A semi-parametric time-frequency domain notch filtering method for weak suppression of broadband interference in SAR systems

Abstract

The present invention discloses a semi-parametric time-frequency domain notch filter suppression method for weakly suppressed broadband interference of a SAR system, comprising: obtaining disturbed original SAR image data, wherein weakly suppressed broadband interference exists on the original SAR image data; performing sparse decomposition on the original SAR image data to obtain a total optimization problem; decomposing the total optimization problem into two sub-optimization problems, wherein the first sub-optimization problem is an optimization problem for solving low-rank interference, and the second sub-optimization problem is an optimization problem for solving sparse useful signals; performing alternating optimization solutions on the first sub-optimization problem and the second sub-optimization problem to obtain sparse items and low-rank items. The present invention can be directly applied to original data of a SAR system or to SAR images.

Description

Technical Field

[0001] The invention relates to the technical field of SAR data processing, and in particular to a semi-parametric time-frequency domain notch filtering suppression method for weakly suppressing broadband interference of a SAR system. Background Art

[0002] Under extreme weather conditions, ordinary optical radar cannot provide clear imaging, while synthetic aperture radar (SAR) is a microwave sensor that can provide high-resolution images of the observed scene. Due to the increasingly crowded spectrum, during the imaging process, the SAR system may be interfered by other systems, such as communication systems and other radar systems, among which there is a large amount of suppressed wideband interference. Therefore, how to remove or mitigate suppressed narrowband interference in SAR systems has attracted great attention from the academic community.

[0003] Currently, there are three traditional methods for removing suppressive broadband interference in SAR systems: non-parametric method, parametric method and semi-parametric method.

[0004] As for non-parametric methods, although traditional non-parametric methods have the advantages of being simple and fast, and not relying on the waveform characteristics of interference, they will fail when the interference is weak. Therefore, this type of algorithm is not robust enough in a time-varying complex interference environment. At the same time, the signal actually received by the radar is a combination of interference, true echo and noise. The non-parametric algorithm only considers the intensity characteristics of the interference and lacks constraints on the true echo, making it impossible to effectively protect it.

[0005] The parametric method can ideally remove the corresponding interference, but it requires the interference to be an isolated single type of interference, and the calculation complexity is high. It also does not take into account the characteristics of the real echo signal and lacks protection for it.

[0006] For the semi-parametric method, it not only uses the optimization model to constrain the interference, but also takes into account the characteristics of the real echo signal. However, it can only target one type of interference at the same time, and the amount of calculation is huge when performing large-scale matrix operations. Summary of the invention

[0007] In view of this, the purpose of the present invention is to provide a semi-parametric time-frequency domain notch filtering suppression method for weak suppression broadband interference in SAR systems, so as to solve the technical problem that there are a large number of weak suppression broadband interferences in current SAR systems and it is difficult to easily remove the weak suppression broadband interferences.

[0008] In order to achieve the above object, the present invention adopts the following technical solution:

[0009] A semi-parametric time-frequency domain notch filtering suppression method for weakly suppressing broadband interference in a SAR system, the method comprising:

[0010] Step S1, obtaining interfered original SAR image data, wherein weakly suppressed broadband interference exists on the original SAR image data;

[0011] Step S2, performing sparse decomposition on the original SAR image data obtained in step S1 to obtain an overall optimization problem, wherein the overall optimization problem is: solving the interference of low-rank terms and the useful signal of L2 sparse terms by using the augmented Lagrangian method on the original matrix data;

[0012] Step S3, decomposing the total optimization problem obtained in step S2 into two sub-optimization problems, wherein the first sub-optimization problem is to solve the optimization problem of low-rank interference, and the second sub-optimization problem is to solve the optimization problem of sparse useful signals;

[0013] Step S4: For the first sub-optimization problem and the second sub-optimization problem obtained in step S3, perform alternating optimization solutions to obtain sparse terms and low-rank terms.

[0014] Furthermore, in the step S1, the original SAR image data is matrix data arranged in the range time domain and the azimuth time domain.

[0015] Furthermore, in step S2, the overall optimization problem is:

[0016]

[0017] In formula (1), R is the original data of the SAR image, A is the interference in the form of a low-rank term matrix, S is the useful signal in the form of a sparse term matrix, λ is a hyperparameter for adjusting the relationship between the solution results A and S, μ is a penalty parameter, Y1 is a Lagrange multiplier, rank(·) represents the rank of the matrix, ‖·‖2 represents the l2 norm of the matrix, ‖·‖ F Represents the F-norm of the matrix.

[0018] Furthermore, in step S3, the method for solving the first sub-optimization problem is: performing a time-frequency domain notch filtering method on the low-rank term A.

[0019] Furthermore, in step S3, the method for solving the second sub-optimization problem is to solve formula (2), specifically:

[0020]

[0021] Furthermore, the step S4 specifically includes:

[0022] Step S401: Initialize algorithm parameters, including: setting P = 0, setting S (p) 、Y1 (P) , μ (p)and A (p) The initial value of , where P represents the number of iterations;

[0023] Step S402: (p) , μ (p) and Y1 (P) Substitute into the first sub-optimization problem and solve the low-rank term A;

[0024] Among them, the low-rank term A is solved by adopting the time-frequency domain notch filtering method, which includes:

[0025] Step S4021: Perform a short-time Fourier transform (STFT) in the distance dimension on each pulse of formula (3) to obtain F i(p) , where i means there are i pulses in total, and after short-time Fourier transform, there are i time-frequency matrices F (p) ;

[0026]

[0027] Step S4022: Setting the threshold T (p) , F i(p) Medium less than T (p) The data is set to 0, and F1 is obtained i(p) ;

[0028] Step S4023: F1 i(p) Perform inverse short-time Fourier transform (ISTFT) in the distance dimension and obtain Z1 (p) ;

[0029] Step S4024: A (p+1) Equal to Z1 (p) ;

[0030] Step S403: Y1 (P) , μ (p) and A (p+1) Substitute it into the second sub-optimization problem, that is, into formula (2), and solve it to obtain the sparse term S (p+1) ;

[0031] Step S404: Execute update Y1 (P+1) , μ (p+1) , the update is performed through the following two formulas:

[0032] Y1 (P+1) =Y1 (P) +μ (p) (RA (p+1) -S (p+1) ) (4)

[0033] μ (p+1) =min{ρμ (p) ,μmax} (5)

[0034] In formula (5), μ max and ρ are fixed values ​​set before the cycle;

[0035] Step S405, execute P=P+1;

[0036] Step S406, loop through steps S402 to S405 until convergence is achieved or the maximum number of iterations is reached, and output the low-rank term A and the sparse term S.

[0037] Furthermore, in step S403, the sparse term S is obtained by solving the following method: (p+1) :

[0038] First, perform compression on the following:

[0039]

[0040] Then, for the obtained K (p) Multiply

[0041] Finally, we get S (p+1) .

[0042] The beneficial effects of the present invention are:

[0043] The present invention can be directly applied to the original data of the SAR system or the SAR image, and can achieve the effect of suppressing weak suppression broadband interference and protecting the real echo in the SAR system. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 A flow chart of a semi-parametric time-frequency domain notch filtering suppression method for weakly suppressing broadband interference in a SAR system provided in Example 1;

[0045] Figure 2 is the Sentinel-1A original image used in Example 1;

[0046] Figure 3 The image is obtained after the method of this embodiment is executed on the Sentinel-1A original image. DETAILED DESCRIPTION

[0047] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0048] Example 1

[0049] See also Figure 1-Figure 3 This embodiment provides a semi-parametric time-frequency domain notch filtering suppression method for weakly suppressing broadband interference in a SAR system. The specific process of the method is as follows: Figure 1 The method specifically comprises the following steps:

[0050] Step S1, obtaining interfered original SAR image data, wherein weakly suppressed broadband interference exists on the original SAR image data;

[0051] Specifically, the original SAR image data of Sentinel-1A is read using MATLAB software. There is suppressed broadband interference in the local area of ​​the image. The original image effect is as follows Figure 2 shown.

[0052] Step S2, performing sparse decomposition on the original SAR image data obtained in step S1 to obtain an overall optimization problem, wherein the overall optimization problem is: solving the interference of low-rank terms and the useful signal of L2 sparse terms by using the augmented Lagrangian method on the original matrix data;

[0053] Specifically, in this embodiment, the overall optimization problem in step S2 is specifically expressed as:

[0054]

[0055] In formula (1), R is the original data of the SAR image, A is the interference in the form of a low-rank term matrix, S is the useful signal in the form of a sparse term matrix, λ is a hyperparameter for adjusting the relationship between the solution results A and S, μ is a penalty parameter, Y1 is a Lagrange multiplier, rank(·) represents the rank of the matrix, ‖·‖2 represents the l2 norm of the matrix, ‖·‖ F represents the F norm of the matrix;

[0056] Step S3, decomposing the total optimization problem obtained in step S2 into two sub-optimization problems, wherein the first sub-optimization problem is to solve the optimization problem of low-rank interference, and the second sub-optimization problem is to solve the optimization problem of sparse useful signals;

[0057] Specifically, in this embodiment, step S3 includes:

[0058] The method to solve the first sub-optimization problem is:

[0059] The low-rank term A is subjected to time-frequency domain notch filtering.

[0060] The method to solve the second sub-optimization problem is:

[0061] Solve formula (2), specifically:

[0062]

[0063] Step S4, performing alternating optimization solutions on the first sub-optimization problem and the second sub-optimization problem obtained in step S3 to obtain a sparse term S and a low-rank term A;

[0064] Specifically, in this embodiment, step S4 includes:

[0065] Step S401: Initialize algorithm parameters, including: setting P = 0, setting S (p) 、Y1 (P) , μ (p) and A (p) The initial value of , where P represents the number of iterations;

[0066] Step S402: (p) , μ (p) and Y1 (P) Substitute into the first sub-optimization problem and solve the low-rank term A;

[0067] More specifically, in this embodiment, the low-rank term A is solved by adopting a time-frequency domain notch filtering method, which includes:

[0068] Step S4021: Perform a short-time Fourier transform (STFT) in the distance dimension on each pulse of formula (3) to obtain F i(p) , where i means there are i pulses in total, and after short-time Fourier transform, there are i time-frequency matrices F (p) ;

[0069]

[0070] Step S4022: Setting the threshold T (p) , F i(p) Medium less than T (p) The data is set to 0, and F1 is obtained i(p) , specifically, the threshold T (p) A suitable value is manually entered for each iteration;

[0071] Step S4023: F1 i(p)Perform inverse short-time Fourier transform (ISTFT) in the distance dimension and obtain Z1 (p) ;

[0072] Step S4024: Set A (p+1) Equal to Z1 (p) ;

[0073] Step S403: Y1 (P) , μ (p) and A (p+1) Substitute it into the second sub-optimization problem, that is, into formula (2), and solve it to obtain the sparse term S (p+1) ;

[0074] More specifically, first, compression processing is performed on the following equation:

[0075]

[0076] Then, for the obtained K (p) Multiply

[0077] Finally, we get S (p+1) .

[0078] Step S404: Execute update Y1 (P+1) , μ (p+1) ;

[0079] Specifically, in this embodiment, the update is performed by the following two formulas:

[0080] Y1 (P+1) =Y1 (P) +μ (p) (RA (p+1) -S (p+1) ) (5)

[0081] μ (p+1) =min{ρμ (p) ,μ max} (6)

[0082] In formula (6), μ max and ρ are fixed values ​​set before the cycle.

[0083] Step S405, execute P=P+1;

[0084] Step S406, loop through steps S402 to S405 until convergence is achieved or the maximum number of iterations is reached, and the low-rank term A and the sparse term S are output. The imaging effect is as follows: Figure 3 shown.

[0085] In summary, the present invention provides a time-frequency domain semi-parametric suppression method for SAR weak broadband interference, designs an algorithm that can suppress weakly suppressed broadband interference, and can be directly applied to the original data of the SAR system or the SAR image, so as to achieve the effect of suppressing weakly suppressed broadband interference in the SAR system and protecting the real echo.

[0086] The matters not described in detail in the present invention are all known technologies to those skilled in the art.

[0087] The preferred specific embodiments of the present invention are described in detail above. It should be understood that a person skilled in the art can make many modifications and changes based on the concept of the present invention without creative work. Therefore, any technical solution that can be obtained by a person skilled in the art through logical analysis, reasoning or limited experiments based on the concept of the present invention on the basis of the prior art should be within the scope of protection determined by the claims.

Claims

1. A semi-parametric time-frequency domain notch filtering method for weakly suppressing broadband interference in a SAR system, characterized in that: The method includes: Step S1, obtaining interfered original SAR image data, wherein weakly suppressed broadband interference exists on the original SAR image data; Step S2: performing sparse decomposition on the original SAR image data obtained in step S1 to obtain an overall optimization problem, wherein the overall optimization problem is: solving the interference of low-rank terms and the useful signal of L2 sparse terms by using the augmented Lagrangian method on the original matrix data; the overall optimization problem is: In formula (1), R is the original data of the SAR image, A is the interference in the form of a low-rank term matrix, S is the useful signal in the form of a sparse term matrix, λ is a hyperparameter for adjusting the relationship between the solution results A and S, μ is a penalty parameter, Y1 is a Lagrange multiplier, rank(·) represents the rank of the matrix, ‖·‖2 represents the l2 norm of the matrix, ‖·‖ F represents the F norm of the matrix; Step S3: Decompose the total optimization problem obtained in step S2 into two sub-optimization problems, wherein the first sub-optimization problem is to solve the optimization problem of low-rank interference, and the second sub-optimization problem is to solve the optimization problem of sparse useful signals; the method for solving the first sub-optimization problem is: performing a time-frequency domain notch filtering method on the low-rank term A; the method for solving the second sub-optimization problem is to solve formula (2), specifically: Step S4: For the first sub-optimization problem and the second sub-optimization problem obtained in step S3, perform alternating optimization solutions to obtain sparse terms and low-rank terms.

2. According to claim 1, a semi-parametric time-frequency domain notch filtering suppression method for weakly suppressing broadband interference in a SAR system is characterized in that: In the step S1, the original SAR image data is matrix data arranged in the range time domain and the azimuth time domain.

3. The semi-parametric time-frequency domain notch filtering suppression method for weakly suppressing broadband interference in a SAR system according to claim 1, characterized in that: The step S4 specifically includes: Step S401: Initialize algorithm parameters, including: setting P = 0, setting S (p) 、Y1 (P) , μ (p) and A (p) The initial value of , where P represents the number of iterations; Step S402: (p) , μ (p) and Y1 (P) Substitute into the first sub-optimization problem and solve the low-rank term A; Among them, the low-rank term A is solved by adopting the time-frequency domain notch filtering method, which includes: Step S4021: Perform a short-time Fourier transform (STFT) in the distance dimension on each pulse of formula (3) to obtain F i(p) , where i means there are i pulses in total, and after short-time Fourier transform, there are i time-frequency matrices F (p) ; Step S4022: Setting the threshold T (p) , F i(p) Medium less than T (p) The data is set to 0, and F1 is obtained i(p) ; Step S4023: F1 i(p) Perform inverse short-time Fourier transform (ISTFT) in the distance dimension and obtain Z1 (p) ; Step S4024: A (p+1) Equal to Z1 (p) ; Step S403: Y1 (P) , μ (p) and A (p+1) Substitute it into the second sub-optimization problem, that is, into formula (2), and solve it to obtain the sparse term S (p+1) ; Step S404: Execute update μ (p+1) , the update is performed through the following two formulas: m (p+1) =min{ρμ (p) ,m max } (5) In formula (5), μ max and ρ are fixed values ​​set before the cycle; Step S405, execute P=P+1; Step S406, loop through steps S402 to S405 until convergence is achieved or the maximum number of iterations is reached, and output the low-rank term A and the sparse term S.

4. The semi-parametric time-frequency domain notch filtering suppression method for weakly suppressing broadband interference in a SAR system according to claim 3, characterized in that: In step S403, the sparse term S is obtained by the following method: (p+1) : First, perform compression on the following: Then, for the obtained K (p) Multiply Finally, we get S (p+1) .

Images