A method for suppressing reinforcement clutter based on distance compensation and low-rank sparse decomposition

Abstract

The application discloses a through-wall radar rebar clutter suppression method based on distance compensation and low-rank sparse decomposition. After removing the wall clutter through mean cancellation, the position of the rebar is located and the propagation speed of the electromagnetic wave in the wall is estimated by using the Hough transform. Then, the offset distance of the rebar clutter is calculated, and it is offset to a horizontal straight line through distance compensation. Based on the low-rank sparse decomposition and the total variation regularization model, the target echo matrix is extracted. Finally, inverse distance compensation is performed to restore the original shape of the target echo. The simulation and experimental results of mixed experimental data show that, compared with the existing SVD method and RPCA method, the proposed method achieves better rebar clutter suppression effect.

Description

Technical Field

[0001] This invention belongs to the field of through-wall radar, and particularly relates to a method for suppressing reinforcement clutter based on range compensation and low-rank sparse decomposition. Background Technology

[0002] Through-the-wall radar (TWR) utilizes the penetrating properties of electromagnetic waves to detect indoor targets and is widely used in counter-terrorism, stability maintenance, urban warfare, law enforcement, and disaster relief. However, the interaction between electromagnetic waves and complex building walls produces effects such as reflection, refraction, and diffraction, resulting in weak radar echo signals and strong wall clutter. Therefore, there is an urgent need to study wall clutter suppression methods to provide technical support for achieving high-resolution target imaging and accurate identification.

[0003] Currently, wall clutter suppression methods mainly include background cancellation, subspace decomposition, and low-rank sparse decomposition. Among these, background cancellation utilizes empty scene data to remove wall clutter. Subspace decomposition, through eigenvalue decomposition, selects the eigenvalues ​​corresponding to the wall clutter and sets them to zero, reconstructing the echo matrix to achieve wall clutter suppression. Low-rank sparse decomposition, based on the low-rank nature of wall clutter and the sparsity of the target, establishes a low-rank-sparse optimization problem, solves it using an inaccurate Lagrange multiplier method, obtains the sparse components, and achieves wall clutter suppression. The main problem is that existing techniques cannot effectively suppress wall reinforcement clutter. Summary of the Invention

[0004] To address the aforementioned issues, this invention proposes a rebar clutter suppression algorithm based on range compensation and low-rank sparse decomposition, which can effectively suppress rebar clutter in through-wall radar data while preserving the target area to a certain extent.

[0005] The reinforcement clutter suppression algorithm based on distance compensation and low-rank sparse decomposition described in this invention is implemented through the following technical solution:

[0006] Step 1: Transfer the through-wall radar data matrix Model as a matrix ,matrix sum matrix The sum of, where the matrix Represents wall clutter data, matrix Represents rebar clutter data, sparse matrix Represents target echo data;

[0007] Step 2: Extract wall clutter data from the through-wall radar echo data. Removed based on mean cancellation method;

[0008] Step 3: Use Hough transform to detect the position of the rebar clutter apex. And estimate the propagation speed of electromagnetic waves within the wall. ;

[0009] Step 4: Based on the obtained parameters, calculate the offset of the rebar clutter in each A-scan, and use distance compensation to offset the rebar clutter into a horizontal straight line;

[0010] Step 5: Transfer the offset through-wall radar data matrix Model as a low-rank matrix sparse matrix The sum of, where the low-rank matrix is Represents rebar clutter data, sparse matrix Represents target echo data;

[0011] Step 6: Apply a low-rank sparse decomposition method to the sparse matrix. Extract;

[0012] Step 7: Restore the sparse matrix using inverse distance compensation The original shape of the target echo in the solution, and the sparse matrix obtained by solving. This is the target data matrix after clutter suppression;

[0013] Beneficial effects:

[0014] 1. This invention provides a method for suppressing rebar clutter in through-wall radar based on range compensation and low-rank sparse decomposition. Range compensation is used to offset the rebar clutter into a horizontal straight line, which improves the low-rank nature of the rebar clutter and makes the signal model more in line with the requirements of low-quality sparse decomposition algorithm, resulting in better clutter suppression effect.

[0015] 2. This invention provides a method for suppressing rebar clutter in through-wall radar based on range compensation and low-rank sparse decomposition. It employs a low-rank sparse decomposition algorithm with total variation, transforming the low-rank sparse decomposition optimization problem into a joint optimization problem based on minimizing the sum of the low-rank matrix rank function, the sparse matrix sparsity, and the sparse matrix total variation regularization term. Minimizing the low-rank matrix rank function suppresses rebar clutter, the sparse matrix sparsity enhances target echoes, and the sparse matrix total variation regularization term further filters out remaining rebar clutter components and background noise. Furthermore, simulation and experimental results demonstrate that this invention can improve the target clutter ratio of through-wall radar data, making it an effective method for suppressing rebar clutter in through-wall radar. Attached Figure Description

[0016] Figure 1 This is the through-wall radar scene signal model on which the method of this invention is based;

[0017] Figure 2 This is a flowchart of the signal processing according to an embodiment of the present invention;

[0018] Figure 3 This is the geometric relationship model of the steel bar echo on which the method of this invention is based;

[0019] Figure 4 This is a scene diagram used in the simulation of the method of this invention;

[0020] Figure 5 These are graphs showing the results of processing simulation data using different methods. (a) represents the original data, (b) represents the result of processing simulation data using the SVD method, (c) represents the result of processing simulation data using the RPCA method, and (d) represents the result of processing simulation data using the method of this invention.

[0021] Figure 6 This is a scene diagram obtained from actual measurements using the method of this invention;

[0022] Figure 7 These are graphs showing the results of processing measured mixed data using different methods. (a) represents the original data, (b) represents the result of processing measured mixed data using the SVD method, (c) represents the result of processing measured mixed data using the RPCA method, and (d) represents the result of processing measured mixed data using the method of this invention.

[0023] Specific implementation process

[0024] The purpose of this invention is to address the problem that existing methods cannot effectively suppress rebar clutter, and to propose a rebar clutter suppression method based on distance compensation and low-rank sparse decomposition. This invention is applicable to homogeneous medium walls with a single layer of uniformly distributed rebar, providing technical support for subsequent high-resolution imaging and accurate identification of targets.

[0025] Figure 1 This invention is based on a through-wall radar signal model, wherein, and This represents the reflected waves from the wall surface. Indicates the echo of the reinforcing bar. This indicates the target echo. Figure 2 This is a flowchart of the signal processing implementation of an embodiment of the present invention. Based on the scenario of a single-transmitter, single-receiver through-wall radar, after acquiring the time-domain echo, the present invention is implemented through the following steps:

[0026] Step 1: Modeling the raw echo signal from the through-wall radar;

[0027] The obtained raw echo data matrix of the through-wall radar Model as a matrix ,matrix sum matrix The sum of, where the matrix Represents wall clutter data, matrix Represents rebar clutter data, sparse matrix This represents the target echo data.

[0028] Step 2: Mean cancellation to remove wall noise;

[0029] The wall clutter signal remains essentially unchanged along the survey line, exhibiting roughly the same waveform, amplitude, and time delay. By averaging the radar data from each A-scan constituting the B-scan along the horizontal direction to obtain the average channel, and then subtracting the individual A-scans, the wall clutter signal is eliminated. This method can be expressed mathematically as follows:

[0030]

[0031] in, These represent B-scan data before and after wall clutter removal, respectively. Indicates the number of time sampling points. This indicates the number of A-scan channels measured by the through-wall radar.

[0032] Step 3: Hough transform, detect the coordinates of the top position of the steel bar echo, and estimate the electromagnetic wave propagation speed;

[0033] like Figure 3 The geometric model of the reinforcing bars shown is an example where the reinforcing bar echo is a hyperbola in B-scan, and its discretized curve equation is expressed as:

[0034]

[0035] in, These represent the row and column coordinates in a B-scan, respectively. These represent the row and column coordinates of the rebar echo vertex, respectively. The range of values ​​is , The range of values ​​is , Indicates the antenna step distance. Indicates the time sampling interval. This indicates the speed at which electromagnetic waves propagate through the wall.

[0036] According to the curve equation Given The range of values ​​for , calculate

[0037] Corresponding hyperbolic feature points coordinates Calculate the statistics of the Hough transform:

[0038] .

[0039] Therefore, the estimated values ​​for the vertex position and velocity of the steel bar echo are:

[0040]

[0041] Step 4: Distance compensation, offset the rebar echo to a horizontal straight line;

[0042] Calculate the number of time delay offset points for each A-scan of the rebar echo. :

[0043]

[0044] in, and They represent The integer part and the fractional part.

[0045] Therefore, the steel bar echo is offset:

[0046]

[0047] in, The length of the interpolation kernel for sinc interpolation.

[0048] Step 5: Modeling the processed radar echo data;

[0049] The offset through-wall radar data matrix Model as a low-rank matrix sparse matrix The sum of, where the low-rank matrix is Represents rebar clutter data, sparse matrix This represents the target echo data.

[0050] Step 6: Low-rank sparse decomposition to obtain the target echo;

[0051] Based on the low-rank sparse decomposition method, the optimization problem is:

[0052]

[0053] in, express nuclear norm number, express of Norm, express Anisotropic total variation norm, and They represent Regularization parameters for norms and anisotropic total variational norms. Establish the optimization function:

[0054]

[0055] in, For Lagrange multipliers, For matrix and The inner product, For matrix of The square of the norm.

[0056] The steps for processing through-wall radar signals based on the above optimization function are as follows:

[0057] 1) Initialization: Iteration parameters low-rank matrix sparse matrix Lagrange multipliers Penalty coefficient .

[0058] 2) Order Optimize the solution of low-rank matrices :

[0059]

[0060]

[0061] in, For SVD decomposition, This is a soft threshold operator.

[0062] 3) Optimize the solution of sparse matrices :

[0063]

[0064] in, The solution is obtained using the fast gradient projection algorithm.

[0065] 4) Update the Lagrange multipliers :

[0066]

[0067] 5) Update the penalty coefficient :

[0068]

[0069] 6) When When the iteration termination condition is met, output... and Otherwise, return to step 2) for the next iteration;

[0070] 7) Output a sparse matrix through iterative loops. .

[0071] Step 7: Perform inverse distance compensation to restore the target echo shape;

[0072] In step 3, the hyperbolic shape of the target echo also changes. Therefore, the inverse range compensation algorithm, which is the exact opposite of step 3, is used to restore the shape of the target echo:

[0073]

[0074] in, This represents the final target echo matrix. Thus, a method for suppressing rebar clutter based on distance compensation and low-rank sparse decomposition is complete.

[0075] To verify the proposed method for suppressing reinforcement clutter based on distance compensation and low-rank sparse decomposition, a simulation experiment was designed for analysis and verification using measured data.

[0076] In this invention, the information density improvement factor used for quantitative analysis is defined as follows:

[0077]

[0078] in, Defined as the target-to-clutter ratio in ground-penetrating radar data. The signal-to-noise ratio after processing by the method. The signal-to-noise ratio before the method processing

[0079]

[0080] In the above formula, and Representing the target area and clutter region The amount of data, For in position p Data values ​​at

[0081] like Figure 4 As shown in Table 1, the simulation was performed using gprMax software to build a wall-penetrating scenario model.

[0082] Table 1 Simulation Parameter Settings

[0083]

[0084] Table 2 shows the improvement factor results of the method of this invention and the other two methods for processing simulation data. As can be seen from Table 3, the clutter suppression effect of the method of this invention is significantly better than that of SVD and RPCA. Figure 5 These are the results of processing simulation data using different methods. (a) shows the original data, (b) shows the result of processing the simulation data using the SVD method, (c) shows the result of processing the simulation data using the RPCA method, and (d) shows the result of processing the simulation data using the method of this invention. The SVD method suppresses some rebar clutter, but leaves a significant amount of residual clutter. The RPCA method has no significant suppression effect on rebar clutter, and the clutter is almost completely preserved. The method proposed in this invention achieves better clutter suppression; almost all rebar clutter is suppressed, with only a very small amount remaining, and the target echo is also completely preserved.

[0085] Table 2 Results of the Improvement Factor (IF) for processing simulation data

[0086]

[0087] Cross-sectional view of the actual test scene as shown Figure 6 As shown, the cross-sectional area of ​​the measured scene is 2.0 m × 1.0 m. The target transmitted waveform uses a Ricker wavelet with a center frequency of 900 MHz, a time window of 25 ns, and 2048 sampling points. The transmitting and receiving antennas are separate, 10 cm from the wall, with a step size of 1 cm. The resulting raw data matrix of the through-wall radar is 2048 × 200. Due to limitations, wall clutter, rebar clutter, and background noise are obtained from the radar, while the target echo data is generated by gprMax software. The two are then mixed, and the final data is defined as the measured mixed data. The target is set as a small metal sphere with a radius of 5 cm, 20 cm from the wall.

[0088] Table 3 shows the improvement factor results of the method of this invention and the other two methods for processing the measured mixed data. As can be seen from Table 3, the improvement factor of the method of this invention is significantly better than that of SVD and RPCA. Figure 5 The figures show the results of processing simulation data using different methods. (a) shows the original data, (b) shows the result of processing mixed data using the SVD method, (c) shows the result of processing mixed data using the RPCA method, and (d) shows the result of processing mixed data using the method of this invention. The SVD and RPCA methods have no significant suppression effect on rebar clutter; the clutter is almost completely preserved. The method proposed in this invention achieves better clutter suppression; rebar clutter is almost completely suppressed with no obvious rebar clutter residue, and the target echo is also well preserved.

[0089] Table 3. Results of the Improvement Factor (IF) for processing simulation data.

[0090]

[0091] This invention proposes a method for suppressing rebar clutter in through-wall radar based on range compensation and low-rank sparse decomposition. After removing wall clutter through mean cancellation, Hough transform is used to locate the rebar position and estimate the propagation velocity of electromagnetic waves in the wall. Then, the offset range of the rebar clutter is calculated and offset into a horizontal straight line through range compensation. Next, the target echo matrix is ​​extracted based on low-rank sparse decomposition and a fully variational regularization model. Finally, inverse range offset is performed to restore the original shape of the target echo. Simulation and experimental results with mixed experimental data show that the proposed method achieves better rebar clutter suppression performance compared to directly using existing standard subspace methods and RPCA methods.

[0092] Of course, the present invention may have other various embodiments. Without departing from the spirit and essence of the present invention, those skilled in the art can make various corresponding changes and modifications according to the present invention, but these corresponding changes and modifications should all fall within the protection scope of the appended claims.

Claims

1. A method for wall penetrating radar rebar clutter suppression based on range compensation and low-rank sparse decomposition, characterized in that, The steps include the following: Step 1: Constructing a matrix of through-the-wall radar data Modeling into a matrix , a matrix and a matrix , where the matrix represents the wall clutter data, the matrix represents the rebar clutter data, and the sparse matrix represents the target echo data; Step 2: Removing wall clutter data in the through-wall radar echo data based on mean cancellation method Step 3: Detecting the position of the rebar clutter vertex using the Hough transform and estimating the speed of propagation of electromagnetic waves within the wall ; Step 4: According to the obtained parameters, the offset of the steel bar clutter in each A-scan is calculated, and the steel bar clutter is offset into a horizontal straight line by distance compensation; Step 5: the offset wall-penetrating radar data matrix is modeled as a low-rank matrix and a sparse matrix , where the low-rank matrix represents the rebar clutter data, and the sparse matrix represents the target echo data; Step 6: Low-rank sparse decomposition method is used to decompose the sparse matrix extraction; Step 7: Reconstruct the sparse matrix by inverse distance compensation The sparse matrix obtained by solving the above equation is The target data matrix after clutter suppression.

2. The method of claim 1, wherein the method is characterized by, The wall clutter is removed by mean cancellation: ； wherein, B-scan data before and after wall clutter removal, respectively, denotes the number of time samples, denotes the number of A-scan channels measured by the through-the-wall radar.

3. The method of claim 1, wherein the method is characterized by Based on the Hough transform, the position coordinates of the steel bar echo vertex are detected, and the electromagnetic wave propagation speed is estimated; The steel bar echo is a hyperbola in the B-scan, and the discretized curve equation is expressed as: ； wherein, denote the row and column coordinates in the B-scan, respectively, denote the row and column coordinates of the rebar echo apex, respectively, the value range of is , the value range of is , denotes the step distance of the antenna, denotes the time sampling interval, denotes the propagation speed of the electromagnetic wave in the wall. According to the curve equation Given The range of values ​​for , calculate Corresponding hyperbolic feature points coordinates ; Calculate the statistics of the Hough transform: Thus, the position and velocity estimates of the rebar echo are: 。 4. The method of claim 1, wherein the method is characterized by Distance compensation is used to offset the steel bar echo into a horizontal straight line, and the low rank of the steel bar clutter is improved; Computing time delay offset points for each A-scan of a rebar echo : ； wherein and denote the integer part and the decimal part, respectively, of the integer part and the decimal part, respectively, of Therefore, the steel bar echo is offset as follows: ； wherein is the interpolation kernel length for the sinc interpolation.

5. The method of claim 1, wherein the method is characterized by, Based on low rank sparse decomposition, the target echo matrix is obtained; The optimization problem is as follows: ； wherein denotes the nuclear norm of denotes the norm of denotes the anisotropic total variation norm of and denote the regularization parameters for the norm and the anisotropic total variation norm, respectively; and establishing an optimization function: ； wherein is a Lagrange multiplier, is a matrix is an inner product of with is a matrix is a norm of the square of The steps of the through-wall radar signal processing based on the above optimization function are as follows: 1) Initialization: iteration parameters , low-rank matrix , sparse matrix , Lagrange multiplier , penalty term coefficient ; 2) Let , optimize solving low-rank matrix : ； ； wherein is the SVD decomposition, is a soft thresholding operator; 3) Optimized solution of sparse matrices : ； wherein, is acquired for the fast gradient projection algorithm; 4) Update Lagrange multipliers : ； 5) updating the penalty coefficient : ； wherein is a convergence parameter; 6) When the iteration termination condition is satisfied, where, is a convergence parameter, output and otherwise, return to step 2) for the next iteration; 7) output the sparse matrix through a loop iteration .

6. The method of wall penetrating radar rebar clutter suppression based on distance compensation and low-rank sparse decomposition according to claim 1, characterized in that, Based on inverse distance compensation, the original shape of the target echo is restored: ； wherein, is the resulting target echo matrix.

7. The method of any one of claims 1-6, wherein the method is a through-the-wall radar rebar clutter suppression method based on distance compensation and low-rank sparse decomposition. The through-wall radar radar data matrix is obtained by The sampling data of each channel is constructed from the depth direction scanning, and the number of sampling points of each channel sampling data is recorded as , and .

Images