An airborne radar clutter training selection method

Abstract

The present application belongs to the technical field of radar system, and discloses a kind of airborne radar clutter training selection method.The present application selects the channel located in edge, corner in data cube, excludes adjacent channel with normal direction as central axis, selects the edge channel data block far from central axis as effective learning sample, if detecting target signal, impact noise or artificial interference from specific edge channel, automatically discards the contaminated channel sample, retains the far end edge channel data sample not contaminated, calculates root mean square value using screened pure clutter sample, and constructs clutter covariance matrix based on this, uses clutter covariance matrix to calculate signal clutter noise ratio SCNR.The present application can obtain the clutter covariance matrix close to real countermeasure environment through the calculation formula of clutter covariance matrix and SCNR, to a certain extent, excludes the influence of impact noise and artificial interference, so as to obtain relatively ideal optimal weighting function and STAP performance.

Description

Technical Field

[0001] The present invention mainly relates to the technical field of radar systems, and in particular to a method for selecting clutter for airborne radar training. Background Art

[0002] Airborne phased array radars often encounter the problem of strong clutter drowning out moving targets when detecting them. Space-Time Adaptive Processing (STAP) is commonly used to suppress clutter. STAP, based on a sampled covariance matrix, uses training samples to estimate the covariance matrix of the target unit and calculate adaptive weights to achieve adaptive filtering. These training samples must satisfy the IID (Independent and Identically Distributed) condition for the target unit. To ensure the accuracy of the estimated covariance matrix, the number of training samples used is typically greater than twice the system's degrees of freedom. Therefore, the training samples may be distributed over an area of ​​tens of kilometers. Such a large area may contain a variety of terrain, such as mountains, rivers, deserts, and grasslands. Radar clutter reflected from this terrain has different characteristics, resulting in the estimated covariance matrix not accurately representing the actual clutter covariance matrix of the target unit, thus degrading STAP performance.

[0003] When detecting close-range targets, the frequency information of clutter depends on the range due to the varying pitch angles at different ranges. This can cause deviations in the estimated clutter covariance matrix and broaden the clutter spectrum, a phenomenon known as clutter range dependence. Using radar operating parameters, detection angle, range, and other information, the Doppler frequency of each range ring in the observation direction and the frequency difference with the target unit are calculated. Doppler frequency compensation (DW) and high-order Doppler compensation are then performed. Azimuth spatial frequency differences also exist between samples at different ranges, leading to the development of angle-Doppler compensation (ADC) algorithms. However, the radar's a priori operating parameters (such as platform velocity and array position) often deviate from the control values, resulting in inaccurate calculated ADC values. STAP suppresses clutter by adaptively forming notches based on the clutter power spectrum in the target unit.

[0004] The plane of a phased array radar usually contains multiple different subarrays and control areas. Each subarray and control area contains dozens or hundreds of transmit / receive channels (TR channels). When a TR channel reaches a load state, the adjacent channel can be switched to a receive state; and vice versa. Through the above analysis, when selecting sample data to solve the clutter covariance matrix, it is usually necessary to carefully select the source of sample data. In traditional phased array radar wave control systems, in order to reduce the amount of calculation and reduce the dimension, the channel corresponding to a certain column or row in the plane array is usually placed in a receive state, and the clutter covariance matrix when the signal + interference + noise exist at this moment is calculated. This method has two more serious defects: (1) The entry angle of the target and interference is unknown, which may cause more serious sample contamination on the selected row or column; (2) As an array antenna using a superheterodyne receiver to implement the receiving channel, its samples near the normal center point are more likely to be affected by the internal local oscillator and the RF leakage of the excitation source. Summary of the Invention

[0005] Traditional phased array radars select random row or column data channels to obtain clutter samples. This invention provides a clutter training and selection method for airborne radars that uses edge and corner sample intervals. When a sample interval is contaminated by impulsive noise or interference, data from other sample intervals is used as learning samples. By using the clutter covariance matrix and SCNR calculation formulas in the technical solution, this invention can obtain a clutter covariance matrix in a near-realistic adversarial environment, eliminating the influence of impulsive noise and interference to a certain extent, thereby achieving a relatively ideal optimal weighting function and STAP performance.

[0006] To achieve the above object, the present invention provides an airborne radar clutter training and selection method, comprising the following steps:

[0007] Step 1: In the radar data cube, select channel data located at the edge or corner of the space as candidate clutter learning samples;

[0008] Step 2: Exclude the adjacent channels with the normal direction as the central axis and select the edge channel data blocks away from the central axis as valid learning samples;

[0009] Step 3: If a target signal, impact noise, or human interference is detected entering from the edge channel in step 2, the contaminated channel sample is automatically discarded and the uncontaminated far-end edge channel data sample is retained;

[0010] Step 4: Calculate the root mean square value using the filtered clean clutter samples and construct the clutter covariance matrix based on it; assume that the velocity direction vector of the radar platform is in the YOZ plane, and its angle with the ground is ;

[0011] Constructing the clutter covariance matrix includes the following steps:

[0012] S1: Let any scattering target point be P, then the echo of the scattering target point P is expressed as ,in, is the amplitude of the scattering target point P, is the integrated steering vector at the scattering target point P, 、 、 are the azimuth spatial frequency, elevation spatial frequency and Doppler frequency of the scattering target point P respectively;

[0013] , , ,in, 、 are the array element spacings in azimuth and elevation dimensions, is the azimuth of the scattering target point P, is the pitch angle of the scattering target point P, is the wavelength, v is the level flight speed of the radar;

[0014] S2: Ground clutter can be equivalent to the synthesis of multiple scattered target point echoes, including the following steps:

[0015] K1: Assume a radar planar array with N columns and M rows. Each radar coherent processing interval CPI accumulates K pulses. The ground is divided into multiple grids according to the range resolution and angle resolution. The range resolution is expressed as ,in, c The speed of light is 3×10 8 m / s, B is the radar's operating bandwidth, and the angular resolution is the 3dB beamwidth of the array element;

[0016] K2: Divide the ground of interest to the radar into distance rings, and each distance ring is divided into Nc parts, then the The i-th clutter plate on the range ring is regarded as a scattering target point, and its echo form after matched filtering is expressed as ,in, Represents tensor product Tensor Product, 、 、 、 Respectively The clutter amplitude, Doppler frequency, azimuth spatial frequency and elevation spatial frequency of the ith clutter plate in the range ring, , , is the corresponding coefficient;

[0017] K3: No. The echo of a range ring is expressed as ,in, Indicates the distance from the 1st to the 2nd on a certain distance ring. N c clutter slices perform summation operation, After expansion in distance, the covariance matrix R of the clutter is obtained ；

[0018] K4: The integrated echo of ground clutter is expressed as ,in, is the index value of the range ring where the scattering target point is located, S is the signal, is the clutter matrix, The mean is 0 and the variance is Gaussian white noise;

[0019] S3: In a radar coherent processing interval CPI, the J pulses received by the Q array elements constitute dimensional data cube, the size of K is determined by the range resolution of the radar and pulse repetition frequency determined;

[0020] Step 5: Use the clutter covariance matrix to calculate the signal-clutter-noise ratio (SCNR) to achieve radar clutter suppression.

[0021] Furthermore, the signal-to-clutter-noise ratio (SCNR) is calculated in step 5, including the following steps:

[0022] T1: Use space-time adaptive processing (STAP) to construct the minimum variance distortion-free response (MVDR) optimization problem and find the space-time filter weights that maximize the output signal-clutter-to-noise ratio (SCNR).

[0023] T2: Define the space-time filter weight vector as w, then the output SCNR value is expressed as ,in, is the expected target amplitude, is the expression of the expected function, is the amplitude of the scattering target point P, is the conjugate transpose operation, is the desired steering vector of the target, corresponding to the azimuth spatial frequency , pitch spatial frequency and Doppler frequency , R is the clutter covariance matrix;

[0024] T3: To make the scattering target point output without loss, you need to use , the optimization problem in step T1 is corrected to , the closed-form solution is ,in, is a non-zero scalar coefficient.

[0025] Beneficial effect: When selecting clutter learning samples, the present invention selects channels located at the edges and corners of the data cube, excludes adjacent channels with the normal direction as the central axis, and selects edge channel data blocks far from the central axis as valid learning samples. If a target signal, impact noise or human interference is detected entering from a specific edge channel, the contaminated channel sample is automatically discarded, and the uncontaminated far-end edge channel data sample is retained. The root mean square value is calculated using the filtered pure clutter samples, and a clutter covariance matrix is ​​constructed based on this, and the signal-to-clutter-noise ratio (SCNR) is calculated using the clutter covariance matrix.

[0026] This invention abandons the traditional phased array radar method of selecting random row or column data channels to obtain clutter samples, and instead adopts an edge and corner sample interval selection method. Specifically, when a sample interval is contaminated by impulse noise or jamming, data from other sample intervals is used as learning samples. By calculating the clutter covariance matrix and SCNR, a clutter covariance matrix close to that in a realistic adversarial environment can be obtained, eliminating the influence of impulse noise and jamming to a certain extent, thereby achieving a more ideal optimal weighting function and STAP performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 1 is a schematic diagram of a clutter ring in an airborne radar clutter training and selection method according to an embodiment of the present invention;

[0028] Figure 2 This is the embodiment of the present invention. Distance rings by angle Equal division map;

[0029] Figure 3 1 is a schematic diagram of a radar echo data cube according to an embodiment of the present invention;

[0030] Figure 4 The impulse noise and human interference involved in the embodiment of the present invention may come from any clutter block or channel schematic diagram;

[0031] Figure 5 is a schematic diagram of selecting a sample interval involved in an embodiment of the present invention;

[0032] Figure 6 The training sample involved in the embodiment of the present invention is a schematic diagram of a row containing a target signal side lobe;

[0033] Figure 7 2 is a schematic diagram of a sample interval in which a training sample involved in an embodiment of the present invention is located at a corner. DETAILED DESCRIPTION

[0034] like Figures 1 to 7 As shown, the present invention provides an airborne radar clutter training selection method.

[0035] The sampling covariance matrix is ​​calculated based on sampled data of random variables and is often used to describe the relationships between variables in a multidimensional dataset. In practical applications, the sampling covariance matrix can help analyze the correlations between different dimensions and is often used in signal processing and statistical analysis.

[0036] Space-Time Adaptive Processing (STAP) is an advanced radar signal processing technology that leverages spatial (antenna array) and temporal (pulse repetition interval) information to detect and suppress clutter, thereby improving the detection of small, low-speed, and low-flying targets. The basic principle of the STAP algorithm is to achieve effective clutter suppression through two-dimensional adaptive filtering in the spatial and temporal domains. In airborne radar systems, due to the spatial and temporal coupling of clutter signals, STAP technology can fully utilize the multiple spatial channel information provided by multi-channel radars and the temporal information provided by coherent pulse trains. STAP's adaptability is reflected in the accurate perception and response to the external clutter environment, which relies on the real-time acquisition of the clutter covariance matrix (CCM) of the range cell under test (RUT).

[0037] In probability and statistics theory, independent and identically distributed (IID) refers to a random process in which the values ​​at any moment are random variables. If these random variables obey the same distribution and are independent of each other, then these variables are randomly and identically distributed.

[0038] The limited observation time, or the limited number of echo pulses, causes the echo spectrum to broaden. Each spectral line forms a spectral lobe, whose width is approximately the inverse of the observation time. The widening of the clutter power spectrum lobe caused by observation time is inversely proportional to the observation time. The motion of the clutter source itself also affects the shape of the clutter spectrum lobe.

[0039] Radar CPI is a performance metric used to measure radar performance. It indicates the time it takes for the radar to complete coherent processing after receiving an echo signal. This metric reflects the radar system's detection capability and anti-interference capabilities.

[0040] The factors that cause clutter non-uniformity include: (1) internal clutter motion, waves, and weather changes, etc., which cause the clutter power spectrum of the sample to change with time, direction, and distance; (2) the training samples contain discrete clutter, singular samples, and even deceptive interference, which leads to serious clutter non-uniformity; (3) changes in terrain, distance, ground undulations, and ground obstruction can cause changes in clutter distribution and amplitude. These non-uniformities make the units to be detected and the training samples not meet the independent and identically distributed conditions, resulting in the estimated covariance matrix being inconsistent with the actual one.

[0041] STAP requires a sufficient number of independent and identically distributed training samples to estimate the clutter statistics of the unit to be detected. If the training samples contain no interference, but the unit to be detected does, STAP cannot suppress the interference in the unit to be detected. Even if some training samples contain forwarding interference, since only some samples contain interference, the estimated interference power will be lower than the true value when estimating the interference covariance matrix. As a result, the system cannot form a notch of sufficient depth and width in the interference direction to completely suppress the interference in the unit to be detected. When the system's degrees of freedom are not high enough, deceptive interference can also degrade STAP's clutter suppression performance. The singular value detection algorithm used by STAP also eliminates a large number of samples with dense interference, resulting in inaccurate estimates.

[0042] A phased array radar plane typically consists of multiple subarrays and control regions. Each subarray and control region contains dozens or even hundreds of transmit / receive channels (TR channels). When a TR channel reaches a load state, it can switch adjacent channels to a receive state, and vice versa. When selecting sample data to solve the clutter covariance matrix, careful selection of the sample data source is often required. In traditional phased array radar beam control systems, to reduce computational complexity and dimensionality, channels corresponding to a specific column or row in the planar array are placed in a receive state. The clutter covariance matrix is ​​then calculated for the presence of signal, interference, and noise at that moment. This approach has two significant drawbacks: First, the target and interference angles are unknown, potentially leading to significant sample contamination in the selected row or column. Second, because the array antenna utilizes a superheterodyne receiver for its receive channel, samples near the center of the normal are more likely to be affected by internal local oscillator (LO) and RF leakage from the excitation source.

[0043] A method for selecting clutter for airborne radar training includes the following steps:

[0044] Step 1: In the radar data cube, select channel data located at the edge or corner of the space as candidate clutter learning samples;

[0045] Step 2: Exclude the adjacent channels with the normal direction as the central axis and select the edge channel data blocks away from the central axis as valid learning samples;

[0046] Step 3: If a target signal, impact noise, or human interference is detected entering from the edge channel in step 2, the contaminated channel sample is automatically discarded and the uncontaminated far-end edge channel data sample is retained;

[0047] Step 4: Calculate the root mean square value using the filtered clean clutter samples and construct the clutter covariance matrix based on it;

[0048] Step 5: Use the clutter covariance matrix to calculate the signal-clutter-noise ratio (SCNR) to achieve radar clutter suppression.

[0049] In step 4, the velocity direction vector of the radar platform is set to be in the YOZ plane, and its angle with the ground is ;

[0050] Constructing the clutter covariance matrix in step 4 includes the following steps:

[0051] (1) Let any scattering target point be P, then the echo of the scattering target point P is expressed as ,in, is the amplitude of the scattering target point P, is the integrated steering vector at the scattering target point P, 、 、 are the azimuth spatial frequency, elevation spatial frequency and Doppler frequency of the scattering target point P respectively;

[0052] , , ,in, 、 are the array element spacings in azimuth and elevation dimensions, is the azimuth of the scattering target point P, is the pitch angle of the scattering target point P, is the wavelength, v is the level flight speed of the radar;

[0053] (2) Ground clutter can be equivalent to the synthesis of multiple scattered target point echoes, including the following:

[0054] Assume a radar plane array with N columns and M rows. Each radar coherent processing interval CPI accumulates K pulses. The ground is divided into multiple grids according to the range resolution and angle resolution. The range resolution is expressed as ,in, c The speed of light is 3×10 8 m / s, B is the radar's operating bandwidth, and the angular resolution is the 3dB beamwidth of the array element;

[0055] Divide the ground of interest to the radar into distance rings, and each distance ring is divided into Nc parts, then the The i-th clutter plate in the range ring is regarded as a scattering target point, and its echo form after matched filtering is expressed as ,in, Represents tensor product Tensor Product, 、 、 、 Respectively l The clutter amplitude, Doppler frequency, azimuth spatial frequency and elevation spatial frequency of the ith clutter plate in the range ring, , , is the corresponding coefficient;

[0056] No. The echo of a range ring is expressed as ,in, Indicates the distance from the 1st to the 2nd on a certain distance ring. N c clutter slices perform summation operation, After expansion in distance, the covariance matrix R of the clutter is obtained ；

[0057] The integrated echo of ground clutter is expressed as ,in, is the index value of the range ring where the scattering target point is located, S is the signal, is the clutter matrix, The mean is 0 and the variance is Gaussian white noise;

[0058] like Figure 1 As shown, is the distance from the radar to a clutter area on the ground, is the radar antenna depression angle in the direction of the clutter area, is the ground-grazing angle. Assume that the ground of interest is divided into Each distance ring is divided into Figure 2 As shown in Nc parts, then the The i-th clutter plate in a range ring can be regarded as a ground scattering point. is the step angle of each clutter interval on a certain range ring, For the The distance interval of a distance ring.

[0059] (3) Figure 3 : is a schematic diagram of a radar echo data cube involved in an embodiment of the present invention. In a radar coherent processing interval CPI, J pulses received by Q array elements constitute dimensional data cube, the size of K is determined by the range resolution of the radar and pulse repetition frequency Determined.

[0060] In step 5, the signal-to-clutter-noise ratio (SCNR) is calculated, which includes the following steps:

[0061] (4) Use space-time adaptive processing (STAP) to construct the minimum variance distortion-free response (MVDR) optimization problem and find the space-time filter weights that maximize the output signal-clutter-to-noise ratio (SCNR);

[0062] (5) Define the space-time filter weight vector as w, then the output SCNR value is expressed as ,in, is the expected target amplitude, is the expression of the expected function, is the amplitude of the scattering target point P, is the conjugate transpose operation, is the desired steering vector of the target, corresponding to the azimuth spatial frequency , pitch spatial frequency and Doppler frequency , R is the clutter covariance matrix;

[0063] (6) If the scattering target point is to have no loss of output, it is necessary to use , the optimization problem in step T1 is corrected to , the closed-form solution is ,in , is a non-zero scalar coefficient. In this formula, it is required that The cross-correlation matrix between the signal and the clutter is known but does not need to be calculated.

[0064] Figure 4 The embodiment of the present invention involves that the impact noise and artificial interference may come from any clutter block or channel schematic diagram, the impact noise and artificial interference may come from any clutter block or channel, if the channel distribution of the array is as follows Figure 4 As shown in the rectangle, the noise and interference that exceeds the CFAR threshold and thus affects the mine detection may be distributed on any random channel, as shown in the dark squares in the figure.

[0065] Figure 5This is a schematic diagram of sample interval selection in accordance with an embodiment of the present invention. Sample intervals can be selected at edges, corners, or locations away from the center. By analyzing the random distribution characteristics of clutter and noise, channels located at the corners of the array (not limited to rectangular shapes; elliptical shapes are also acceptable) are less likely to be affected by clutter interference and are less affected by excitation signal leakage. The number of channels included in sample intervals 1-4 can vary and is not necessarily the 8 shown. For a total of 1000 channels, the number of channels in a sample interval can be selected between 5 and 20. The criteria for channel selection within a sample interval are that the channel amplitude deviation is within ±15% of the mean of all channel noise, the amplitude cannot be too large or too small, and each selected sample amplitude must be within a certain range of the mean. If the amplitude is outside this range, the abnormally large or small amplitude is discarded during calculation, or the channel sample is reselected. Therefore, the selection of sample intervals requires multiple calculations and iterations.

[0066] Figure 6 and Figure 7 The UV lobe patterns are generated using different training samples. U and V represent the antenna's azimuth and elevation directions, respectively. Warm red represents areas with high energy, while cool blue represents areas with low energy. Figure 6 1 is a schematic diagram of a row where a training sample according to an embodiment of the present invention includes a side lobe of a target signal, indicating that the training sample includes a row where an excitation signal leaks; Figure 7 This is a schematic diagram of the sample interval where the training samples involved in the embodiment of the present invention are located at the corners, indicating that the training samples are located at the corners. For the same darkroom test environment, the same excitation signal leakage, the same interference and environmental reflection scene, Figure 7 The lobe pattern signal in is more prominent.

[0067] The present invention proposes a clutter selection training method for airborne radars. This method abandons the conventional method of selecting random row or column data channels to obtain clutter samples, as used in conventional phased array radars. Instead, it adopts a method for selecting edge and corner sample intervals. Specifically, when a sample interval is contaminated by impulsive noise or interference, data from other sample intervals is used as learning samples. By using the clutter covariance matrix and the SCNR calculation formula, a clutter covariance matrix close to that in a realistic adversarial environment is obtained, which, to a certain extent, eliminates the influence of impulsive noise and interference, thereby achieving a relatively ideal optimal weighting function and STAP performance.

[0068] Finally, it should be noted that the above are only preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or make equivalent replacements for some of the technical features therein. However, any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for selecting clutter for airborne radar training, characterized in that: The following steps are involved: Step 1: In the radar data cube, select channel data located at the edge or corner of the space as candidate clutter learning samples; Step 2: Exclude the adjacent channels with the normal direction as the central axis and select the edge channel data blocks away from the central axis as valid learning samples; Step 3: If a target signal, impact noise, or human interference is detected entering from the edge channel in step 2, the contaminated channel sample is automatically discarded and the uncontaminated far-end edge channel data sample is retained; Step 4: Calculate the root mean square value using the filtered clean clutter samples and construct the clutter covariance matrix based on it; assume that the velocity direction vector of the radar platform is in the YOZ plane, and its angle with the ground is ; Constructing the clutter covariance matrix includes the following steps: S1: Let any scattering target point be P, then the echo of the scattering target point P is expressed as ,in, is the amplitude of the scattering target point P, is the integrated steering vector at the scattering target point P, 、 、 are the azimuth spatial frequency, elevation spatial frequency and Doppler frequency of the scattering target point P respectively; , , ,in, 、 are the array element spacings in azimuth and elevation dimensions, is the azimuth of the scattering target point P, is the pitch angle of the scattering target point P, is the wavelength, v is the level flight speed of the radar; S2: Ground clutter can be equivalent to the synthesis of multiple scattered target point echoes, including the following steps: K1: Assume a radar planar array with N columns and M rows. Each radar coherent processing interval CPI accumulates K pulses. The ground is divided into multiple grids according to the range resolution and angle resolution. The range resolution is expressed as ,in, c The speed of light is 3×10 8 m / s, B is the radar's operating bandwidth, and the angular resolution is the 3dB beamwidth of the array element; K2: Divide the ground of interest to the radar into distance rings, and each distance ring is divided into Nc parts, then the The i-th clutter plate on the range ring is regarded as a scattering target point, and its echo form after matched filtering is expressed as ,in, Represents tensor product Tensor Product, 、 、 、 Respectively The clutter amplitude, Doppler frequency, azimuth spatial frequency and elevation spatial frequency of the ith clutter plate in the range ring, , , is the corresponding coefficient; K3: No. The echo of a range ring is expressed as ,in, Indicates the distance from the 1st to the 2nd on a certain distance ring. N c clutter slices perform summation operation, After expansion in distance, the covariance matrix R of the clutter is obtained ； K4: The integrated echo of ground clutter is expressed as ,in, is the index value of the range ring where the scattering target point is located, S is the signal, is the clutter matrix, The mean is 0 and the variance is Gaussian white noise; S3: In a radar coherent processing interval CPI, the J pulses received by the Q array elements constitute dimensional data cube, the size of K is determined by the range resolution of the radar and pulse repetition frequency determined; Step 5: Use the clutter covariance matrix to calculate the signal-clutter-noise ratio (SCNR) to achieve radar clutter suppression.

2. The airborne radar clutter training and selection method according to claim 1, characterized in that: In step 5, the signal-to-clutter-noise ratio (SCNR) is calculated, which includes the following steps: T1: Use space-time adaptive processing (STAP) to construct the minimum variance distortion-free response (MVDR) optimization problem and find the space-time filter weights that maximize the output signal-clutter-to-noise ratio (SCNR). T2: Define the space-time filter weight vector as w, then the output SCNR value is expressed as ,in, is the expected target amplitude, is the expression of the expected function, is the amplitude of the scattering target point P, is the conjugate transpose operation, is the desired steering vector of the target, corresponding to the azimuth spatial frequency , pitch spatial frequency and Doppler frequency , R is the clutter covariance matrix; T3: To make the scattering target point output without loss, you need to use , the optimization problem in step T1 is corrected to , the closed-form solution is ,in, is a non-zero scalar coefficient.

Images