Radar clutter suppression method based on alternating projection

Abstract

The present application belongs to the technical field of external radiation source radar signal processing, and particularly relates to a radar clutter suppression method based on alternating projection, aiming at the coupling problem of clutter suppression performance and memory space requirement of existing algorithms, realizing an iterative optimization scheme, and different from traditional iterative clutter suppression schemes (CLEAN algorithm, etc.), the radar clutter suppression method based on alternating projection proposed by the present application can guarantee that the algorithm converges linearly to the overall least square solution in norm, that is, there is theoretical guarantee and actual improvement in convergence and convergence speed.

Description

Technical Field

[0001] This invention belongs to the field of external radiation source radar signal processing technology, specifically relating to a radar clutter suppression method based on alternating projection. Background Technology

[0002] Clutter suppression techniques aim to reveal targets obscured by clutter by suppressing clutter in echo signals. Classical clutter suppression techniques employ weighted tapped delay lines to construct adaptive filters, including: Least Mean Square (LMS) adaptive filters, Normalized LMS (NLMS) adaptive filters, Recursive Least Square (RLS) adaptive filters, and many other adaptive filters. In 2012, James Palmer and Stephen Searle compared and analyzed various classical adaptive filters based on their zero-frequency clutter suppression performance and computational complexity. However, these adaptive filters suffer from slow convergence speed, high filter order, and high computational cost. When clutter is distributed over a wide area, the adaptive filter algorithms struggle to effectively handle real-world data.

[0003] Therefore, clutter suppression algorithms based on the "cancellation" concept have gained widespread attention, gradually forming a processing flow encompassing clutter parameter estimation, clutter template construction, clutter amplitude estimation, and cancellation. In related research, F. Colone et al. first proposed the extensive cancellation algorithm (ECA). However, in real-world data, clutter is continuously distributed, making effective suppression difficult through "discrete" algorithms. To address this problem, Olivier Rabaste et al. proposed the concept of infinite interpolation and achieved "continuous suppression" in the Doppler dimension. Subsequently, Brian Day et al. extended infinite interpolation theory to non-periodic continuous waves, simultaneously achieving continuous suppression in both the range and Doppler dimensions.

[0004] However, as clutter suppression algorithms continue to achieve breakthroughs in suppression performance, their computational complexity and resource requirements also increase, particularly in terms of memory demands. To address this excessive memory requirement, various methods have been proposed, including Extensive Cancellation Algorithm-batches (ECA-B), Sliding Extensive Cancellation Algorithm (ECA-S), Subcarrier Domain Clutter Suppression (ECA-C), Matching Pursuit (MP), and Generalized Subband Cancellation (GSC). However, while existing methods can reduce algorithm complexity, they often result in a loss of clutter suppression performance. Therefore, this invention proposes a radar clutter suppression method based on alternating projection, aiming to reduce complexity while maintaining clutter suppression performance. Verification shows that this method effectively reduces memory requirements and achieves clutter suppression performance superior to the aforementioned improved methods. Summary of the Invention

[0005] The purpose of this invention is to address the coupling problem between clutter suppression performance and memory space requirements in existing algorithms by implementing an iterative optimization scheme. Unlike traditional iterative clutter suppression schemes (such as the CLEAN algorithm), the radar clutter suppression method based on alternating projection proposed in this invention guarantees that the algorithm converges linearly with respect to the global least squares solution, thus providing both theoretical guarantees and practical improvements in convergence and convergence speed. The technical solution adopted to achieve the purpose of this invention is as follows:

[0006] A radar clutter suppression method based on dynamic alternating projection, characterized by the following steps:

[0007] Step 1: Acquire echo signals and reference signals from the radar separately, obtain the range-Doppler spectrum through pulse-Doppler processing, identify clutter based on threshold detection, extract the time delay-Doppler parameters of all clutter, and thus obtain the clutter time delay-Doppler parameter set.

[0008] Where, N c τ represents the clutter number, i represents the clutter index, and τ represents the clutter index. i f i These are the time delay and Doppler frequency of the i-th clutter, respectively;

[0009] Step 2: Construct the clutter template matrix Φ based on the range-Doppler parameter set;

[0010] in, x(τ i ,f i ) represents a time delay of τ i The Doppler frequency is f i Clutter template;

[0011] Step 3: Randomly select each column of the clutter template matrix to form Q clutter template submatrices with the same number of rows and columns and no repeated column vectors.

[0012] Where h is the current number of dynamic adjustments, H is the total number of dynamic adjustments, and 1≤h≤H; q is the index of the clutter template submatrix, and Q is the total number of clutter template submatrixes;

[0013] Step 4: Based on Step 3, perform m-processing on the echo signal. h By alternating projections, the time-domain vector of the digital echo signal after the h-th dynamic adjustment is obtained. Here, y0 is defined as the time-domain vector of the digital echo signal when h = 0, which is the original time-domain vector;

[0014] Step 5: Let h = h + 1; when h > H, end the process and output the time-domain vector y of the echo signal after clutter suppression. H If h ≤ H, continue execution.

[0015] Step 6: Based on the time-domain vector y obtained in Step 4 h The partitioning scheme of the clutter template matrix Φ is adaptively and dynamically adjusted so that the Q clutter subspaces corresponding to the dynamically adjusted Q clutter template submatrices become more orthogonal.

[0016] Step 4 of this invention specifically includes the following steps:

[0017] Step 4(a): Generate an orthogonal projection matrix from the echo space to the corresponding clutter subspace based on the clutter template submatrix.

[0018] in, From echo space to clutter template submatrix Φ h,q The orthogonal projection matrix of the corresponding clutter subspace's orthogonal complement space;

[0019] Step 4(b): In y h-1 Based on this, repeat m h Alternating projection of wheels, i.e.

[0020] Step 6 of this invention specifically includes the following steps:

[0021] Step 6(a): For the time-domain vector y obtained in step 4 hPulse-Doppler processing is performed to obtain the corresponding range-Doppler spectrum, and peak extraction is completed based on the clutter delay-Doppler parameter set to estimate N. c The remaining energy of each clutter, i.e., obtaining the set of clutter remaining energy.

[0022] Step 6(b): Set the clutter residual energy collection according to E i Sort the clutter template matrix in descending order of size, and then re-divide the clutter template matrix into Q clutter template sub-matrices according to the re-sorted order;

[0023] Step 6(c): Jump to step 4.

[0024] This invention provides a radar clutter suppression method based on alternating projection, which has the following advantages compared with existing technologies:

[0025] 1. The trade-off between clutter suppression performance and memory space requirements is transformed into a trade-off between computation time and memory space requirements, ensuring that clutter suppression performance is not limited by the size of the computer's memory space;

[0026] 2. Compared to traditional iterative clutter suppression methods, the method provided in this invention can linearly converge to the global least squares solution with norm. Furthermore, the dynamic adjustment mechanism provided in this invention ensures a constant-level improvement in convergence speed during linear convergence. In short, the radar clutter suppression method based on alternating projection can efficiently and accurately obtain the least squares solution, i.e., [I-Φ(Φ H Φ) -1 Φ H ]y0. Attached Figure Description

[0027] Figure 1 This is a schematic flowchart of a radar clutter suppression method based on alternating projection provided by the present invention;

[0028] Figure 2 This invention provides a range-Doppler spectrum generated from OFDM external radiation source radar simulation data.

[0029] Figure 3 The range-Doppler spectrum is generated after clutter suppression of the simulation data provided by this invention using the ECA-C method;

[0030] Figure 4 The range-Doppler spectrum is generated after clutter suppression of the simulation data provided by this invention using the ECA-S method;

[0031] Figure 5 The range-Doppler spectrum is generated after clutter suppression of the simulation data provided by this invention using the GSC method;

[0032] Figure 6 The range-Doppler spectrum is generated after the simulation data provided by this invention is subjected to clutter suppression by a radar clutter suppression method based on alternating projection provided by this invention.

[0033] Figure 7 It is the Doppler slice curve of the target position in the simulation data provided by the present invention, which includes the results of existing methods and the method described in the present invention. Detailed Implementation

[0034] The radar clutter suppression method based on alternating projection described in this invention will be further described below with reference to specific embodiments and accompanying drawings, but the implementation of this invention is not limited thereto.

[0035] In this embodiment, experimental verification is performed using OFDM external radiation source radar simulation data. Figure 1 As shown, the radar clutter suppression method based on alternating projection provided by the present invention includes:

[0036] Step 1: Acquire echo signals and reference signals from the OFDM external radiation source radar, obtain the time-delay-Doppler spectrum through pulse-Doppler processing, identify clutter based on threshold detection, extract the time-delay-Doppler parameters of all clutter, and thus obtain the clutter time-delay-Doppler parameter set. Specifically:

[0037] Step 1(a): Based on the detection environment, simulate the radar system, construct the echo model, and then generate the simulated digital echo signal y and reference signal x;

[0038] The radar system parameters include the carrier frequency of the transmitted signal, the array element arrangement and number of the receiving array; the echo model includes clutter, targets and noise, and needs to consider the position distribution of clutter and targets, signal-to-clutter ratio, and signal-to-noise ratio; the continuously distributed clutter in the detection environment can be equivalent to the superposition of multiple discrete clutter scattering points, and the noise uses additive white Gaussian noise.

[0039] Step 1(b): Obtain the time-delay-Doppler spectrum χ(τ,f) through pulse-Doppler processing, identify clutter based on threshold detection, and extract the time-delay-Doppler parameters of all clutter, i.e.

[0040] Where ε0 is the threshold.

[0041] Step 2: Construct the clutter template matrix Φ based on the time delay-Doppler parameter set, specifically:

[0042] Step 2(a): Denote the reference signal at sampling time t as... Among them, f sLet T be the sampling frequency. Assuming T is the coherent processing time, we can construct a reference signal vector x = [x(0), x(1), ..., x(N-1)]. T Where, N = Tf s .

[0043] Step 2(b): Given the time delay τ i A time-shift transform is performed on the reference signal vector x. To ensure that when... When time shift transformation can obtain the error-free true value, this invention utilizes the cyclic prefix of the OFDM signal for time shift transformation. Let τ... i f s =n a +n b ;in, n is an integer displacement. b ∈(0,1) represents the decimal shift.

[0044] Step 2(b).1: Perform a time-shift transformation on x using integer sampling points, and denote the temporary variable as x. (a) Then there is,

[0045] x (a) =[x(-n a ),x(-n a +1),…,x(Nn a -1)] T

[0046] At the same time, through sign synchronization and appropriate truncation, x(-n) is guaranteed to be... a The sampling point corresponding to the start time of the OFDM symbol, and x (a) The sampling points completely contain L OFDM symbols. Therefore, x (a) It can also be expressed as,

[0047]

[0048] Where, x l Let represent the time-domain vector of the l-th OFDM symbol, and satisfy .

[0049] x l =[x l (0),x l (1),…x l (N u +N g -1)] T

[0050]

[0051]

[0052] xl,g (n)=x l,u (N u -N g +n), 0≤n≤N g -1

[0053] Where, N g N is the sampling length of the cyclic prefix. u It is the sampling length of the effective symbol portion, K is the number of subcarriers, and s l,k It is the complex amplitude of the k-th subcarrier of the l-th OFDM symbol.

[0054] Step 2(b).2: For x (a) Perform a time-shift transformation on non-integer sampling points, with the temporary variable denoted as x. (b) Then there is,

[0055]

[0056]

[0057]

[0058] in,

[0059] x l,u =[x l,u (0),…,x l,u (N u -1)] T

[0060] x l,u (-n b )=[x l,u (-n b ),…,x l,u (N u -n b -1)] T

[0061]

[0062] Step 2(c): Given the Doppler frequency f i For x (b) The final result after frequency transformation and time-delay-Doppler transformation is...

[0063] in,

[0064]

[0065] Step 2(d): For all i = 1, ..., N c Repeat steps 2(b)-2(c) to obtain the following result:

[0066]

[0067] Step 3: Divide the clutter template matrix into Q clutter template sub-matrices, denoted as... Where h = 1, 2, ..., H represents the current number of dynamic adjustments, and H represents the total number of dynamic adjustments. Specifically: define r(i) → {q, j} as the clutter template submatrix from column index i to the q-th clutter template submatrix Φ. h,q A bijection of column index j; and define r -1 (q,j)→i is the inverse mapping of r(i); furthermore, it is defined as follows: Then there is,

[0068]

[0069] The bijective r(i) can have any form.

[0070] Step 4: Based on Step 3, perform m-processing on the echo signal. h By alternating projections, the time-domain vector y of the digital echo signal after the h-th dynamic adjustment is obtained. h Specifically:

[0071] Step 4(a): Modify the clutter template submatrix Φ h,q The corresponding clutter subspace is denoted as S. h,q Its orthogonal complement space is denoted as Will denoted as from Hilbert space arrive The projection matrix, i.e. Then there is,

[0072]

[0073] Step 4(b): Definition Echo signal sequence; where y0 = y is the time-domain vector of the input echo signal, y H Let be the time-domain vector of the output echo signal. After the h-th dynamic adjustment, in y h-1 Based on this, repeat m times. h Alternating projection of wheels, i.e.

[0074] Step 5: Let h = h + 1; when h > H, end the process and output the time-domain vector y of the echo signal after clutter suppression. H If h ≤ H, continue execution.

[0075] Step 6: Based on the time-domain vector y obtained in Step 4 h The clutter template matrix Φ is adaptively and dynamically adjusted to optimize its partitioning scheme as much as possible. They are mutually orthogonal, specifically:

[0076] Step 6(a): For the time-domain vector y obtained in step 4 h Pulse-Doppler processing is performed to obtain the corresponding time-delay-Doppler spectrum, and peak extraction is completed based on the clutter time-delay-Doppler parameter set to estimate N. c The remaining energy of each clutter, i.e., obtaining the set of clutter remaining energy.

[0077] Step 6(b): Similar to step 3, define e(q,j)→i as the submatrix Φ from the q-th clutter template. h,q The bijection from column index j to column index i of clutter template matrix Φ; for the j1st column of the q1th clutter template submatrix and for the j2nd column of the q2th clutter template submatrix, the following condition is satisfied when q1≤q2 and j1≤j2.

[0078]

[0079] Construct the clutter template submatrix using the bijective e(q,j) as the mapping function, i.e.

[0080] x q,j =x(τ) e(q,j) ,f e(q,j) )

[0081] Step 6(c): Jump to step 4.

[0082] The simulation signal used in this example is an OFDM signal with a bandwidth of 10 kHz, sampled at a sampling frequency of 48 kHz. It is assumed that the reference signal x is interference-free and noise-free, the echo signal y contains ground clutter, sea clutter, and ionospheric clutter, and the target is a moving target. Specific parameters are shown in Table 1.

[0083] Table 1 Simulation Target and Clutter Parameters

[0084]

[0085] like Figure 2 As shown, pulse-Doppler processing is performed on the echo signal y, and the resulting time-delay-Doppler spectrum has a uniform and random substrate throughout the entire detection plane. This substrate is generated by clutter and is much higher than the noise substrate, which is the main factor in masking the target. Figure 6 a is the range-Doppler spectrum obtained after 12 rounds of alternating projection in a radar clutter suppression method based on alternating projection provided by this invention. Figure 6 b is the range-Doppler spectrum obtained after 24 rounds of alternating projection; obviously, after 12 rounds of alternating projection, the ionospheric clutter and target obscured by the sea clutter are revealed. Figure 6a) After 12 rounds of alternating projection, ionospheric clutter is further suppressed, and the signal-to-clutter ratio of the target is further improved. Figure 6 b). In addition, in comparison Figures 3 to 6 The range-Doppler spectrum output by the algorithm reveals that the radar clutter suppression method based on alternating projection provided by this invention can achieve the maximum output signal-to-clutter ratio.

[0086] exist Figure 7 In the image, the vertical gray dashed lines mark the locations of targets in the Doppler slices, while the remaining five horizontal curves represent Doppler slices obtained based on the original echo signal vector and the output vectors of different clutter suppression methods. Figure 7 It can be clearly observed that the radar clutter suppression method based on alternating projection has the lowest basis, the least target energy loss, and the largest output signal-to-clutter ratio.

[0087] To illustrate the content and implementation method of this invention, this specification provides a typical embodiment. The details introduced in the embodiment are not intended to limit the scope of the claims, but rather to aid in understanding the method described in this invention. Those skilled in the art should understand that various modifications, variations, deductions, or substitutions to the steps of the typical embodiment, without departing from the concept of this invention, should be considered within the scope of protection of this invention.

Claims

1. A radar clutter suppression method based on alternating projection, characterized in that, Includes the following steps: Step 1: Acquire digitized echo signals and reference signals from the external radiation source radar, obtain the range-Doppler spectrum through pulse-Doppler processing, identify clutter based on threshold detection, and extract the time delay-Doppler parameters of all clutter, thus obtaining the clutter time delay-Doppler parameter set. ; in, For clutter number, For clutter indexing, , The first The time delay and Doppler frequency of the clutter; Step 2: Construct clutter template matrix based on range-Doppler parameter set ; in, , Indicates the delay as Doppler frequency is Clutter template; Step 3: Randomly select and arrange the columns of the clutter template matrix to form... A clutter template submatrix with the same number of rows and columns and no repeating column vectors. ; in, This is the current number of times the system can be dynamically adjusted. The total number of times is dynamically adjusted, and has ; The index of the clutter template submatrix. This represents the total number of clutter template submatrices; Step 4: Based on step 3, that is, the... After the dynamic adjustment, the echo signal is... Alternating projections are used to obtain the first... The time-domain vector of the digital echo signal after dynamic adjustment ; Among them, the definition for The digital echo signal time-domain vector, i.e., the original time-domain vector; Step 5: Let ;when When the clutter-suppressed echo signal time-domain vector is output, the process ends. ;when Continue execution if necessary; Step 6: Based on the time-domain vector obtained in Step 4 Adaptive dynamic adjustment of clutter template matrix The partitioning scheme allows for dynamic adjustments. The clutter template submatrix corresponding to The clutter subspaces further tend to be orthogonal; proceed to step 4.

2. The radar clutter suppression method based on alternating projection according to claim 1, characterized in that, The echo signal and the reference signal are in the form of OFDM signal.

3. The radar clutter suppression method based on alternating projection according to claim 1, characterized in that, Step 4 specifically includes: Step 4(a): Generate an orthogonal projection matrix from the echo space to the corresponding clutter subspace based on the clutter template submatrix; in, From echo space to clutter template submatrix The orthogonal projection matrix of the corresponding clutter subspace's orthogonal complement space; Step 4(b): In Based on this, repeat the process. Alternating projection of wheels, i.e. .

4. The radar clutter suppression method based on alternating projection according to claim 1, characterized in that, Step 6 specifically includes: Step 6(a): The time-domain vector obtained in Step 4 Pulse-Doppler processing is performed to obtain the corresponding range-Doppler spectrum, and peak extraction is completed based on the clutter delay-Doppler parameter set to estimate the range-Doppler spectrum. The remaining energy of each clutter, i.e., obtaining the set of clutter remaining energy. ; Step 6(b): Sort the clutter residual energy set in descending order of size, and re-divide the clutter template matrix according to the re-sorted order. A clutter template submatrix; Step 6(c): Jump to step 4.

5. The radar clutter suppression method based on alternating projection according to any one of claims 3 and 4, characterized in that, In terms of convergence, it converges with respect to the norm. It has at least linear convergence speed, and can guarantee that the linear convergence rate can be continuously improved at the constant level.

Images