A Range Estimation Method for Skywave Over-the-Horizon Radar Based on Sparse Bayesian Algorithm

Abstract

This invention proposes a range estimation method for skywave over-the-horizon radar based on a sparse Bayesian algorithm. The implementation steps are as follows: beamforming is performed on the data received by the skywave over-the-horizon radar antenna array; pulse compression and moving target detection are performed on the echo data from the target's azimuth to obtain the range-Doppler matrix; two-dimensional constant false alarm rate (CFAR) detection is performed on the range-Doppler matrix; and the target's range information is obtained by sparse recovery of the echo data from the Doppler channel using the sparse Bayesian algorithm. This invention can improve the range resolution and range estimation accuracy of skywave over-the-horizon radar; it reduces the computational load and system complexity of the algorithm, making it more suitable for engineering applications.

Description

Technical Field

[0001] This invention belongs to the field of radar technology, and more specifically relates to a range estimation method for skywave over-the-horizon radar (OTHR) applied to the field of radar signal processing technology. This invention can be used to estimate parameters from echo data received by skywave radar, achieving range super-resolution beyond the Rayleigh limit even with limited radar bandwidth, thus improving target range estimation accuracy. Background Technology

[0002] Skywave over-the-horizon (OTHR) radar operates in the high-frequency (HF) band between 3-30 MHz. Utilizing the ionosphere's reflection of high-frequency signals, OTHR detects targets from above, thus overcoming the limitations of the Earth's curvature and detecting targets at extremely long distances (1000-4000 km) beyond the line-of-sight range of conventional radar. However, the high-frequency band has limited spectrum resources and is subject to various interferences, resulting in a limited operating bandwidth and low range resolution, typically a few kilometers to tens of kilometers. Due to the limited transmission signal bandwidth, the range resolution of OTHR is limited, leading to a significant decrease in ranging accuracy in multi-target situations because it cannot correctly distinguish between targets.

[0003] The University of Electronic Science and Technology of China (UESTC) proposed a radar range super-resolution calculation method based on the OMP and DPL1 algorithms in its patent application, "A Radar Range Super-Resolution Calculation Method Based on OMP and DPL1 Algorithms" (Patent Application No.: 202210626515.7, Publication No.: CN 115564645A). The implementation steps are as follows: First, pulse compression is performed on the radar echo signal to determine the radar signal segment of the target group; then, frequency de-chewing is performed on the radar signal segment of the target group to obtain a single-frequency signal; next, the OMP algorithm is used to obtain the approximate position of the target; finally, the L1 regularization algorithm is used to further process the obtained frequency position for super-resolution to obtain the precise frequency position of the target group, and the frequency information is converted into range information to achieve range super-resolution. The shortcomings of this method are that digital de-chewing of the linear frequency modulated continuous wave signal requires subsequent processing of all echo signals received by the radar, resulting in a large computational load and poor real-time performance. Furthermore, this method requires two steps, leading to high complexity. At the same time, the OMP algorithm is prone to overmatching. If the first step is estimated incorrectly, the error from the previous step will accumulate in every subsequent step, resulting in poor stability of the algorithm.

[0004] The University of Electronic Science and Technology of China (UESTC) proposed a method for estimating OTHR maneuvering target parameters based on sparse decomposition of the instantaneous autocorrelation matrix in its patent application, "A Method for Estimating OTHR Maneuvering Target Parameters Based on Sparse Decomposition of Instantaneous Autocorrelation Matrix" (Patent Application No.: CN201711068233.5, Publication No.: CN107861115A). The implementation steps are as follows: first, an instantaneous autocorrelation transform is performed on the echo; then, cross-term suppression is applied to the instantaneous autocorrelation matrix; next, sparse decomposition is performed on the instantaneous autocorrelation matrix; finally, a Hough transform is performed on the decomposed sparse matrix to obtain the instantaneous frequency of the signal, thereby estimating the target motion parameters. The drawback of this method is that although it improves the estimation accuracy of Doppler frequency and angle in OTHR under multi-target conditions, it does not solve the problem of insufficient accuracy in estimating target range parameters due to the low range resolution of OTHR. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of the existing technologies mentioned above by proposing a high-precision estimation method for skywave radar parameters. This method aims to solve the problem of insufficient parameter estimation accuracy in OTHR due to poor range resolution, as well as the problems of large computational load and poor real-time performance of existing frequency deskewing range super-resolution methods.

[0006] The technical solution to achieve the objective of this invention is that it constructs a basis matrix for the sparse recovery problem by compressing the data delay of the transmitted linear frequency modulated continuous wave signal pulse. Each column of the basis matrix represents the pulse compression waveform data of the linear frequency modulated continuous wave signal under different delays. By mapping the delay to the range, the pulse compression waveform data at different ranges can be obtained. The distance interval between each column of the basis matrix is ​​less than one range resolution unit. By matching the observation vector with each column of the basis matrix, a sparse recovery vector is obtained. Then, the position of each element in the sparse vector is converted into the range position of the target, which can overcome the problem of poor range resolution of skywave over-the-horizon radar in the prior art. This invention utilizes the signal received by a skywave over-the-horizon radar antenna array for beamforming, accumulating energy in space. Then, pulse compression is applied to the signal at the target's azimuth. Next, a Directed Fourier Transform (DFT) is performed on the echo data within a coherent accumulation time of the same range cell to obtain the target's range-Doppler matrix. Two-dimensional constant false alarm rate (CFAR) detection is then performed on the range-Doppler matrix. Finally, sparse recovery is performed on the pulse-compressed data of the Doppler channel containing the detected target. Based on the pulse compression result, only the portion of the pulse compression signal containing target information is extracted as the observation vector, effectively reducing the computational load of the entire algorithm and overcoming the problems of high computational load and poor real-time performance in existing frequency "de-skewing" range super-resolution methods. This invention employs a sparse Bayesian algorithm for sparse recovery of the observation vector. This algorithm can process single-shot data and does not require setting regularization parameters. Even at low signal-to-noise ratios, it maintains good resolution and stability, overcoming the limitations of existing subspace-based algorithms that can only handle incoherent sources and require multiple-shot data, as well as the over-matching tendency of greedy sparse recovery algorithms. This significantly improves the range resolution capability of skywave over-the-horizon radar.

[0007] The implementation steps of this invention are as follows:

[0008] Step 1: Beamforming is performed on the data received by the skywave over-the-horizon radar antenna array to obtain the echo data of the target's azimuth.

[0009] Step 2: Perform pulse compression on the echo data received by the skywave over-the-horizon radar at the target's location to obtain pulse-compressed echo data.

[0010] Step 3: Perform moving target detection on the pulse-compressed echo signal from the skywave over-the-horizon radar.

[0011] The range-Doppler matrix of the target is obtained by performing Discrete Fourier Transform (DFT) on the pulse compression echo data within a coherent accumulation time of the same range cell.

[0012] Step 4: Perform two-dimensional constant false alarm rate (CFAR) detection on the target's range-Doppler matrix to obtain the range cell and Doppler cell where the target is located;

[0013] Step 5: Use the sparse Bayesian algorithm to sparsely recover the signal of the Doppler cell containing the target:

[0014] By compressing the pulse delay of the linear frequency modulated continuous wave signal pulse transmitted by the skywave over-the-horizon radar, a measurement matrix is ​​constructed according to the dimension of the observation vector. The sparse Bayes algorithm is used to perform sparse recovery on the observation vector to obtain a sparse signal vector. The position of each element in the sparse vector is converted into the target distance value.

[0015] Compared with the prior art, the present invention has the following advantages:

[0016] First, because the present invention uses a sparse Bayesian algorithm to estimate the range parameters of skywave over-the-horizon radar, it overcomes the problem of poor range resolution of skywave over-the-horizon radar in the prior art, enabling the present invention to achieve a range resolution of more than three times higher than conventional range parameter estimation methods for skywave over-the-horizon radar.

[0017] Secondly, since this invention selects the pulse-compressed data of the Doppler channel where the target is located for range super-resolution processing, and extracts a portion of the pulse-compressed signal as the observation vector based on the pulse compression result, it overcomes the problems of large computational load and poor real-time performance of the existing frequency "de-skewing" range super-resolution method. This invention greatly reduces the computational load and system complexity of the algorithm, making it more suitable for engineering applications.

[0018] Third, since this invention uses a sparse Bayesian algorithm to sparsely recover the observation vector, this algorithm can process single snapshot data and does not require setting regularization parameters. Even when the signal-to-noise ratio is low, it still has good resolution and stability. It overcomes the problems of existing subspace-type algorithms that can only handle incoherent sources and require multiple snapshot data, as well as the overmatching shortcomings of greedy sparse recovery algorithms. This makes the range parameter estimation of skywave over-the-horizon radar more accurate and more stable, thereby improving the range resolution capability of skywave over-the-horizon radar. Attached Figure Description

[0019] Figure 1 This is a flowchart of the present invention;

[0020] Figure 2 This is a diagram showing the results of processing skywave over-the-horizon radar echo data using the method of this invention under simulation condition 1.

[0021] Figure 3 This is a diagram showing the results of processing skywave over-the-horizon radar echo data using the method of this invention under simulation condition 2.

[0022] Figure 4 This is a diagram showing the results of processing skywave over-the-horizon radar echo data using the method of this invention under simulation condition 3. Detailed Implementation

[0023] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments.

[0024] Reference Figure 1 The specific implementation steps of the embodiments of the present invention will be described in further detail.

[0025] Step 1: Beamforming is performed on the data received by the skywave over-the-horizon radar antenna array to accumulate energy in space and obtain echo data of the target's location.

[0026] Step 2: Perform pulse compression on the echo data received by the skywave over-the-horizon radar at the target's location to obtain pulse-compressed echo data.

[0027] The steps for pulse compression of the echo data received by the skywave over-the-horizon radar are as follows:

[0028] The first step is to construct the impulse response of the matched filter based on the transmitted signal of the skywave over-the-horizon radar. Where s represents the discrete sequence obtained after discrete sampling of the linear frequency modulated continuous wave signal s(t) transmitted by the skywave over-the-horizon radar within one pulse repetition period, s=[s1,s2,...,s G ], G represents the number of sampling points of the discrete sequence within one pulse repetition period, s g Let represent the g-th element of s, where g = 1, 2, ..., G, (·). * Indicates the conjugate operation;

[0029] The second step is to convolve the skywave over-the-horizon radar echo data with the impulse response of the matched filter to obtain pulse-compressed echo data.

[0030] Step 3: Perform moving target detection on the pulse-compressed echo signal from the skywave over-the-horizon radar.

[0031] The range-Doppler matrix of the target is obtained by performing Discrete Fourier Transform (DFT) on the pulse compression echo data within a coherent accumulation time of the same range cell.

[0032] Step 4: Perform two-dimensional constant false alarm rate (CFAR) detection on the target's range-Doppler matrix to obtain the range cell and Doppler cell where the target is located.

[0033] The steps of the two-dimensional constant false alarm rate (CFAR) detection are as follows:

[0034] The first step is to select unselected elements in the distance-Doppler matrix as the units to be detected;

[0035] The second step involves selecting N reference units on both sides of the unit to be detected, where N is an integer power of 2. The average value of these 2N reference units is then calculated and used as the power estimate Z of the background noise. ca ;

[0036] The third step is to estimate the power value Z of the background noise. ca Multiply by the threshold product factor k to obtain the threshold value s of the cell to be detected;

[0037] The fourth step is to compare the signal value of the unit to be detected with the threshold value s. If the signal value of the unit to be detected is greater than the threshold value, the signal is determined to be the target signal and assigned a value of 1; otherwise, the signal is determined to be a noise signal and assigned a value of 0.

[0038] The fifth step is to determine whether all elements in the range-Doppler matrix have been selected. If so, the coordinates of the elements identified as target signals are mapped into the range-Doppler matrix to obtain the range cell and Doppler cell where the target is located; otherwise, proceed to the first step.

[0039] Step 5: Use the sparse Bayesian algorithm to sparsely recover the signal of the Doppler cell containing the target:

[0040] By compressing the pulse delay of the linear frequency modulated continuous wave signal pulse transmitted by the skywave over-the-horizon radar, a measurement matrix is ​​constructed according to the dimension of the observation vector. The sparse Bayes algorithm is used to perform sparse recovery on the observation vector to obtain a sparse signal vector. The position of each element in the sparse vector is converted into the target distance value.

[0041] The observation vector refers to the data formed by selecting pulse-compressed echo data from the Doppler unit where the target is located, and extracting the width of the main lobe of the echo data.

[0042] The steps for constructing the measurement matrix are as follows:

[0043] The first step is to construct a measurement matrix with dimensions P×M and initial values ​​of all elements being 0; the value of P is equal to the length of the observation vector, M = KP, and the value of K is the ratio of the sampling frequency of the linear frequency modulated signal to the sampling frequency of the observation vector. In this embodiment of the invention, K = 5.

[0044] The second step involves pulse compression of the linear frequency modulated continuous wave signal and the impulse response of the matched filter to obtain the basis vector. The peak position q of the basis vector is found, and the qb to q+b elements of the main lobe of the basis vector are extracted, where b is half the number of sampling points of the main lobe of the basis vector. All elements of the main lobe of the basis vector are placed in the first row to the 2b row of the first column of the measurement matrix, and so on, placing all elements of the main lobe of the basis vector in the k row to the k+2b-1 row of the k column of the measurement matrix, k = 1, 2, ..., M.

[0045] The third step is to extract rows b to P+b-1 of the measurement matrix. The first column of the measurement matrix represents the pulse compression waveform data of the linear frequency modulated continuous wave signal at time 0, the kth column represents the pulse compression waveform data of the linear frequency modulated continuous wave signal at time k×Δt, and Δt represents the time interval between each column of the measurement matrix.

[0046] The fourth step is to extract all column vectors of the measurement matrix by a factor of K, so that the dimension of the pulse compression waveform data of the linear frequency modulated continuous wave signal is consistent with the observation vector.

[0047] The specific steps for using the sparse Bayesian algorithm to sparsely recover the observation vector and obtain the sparse signal vector are as follows:

[0048] The first step is to initialize the hyperparameters α and α0, where α = (α1, α2, ..., α0). M α represents the hyperparameters of the sparse signal vector, and each element of α is initialized to 1; α0 represents the noise variance, which is initialized to 1.

[0049] The second step is to calculate the variance matrix R and mean vector μ of the posterior probability density function of the sparse signal vector to be determined according to the following formula.

[0050] R=(α0Φ T Φ+Λ) -1

[0051] μ=α0RΦ T y

[0052] Where Φ represents the measurement matrix, (·) T Let represent the transpose operation, Λ represent the diagonal matrix of the hyperparameter α, and y represent the observation vector. -1 This indicates the inverse operation;

[0053] Third, update the hyperparameters according to the following formula. and α new ;

[0054]

[0055]

[0056] in, Represents α new The i-th column vector, γ i γ represents the i-th element of the quantization factor. i =1-α i R ii ∑ represents the summation operation, i = 1, 2, ..., M;

[0057] The fourth step is to determine whether the root mean square error of the current iteration satisfies the convergence condition or reaches the maximum number of iterations. If yes, proceed to the fifth step; otherwise, proceed to the second step. The convergence condition is that the root mean square error between the mean vector obtained in the current iteration and the mean vector obtained in the previous iteration is less than e. -4 The maximum number of iterations is 1000.

[0058] The fifth step is to use the mean vector obtained in the current iteration as the sparse signal vector.

[0059] The process of converting the positions of each element in a sparse vector into distance values ​​from the target refers to using the formula... The positions of each element in the sparse vector are converted into distance values ​​to the target; where c represents the speed of light, L represents the starting coordinates of the pulse-compressed echo data extracted from the Doppler cell containing the target, and f s This represents the sampling frequency of the observation vector.

[0060] The effects of this invention will be further illustrated below with simulation experiments:

[0061] 1. Simulation experiment conditions:

[0062] The simulation experiment platform of this invention is: Windows 10 operating system and Matlab R2020b.

[0063] Simulation Experiment 1: Time Width T of Linear Frequency Modulated Continuous Wave Signal Transmitted by Radar p =15×10 -3 s, bandwidth B is 1×10 4 The standard pulse compression range resolution is ΔR = c / 2B = 15000m. The number of receiving antennas is 32, corresponding to an angular resolution of 3.169. The echo signal contains target information and Gaussian white noise; the sampling frequency of the echo signal is f. s =2×10 4 The signal-to-noise ratios (SNR) of the two targets are SNR1 = 10 dB and SNR2 = 10 dB, respectively; the targets are located at azimuths of 1° and 1.5°, respectively; the distances are R1 = 313500 m and R2 = 318000 m, respectively; the target velocities are v1 = 39 m / s and v2 = 39 m / s, respectively, corresponding to Doppler frequencies f1 and f2. d 1 = 2.6 Hz, f d 2 = 2.6 Hz.

[0064] Simulation Experiment 2: The target velocities are v1 = 39 m / s and v2 = 62 m / s, respectively, and the corresponding Doppler frequencies are f1 = 39 m / s and f2 = 62 m / s. d 1 = 2.6 Hz, f d 2 = 4.13 Hz, and the other parameters are the same as those in simulation experiment 1.

[0065] Simulation Experiment 3: The distances in Simulation Experiment 3 are R1 = 310500m and R2 = 318000m; the target velocities are v1 = 39m / s and v2 = 62m / s, corresponding to Doppler frequencies f1 and f2 respectively. d 1 = 2.6 Hz, f d 2 = 4.13 Hz. The remaining parameters are the same as in simulation experiment 1.

[0066] 2. Simulation content and result analysis:

[0067] The simulation experiments of this invention are three.

[0068] Simulation Experiment 1 illustrates that when the Doppler velocities of two targets are the same and the distance between them is less than one range resolution unit, it is impossible to distinguish between the two targets. Conventional beamforming is then performed on the echo signal received by the antenna to obtain the target's azimuth. Figure 2 As shown in (a). Figure 2 In (a), the horizontal axis represents angle, and the vertical axis represents normalized amplitude. It can be seen that conventional beamforming cannot separate two targets in terms of azimuth. Pulse compression processing is performed on the echo data after beamforming; the pulse compression result is as follows... Figure 2 As shown in (b) Figure 2 (b) The horizontal axis represents the distance value, and the vertical axis represents the normalized amplitude, yielding the target's distance information. Since the distance interval between two targets is less than one range resolution cell, only one distance value of 315 km is estimated, making it impossible to distinguish the two targets based on distance alone. A DFT is performed on the echo data within a coherent accumulation time of the same range cell, and constant false alarm rate (CFAR) detection is applied to obtain the target's range-Doppler image as shown below. Figure 2 As shown in (c) Figure 2 (c) The x-axis represents distance, the y-axis represents Doppler frequency, and the z-axis represents amplitude. It can be seen that the Doppler frequency value is 2.60417Hz, which matches the actual value. However, since the targets are within the same range resolution cell, the distance information between the two targets cannot be distinguished, hence the two targets cannot be identified. The results of two-dimensional CFAR detection on the range-Doppler map are as follows: Figure 2 As shown in (d), Figure 2 (d) The x-axis represents distance, the y-axis represents Doppler frequency, and the z-axis represents the discriminant value. The distance parameter is estimated using a sparse Bayesian algorithm on the pulse-compressed echo data in the Doppler channel with a discriminant value of 1 in the two-dimensional CFAR detection results, as shown below. Figure 2 As shown in (e), the dashed line represents the true distance value, and the peak of the curve represents the target distance value estimated using the present invention. It can be seen that the distance information of the two targets can be distinguished, and the obtained distance values ​​are 313.5km and 318km, respectively, which are consistent with the actual values.

[0069] Simulation Experiment 2 illustrates a situation where, although the target is distinguishable in the Doppler domain, the estimable range value is affected by the radar echo signal sampling rate, meaning only the range value at the sampling point can be estimated, resulting in the range dimension remaining indistinguishable. Conventional beamforming is then performed on the echo signal received by the antenna to obtain the target's azimuth, as shown below. Figure 3 As shown in (a). Figure 3 In (a), the horizontal axis represents angle, and the vertical axis represents amplitude. It can be seen that conventional beamforming cannot separate the azimuth angles of two targets. Pulse compression processing is performed on the echo data after beamforming; the pulse compression result is as follows... Figure 3 As shown in (b) Figure 3 (b) The horizontal axis represents the distance value, and the vertical axis represents the normalized amplitude. The target's distance information is 315km. It can be seen that the two targets cannot be distinguished from each other based on distance alone. A DFT is performed on the echo data within a coherent accumulation time of the same range cell, and constant false alarm rate (CFAR) detection is applied to obtain the target's range-Doppler image as shown below. Figure 3 As shown in (c) Figure 3 (c) The x-axis represents distance, the y-axis represents Doppler frequency, and the z-axis represents amplitude. It can be seen that the Doppler frequency values ​​are 2.60417Hz and 4.16667Hz, which are consistent with the actual values. Although the Doppler frequency is distinguishable, only the distance value corresponding to the sampling point can be estimated. When the distance interval between two targets is small, the distance values ​​of the two targets are estimated as the value at the same sampling point, both estimated as 315km. Therefore, it is still impossible to distinguish the specific distance information of the two targets. The results of two-dimensional CFAR detection on the range-Doppler map are as follows: Figure 3 As shown in (d), Figure 3 (d) The x-axis represents distance, the y-axis represents Doppler frequency, and the z-axis represents the discriminant value. The distance parameter is estimated using a sparse Bayesian algorithm on the pulse-compressed echo data in the Doppler channel with a discriminant value of 1 in the two-dimensional CFAR detection results, as shown below. Figure 3 As shown in (e), the dashed line represents the true distance value, and the peak of the curve represents the distance value estimated using the present invention. It can be seen that the distance information of the two targets can be distinguished, and the obtained distance values ​​are 313.5km and 318km, respectively, which are consistent with the actual values. It can be seen that this method can improve the distance estimation accuracy of skywave over-the-horizon radar.

[0070] Simulation Experiment 3 illustrates a situation where, although the target is resolvable in the Doppler domain, the estimated distance is only possible at grid points due to the sampling rate limitation. While the range dimension is resolvable, the estimated distance differs from the actual value. Conventional beamforming is then performed on the echo signal received by the antenna to obtain the target's azimuth. Figure 4 As shown in (a). Figure 4In (a), the horizontal axis represents angle, and the vertical axis represents amplitude. It can be seen that conventional beamforming cannot separate two targets in terms of azimuth. Pulse compression processing is performed on the echo data after beamforming; the pulse compression result is as follows... Figure 4 As shown in (b) Figure 4 (b) The horizontal axis represents the distance value, and the vertical axis represents the normalized amplitude. The target's distance information is 315km. The two targets cannot be distinguished solely from the distance dimension. A DFT is performed on the echo data within a coherent accumulation time of the same range cell, and constant false alarm rate (CFAR) detection is applied to obtain the target's range-Doppler image as shown below. Figure 4 As shown in (c) Figure 4 (c) The x-axis represents distance, the y-axis represents Doppler frequency, and the z-axis represents amplitude. It can be seen that the Doppler frequency values ​​are 2.60417Hz and 4.16667Hz, which are consistent with the actual values. Although the Doppler frequency and distance dimensions are separable, the distance values ​​of the two targets are estimated as the values ​​at the grid points closest to the true values, which are 307.5km and 315km respectively, inconsistent with the actual distance values. The results of two-dimensional CFAR detection on the distance-Doppler map are as follows... Figure 4 As shown in (d), Figure 4 (d) The x-axis represents distance, the y-axis represents Doppler frequency, and the z-axis represents the discriminant value. Distance estimation is performed on the pulse-compressed echo data in the Doppler channel with a discriminant value of 1 in the two-dimensional CFAR detection results using a sparse Bayesian algorithm. The results are as follows: Figure 4 As shown in (e), the dashed line represents the true distance value, and the peak of the curve represents the distance value estimated using the present invention. It can be seen that the distance between the two targets can be distinguished, and the obtained distance values ​​are 310.5km and 318km, respectively, which are consistent with the actual values. It can be seen that this method can improve the distance estimation accuracy of skywave over-the-horizon radar.

[0071] The simulation experiments above show that, in order to address the problems of limited operating bandwidth, low range resolution, and insufficient range estimation accuracy of skywave over-the-horizon radar, this invention proposes a range estimation method for skywave over-the-horizon radar based on the sparse Bayesian algorithm. Compared with the conventional range parameter estimation method for skywave over-the-horizon radar, the range estimation accuracy can be improved by more than three times.

Claims

1. A method for sky-wave over-the-horizon radar range estimation based on a sparse Bayesian algorithm, characterized in that, The linear frequency modulation continuous wave signal transmitted by the sky wave over-the-horizon radar is used to construct a measurement matrix, and the sparse Bayesian algorithm is used to perform sparse recovery on the pulse compression data of the Doppler channel where the target is located. Step 1, beamforming is performed on the data received by the antenna array of the sky wave over-the-horizon radar to obtain echo data in the direction where the target is located. Step 2, pulse compression is performed on the echo data received by the sky wave over-the-horizon radar in the direction where the target is located to obtain pulse-compressed echo data. Step 3, moving target detection is performed on the pulse-compressed echo signal of the sky wave over-the-horizon radar: Discrete Fourier transform (DFT) is performed on the pulse-compressed echo data in a coherent accumulation time for a distance unit to obtain a range-Doppler matrix of the target. Step 4, two-dimensional constant false alarm detection is performed on the range-Doppler matrix of the target to obtain the distance unit and the Doppler unit where the target is located. Step 5, the sparse Bayesian algorithm is used to perform sparse recovery on the signal in the Doppler unit where the target is located: The time delay of the pulse-compressed signal of the linear frequency modulation continuous wave signal transmitted by the sky wave over-the-horizon radar is constructed according to the dimension of the observation vector to construct a measurement matrix. The sparse Bayesian algorithm is used to perform sparse recovery on the observation vector to obtain a sparse signal vector, and the positions of the elements in the sparse vector are converted into the distance values of the target.

2. The sky-wave over-the-horizon radar range estimation method based on sparse Bayesian algorithm according to claim 1, characterized in that: The pulse compression of the echo data received by the sky wave over-the-horizon radar in step 2 is as follows: The first step is to construct the impulse response of the matched filter according to the transmitting signal of the sky-wave over-the-horizon radar Wherein, s represents the discrete sequence obtained after the discrete sampling of the linear frequency modulation continuous wave signal s(t) transmitted by the sky-wave over-the-horizon radar in one pulse repetition period, s=[s1,s2,...,s G ], G represents the sampling point number of the discrete sequence in one pulse repetition period, s g g represents the gth element of s, g=1,2,G,(·) * represents the conjugate operation; Second, the echo data of the sky wave over-the-horizon radar is convolved with the impulse response of the matched filter to obtain pulse-compressed echo data. 3.The sky-wave over-the-horizon radar range estimation method based on sparse Bayesian algorithm according to claim 1, characterized in that: The steps of the two-dimensional constant false alarm detection in step 4 are as follows: First, select an element in the range-Doppler matrix that has not been selected as a detection unit. Secondly, N reference cells are selected on both sides of the cell to be detected, N is an integer power of 2, and the average of the 2N reference cells is calculated, which is taken as the power estimation value Z of the background noise ca ; Thirdly, the power estimation value Z of the background noise is multiplied by a threshold product factor k to obtain the threshold value s of the cell to be detected. ca Thirdly, the power estimation value Z of the background noise is multiplied by a threshold product factor k to obtain the threshold value s of the cell to be detected. Fourth, compare the signal value of the detection unit with the threshold value s. If the signal value of the detection unit is greater than the threshold value, it is determined that the signal is a target signal and is assigned a value of 1. Otherwise, it is determined that the signal is a noise signal and is assigned a value of 0. Fifth, determine whether all elements in the range-Doppler matrix have been selected. If so, map the coordinates of the elements determined to be target signals to the range-Doppler matrix to obtain the distance unit and the Doppler unit where the target is located. Otherwise, perform the first step.

4. The sparse Bayesian algorithm based over-the-horizon radar range estimation method of claim 1, wherein: The observation vector in step 5 refers to selecting the pulse-compressed echo data in the Doppler unit where the target is located, and selecting a group of data with a main lobe width to form an observation vector.

5. The sparse Bayesian algorithm based over-the-horizon radar range estimation method of claim 1, wherein: The steps of constructing the measurement matrix in step 5 are as follows: First, construct a measurement matrix with a dimension of P x M, and the initial value of each element is 0. The value of P is equal to the length of the observation vector, and M = KP, where K is the ratio of the sampling frequency of the linear frequency modulation signal to the sampling frequency of the observation vector. Secondly, the linear frequency modulation continuous wave signal is pulse compressed with the matched filter impulse response to obtain a base vector, a peak point position q of the base vector is found, q-b to q+b elements of a main lobe of the base vector are intercepted, b is half of the number of sampling points of the main lobe of the base vector, all elements in the main lobe of the base vector are placed into the first row to the 2b row of the first column of the measurement matrix, and all elements in the main lobe of the base vector are placed into the kth row to the k+2b-1 row of the kth column of the measurement matrix, k=1, 2,..., M; Thirdly, the bth to the P+b-1th row of the measurement matrix is intercepted, the first column of the measurement matrix represents pulse compression waveform data of the linear frequency modulation continuous wave signal at the 0th moment, the kth column represents pulse compression waveform data of the linear frequency modulation continuous wave signal at the k*Δt moment, and Δt represents a time interval between each column of the measurement matrix; Fourthly, all column vectors of the measurement matrix are K times decimated, so that the dimension of the pulse compression waveform data of the linear frequency modulation continuous wave signal is consistent with the observation vector.

6. The sparse Bayesian algorithm based over-the-horizon radar range estimation method of claim 1, wherein: The specific steps of the step 5 of obtaining the sparse signal vector by sparsely recovering the observation vector by using the sparse Bayesian algorithm are as follows: First, initialize hyperparameters α, α0, α = (α1, α2, …, α M ) represents the hyperparameters of the sparse signal vector, each element of α is initialized to 1; α0represents the noise variance, which is initialized to 1; Secondly, the variance matrix R and the mean vector μ of the posterior probability density function of the to-be-solved sparse signal vector are calculated according to the following formula; R = (a0Φ T Φ + Λ) -1 μ = a0RΦ T y where Φ denotes the measurement matrix, (·) T denotes the transpose operation, Λ denotes the diagonal matrix of the hyperparameter α, y denotes the observation vector, (·) -1 denotes the inverse operation; Step 3. Update the hyperparameters according to the following formula and a new ; wherein denotes a new ith column vector of a i denotes the i i th element of a i R ii denotes a summation operation, i = 1, 2, …, M; In the fourth step, it is judged whether the root mean square error of the current iteration meets the convergence condition or reaches the maximum iteration number, if yes, the fifth step is executed, otherwise, the second step is executed; the convergence condition is that the root mean square error of the mean vector obtained in the current iteration and the mean vector obtained in the previous iteration is less than e -4 , and the maximum iteration number is 1000. Fifthly, the mean vector obtained in the current iteration is taken as the sparse signal vector.

7. The sparse Bayesian algorithm based over-the-horizon radar range estimation method of claim 5, wherein: The conversion of the position of each element in the sparse vector into the distance value of the target in Step 5 refers to: using the formula The position of each element in the sparse vector is converted into the distance value of the target; wherein c represents the speed of light, L represents the starting coordinate value in the pulse compressed echo data in the Doppler unit where the target is intercepted, f s represents the sampling frequency of the observation vector.

Images