A Local Joint Spatiotemporal Adaptive Clutter Suppression Method Based on Beam Domain Weighting

Abstract

This invention provides a local joint space-time adaptive clutter suppression method based on beam domain weighting, comprising: S1: processing data using subarray partitioning combined with difference beamforming to obtain beam domain weighting values; S2: constructing a weighted space-time steering vector based on the beam domain weighting values ​​calculated in S1, collecting training samples in the range dimension, calculating the covariance matrix containing clutter information using a local joint dimensionality reduction space-time adaptive method, and then suppressing ionospheric clutter. For small-aperture high-frequency ground wave radar, in complex electromagnetic environments, insufficient array degrees of freedom can lead to reduced clutter suppression performance, insufficient signal-to-clutter ratio of the target, and inability to form a clear peak at the angle. This invention addresses this problem, enabling small-aperture high-frequency ground wave radar to improve the target signal-to-clutter ratio and achieve higher angular accuracy after clutter suppression.

Description

Technical Field

[0001] This invention belongs to the field of radar clutter suppression, specifically relating to a local joint space-time adaptive clutter suppression method based on beam domain weighting. Technical Background

[0002] High-frequency ground-wave over-the-horizon radar possesses unique advantages in anti-stealth, anti-low-altitude penetration, and over-the-horizon detection. Currently, most existing shore-based high-frequency ground-wave radar systems primarily used for target detection employ large array structures. However, as the ocean plays an increasingly important role in the national economy, the large-scale occupation of scarce coastal resources has become a significant factor limiting the development of ground-wave radar. Therefore, there is an urgent need to develop a high-frequency ground-wave radar system with a small aperture array primarily for target detection.

[0003] While small-aperture high-frequency ground-wave radar systems significantly reduce the footprint, they also present substantial challenges to signal processing methods. Especially in complex clutter environments, small-aperture arrays have fewer degrees of freedom and wider main lobe beams. The broadened radio station and clutter angular spectrum makes targets more easily obscured and difficult to detect. Clutter suppression requires training samples with identically distributed (i.i.d.) characteristics to construct the covariance matrix. Beam broadening severely impacts the performance of space-time adaptive clutter suppression methods. Furthermore, the reduced degrees of freedom diminish the i.i.d. characteristics of ionospheric clutter data, leading to severely inaccurate estimation of the covariance matrix carrying clutter information and ultimately compromising the target information after clutter suppression. Summary of the Invention

[0004] The purpose of this invention is to solve the problem of insufficient degrees of freedom leading to a widened main beam and low clutter suppression performance in small-aperture high-frequency ground wave radars during clutter suppression. This results in a significant improvement in the accuracy of the signal-to-clutter ratio and angle information for the target.

[0005] To achieve the above objectives, the present invention adopts the following technical solution:

[0006] A local joint space-time adaptive clutter suppression method based on beam domain weighting includes the following steps:

[0007] S1: The data is processed using a combination of subarray partitioning and sum-difference beamforming to obtain beam domain weights;

[0008] S2: Based on the beam domain weights calculated in S1, a weighted space-time steering vector is constructed. Training samples are collected in the distance dimension. A local joint dimensionality reduction space-time adaptive method is used to calculate the covariance matrix containing clutter information. This covariance matrix is ​​then used to suppress ionospheric clutter.

[0009] Preferably, S1 includes the following sub-steps:

[0010] S11: Set the receiving antenna array to include N e A uniform linear array of n elements with an array spacing of d; assume the incident signal is a far-field plane wave; take the leftmost element as the reference element, and the direction perpendicular to the array as the normal direction; the pulse repetition period is N. p one, f s The sampling frequency is M = N for the l-th distance unit. e N p Dimensional data is represented as:

[0011]

[0012]

[0013] in[·] T As a matrix transpose, the spacetime steering vector of the echo is defined as:

[0014]

[0015] in, For the Kronecker product; a(φ) t ) and b(f t These are the spatial steering vector and the temporal steering vector, respectively.

[0016]

[0017]

[0018] Where, φ t f is the azimuth of the target echo. t The radial Doppler frequency of the target;

[0019] The received data is represented as the sum of the target signal, external interference, and noise:

[0020] X = ξ t v(f t ,φ t )+c+n

[0021] Where ξ is the signal amplitude of the target, c is the clutter signal, and n is the noise signal;

[0022] S12: Set the spacetime domain to K=N s N d A uniform discrete grid of points,

[0023] Where, N s and N d These represent the number of spatial elements and the number of Doppler elements, respectively. Furthermore, each point on the mesh corresponds to a spacetime steering vector: v k,k=1,...,K.

[0024] S13: As Figure 1 As shown, the receiving array is divided into two equal subarrays, left and right. Assuming the number of array elements is even, the left subarray consists of elements from 1 to N. e It consists of 2 array elements, with the right subarray being the Nth element. e / 2+1 to the Nth e Composed of several array elements, each performing conventional beamforming on the left and right subarrays, a beam pointing towards φ can be obtained. t Left beam R L and R R Right beam:

[0025]

[0026]

[0027] Where I n Given the signal amplitude of the nth array element, the left and right beam outputs are then used to construct the sum beam Sum(θ) and difference beam Diff(θ) according to the following formulas:

[0028] Sum(φ t )=|R L |+|R R |

[0029] Diff(φ t )=|R L -R R |

[0030] S14: Construct beam coefficients from the obtained sum and difference beams:

[0031] AF Hyper ={(|R L |+|R R |) u -(|R L -R R |) u} 1 / u

[0032] Where μ∈(0.3,1] is the modulation coefficient. The obtained superbeam coefficients are then combined with the spatial steering vector as a weight vector to obtain the superbeam spatial steering vector, which can be expressed as:

[0033] a Hyper (φ t ) = AF Hyper ·a(φ t )

[0034] Therefore, the reconstructed weighted beam space-time steering vector expression is:

[0035]

[0036] v Hyper (f t ,φ t This refers to the beam-domain weighted space-time steering vector, which can replace the general space-time steering vector in subsequent space-time adaptive algorithms. The beam-domain weighted data better conforms to the independent and identically distributed characteristics in the distance dimension, which can improve the performance of clutter suppression.

[0037] Preferably, S2 includes the following sub-steps:

[0038] S21: Space-time adaptive processing involves finding an optimal weight vector to process the data, which can be represented as:

[0039] y = w H X l

[0040] in[·] H Let y be the transpose of the matrix, which is the result of the space-time adaptive processing.

[0041] For beam-domain weighted spatiotemporal adaptive processing, the optimal weight vector can be expressed as:

[0042] w = R -1 v Hyper

[0043] Where R is the covariance matrix containing noise and clutter information. For practical systems, the covariance matrix R needs to be constructed by obtaining training samples along the distance dimension:

[0044]

[0045] Where L is the number of training samples, and l = [0, 1, ..., L-1]. The problem with this all-space-time adaptive algorithm is the difficulty in obtaining prior knowledge of clutter, and it requires at least 2M independent and identically distributed sample data to accurately estimate the covariance matrix. However, for practical high-frequency ground wave radar systems, the number of samples is far from sufficient. To address this issue, a local joint dimensionality reduction space-time adaptive (JDL) method is used.

[0046] S22: The JDL algorithm performs local joint processing on M×1 training samples, transforming them into η. a η d ×1 dimensional space-time data, beam-domain weighted space-time transformation matrix T Hyper Set to:

[0047]

[0048] Where, ηd and η a These represent the number of Doppler resolution units and the number of beams after dimensionality reduction, respectively.

[0049] The dimensionality-reduced data and the space-time steering vector can then be represented as:

[0050]

[0051]

[0052] The covariance matrix can then be estimated from the training samples:

[0053]

[0054] Then the adaptive weight vector for local joint dimensionality reduction processing based on beam domain weighting is:

[0055]

[0056] S23: The output result can be obtained by using the adaptive weight vector obtained by local joint dimensionality reduction processing based on beam domain weighting to perform clutter suppression processing on the data:

[0057] y = w Hyper H X l

[0058] In step S22, the angle and Doppler data of the l-th range cell are processed to obtain:

[0059]

[0060] Where, N d and N a These represent the number of Doppler resolution units and the number of beams selected for the data, respectively; Y l This is the clutter suppression output obtained by processing the l-th distance cell using this method.

[0061] S24: Repeat step S23 to calculate all distance cells Y. l The result.

[0062] The technical solution provided by this invention can enable small array high-frequency ground wave radar to obtain a narrower main lobe width and can effectively suppress ionospheric clutter, making up for the performance degradation of traditional space-time adaptive processing in the case of small arrays, thereby significantly improving the signal-to-clutter ratio of the target and having a more accurate peak value at the angle. Attached Figure Description

[0063] Figure 1 This is a flowchart of the present invention;

[0064] Figure 2 Distance-Doppler spectrum of measured echo data.

[0065] Figure 3 The angle-Doppler spectrum of the distance cell after adding the simulation target.

[0066] Figure 4 A bi-subarray model of a uniform linear array.

[0067] Figure 5 Range-Doppler spectrum after clutter suppression processing.

[0068] Figure 6 Angle-Doppler spectrum of the distance cell after clutter suppression processing.

[0069] Figure 7 Doppler profile of the target location.

[0070] Figure 8 Angle profile of the target location. Detailed Implementation

[0071] The specific embodiments of the present invention are described below to enable those skilled in the art to understand the present invention. However, it should be understood that the present invention is not limited to the scope of the specific embodiments. For those skilled in the art, various changes are obvious as long as they are within the spirit and scope of the present invention as defined and determined by the appended claims. All inventions utilizing the concept of the present invention are protected.

[0072] like Figure 1 As shown, a local joint space-time adaptive clutter suppression method based on beam domain weighting includes:

[0073] S1: The data is processed using a subarray partitioning combined with sum-difference beamforming to obtain beam domain weights; the specific sub-steps are as follows:

[0074] S11: Set the high-frequency ground wave radar to N. e =A uniform linear array of 8 units, with an array spacing of d = 14.5 meters, and the signal contains N p = 2048 pulse repetition cycles, the signal sampling frequency is f s Thus, we get M = N e N p The data has dimensions, where the i-th distance unit is represented as:

[0075]

[0076]

[0077] in[·] TAs a matrix transpose, the spacetime steering vector of the data with the target is represented as:

[0078]

[0079] Where, a(φ) t ) and b(f t These are the spatial steering vector and the temporal steering vector, respectively.

[0080]

[0081]

[0082] in, Indicates the Kronecker product. λ = 65.8 meters is the wavelength of the signal, φ t f is the azimuth of the target echo. t f is the radial Doppler frequency of the target. R The pulse repetition frequency of the signal;

[0083] S12: Set the spacetime domain to K=N s N d A uniform discrete grid of points,

[0084] Where, N s and N d Representing the number of spatial elements and the number of Doppler elements respectively, N is chosen. s =61, N d =251. And each point on the grid corresponds to a spacetime steering vector, with an angular interval of 3° and a Doppler interval of 0.0093Hz: v k k = 1, ..., K. Preferably, a simulated target with a signal-to-noise ratio of 0 dB, a target angle of 12°, and a Doppler frequency of -0.18 Hz is added to the measured echo data. The distance-Doppler spectrum and angle-Doppler spectrum of the processed measured echo data are as follows: Figure 2 , Figure 3 As shown. Due to the presence of strong ionospheric clutter, the target cannot be detected.

[0085] S13: As Figure 4 As shown, the receiving array is divided into two equal subarrays, left and right. At this time, the number of array elements is even. The left subarray consists of elements from 1 to N. e The array is composed of 4 array elements, with the right subarray being the Nth element. e / 2+1=5 to the Nth e =Composed of 8 array elements, respectively performing conventional beamforming on the left and right subarrays, a beam pointing to φ can be obtained. t Left beam R L and R R Right beam:

[0086]

[0087]

[0088] Where I n Given the signal amplitude of the nth array element, the left and right beam outputs are then used to construct the sum beam Sum(θ) and difference beam Diff(θ) according to the following formulas:

[0089] Sum(φ t )=|R L |+|R R |

[0090] Diff(φ t )=|R L -R R |

[0091] Then, the obtained sum and difference beams are used to construct beam domain weighting coefficients:

[0092] AF Hyper ={(|R L |+|R R |) u -(|R L -R R |) u} 1 / u

[0093] Where μ = 0.8, μ∈(0.3,1] is the modulation coefficient. The obtained beam-domain weighting coefficients are then combined with the spatial steering vector as a weight vector to obtain the beam-weighted spatial steering vector, which can be expressed as:

[0094] a Hyper (φ t ) = AF Hyper ·a(φ t )

[0095] Therefore, the reconstructed beam-weighted space-time steering vector expression is:

[0096]

[0097] v Hyper (f t ,φ t This is the beam-weighted space-time steering vector, which can replace the general space-time steering vector in subsequent space-time adaptive algorithms. The beam-domain weighted data better conforms to the independent and identically distributed characteristics in the range dimension, which can improve the performance of clutter suppression.

[0098] S2: Based on the beam domain weights calculated in S1, a weighted space-time steering vector is constructed. Training samples are collected in the distance dimension. A local joint dimensionality reduction space-time adaptive method is used to calculate the covariance matrix containing clutter information. This covariance matrix is ​​then used to suppress ionospheric clutter.

[0099] S2 includes the following sub-steps:

[0100] S21: Space-time adaptive processing involves finding an optimal weight vector to process the data, which can be represented as:

[0101] y = w H X l

[0102] in[·] H Let y be the transpose of the matrix, which is the result of the space-time adaptive processing.

[0103] For all-space-time adaptive processing, the optimal weight vector can be expressed as:

[0104] w = R -1 v Hyper

[0105] Where R is the covariance matrix containing noise and clutter information. For practical systems, the covariance matrix R needs to be constructed by obtaining training samples along the distance dimension:

[0106]

[0107] Where L is the number of training samples. In this example, the number of samples is 30, and l = [0, 1, ..., L-1].

[0108] S22: The JDL algorithm performs local joint processing on M×1 training samples, transforming them into η. a η d ×1 dimensional spacetime data. Then the spacetime transformation matrix T Hyper Represented as:

[0109]

[0110] Where, η d and η a These represent the number of Doppler resolution units and the number of beams after dimensionality reduction, respectively, where η is set. a =5, η d =3.

[0111] The dimensionality-reduced data and the space-time steering vector can then be represented as:

[0112]

[0113]

[0114] The covariance matrix can then be estimated from the training samples:

[0115]

[0116] Then the adaptive weight vector for local joint dimensionality reduction processing based on beam domain weighting is:

[0117]

[0118] S23: The output result can be obtained by using the adaptive weight vector obtained by local joint dimensionality reduction processing based on beam domain weighting to perform clutter suppression processing on the data:

[0119] y = w Hyper H X l

[0120] In step S22, the angle and Doppler data of the l-th range cell are processed to obtain:

[0121]

[0122] Where, N d and N a These represent the number of Doppler resolution units and the number of beams selected for the data, respectively; Y l This is the clutter suppression output obtained by processing the l-th distance cell using this method.

[0123] S24: Repeat step S23 to calculate all distance cells Y. l The result.

[0124] The distance-Doppler spectrum and angle-Doppler spectrum after processing by the clutter suppression algorithm proposed in this invention are as follows: Figure 5 and Figure 6 As shown in the diagram, the circled area indicates the location where the target was added. It can be observed that the target is clearly highlighted after clutter suppression.

[0125] like Figure 7 and Figure 8 The figure shows the Doppler frequency profile and angle profile of the processed results using the proposed algorithm. For comparison, a profile before clutter suppression is also included. The dashed lines in the figure represent the Doppler frequency and angle of the target. In the Doppler frequency profile, it can be seen that after clutter suppression processing using the proposed algorithm, the target's Doppler frequency forms a distinct peak. In the angle profile, the unprocessed data is high-amplitude, indistinguishable clutter, while the processed data forms a sharp peak in the target's direction. The signal-to-clutter ratio of the target is improved by 17.9 dB, making the target easily detectable.

[0126] Based on the above results, it can be concluded that the beam domain weighted local joint space-time adaptive clutter suppression method proposed in this invention can effectively process complex clutter under small arrays, preserve target information as completely as possible, and significantly improve the signal-to-clutter ratio of the target. In particular, it can form sharp and accurate peaks in terms of angle, which can effectively help subsequent target detection.

Claims

1. A local joint space-time adaptive clutter suppression method based on beam domain weighting, characterized in that, Includes the following steps: S1: The data is processed using a combination of subarray partitioning and sum-difference beamforming to obtain beam domain weights; Includes the following sub-steps: S11: Set the receiving antenna array to be used, including... A uniform linear array with n elements and an array spacing of n. Assume the incident signal is a far-field plane wave; take the leftmost element as the reference element, and the direction perpendicular to the array as the normal direction; the pulse repetition period is... indivual, The sampling frequency is the frequency of the l-th distance unit. Dimensional data is represented as: , ， in As a matrix transpose, the spacetime steering vector of the echo is defined as: , in, For Kronecker product; and These are the spatial steering vector and the temporal steering vector, respectively. ， ， in, The azimuth of the target echo. The radial Doppler frequency of the target; The received data is represented as the sum of the target signal, external interference, and noise: ， in, The signal amplitude of the target It is a clutter signal. This is a noise signal; S12: Set the space-time domain to A uniform discrete grid of points, in, and These represent the number of spatial elements and the number of Doppler elements, respectively; and each point on the grid corresponds to a spacetime steering vector: ; S13: Divide the receiving array into two equal subarrays, left and right. Assuming the number of array elements is even, the left subarray is divided into two subarrays from the first to the second element. The array is composed of several elements, with the right subarray being the first... To the Composed of several array elements, each element performs conventional beamforming on the left and right subarrays to obtain a directional beam. left beam and Right beam: ， ， in For the first The signal amplitude of each array unit is then used to construct the beam using the left and right beam outputs according to the following formula. Sum and difference beam : ， ， S14: Construct beam coefficients from the obtained sum and difference beams: ， in The modulation coefficients are then used as the weighting coefficients, which are combined with the spatial steering vector to obtain the beam spatial steering vector, expressed as: ， Therefore, the reconstructed weighted post-space-time steering vector expression is: ， This is the beam-domain weighted space-time steering vector, which replaces the general space-time steering vector in subsequent space-time adaptive algorithms; the beam-domain weighted data is more consistent with the independent and identically distributed characteristics in the distance dimension, thus improving the performance of clutter suppression. S2: Construct a weighted space-time steering vector based on the beam domain weights calculated in S1, collect training samples in the distance dimension, use a local joint dimensionality reduction space-time adaptive method to calculate the covariance matrix containing clutter information, and use the covariance matrix to suppress ionospheric clutter.

2. The method for suppressing clutter based on beam domain weighting in a local joint space-time adaptive manner according to claim 1, characterized in that, In S11, and Specifically, it is expressed as follows: ， ， in, , , The wavelength of the signal. The azimuth of the target echo. The radial Doppler frequency of the target is the pulse repetition frequency of the signal.

3. The method for suppressing clutter based on beam domain weighting in a local joint space-time adaptive manner according to claim 1, characterized in that, S2 includes the following sub-steps: S21: Space-time adaptive processing involves solving for an optimal weight vector to process the data, expressed as: ， in This is the transpose of the matrix. That is, the result of space-time adaptive processing; For beam-domain weighted spatiotemporal adaptive processing, the optimal weight vector is expressed as: ， in It is the covariance matrix containing noise and clutter information; Then, the local joint dimensionality reduction space-time adaptive JDL method is used; S22: The JDL algorithm is a method for calculating dimensions... The training samples are subjected to local joint processing and converted into Dimensional space-time data, weighted space-time transformation matrix Set to: ， in, and These represent the number of Doppler resolution units and the number of beams after dimensionality reduction, respectively. The dimensionality-reduced data and the space-time steering vector are then represented as: ， ， The covariance matrix is ​​then estimated from the training samples: ， Then the adaptive weight vector for local joint dimensionality reduction processing based on beam domain weighting is: ， S23: The output result is obtained by applying an adaptive weight vector based on beam-domain weighted local joint dimensionality reduction processing to suppress clutter in the data. ， In step S22, the angle and Doppler data of the l-th range cell are processed to obtain: ， in, and These represent the number of Doppler resolution units and the number of beams selected for the data, respectively. The clutter suppression output is obtained by processing the l-th range cell using this method. S24: Repeat step S23 to calculate all distance cells. The result.

4. The method for suppressing clutter based on beam domain weighting with local joint space-time adaptive clutter according to claim 3, characterized in that, In S21, the covariance matrix The training samples are constructed by obtaining them along the distance dimension: ， in The number of training samples, and .

Images