A method for counteracting a towed jamming based on spatial form features

Abstract

The present application relates to a kind of based on spatial form feature's towed jamming countermeasure method, comprising: obtaining and channel, azimuth difference channel and the echo data of pitch difference channel, and utilize single pulse three-dimensional imaging technology to the three-dimensional imaging of radar forward-looking area to echo data, obtain three-dimensional point cloud;Utilize the spatial filtering algorithm based on density clustering to cluster clustering and noise rejection of three-dimensional point cloud, obtain several point cloud clusters;Based on the spatial covariance eigenvalue of point cloud cluster, the spatial form feature descriptor of each point cloud cluster is calculated;Based on the difference of target and jamming in spatial form, utilize soft interval linear SVM target discriminator to each point cloud cluster in point cloud cluster sample set Target identification and jamming suppression, obtain the target imaging result after anti-jamming.This method makes full use of the difference of jamming and target in spatial form feature realizes towed jamming countermeasure, jamming discrimination accuracy is higher, and the quality of target imaging result after anti-jamming is higher.

Description

Technical Field

[0001] This invention belongs to the field of radar technology, specifically relating to a dragged jamming countermeasure method based on spatial morphological characteristics. Background Technology

[0002] As a main lobe jamming technique, towed decoys deceive guidance radars by emitting high-power jamming signals, thereby protecting the target aircraft. Thanks to their low cost, ease of installation, and simple operation, towed decoys have become a commonly used and important technology for self-defense jamming of aircraft.

[0003] Target tracking under towed decoy jamming conditions has always been a research hotspot both domestically and internationally. Due to its advantages such as low computational complexity, high tracking accuracy, and simple engineering implementation, monopulse measurement technology is widely used in terminal guidance. Addressing the drawback of radar being easily deceived by decoy signals, leading to radar lock-on failure or decoy tracking, Chen Boxiao et al. utilized the differences in the corresponding polarization channels of the jamming signal and the target echo to achieve jamming presence detection and suppression. Chen Anna proposed a towed decoy detection method based on adaptive smoothing (Wigner-Hough transform), which detects the presence of jamming by detecting the difference in peak characteristics before and after jamming release. Yang Cheng et al. proposed using monopulse signals from the angle difference channel to identify the presence of decoys and implemented target angle measurement through an adaptive one-dimensional conformal angle measurement algorithm. K.Yu et al. implemented towed jamming countermeasures under EPC radar systems through joint element-pulse coding.

[0004] Single-pulse 3D imaging technology, as a forward-looking imaging technique, fills the blind spot where synthetic aperture technology cannot operate in forward-looking scenarios. Diao Guijie et al. used single-pulse 3D imaging technology to reconstruct the strong scattering center point cloud of ship targets. Chen Boxiao et al. used single-pulse 3D imaging technology to counteract external active jamming and cross-eye jamming, ensuring stable target tracking.

[0005] However, for a long time, research on countermeasures against towed jamming has mostly been based on the premise that radar transmits narrowband signals, and there has been little research on countermeasures against radar transmitting broadband signals in the context of towed jamming. Summary of the Invention

[0006] To address the aforementioned problems in the existing technology, this invention provides a drag-and-drop interference countermeasure method based on spatial morphological characteristics. The technical problem to be solved by this invention is achieved through the following technical solution:

[0007] This invention provides a drag-and-go interference countermeasure method based on spatial morphological features, comprising the following steps:

[0008] Echo data from the azimuth, pitch, and azimuth channels are acquired, and the echo data are used to perform three-dimensional imaging of the radar forward-looking area using single-pulse three-dimensional imaging technology to obtain a three-dimensional point cloud.

[0009] The three-dimensional point cloud is clustered and noise is removed using a density-based spatial filtering algorithm to obtain several point cloud clusters.

[0010] Based on the spatial covariance eigenvalues ​​of point cloud clusters, the spatial morphological feature descriptor of each point cloud cluster is calculated, and the spatial morphological features of multiple point cloud clusters constitute a point cloud cluster sample set.

[0011] Based on the spatial morphological differences between the target and the interference, a soft-spaced linear SVM target discriminator is used to perform target identification and interference suppression on each point cloud cluster in the point cloud cluster sample set, resulting in anti-interference target imaging results.

[0012] In one embodiment of the present invention, echo data from the azimuth, elevation, and pitch difference channels are acquired, and the echo data are used to perform three-dimensional imaging of the radar forward-looking area using single-pulse three-dimensional imaging technology to obtain a three-dimensional point cloud, including:

[0013] Acquire echo data for the azimuth, azimuth, and pitch channels;

[0014] The echo data is subjected to range-Doppler high-resolution processing to obtain high-resolution processing results for the sum channel, azimuth difference channel, and elevation difference channel;

[0015] Two-dimensional constant false alarm rate (CFAR) detection is performed on the high-resolution processing results of the channels to extract the distance information corresponding to several resolution units occupied by strong scattering points.

[0016] Using the high-resolution processing results of the sum channel, azimuth difference channel, and elevation difference channel, the angle is measured at the amplitude of the over-detection threshold cell to obtain the azimuth and elevation angles of each strong scattering point resolution cell relative to the radar.

[0017] Combining the distance information, the azimuth and elevation angles of each strong scattering point resolution unit relative to the radar are converted into azimuth-dimensional relative distance and elevation-dimensional relative distance to obtain the three-dimensional point cloud.

[0018] In one embodiment of the present invention, the relative distance in the azimuth dimension and the relative distance in the pitch dimension are respectively:

[0019]

[0020] Where, d az The relative distance is in the azimuth dimension. Let R = [R1, R2, ..., R] be the azimuth angle of each strong scattering point resolution element relative to the radar.P ] represents the distance information corresponding to several resolution cells occupied by a strong scattering point, d el Let θ be the relative distance in the pitch dimension, where θ = [θ1, θ2, ..., θ] P [ ] represents the elevation angle of each strong scattering point resolution unit relative to the radar.

[0021] In one embodiment of the present invention, a spatial filtering algorithm based on density clustering is used to perform clustering and noise removal on the three-dimensional point cloud to obtain several point cloud clusters, including:

[0022] Calculate the distance between each pair of scattering points in the three-dimensional point cloud to obtain the distance matrix;

[0023] Choose any scattering point as the reference point for the new cluster. Count the number of first scattering points within a preset neighborhood radius of the reference point. If the number of first scattering points is less than the minimum number of elements contained in the preset neighborhood, the reference point is discarded as noise. If the number of first scattering points is greater than or equal to the minimum number of elements contained in the preset neighborhood, the scattering points within the preset neighborhood radius are added to the current cluster.

[0024] Select a scattering point within the current cluster that has not been used as a reference point as a new reference point;

[0025] The number of second scattering points within the preset neighborhood radius of the new reference point is counted. If the number of second scattering points is less than the minimum number of elements contained in the preset domain, a new reference point is selected again in the current cluster for judgment. If the number of second scattering points is greater than or equal to the minimum number of elements contained in the preset domain, the scattering points within the preset neighborhood radius of the new reference point are included in the current cluster.

[0026] Iterate through all scattering points within the current cluster;

[0027] When there are unclustered points, a scattering point is selected from the unclustered points as a reference point for the new cluster and either removed or clustered to obtain a new point cloud cluster; when there are no unclustered points, the spatially filtered point cloud clusters are output.

[0028] In one embodiment of the present invention, based on the spatial covariance feature value of point cloud clusters, a spatial morphological feature descriptor for each point cloud cluster is calculated, and a point cloud cluster sample set is constituted by the spatial morphological features of multiple point cloud clusters, including:

[0029] The spatial covariance matrix of the point cloud cluster is defined based on the positions of scattering points within the cluster and the cluster center position:

[0030]

[0031] Where K is the number of scattering points within the point cloud cluster, p kcp represents the spatial position of the k-th scattering point relative to the cluster center. c The location of the cluster center p k p is the position of the kth scattering point. k =(x k ,y k ,z k ) T (k = 1, ..., K), x k ,y k ,z k Indicates the coordinates of the k-th scattering point;

[0032] The spatial covariance matrix is ​​subjected to eigenvalue decomposition to obtain the spatial covariance eigenvalues. The expression for the eigenvalue decomposition is as follows:

[0033]

[0034] Where Λ=diag(λ1,λ2,λ3) is a diagonal matrix of spatial covariance eigenvalues ​​arranged in descending order. The eigenvector matrix corresponding to the three spatial covariance eigenvalues ​​λ1, λ2, λ3 is V. LRF = [v1,v2,v3], where v1, v2, and v3 are feature column vectors;

[0035] The spatial morphological feature descriptor of each point cloud cluster is calculated using the spatial covariance eigenvalues, thus obtaining the spatial morphological features of each point cloud cluster:

[0036] x i =[s 1,i ,s 2,i ,…,s N,i ] T

[0037] Where, x i s represents the N-dimensional feature vector formed by the spatial morphological features of the i-th point cloud cluster. 1,i ,s 2,i ,…,s N,i This represents the N spatial morphological feature descriptors corresponding to the i-th point cloud cluster;

[0038] The point cloud cluster sample set is composed of the spatial morphological features of all point cloud clusters:

[0039]

[0040] Where M represents the number of point cloud clusters, It is an N-dimensional real number space.

[0041] In one embodiment of the present invention, the spatial morphological feature descriptor of each point cloud cluster includes anisotropy, sum of eigenvalues, linearity, flatness, sphericity, feature entropy, curvature, and total variance.

[0042] In one embodiment of the present invention, the formula for calculating the anisotropy is:

[0043] The formula for calculating the sum of the eigenvalues ​​is: s2 = λ1 + λ2 + λ3;

[0044] The formula for calculating the linearity is:

[0045] The formula for calculating the flatness is:

[0046] The formula for calculating the sphericity is:

[0047] The formula for calculating the characteristic entropy is: s6=-λ1lnλ1-λ2lnλ2-λ3lnλ3;

[0048] The formula for calculating the curvature is:

[0049] The formula for calculating the total variance is as follows:

[0050] In one embodiment of the present invention, the design method of the soft-spaced linear SVM target discriminator includes the following steps:

[0051] Obtain a point cloud cluster sample set and its corresponding label set;

[0052] The target identification problem in the point cloud cluster sample set is transformed into a convex quadratic programming binary classification problem:

[0053]

[0054] in, To separate the hyperplane w T x i The normal vector of +b=0, b is the offset of the linear classifier, C is the penalty factor satisfying C>0, ζ i Let M be the number of point cloud clusters in the sample set, and x be a slack variable. i =[s 1,i ,s 2,i ,…,s N,i ] T Let N be the feature vector of the i-th point cloud cluster. Given a set of M point cloud clusters, It is an N-dimensional real number space;

[0055] By introducing Lagrange multipliers to construct a Lagrange function and combining it with strong duality, the convex quadratic programming binary classification problem is transformed into a dual classification problem:

[0056]

[0057] Where, λ i μ is the first Lagrange multiplier of the i-th point cloud cluster. i Let i be the second Lagrange multiplier of the i-th point cloud cluster. This is the label set corresponding to the point cloud cluster sample set, where -1 represents "decoy" and "1" represents "target";

[0058] Solve the dual-form classification problem and combine it with the convex quadratic programming binary classification problem to obtain the hyperplane. and the corresponding decision function The soft-spaced linear SVM target discriminator is obtained.

[0059] In one embodiment of the present invention, based on the spatial morphological differences between the target and the interference, a soft-spaced linear SVM target discriminator is used to perform target discrimination and interference suppression on each point cloud cluster in the point cloud cluster sample set, to obtain an anti-interference target imaging result, including:

[0060] The point cloud cluster samples in the point cloud cluster sample set are sequentially input into the interval linear SVM target discriminator. When the output of the interval linear SVM target discriminator is 1, the point cloud cluster sample is the target. When the output of the interval linear SVM target discriminator is -1, the point cloud cluster sample is interference. The target imaging result is formed from the point cloud cluster samples whose output is the target.

[0061] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0062] The dragging jamming countermeasure method based on spatial morphology features provided by this invention achieves noise removal and strong scattering point cloud clustering by performing spatial filtering on three-dimensional point clouds, and extracts spatial morphology features of each point cloud cluster. Target identification and jamming suppression are achieved through a soft-interval linear SVM classifier, resulting in a three-dimensional image of the target after jamming. Taking advantage of the fact that jamming signals emitted by decoys under broadband conditions cannot completely simulate the spatial morphology of the target in three-dimensional imaging space, and in cases where it is difficult to distinguish between the target and the jamming in one-dimensional range image, the dragging jamming countermeasure fully utilizes the differences in spatial morphology features between the jamming and the target to achieve dragging jamming countermeasure. The jamming identification accuracy is high, and the quality of the target imaging result after jamming is high. Attached Figure Description

[0063] Figure 1 A flowchart illustrating a dragging interference countermeasure method based on spatial morphological features provided in an embodiment of the present invention;

[0064] Figure 2 A flowchart illustrating another dragging interference countermeasure method based on spatial morphological features provided in an embodiment of the present invention.

[0065] Figure 3 This is a schematic diagram of the single-pulse three-dimensional imaging process provided in an embodiment of the present invention;

[0066] Figure 4 This is a flowchart illustrating the spatial filtering algorithm based on density clustering provided in an embodiment of the present invention.

[0067] Figure 5 This is a diagram showing the spatial relationship and signal transmission of a dragging interference scenario provided in an embodiment of the present invention.

[0068] Figures 6a-6d The three-dimensional imaging results of the target under dragged interference conditions provided in the embodiments of the present invention are shown in three-dimensional view and ISAR image.

[0069] Figures 7a-7c The above are three views of the point cloud clustering results provided in this embodiment of the invention.

[0070] Figures 8a-8c Three-view diagrams of the target imaging result after interference suppression provided in an embodiment of the present invention. Detailed Implementation

[0071] The present invention will be further described in detail below with reference to specific embodiments, but the implementation of the present invention is not limited thereto.

[0072] Example 1

[0073] Please see Figure 1 and Figure 2 , Figure 1 This is a flowchart illustrating a drag-and-go interference countermeasure method based on spatial morphological features, provided in an embodiment of the present invention. Figure 2 This is a flowchart illustrating another drag-and-drop jamming countermeasure method based on spatial morphology features provided in an embodiment of the present invention. The drag-and-drop jamming countermeasure method based on spatial morphology features includes the following steps:

[0074] S1. Acquire echo data from the azimuth, pitch, and azimuth channels, and use single-pulse 3D imaging technology to perform 3D imaging of the radar forward-looking area from the echo data to obtain a 3D point cloud.

[0075] Please see Figure 3 , Figure 3 This is a schematic diagram of the single-pulse three-dimensional imaging process provided in an embodiment of the present invention. Step S1 specifically includes:

[0076] S11, echo data from radar receiving and channeling, azimuth difference channel, and elevation difference channel.

[0077] S12. Perform range-Doppler high-resolution processing on the echo data to obtain high-resolution processing results for the sum, azimuth, and elevation difference channels, i.e., the focused ISAR image. Specifically, range-Doppler high-resolution processing includes stretch processing, envelope alignment, phase focusing, and inter-pulse FFT.

[0078] S13. Perform two-dimensional constant false alarm rate (2-D CFAR) detection on the high-resolution processing results of the channel and extract the distance information corresponding to several (P) resolution cells occupied by the strong scattering points.

[0079] S14. Using the high-resolution processing results of the sum, azimuth, and elevation difference channels, angles are measured at the amplitude at the over-detection threshold cell to obtain the azimuth angle of each strong scattering point resolution cell relative to the radar. And pitch angle θ=[θ1,θ2,...,θ P ].

[0080] S15. Combining the range information, the azimuth and elevation angles of each strong scattering point resolution unit relative to the radar are converted into azimuth-dimensional relative distance and elevation-dimensional relative distance to obtain a three-dimensional point cloud, thereby realizing three-dimensional imaging of the forward-looking area.

[0081] Specifically, the relative distances in the azimuth and elevation dimensions are as follows:

[0082]

[0083] Where, d az Let d be the relative distance in the orientation dimension. el The relative distance is for pitch.

[0084] S2. Use a density-based spatial filtering algorithm to perform clustering and noise removal on the 3D point cloud to obtain several point cloud clusters.

[0085] Angular scintillation is a phase wavefront distortion of the echo caused by the superposition of random variations in scattering intensity and phase at different parts of a complex target, manifesting as angular measurement anomalies. In the 3D image space, there are angular scintillation points that are severely deviated from the point cloud clusters and angular measurement singularities caused by noise. In this embodiment, these are uniformly classified as noise points and removed during spatial filtering to avoid negatively impacting subsequent processing. Based on the idea of ​​density clustering, this embodiment proposes a spatial filtering algorithm to achieve 3D point cloud clustering and noise removal.

[0086] Please see Figure 4 , Figure 4 This is a flowchart illustrating a density-based clustering-based spatial filtering algorithm provided in an embodiment of the present invention. The density-based clustering-based spatial filtering algorithm includes the following steps:

[0087] S21. Input the 3D point cloud obtained from single-pulse 3D imaging processing, complete the algorithm initialization, and pre-set the algorithm hyperparameters based on prior knowledge: neighborhood radius ξ, minimum number of elements N in the neighborhood. min ;

[0088] S22. Calculate the distance between any two scattering points in the 3D point cloud to obtain the distance matrix D; denote the matrix element D(i,p) in the distance matrix D as the distance between the i-th scattering point and the p-th scattering point (p=1,2,…,P).

[0089] S23. Select any scattering point q as the reference point of the new cluster, and count the number N of the first scattering points within the preset neighborhood radius (D(q,p)≤ξ) of the reference point. near When the number of first scattering points N near Less than the minimum number of elements N contained in the preset domain min N near <N min When the reference point q is considered a noise point, it is included in the noise cluster and removed. When the number of the first scattering points is N... near Greater than or equal to the minimum number of elements N contained in the preset domain min N near ≥N min If the condition is met, then the current cluster is established, and the scattering points within the preset neighborhood radius are assigned to the current cluster.

[0090] S24. Select a scattering point m within the current cluster that has not been used as a reference point as a new reference point.

[0091] S25. Count the number of second scattering points N within the preset neighborhood radius of the new reference point m. near When the number of second scattering points is less than the minimum number of elements N contained in the preset domain min N near <N min If the number of second scattering points N is N, then repeat step S24 to select a new reference point within the current cluster for judgment; near Greater than or equal to the minimum number of elements N contained in the preset domain min N near ≥N min When the time comes, the scattering points within the preset neighborhood radius of the new reference point are assigned to the current cluster, and the cluster is updated.

[0092] S26. Repeat steps S24-S25 until all scattering points within the current cluster have been traversed.

[0093] S27. Determine if there are still un-clustered points. If there are un-clustered points, select a scattering point among the un-clustered points as the reference point of the new cluster. Repeat steps S23-S26 to remove or cluster the reference point of the new cluster to form a new point cloud cluster.

[0094] S28. When there are no un-clustered points, output several point cloud clusters after spatial filtering.

[0095] The spatial filtering algorithm in this embodiment does not require prior knowledge of the number of target types to cluster point clouds, and can remove noise points that are outside the clusters.

[0096] S3. Based on the spatial covariance eigenvalues ​​of point cloud clusters, calculate the spatial morphological feature descriptor of each point cloud cluster, and the spatial morphological features of multiple point cloud clusters constitute the point cloud cluster sample set.

[0097] Step S3 further extracts the spatial morphological features of point cloud clusters to identify target clusters and suppress interfering clusters, specifically including:

[0098] S31. Define the spatial covariance matrix of the point cloud cluster based on the positions of scattering points and the cluster center within the point cloud cluster.

[0099] Specifically, in three-dimensional space, the positions of scattering points within a point cloud cluster P containing K scattering points can be represented as p k =(x k ,y k ,z k ) T (k=1,…,K), where x k ,y k ,z k Let p represent the coordinates of the k-th scattering point, then the location of the cluster center is p. c Represented as:

[0100]

[0101] Define the spatial covariance matrix of this point cloud cluster as:

[0102]

[0103] Where, p kc Let be the spatial position of the k-th scattering point relative to the cluster center.

[0104] S32. Perform eigenvalue decomposition on the spatial covariance matrix to obtain the spatial covariance eigenvalues. The expression for eigenvalue decomposition is:

[0105]

[0106] Where Λ=diag(λ1,λ2,λ3) is a diagonal matrix of spatial covariance eigenvalues ​​arranged in descending order. The eigenvector matrix corresponding to the three spatial covariance eigenvalues ​​λ1, λ2, λ3 is V. LRF = [v1, v2, v3], where v1, v2, and v3 are feature column vectors. These feature column vectors are orthogonal and can serve as the three coordinate axes of the Local Reference Frame (LPF). Point cloud feature descriptions based on the LPF are only related to the geometry of the target and are independent of its position and orientation. The eigenvalues ​​of the spatial covariance matrix demonstrate a strong ability to describe complex geometric structures.

[0107] S33. Calculate the spatial morphological feature descriptor for each point cloud cluster using the spatial covariance eigenvalues ​​to obtain the spatial morphological features of each point cloud cluster:

[0108] x i =[s 1,i ,s 2,i ,…,s N,i ] T

[0109] Where, x i s represents the N-dimensional feature vector formed by the spatial morphological features of the i-th point cloud cluster. 1,i ,s 2,i ,…,s N,i This represents the N spatial morphological feature descriptors corresponding to the i-th point cloud cluster.

[0110] Specifically, the spatial morphological feature descriptor for each point cloud cluster includes anisotropy, sum of eigenvalues, linearity, flatness, sphericity, feature entropy, curvature, and total variance. Please refer to Table 1, which provides several spatial morphological feature descriptors for this embodiment.

[0111] Table 1. Spatial Morphological Feature Descriptors

[0112]

[0113] The spatial morphological feature descriptor of point cloud clusters can characterize point cloud clusters of different physical sizes and geometric shapes, and the features remain robust after the point cloud clusters have undergone translation and rotation transformations.

[0114] S34. A point cloud cluster sample set is formed from the spatial morphological features of all point cloud clusters:

[0115]

[0116] Where M represents the number of point cloud clusters, It is an N-dimensional real number space.

[0117] S4. Based on the differences in spatial morphology between the target and the interference, a soft-spaced linear SVM target discriminator is used to perform target discrimination and interference suppression on each point cloud cluster in the point cloud cluster sample set, and the anti-interference target imaging result is obtained.

[0118] The spatial morphology of a 3D point cloud reflects the physical size and shape information of the target. Under broadband conditions, the distribution of point cloud clusters in the 3D image signal space of the interference signal emitted by the towed decoy and the target echo shows significant differences. The point cloud generated by the towed decoy signal is characterized by high energy, small physical size, and concentrated point cloud distribution; while the target's scattered point cloud clusters exhibit obvious three-dimensional geometric features, larger physical size, wider scattering point distribution range, and significant anisotropy.

[0119] Based on spatial morphological feature descriptors, this embodiment proposes a multi-feature soft-margin linear SVM target discriminator to distinguish between decoys and targets. The design method of this soft-margin linear SVM target discriminator includes the following steps:

[0120] 1) Obtain the point cloud cluster sample set and its corresponding label set.

[0121] Specifically, let the input be a set of M point cloud cluster samples. x i =[s 1,i ,s 2,i ,…,s N,i ] T Let s be the N-dimensional feature vector of the i-th point cloud cluster. 1,i ,s 2,i ,…,s N,i These are the N spatial morphological feature descriptors corresponding to the i-th point cloud cluster. This is the label set for the corresponding input feature vector. Let N be a real number space, where -1 represents "decoy" and "1" represents "target".

[0122] 2) Transform the target identification problem in point cloud cluster samples into a convex quadratic programming binary classification problem:

[0123]

[0124] in, To separate the hyperplane w T x i The normal vector of +b=0, b is the offset of the linear classifier, C is the penalty factor satisfying C>0, ζ i These are slack variables.

[0125] 3) By introducing Lagrange multipliers to construct Lagrange functions and combining them with strong duality, the convex quadratic programming binary classification problem is transformed into a dual classification problem.

[0126] Specifically, by introducing the Lagrange multiplier λ i μ i Construct the Lagrange function:

[0127]

[0128] stλ i ≥0, μ i ≥0, ζ i ≥0

[0129] The outpost meets the following requirements:

[0130]

[0131] Substituting the conditions satisfied at the stationary points into the Lagrangian function, and leveraging strong duality, the convex quadratic programming binary classification problem is transformed into a more computationally efficient dual classification problem:

[0132]

[0133] Where, λ i μ is the first Lagrange multiplier of the i-th point cloud cluster. i It is the second Lagrange multiplier for the i-th point cloud cluster.

[0134] 4) Solve the dual form of the classification problem, and combine it with the convex quadratic programming binary classification problem to obtain the hyperplane and the corresponding decision function, thus obtaining the soft-margin linear SVM object discriminator.

[0135] Specifically, the Sequential Minimal Optimization (SMO) algorithm is used to solve the dual form of the solution. Let the vector be formed by the Lagrange multipliers of multiple point cloud clusters, and then... Substituting into the convex quadratic programming binary classification problem, we can obtain the solution to this problem. and Finally, the separating hyperplane is obtained. and the corresponding decision function Obtain a soft-spaced linear SVM target discriminator.

[0136] Furthermore, the point cloud cluster sample set obtained after steps S1, S2, and S3 is input into the soft-interval linear SVM target discriminator, which outputs the target discrimination result. When the output of the soft-interval linear SVM target discriminator is 1, the point cloud cluster sample is the target; when the output of the soft-interval linear SVM target discriminator is -1, the point cloud cluster sample is interference. The target imaging result is formed from the point cloud cluster sample whose output is the target.

[0137] The dragging jamming countermeasure method based on spatial morphology features provided in this embodiment achieves noise removal and strong scattering point cloud clustering by performing spatial filtering on the three-dimensional point cloud, and extracts spatial morphology features of each point cloud cluster. Target identification and jamming suppression are achieved through a soft-interval linear SVM classifier, resulting in a three-dimensional image of the target after jamming. Taking advantage of the fact that the jamming signal emitted by the decoy under broadband conditions cannot completely simulate the spatial morphology of the target in the three-dimensional imaging space, and in cases where it is difficult to distinguish between the target and the jamming in one-dimensional range image, the dragging jamming countermeasure fully utilizes the differences in spatial morphology features between the jamming and the target to achieve dragging jamming countermeasure. The jamming identification accuracy is high, and the quality of the target imaging result after jamming is high.

[0138] Furthermore, to verify the effectiveness of the above method, this embodiment uses a single-pulse terminal guidance and tracking system as an example to simulate a typical towed jamming scenario. The main simulation parameters are shown in Table 2.

[0139] Table 2 Main Simulation Parameters

[0140]

[0141] dragged decoy interference scenarios such as Figure 5 As shown, Figure 5 This diagram illustrates the spatial relationships and signal transmission in a towed jamming scenario provided by an embodiment of the present invention. The radar is located at the origin O(0,0,0). Initially, the radar beam center illuminates the target's centroid, and the beam center's direction forms a 30° horizontal angle with the target's flight direction. The towed decoy is located directly behind the target. The jammer transmits modulated false target signals.

[0142] By performing stretch processing on the received signals from the three channels of azimuth, pitch difference, and slew rate, it can be seen from the processing results that the modulated forwarding jamming signal emitted by the decoy jammer occupies multiple range cells and has an amplitude higher than that of the real target, thus exhibiting good range deception characteristics. It is difficult to distinguish the target from the jammer from the one-dimensional range profile.

[0143] After further motion compensation, single-pulse three-dimensional imaging was performed on the target under dragged interference conditions. According to the angle measurement results of the three-dimensional imaging, the measurement singularity caused by the interference and the strong scattering point of the target are both located within the main lobe beam, which seriously affects guidance and tracking.

[0144] Please see Figures 6a-6d , Figures 6a-6d The three-view diagram and ISAR image of the target under dragged interference conditions provided in the embodiments of the present invention are shown below. Figure 6a This is a front view. Figure 6b This is a side view. Figure 6c This is a top view. Figure 6dThis is an ISAR image. The angle measurement results, combined with distance information, are converted into a 3D image of the forward-looking scene using the calculation formulas for the relative distance in the azimuth and pitch dimensions, as shown below. Figures 6a-6c As shown in the 3D image, the forward-looking region contains two point cloud clusters with significantly different shapes, formed by strong scattered point echoes. These clusters represent the real target and the false target generated by the interference signal, respectively. The 6D ISAR imaging results also reveal strong interference with amplitudes far exceeding the target echo size, exceeding the target's dimensions. Since the target and decoy velocities are much greater than the velocity measurement range, velocity ambiguity exists, a phenomenon also reflected in the ISAR results.

[0145] Furthermore, spatial noise removal and point cloud clustering are performed on the imaging results to obtain point cloud cluster division results, such as... Figures 7a-7c As shown, Figures 7a-7c These are three views of the point cloud clustering results provided in this embodiment of the invention. Figure 7a This is a front view. Figure 7b This is a side view. Figure 7c This is a top view. As can be seen from Figure 7, the false target is generated by the interference signal. In terms of spatial morphology, it exhibits high density and linear features, and there is a significant difference in three-dimensional shape features between it and the real target in three-dimensional image space.

[0146] Under the current simulation parameters, 1000 Monte Carlo experiments were conducted to obtain clustered point cloud sets. Feature extraction and classifier training were performed on the clustered point cloud sets generated from 800 of these experiments, resulting in the designed discriminator decision function parameters:

[0147]

[0148] The target identification test was conducted using the point cloud clusters obtained from the remaining 200 experiments. The results show that the decoy's interference signal and the target echo exhibit significant differences in spatial morphological characteristics after reconstruction in 3D imaging space. The interference signal in 3D space exhibits higher linearity, smaller size, higher aggregation, and weaker irregularity, while the target displays obvious shape characteristics, with a larger volume, higher flatness, more dispersed scattering point distribution, and stronger irregularity. The test results indicate that the designed classifier achieves a 100% accuracy rate in distinguishing between the target and the interference.

[0149] The target imaging results after interference suppression are as follows Figures 8a-8c As shown, Figures 8a-8c These are three views of the target imaging result after interference suppression provided in an embodiment of the present invention. Figure 8a This is a front view; 8b is a side view. Figure 8c This is a top view. As shown in Figure 8, the three-view diagram of the target after interference suppression is basically consistent with the three-view diagram of the real target.

[0150] This embodiment addresses the problem of terminal guidance radar being easily deceived by towed decoys, leading to strike failure. It proposes a method for countering towed decoys in terminal guidance radar based on spatial morphology features. This method utilizes single-pulse 3D imaging technology to achieve 3D imaging of the forward-looking area; spatial filtering is used to cluster and remove noise from strong scattering point cloud clusters; and by extracting the spatial morphology features of each point cloud cluster, a soft-interval linear SVM classifier is used to identify the target and remove interference points, resulting in a true 3D image of the target. Simulation results verify the effectiveness of this method against towed decoys.

[0151] The above description, in conjunction with specific preferred embodiments, provides a further detailed explanation of the present invention. It should not be construed that the specific implementation of the present invention is limited to these descriptions. For those skilled in the art, various simple deductions or substitutions can be made without departing from the concept of the present invention, and all such modifications and substitutions should be considered within the scope of protection of the present invention.

Claims

1. A drag-and-drop interference countermeasure method based on spatial morphological characteristics, characterized in that, Including the following steps: Echo data from the azimuth, pitch, and azimuth channels are acquired, and the echo data are used to perform three-dimensional imaging of the radar forward-looking area using single-pulse three-dimensional imaging technology to obtain a three-dimensional point cloud. The three-dimensional point cloud is clustered and noise is removed using a density-based spatial filtering algorithm to obtain several point cloud clusters. Based on the spatial covariance eigenvalues ​​of point cloud clusters, a spatial morphological feature descriptor for each point cloud cluster is calculated, and a point cloud cluster sample set is formed by the spatial morphological features of multiple point cloud clusters. The spatial morphological feature descriptor for each point cloud cluster includes anisotropy, sum of eigenvalues, linearity, flatness, sphericity, feature entropy, curvature, and total variance. The formula for calculating anisotropy is as follows: ; The formula for calculating the sum of the eigenvalues ​​is: ; The formula for calculating the linearity is: ; The formula for calculating the flatness is: ; The formula for calculating the sphericity is: ; The formula for calculating the feature entropy is: ; The formula for calculating the curvature is: ; The formula for calculating the total variance is as follows: ; in, These are three spatial covariance eigenvalues. ; Based on the spatial morphological differences between the target and the interference, a soft-spaced linear SVM target discriminator is used to perform target identification and interference suppression on each point cloud cluster in the point cloud cluster sample set, resulting in anti-interference target imaging results.

2. The dragging interference countermeasure method based on spatial morphological characteristics according to claim 1, characterized in that, Echo data from the azimuth, elevation, and pitch difference channels are acquired, and the echo data is used to perform three-dimensional imaging of the radar forward-looking area using single-pulse three-dimensional imaging technology to obtain a three-dimensional point cloud, including: Acquire echo data for the azimuth, azimuth, and pitch channels; The echo data is subjected to range-Doppler high-resolution processing to obtain high-resolution processing results for the sum channel, azimuth difference channel, and elevation difference channel; Two-dimensional constant false alarm rate (CFAR) detection is performed on the high-resolution processing results of the channels to extract the distance information corresponding to several resolution units occupied by strong scattering points. Using the high-resolution processing results of the sum channel, azimuth difference channel, and elevation difference channel, the angle is measured at the amplitude of the over-detection threshold cell to obtain the azimuth and elevation angles of each strong scattering point resolution cell relative to the radar. Combining the distance information, the azimuth and elevation angles of each strong scattering point resolution unit relative to the radar are converted into azimuth-dimensional relative distance and elevation-dimensional relative distance to obtain the three-dimensional point cloud.

3. The dragging interference countermeasure method based on spatial morphological characteristics according to claim 2, characterized in that, The relative distances in the azimuth dimension and the relative distances in the pitch dimension are respectively: in, The relative distance is in the azimuth dimension. The azimuth angle of each strong scattering point resolution element relative to the radar. This refers to the distance information corresponding to several resolution cells occupied by a strong scattering point. The relative distance is in the pitch dimension. The elevation angle of each strong scattering point resolution unit relative to the radar.

4. The dragging interference countermeasure method based on spatial morphological characteristics according to claim 1, characterized in that, A density-based spatial filtering algorithm is used to perform clustering and noise removal on the 3D point cloud, resulting in several point cloud clusters, including: Calculate the distance between each pair of scattering points in the three-dimensional point cloud to obtain the distance matrix; Choose any scattering point as the reference point for the new cluster. Count the number of first scattering points within a preset neighborhood radius of the reference point. If the number of first scattering points is less than the minimum number of elements contained in the preset neighborhood, the reference point is discarded as noise. If the number of first scattering points is greater than or equal to the minimum number of elements contained in the preset neighborhood, the scattering points within the preset neighborhood radius are added to the current cluster. Select a scattering point within the current cluster that has not been used as a reference point as a new reference point; The number of second scattering points within the preset neighborhood radius of the new reference point is counted. If the number of second scattering points is less than the minimum number of elements contained in the preset domain, a new reference point is selected again in the current cluster for judgment. If the number of second scattering points is greater than or equal to the minimum number of elements contained in the preset domain, the scattering points within the preset neighborhood radius of the new reference point are included in the current cluster. Iterate through all scattering points within the current cluster; When there are unclustered points, a scattering point is selected from the unclustered points as a reference point for the new cluster and either removed or clustered to obtain a new point cloud cluster; when there are no unclustered points, the spatially filtered point cloud clusters are output.

5. The dragging interference countermeasure method based on spatial morphological characteristics according to claim 1, characterized in that, Based on the spatial covariance eigenvalues ​​of point cloud clusters, a spatial morphological feature descriptor for each point cloud cluster is calculated. A point cloud cluster sample set is constructed from the spatial morphological features of multiple point cloud clusters, including: The spatial covariance matrix of the point cloud cluster is defined based on the positions of scattering points within the cluster and the cluster center position: in, This represents the number of scattering points within a point cloud cluster. For the first The spatial position of each scattering point relative to the cluster center The location of the cluster center , For the first The location of each scattering point , , Indicates the first The coordinates of the scattering points; The spatial covariance matrix is ​​subjected to eigenvalue decomposition to obtain the spatial covariance eigenvalues. The expression for the eigenvalue decomposition is as follows: in, A diagonal matrix of spatial covariance eigenvalues ​​arranged in descending order. Three spatial covariance eigenvalues The corresponding eigenvector matrix is , , , For feature column vectors; The spatial morphological feature descriptor of each point cloud cluster is calculated using the spatial covariance eigenvalues, thus obtaining the spatial morphological features of each point cloud cluster: in, Indicates the first The N-dimensional feature vector is formed by the spatial morphological features of a point cloud cluster. Indicates the first N spatial morphological feature descriptors corresponding to a point cloud cluster; The point cloud cluster sample set is composed of the spatial morphological features of all point cloud clusters: in, Indicates the number of point cloud clusters. It is an N-dimensional real number space.

6. The dragging interference countermeasure method based on spatial morphological characteristics according to claim 1, characterized in that, The design method of the soft-spaced linear SVM target discriminator includes the following steps: Obtain a point cloud cluster sample set and its corresponding label set; The target identification problem in the point cloud cluster sample set is transformed into a convex quadratic programming binary classification problem: in, To separate the hyperplane The normal vector, This is the offset of the linear classifier. For the penalty factor to be satisfied , As slack variables, The number of point cloud clusters in the sample set. For the first N-dimensional feature vectors of a point cloud cluster for A set of point cloud cluster samples, It is an N-dimensional real number space; By introducing Lagrange multipliers to construct a Lagrange function and combining it with strong duality, the convex quadratic programming binary classification problem is transformed into a dual classification problem: in, For the first The first Lagrange multiplier of a point cloud cluster, For the first The second Lagrange multiplier of a point cloud cluster, This is the label set corresponding to the point cloud cluster sample set, where -1 represents "decoy" and "1" represents "target"; Solve the dual-form classification problem and combine it with the convex quadratic programming binary classification problem to obtain the hyperplane. and the corresponding decision function The soft-spaced linear SVM target discriminator is obtained.

7. The dragging interference countermeasure method based on spatial morphological characteristics according to claim 6, characterized in that, Based on the spatial morphological differences between the target and the interference, a soft-spaced linear SVM target discriminator is used to perform target discrimination and interference suppression on each point cloud cluster in the point cloud cluster sample set, resulting in anti-interference target imaging results, including: The point cloud cluster samples in the point cloud cluster sample set are sequentially input into the interval linear SVM target discriminator. When the output of the interval linear SVM target discriminator is 1, the point cloud cluster sample is the target. When the output of the interval linear SVM target discriminator is -1, the point cloud cluster sample is interference. The target imaging result is formed from the point cloud cluster samples whose output is the target.

Images