as 2). This threshold corresponds to the lower limit of the cruise speed of a typical civil aircraft and can be adjusted within the range of 80~120 m / s according to the characteristics of the aircraft type.

[0096] (2) Abnormal Change Judgment: When the instantaneous ground speed of adjacent track points exceeds 500 m / s, the point pair is determined to be a track jump point. In this point pair, the last normal track point before the jump is marked as the boundary point before entering the deception range (attribute marked as 3), and the first false track point after the jump is marked as a controlled false track point (attribute marked as 2). To adapt to the performance differences of different aircraft models, the mutation threshold can be adjusted within the range of 400~600 m / s. After the initial screening, the controlled false track points marked as 2 are used for subsequent deception target area identification, and the boundary points marked as 3 are used as candidate edge nodes to enter the re-screening stage.

[0097] S202.2: Secondary screening stage: composite screening based on four spatiotemporal correlations.

[0098] As one possible implementation method, in this embodiment, a four-fold spatiotemporal correlation constraint is constructed, and candidate edge nodes are finely screened layer by layer using a software program:

[0099] (1) Deception Area Constraints: For all controlled false track points marked as 2, the DBSCAN (Density-Based Spatial Clustering of Applications with Noise) algorithm is used for spatial density clustering to determine the latitude and longitude range of each deception target area. The clustering parameters are set as follows: neighborhood radius epsilon = 0.05° (approximately 5.5km), minimum number of points minPts = 30. After clustering, each target area is numbered sequentially (e.g., 1, 2, ..., N), and the center coordinates and boundary range of each target area are recorded. For each candidate edge node, it is determined which target area its corresponding false track point (i.e., the convergence point after the jump) falls into. The track point is re-marked according to the rule of "region number multiplied by 10 plus attribute value" (e.g., the false track point of region 1 is marked as 12, and the entry boundary point is marked as 13). The entry boundary point, exit boundary point, and transition point of its corresponding region are marked simultaneously. Boundary points that do not fall into any target area are judged as stray false anomalies and are removed.

[0100] (2) Reacquisition time constraint: The reacquisition time interval threshold after deception is set to 900s (which can be adjusted within the range of 600~1200s according to the characteristics of the model). Calculate the time difference between each entry boundary point and its corresponding exit boundary point. If the time difference exceeds 900s, the entry boundary point is determined to have deviated from the true edge node circle due to excessive reacquisition delay and is therefore removed. This threshold is set based on the physical characteristic that the difference between the power threshold of the receiver's loss of lock and acquisition is about 5dB. Excessive reacquisition delay will cause the exit boundary point to deviate from the true circle by tens of kilometers.

[0101] (3) Deception Period Constraint: For each deception target area, extract the UTC time of all controlled false track points within that area and convert it into a daily second sequence (0~86400s) in seconds. Set the minimum start-stop interval threshold for deception sources to 600s (which can be adjusted within the range of 300~900s depending on the intensity of the conflict). If the time interval between consecutive controlled false track points is less than this threshold, it is determined to be the same consecutive deception period; if the interval exceeds this threshold, it is determined to be a deception interruption period. Only entry boundary points located within the effective deception period are retained, and non-edge jump points caused by intermittent start-stop of deception sources are removed.

[0102] (4) Cruise elevation constraint: Set the cruise elevation threshold for edge nodes to 9000m (which can be adjusted within the range of 8000~10000m based on the average terrain elevation of the region). Only retain the entry boundary points and corresponding transition points with an air pressure altitude greater than 9000m, and remove the boundary points that are affected by the curvature of the earth and terrain obstruction during the low-altitude take-off and landing phase, and whose deception distance is limited by the line-of-sight conditions rather than the actual coverage range of the deception source.

[0103] After four-fold constraint screening, only the entry boundary points labeled with attributes 13, 23, ..., N3 are retained as valid edge nodes. These are then aggregated and classified by region to generate a standardized partitioned edge node dataset. Since the power threshold difference between receiver loss of lock and acquisition is approximately 5 dB, the entry boundary points are concentrated on the equal power circumference, while the exit boundary points deviate from the true circumference. Therefore, subsequent localization only uses the entry boundary points.

[0104] S203: Two-stage DBSCAN clustering and centroid merging.

[0105] For the partition edge nodes extracted by S202, a two-stage DBSCAN clustering strategy combined with a centroid merging method is used to separate the sets of interference source points. As an implementation method, in this embodiment, all clustering operations are automatically executed by software program.

[0106] The first stage is clustering at convergence locations: the controlled spurious track points (i.e., the spurious locations that have been deceived and have converged stably) corresponding to each edge node are used as clustering objects, and the DBSCAN algorithm is used for clustering. The distance between two convergence points is calculated using the Haversine great circle distance formula. As shown in formula (2):

[0107]

[0108] The clustering radius epsilon1 is set to 0.01° (approximately 1.1 km), corresponding to the typical deception target point location error; the minimum number of points minPts1 is set to 80. This stage ensures that outliers generated by the same deception source are aggregated into the same candidate cluster.

[0109] The second stage is spatial location clustering: For each edge node within a convergent cluster, its position before the jump before entering the deception zone is taken as the clustering object, and the DBSCAN algorithm is used again for clustering, as shown in formula (3):

[0110]

[0111] The cluster radius epsilon2 was set to 1.8° (approximately 190km), corresponding to a maximum effective radius of approximately 120km for deceptive interference sources, while also considering the impact of anomalous jitter in ADS-B track data; the minimum number of points minPts2 was also set to 80. In this stage, point sets from different spatial orientations within the same convergence cluster were further subdivided into subclusters corresponding to different deceptive sources.

[0112] After completing the two-stage clustering, centroid merging is performed, which involves calculating the average latitude and longitude of all edge nodes within each sub-cluster as the cluster centroid. If the Haversine distance between the centroids of two clusters is less than the merging threshold of 0.5° (approximately 55km), the Union-Find algorithm is used to merge them into the same cluster. This merging threshold can be adjusted within the range of 0.3 to 0.8° based on the actual data distribution characteristics. Ultimately, each cluster corresponds to an independent deception interference source.

[0113] After clustering is completed, the edge nodes of each cluster are used for subsequent robust circular arc fitting.

[0114] S204: Three-step hybrid robust least squares circular arc fitting and interference source localization.

[0115] For each cluster obtained from S203, the latitude and longitude coordinates of all edge nodes are converted to station-centered Cartesian coordinates (ENU coordinate system) with the center of the deception area as the reference point. The conversion formula is as follows: Let the latitude and longitude of the reference point be (φ0, λ0), and the latitude and longitude of the edge nodes be (φ0, λ0, ..., λ0, ...). i , λ i ), then the station center coordinates (E) i N i The calculation of ) uses the standard ENU projection formula.

[0116] Based on this, a three-step hybrid robust circular arc fitting is implemented, and all fitting operations are automatically executed in a numerical iteration manner.

[0117] S204.1: Solving for robust initial center of the Tukey double-weighted M-estimation.

[0118] Using the planar coordinates (E) of all points within the cluster i N i Using the median of ) as the initial center, as shown in formula (4), perform iterative weighted updates:

[0119]

[0120] (a) Calculate the Euclidean distance d from each point to the current center of the circle. i As shown in formula (5);

[0121]

[0122] (b) Estimate the robustness scale σ using the median absolute deviation (MAD), as shown in Equation (6);

[0123]

[0124] (c) Calculate the standardized residuals as shown in Formula 7;

[0125]

[0126] (d) Take the standard cutoff constant B = 4.685 for 95% asymptotic efficiency under normal distribution, and calculate the weight of each point using the Tukey double weight function, as shown in formula (8);

[0127]

[0128] (e) Update the center coordinates with the weighted median, as shown in formula (9);

[0129]

[0130] (f) Convergence determination: The iteration terminates when the center displacement of two adjacent iterations is less than a set threshold (preferably 1m), and a robust initial center is output. Tukey's dual-weight estimation has a collapse rate of 50%, which can effectively resist the contamination of the initial center by gross errors.

[0131] S204.2: Gross removal based on interquartile range (IQR) criteria.

[0132] Referring to the initial robust circle center, calculate the Euclidean distance sequence from each edge node to that circle center. Take the first quartile Q1 and the third quartile Q3 of the distance sequence, and calculate the interquartile range IQR = Q3 - Q1. The rejection threshold is defined as:

[0133]

[0134] The multiplier k is set to 3.0 to achieve stricter removal of extreme outliers (k can be adjusted within the range of 2.5 to 3.5). Points whose distance exceeds this threshold range are identified as extreme outliers and removed.

[0135] After removing extreme outliers using the IQR criterion, the remaining valid point set is used for subsequent fine fitting.

[0136] S204.3: Robust iterative fine fitting of IGGIII equivalent weight functions.

[0137] A robust iterative fine-fit is performed on the retained points based on the IGGIII equivalent weight function. First, the Levenberg-Marquardt algorithm is used to solve the nonlinear least squares problem:

[0138]

[0139] Initialize the center and radius of the circle using the median. After fitting, calculate the radius residuals at each point:

[0140]

[0141] Standardized residuals were estimated using the MAD robust metric:

[0142]

[0143] According to the IGGIII three-segment weight function

[0144]

[0145] Set the upper bound for normal observations to k0 = 1.5 and the upper bound for rejection to k1 = 2.5 (k0 can be adjusted within the range of 1.0 to 1.5, and k1 can be adjusted within the range of 2.5 to 3.0). Classify the observation points: points with standardized residuals less than k0 retain unit weight; points between k0 and k1 are smoothly weighted down; points greater than k1 are completely rejected. Re-fit the points with the updated weights, iterating until the change in weights for all points is less than 0.01 or the maximum number of iterations (50) is reached. Finally, perform a final refinement fit using the converged weights to obtain the center coordinates (E0, N0) and radius R0 of the fitted circle. Calculate the center coordinates back to latitude and longitude to obtain the estimated location of the deceptive interference source.

[0146] S205: Positioning accuracy assessment and cross-time period spatiotemporal stability verification.

[0147] S205.1: Internal compliance accuracy assessment.

[0148] Calculate the residual mean square error of the effective edge nodes involved in the fitting.

[0149]

[0150] Simultaneously, the eastward and northward mean square errors are evaluated. An angular coverage index is also introduced, which calculates the azimuth of each effective point with the fitted circle center as a reference. The coverage is defined as 360° minus the maximum gap angle, with 120° as the minimum coverage angle threshold, ensuring that the fitted arc has good geometric constraints.

[0151] S205.2: Verification of spatiotemporal stability across time periods.

[0152] Two verification methods are introduced:

[0153] (1) 95% confidence ellipse analysis: The full-time data is divided into multiple sub-time periods in groups of 5 days. A 95% confidence ellipse is fitted to the edge nodes of each sub-time period. By comparing the degree of overlap and centroid clustering of the core regions of the confidence ellipse in each time period, the stability of the location of the interference source in the time dimension is judged.

[0154] (2) Two-way overlap rate matrix analysis: Using 11.1km (0.1°) as the distance threshold, the two-way overlap rate of edge nodes in each time period is calculated to form an overlap rate matrix.

[0155] The key technical points and areas to be protected in this invention are as follows:

[0156] (1) Track anomaly jump point detection mechanism based on multi-layer joint discrimination. This invention does not rely on navigation quality parameters such as NACp and NIC, but establishes a two-layer progressive extraction architecture of "initial screening + secondary screening" by deeply mining the kinematic characteristics of the ADS-B track itself: the initial screening stage realizes the preliminary separation of track anomaly points based on velocity thresholds (anomaly steady state judgment threshold of 100m / s, anomaly sudden change judgment threshold of 500m / s); the secondary screening stage constructs four spatiotemporal correlation criteria of deception area constraint, recapture duration constraint, deception time period constraint, and cruise elevation constraint, and removes various non-edge pseudo-anomaly points layer by layer, effectively distinguishing between the boundary points entering the deception area and the boundary points leaving the deception area, and only the boundary points entering the deception area are used as effective edge nodes to participate in subsequent positioning.

[0157] (2) A two-stage DBSCAN clustering and centroid merging method for separating multiple interference source point sets. First, clustering is performed on the convergence position of the false position that has been deceived and has converged stably. Then, spatial position clustering is performed on the position before the jump. Combined with centroid merging and disjoint-set data structure algorithms, the effective separation of the mixed point sets of each source in the scenario where multiple deception sources coexist can be achieved.

[0158] (3) Three-step hybrid robust least squares circular arc fitting localization method. Based on the geometric constraints that the edge nodes are distributed on the circular arc with the interference source as the center, a three-step hybrid robust strategy is constructed, which is "Tukey double weight M estimation robust initial circle center solution + interquartile range criterion gross error elimination + IGGIII equivalent weight function robust iterative fine fitting", to realize the robust solution of the interference source position under the condition of gross error.

[0159] (4) Cross-time period spatiotemporal stability verification method. The spatiotemporal stability of the positioning results within a long time window is systematically verified by two methods: 95% confidence elliptic analysis and two-way coincidence matrix analysis, to determine the continuous working status and positional consistency of the interference source.

[0160] Based on the same inventive concept as the above-described method embodiments, such as Figure 5 As shown, this invention also proposes a GNSS spoofing interference detection and positioning device based on ADS-B open-source data, comprising:

[0161] The data acquisition module is used to acquire ADS-B track data, which includes at least flight identifier, timestamp, and location coordinates;

[0162] A dual-layer extraction module is used to initially screen waypoints based on a velocity threshold, classifying and labeling them as normal waypoints, controlled false waypoints, entry boundary points, and exit boundary points. The controlled false waypoints are those with velocities below a first set threshold, while the entry and exit boundary points are those with velocities above a second set threshold. The entry and exit boundary points are then further screened, and invalid points are removed using spatiotemporal correlation constraints to obtain valid edge nodes.

[0163] The clustering module is used to cluster the effective edge nodes, grouping edge nodes belonging to the same deception interference source into the same cluster;

[0164] The localization module is used to perform circular arc fitting on the edge nodes within each cluster, and the center position of the fitted circle is used as the localization result of the corresponding deception interference source.

[0165] Furthermore, such as Figure 6 As shown, it also includes:

[0166] The spatiotemporal stability assessment module is used to divide the full-time data into multiple sub-time periods according to time, fit 95% confidence ellipses to the edge nodes of each sub-time period, and judge the stability of the location of the interference source by comparing the degree of overlap and centroid clustering of the confidence ellipses of each time period; and / or, construct a two-way overlap rate matrix, calculate the two-way overlap rate of edge nodes between each time period, and quantitatively assess the consistency of the spatial distribution of edge nodes in different time periods.

[0167] The specific functions of each module mentioned above correspond exactly to the steps in the aforementioned method embodiments, and will not be repeated here.

[0168] Based on the same inventive concept as the above-described method embodiments, the present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the GNSS spoofing interference detection and location method based on ADS-B open-source data as described in any of the above claims.

[0169] The computer-readable storage medium can be any tangible medium capable of storing program code, including but not limited to: read-only memory (ROM), random access memory (RAM), magnetic disk, optical disk (such as CD-ROM, DVD), flash drive, portable hard disk, magneto-optical disk, etc. The processor can be a central processing unit (CPU), graphics processing unit (GPU), digital signal processor (DSP), application-specific integrated circuit (ASIC), field-programmable gate array (FPGA), or other programmable logic device.

[0170] Those skilled in the art will understand that all or part of the processes in the above method embodiments can be implemented by hardware related to computer program instructions. The computer program can be stored in a computer-readable storage medium, and when executed, it can include the processes described in the above method embodiments. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0171] Since the program flow executed in this storage medium embodiment is exactly the same as that in the aforementioned method embodiment, the specific steps will not be repeated here.

[0172] The above description is only a preferred embodiment of the present invention. It should be noted that those skilled in the art can make several improvements and modifications without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.< / gx:coord> < / when>

Claims

1. A method for detecting and locating GNSS spoofing interference based on ADS-B open-source data, characterized in that, include: Acquire ADS-B track data, which includes at least flight identifier, timestamp, and location coordinates; Initial screening of waypoints is performed based on speed thresholds, classifying them into normal waypoints, controlled false waypoints, entry boundary points, and exit boundary points. Controlled false waypoints are those with speeds below a first set threshold, while entry and exit boundary points are those with speeds above a second set threshold. Entry and exit boundary points are then further screened, and invalid points are removed using spatiotemporal correlation constraints to obtain valid edge nodes. Cluster the effective edge nodes, grouping edge nodes belonging to the same deception interference source into the same cluster; For each cluster, the edge nodes are fitted with circular arcs, and the center position of the fitted circle is used as the localization result of the corresponding deception interference source.

2. The GNSS spoofing interference detection and localization method based on ADS-B open-source data according to claim 1, characterized in that, The spatiotemporal correlation constraints include at least one of the following: deception region constraints, recapture duration constraints, deception time period constraints, and cruise elevation constraints. The deception region constraint is to assign a region to each boundary point and remove stray points based on the spatial density clustering results of the controlled false track points. The recapture duration constraint is based on a set threshold for the recapture time interval after deception, which excludes timeout jump points. The deception period constraint is based on the temporal distribution of controlled false track points to determine the effective deception period and remove boundary points outside the period. The cruise elevation constraint is based on a set elevation threshold to exclude boundary points during low-altitude take-off and landing phases.

3. The GNSS spoofing interference detection and localization method based on ADS-B open-source data according to claim 1, characterized in that, The clustering adopts a two-stage clustering strategy: In the first stage, spatial clustering is performed using the location of the controlled spurious track points corresponding to each edge node as the clustering object, and edge nodes with similar convergence locations are aggregated into the same convergence cluster. In the second stage, for each edge node within a convergence cluster, spatial clustering is performed again using its position before entering the deception zone as the clustering object. This further subdivides the point sets from different spatial orientations within the same convergence cluster into subclusters corresponding to different deception sources.

4. The GNSS spoofing interference detection and localization method based on ADS-B open-source data according to claim 3, characterized in that, After completing the two-stage clustering, the process also includes: calculating the average position of all edge nodes within each sub-cluster as the cluster centroid, and merging sub-clusters whose centroid distance is less than the merging threshold.

5. The GNSS spoofing interference detection and localization method based on ADS-B open-source data according to claim 1, characterized in that, The circular arc fitting employs a three-step hybrid robustness strategy: The first step is to solve for the robust initial center of the circle based on Tukey's two-weight M-estimation; The second step is to eliminate extreme outliers relative to the initial center of the circle based on the interquartile range criterion. The third step is to perform robust iterative fine fitting based on the IGGIII equivalent weight function, and use the center of the fitted circle after convergence as the localization result of the deception interference source.

6. The GNSS spoofing interference detection and localization method based on ADS-B open-source data according to claim 1, characterized in that, The first set threshold is 100 m / s, and the second set threshold is 500 m / s.

7. The GNSS spoofing interference detection and localization method based on ADS-B open-source data according to claim 1, characterized in that, Also includes: The full-time data is divided into multiple sub-time periods according to time. A 95% confidence ellipse is fitted to the edge nodes of each sub-time period. The stability of the location of the interference source is judged by comparing the degree of overlap and centroid clustering of the confidence ellipses of each time period. And / or, a two-way overlap rate matrix is ​​constructed to calculate the two-way overlap rate of edge nodes between each time period, and the consistency of the spatial distribution of edge nodes in different time periods is quantitatively evaluated.

8. A GNSS spoofing interference detection and localization method based on ADS-B open-source data according to any one of claims 1-7, characterized in that, After acquiring the ADS-B track data, the process also includes: performing an integrity check on the data, removing invalid records with missing timestamps or abnormal coordinates, and performing trajectory splicing and repair on the segmented tracks to reconstruct a complete track sequence.

9. A GNSS spoofing interference detection and positioning device based on ADS-B open-source data, characterized in that, include: The data acquisition module is used to acquire ADS-B track data, which includes at least flight identifier, timestamp, and location coordinates; A dual-layer extraction module is used to initially screen waypoints based on a velocity threshold, classifying and labeling them as normal waypoints, controlled false waypoints, entry boundary points, and exit boundary points. The controlled false waypoints are those with velocities below a first set threshold, while the entry and exit boundary points are those with velocities above a second set threshold. The entry and exit boundary points are then further screened, and invalid points are removed using spatiotemporal correlation constraints to obtain valid edge nodes. The clustering module is used to cluster the effective edge nodes, grouping edge nodes belonging to the same deception interference source into the same cluster; The localization module is used to perform circular arc fitting on the edge nodes within each cluster, and the center position of the fitted circle is used as the localization result of the corresponding deception interference source.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1 to 8.

Images