Isar imaging method based on sparse low-rank decomposition and frequency coefficient redistribution

Abstract

This invention discloses an ISAR imaging method based on sparse low-rank decomposition and frequency coefficient redistribution. The method includes: acquiring an initial image; constructing a target region based on the initial image using a sparse low-rank decomposition method; and redistributing the target frequency coefficients based on a RID imaging framework, using the target region as prior knowledge, to obtain the final image. By using this invention, the technical problem of low ISAR imaging quality caused by clutter in existing ISAR imaging methods can be solved, improving target resolution while suppressing clutter. This invention can be widely applied in the field of image processing.

Description

Technical Field

[0001] This invention relates to the field of image processing, and particularly to an ISAR imaging method based on sparse low-rank decomposition and frequency coefficient redistribution. Background Technology

[0002] Synthetic Aperture Radar (SAR) can acquire high-resolution images for Earth observation under all-weather conditions. However, due to target motion and the presence of noise or clutter background, it is difficult to obtain high-quality SAR images of non-cooperative targets. Inverse SAR (ISAR) provides an alternative signal processing method to obtain high-resolution images of non-cooperative targets. SAR obtains the synthetic aperture through the motion of the radar platform while assuming the imaging target is stationary. ISAR, on the other hand, uses target motion for azimuth imaging during the synthetic aperture formation process. To obtain high-quality images of moving targets, we need to cut target image slices around the target and then obtain ISAR data through inverse SAR imaging algorithms. Next, ISAR signal processing methods are used to refocus the moving target in the SAR image. However, clutter appearing in the defocused SAR target image significantly degrades the quality of ISAR imaging. Summary of the Invention

[0003] In view of this, in order to solve the technical problem of low ISAR imaging quality caused by clutter in existing ISAR imaging methods, in a first aspect, the present invention proposes an ISAR imaging method based on sparse low-rank decomposition and frequency coefficient redistribution, the method comprising the following steps:

[0004] Get the initial image;

[0005] Based on the initial image, the target region is constructed using a sparse low-rank decomposition method;

[0006] Based on the RID imaging framework, the target region is used as prior knowledge to redistribute the target frequency fraction coefficients to obtain the final image.

[0007] The initial image is generated by the RD imaging algorithm.

[0008] Optionally, the step of constructing the target region based on the initial image using a sparse low-rank decomposition method specifically includes:

[0009] The problem of separating the defocus component and the strong scattering point region of the target in the initial image is formulated as a low-rank sparse l0 norm minimization problem.

[0010] The low-rank sparse l0 norm minimization problem is solved, and the target region is separated by utilizing the low rank of the defocus component and the sparsity of the target, thus obtaining the target region and the clutter region.

[0011] This optimization step separates the target region from the clutter region, making it easier to subsequently redistribute only the time-frequency representation components belonging to the target.

[0012] Optionally, the low-rank sparse l0 norm minimization problem is expressed by the following formula:

[0013]

[0014] stX=L+S

[0015] Where X represents the initial image, L represents the low-rank part, S represents the sparse part, λ represents the weight between the low-rank part and the sparse part, and ||·||0 represents the l0 norm.

[0016] Optionally, the step of reallocating the target frequency fraction coefficients based on the RID imaging framework, using the target region as prior knowledge, to obtain the final image, specifically includes:

[0017] Based on the RID imaging framework, a frequency coefficient rearrangement operator is constructed using the target region as prior knowledge.

[0018] Based on the frequency coefficient rearrangement operator, the final image is obtained by iteratively redistributing the target frequency fraction coefficients.

[0019] Through this preferred step, a frequency coefficient rearrangement operator is constructed to perform frequency coefficient redistribution on the target region.

[0020] In some embodiments, the frequency coefficient rearrangement operator is represented as follows:

[0021]

[0022]

[0023] Among them, R τ,tar get R represents the target region of the τ-th distance cell. τ,background Let ω1(t,ω) represent the clutter region of the τ-th distance cell, ω0(t,ω) represent the frequency coefficient rearrangement operator, ω0(t,ω) represent the original redistribution operator, t represent the time point, ω represent the preset frequency variable, G represent the result of STFT, and g' represent the first derivative of the window function g.

[0024] This step indicates that:

[0025] If ω0(t,ω) belongs to the target region of the distance cell, then

[0026] If ω0(t,ω) belongs to the clutter region of the range cell, it will not be redistributed, i.e., ω1(t,ω)=ω.

[0027] Optionally, the process of iteratively redistributing the target frequency fraction coefficients is expressed by the following formula:

[0028]

[0029] Where N represents the number of iterations, and η represents the transformed frequency variable.

[0030] In this step, the frequency coefficient rearrangement operator is composited N times, and the expression is rewritten.

[0031] Secondly, this invention also proposes an ISAR imaging system based on sparse low-rank decomposition and frequency coefficient redistribution, the system comprising:

[0032] The image acquisition module is used to acquire the initial image;

[0033] The sparse low-rank decomposition module constructs the target region based on the initial image using a sparse low-rank decomposition method.

[0034] The frequency coefficient redistribution module, based on the RID imaging framework, uses the target region as prior knowledge to redistribute the target frequency coefficients to obtain the final image.

[0035] This invention also proposes an ISAR imaging device based on sparse low-rank decomposition and frequency coefficient redistribution, comprising:

[0036] At least one processor;

[0037] At least one memory for storing at least one program;

[0038] When the at least one program is executed by the at least one processor, the at least one processor implements an ISAR imaging method based on sparse low-rank decomposition and frequency coefficient redistribution as described above.

[0039] Based on the above scheme, this invention provides an ISAR imaging method based on sparse low-rank decomposition and frequency coefficient redistribution. First, sparse low-rank decomposition is used to obtain the target region as prior information for range instantaneous Doppler (RID) imaging. Then, during RID imaging, a target-oriented multisynchrosqueezing transform (TMSST) method is proposed. This time-frequency analysis method is based on synchronous compression transform (SST), which inherits the main advantages of short-time Fourier transform (STFT) and has no cross terms. This method, combined with the RID imaging framework, utilizes the previously extracted target region as prior information and proposes a target frequency coefficient rearrangement operator, which improves target resolution while suppressing clutter, thereby obtaining a high-quality image. Attached Figure Description

[0040] Figure 1 This is a flowchart of the steps of an ISAR imaging method based on sparse low-rank decomposition and frequency coefficient redistribution according to the present invention.

[0041] Figure 2 This is the SST reallocation method in a specific embodiment of the present invention;

[0042] Figure 3 This is a redistribution method for the target-oriented synchronous compression transform (TSST) in a specific embodiment of the present invention;

[0043] Figure 4 This is a schematic diagram of the frequency coefficient redistribution operator of SST in a specific embodiment of the present invention;

[0044] Figure 5 This is a schematic diagram of the frequency coefficient redistribution operator of TSST in a specific embodiment of the present invention;

[0045] Figure 6 This is a simulation experiment in a specific embodiment of the present invention, showing the instantaneous frequency estimation result of the signal of the 120th distance unit. Detailed Implementation

[0046] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0047] It should be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings. Unless otherwise specified, the embodiments and features described in this application can be combined with each other.

[0048] It should be understood that the terms "system," "apparatus," "unit," and / or "module" used in this application are a method of distinguishing different components, elements, parts, sections, or assemblies at different levels. However, if other terms can achieve the same purpose, they may be replaced by other expressions.

[0049] As indicated in this application and claims, unless the context clearly indicates otherwise, the words "a," "an," "a," and / or "the" are not specifically singular and may include the plural. Generally, the terms "comprising" and "including" only indicate the inclusion of expressly identified steps and elements, which do not constitute an exclusive list, and the method or apparatus may also include other steps or elements. An element defined by the phrase "comprising an..." does not exclude the presence of other identical elements in the process, method, product, or apparatus that includes the element.

[0050] In the description of the embodiments of this application, "a plurality of" refers to two or more. The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature.

[0051] Furthermore, flowcharts are used in this application to illustrate the operations performed by the system according to embodiments of this application. It should be understood that the preceding or following operations are not necessarily performed precisely in sequence. Instead, the steps can be processed in reverse order or simultaneously. Additionally, other operations can be added to these processes, or one or more steps can be removed from them.

[0052] Reference Figure 1 The diagram below illustrates an optional example of the ISAR imaging method based on sparse low-rank decomposition and frequency coefficient redistribution proposed in this invention. This method can be applied to computer devices, and the imaging method proposed in this embodiment may include, but is not limited to, the following steps:

[0053] Step S1: Obtain the initial image;

[0054] Step S2: Based on the initial image, construct the target region using a sparse low-rank decomposition method;

[0055] Step S3: Based on the RID imaging framework, using the target region as prior knowledge, the target frequency fraction coefficients are reallocated to obtain the final image.

[0056] In this embodiment, compared to the traditional RID algorithm, we first use the RD imaging algorithm to generate a preliminary focused image, and then use the sparse low-rank decomposition method to decompose the sparse matrix to obtain the focused target region mask. When performing time-frequency analysis on a range-gated signal, we use the TMSST algorithm proposed in this method to obtain the time-frequency representation of each range cell and generate a three-dimensional time-range-Doppler image containing the focused RD image at each slow time. By fixing t = t0, the image at that time can be extracted.

[0057] In cluttered environments, achieving good imaging of ship targets using traditional time-frequency analysis-based RID algorithms remains a challenge. To achieve autofocus on maneuvering targets while suppressing clutter, we propose a target-oriented MSST (TMSSTRID)-based RID imaging method. Compared to traditional RD algorithms, this method is more suitable for imaging irregularly rotating maneuvering targets. Unlike traditional time-frequency analysis-based RID ISAR imaging methods, the TMSSTRID ISAR imaging method can achieve high resolution while enhancing the signal of the target of interest and suppressing clutter interference.

[0058] TF-based RID generates a three-dimensional time-range-Doppler image, instead of the two-dimensional range-Doppler image generated by the Fourier-based RD method. The three-dimensional time-range-Doppler image contains the focused image at each slow time t = t0. Fixing t = t0, we can extract the instantaneous range-Doppler image at that moment. Images at different times contain different target poses. In the RID-based ISAR imaging framework, different TF transform methods can be used. STFT is a linear TF method with lower resolution. WVD-based RID can achieve higher image resolution. However, due to the nonlinearity of the WVD transform, unwanted cross terms significantly degrade image quality. In this method, we use the proposed TMSST-based RID imaging method, which does not introduce cross terms and achieves high resolution through frequency coefficient redistribution. Simultaneously, a novel target-oriented frequency coefficient redistribution operator is designed to enhance the target and suppress clutter. Assigning target frequency coefficients is a process within the TF transform. SST, TSST, and TMSST all belong to a class of TF transform methods. Our method is an improved TF transform.

[0059] The effectiveness of the TMSST-based RID method was verified through simulation experiments and the results of measured data processing.

[0060] In some feasible embodiments, step S2 specifically includes:

[0061] To separate the target region by utilizing the low rank of the defocus component in the imaging result X and the sparsity of the target, the problem of separating the defocus component and the strong scattering point region of the target is formulated as the following joint low-rank sparsity l0 norm minimization problem:

[0062]

[0063] stX=L+S

[0064] Where X represents the initial image, L represents the low-rank component, S represents the sparse component, λ represents the coefficient between the low-rank component and the sparse component, and ||·||0 represents the l0 norm.

[0065] Since the joint recovery problem in the above equation is an NP-hard problem, the l0-norm minimization problem and the rank minimization problem are transformed into the l1-norm minimization problem and the nuclear norm minimization problem, respectively, as given by the following equation:

[0066]

[0067] stX=L+S

[0068] In the formula || || * Let denote the nuclear norm of a matrix, which is the sum of its singular values. Next, we obtain the augmented Lagrangian function:

[0069]

[0070] Where Q is the Lagrange multiplier, <·,·> is the inner product, and ρ is the penalty coefficient. Let F be the norm of the matrix. For simplicity, a scaling matrix Q' = Q / ρ is defined. ADMM optimizes L, S, and Q' alternately by defining the following subproblems:

[0071]

[0072]

[0073] Q'←Q'+L+SX

[0074] The subproblem can be solved using the following iterative method:

[0075] A k =L k +μ1(L k +S k -X+Q k ')

[0076]

[0077] Among them, Ak L represents the intermediate variable in the k-th iteration. k S represents the low-rank component in the k-th iteration. k Let Q' represent the sparse component in the k-th iteration. k Let X = U∑V be the auxiliary scaling matrix for the k-th iteration, μ1 represent the threshold coefficients, D represent the singular value soft thresholding operator, U represent the left singular matrix of X, V represent the right singular matrix of X, H represent the conjugate transpose of the matrix, and M represent the soft thresholding operator X = U∑V. H M τ (∑)=diag(max(0,σ i -τ)).

[0078] The iterative solution is:

[0079] B k =S k +μ2(L k+1 +S k -X+Q k ')

[0080]

[0081] Q' k+1 =Q' k +L k+1 +S k+1 -X

[0082] Among them, B k Let represent the intermediate variable in the k-th iteration, sign represent the sign function, and μ2 represent the threshold coefficient.

[0083] To extract the target region and clutter region of the τ-th range cell, we use the method described above to obtain the target region R. τ,target and clutter background region R τ,background The process involves dividing the image into parts. First, an initial image is generated using the ISARRD method. Then, the defocused portion is removed using the sparse low-rank decomposition method, and the resulting sparse components are used to generate the target mask.

[0084] In some feasible embodiments, step S3 specifically includes:

[0085] S3.1 Based on the RID imaging framework, using the target region as prior knowledge, a frequency coefficient rearrangement operator is constructed;

[0086] For a target with multiple scatterers, the azimuth radar signal within a certain range cell can be viewed as a combination of multiple polynomial phase signal components. Traditional Fourier transform (FT) is not suitable for processing such time-varying phase terms. To illustrate the principle of time-frequency analysis, we first consider a single-component signal model:

[0087]

[0088] Where A(t) represents the instantaneous amplitude, e represents the natural exponent, and j represents the imaginary unit. Indicates the instantaneous phase.

[0089] The STFT representation of signal s(t) is as follows:

[0090]

[0091] Where |G(t,ω)| is the spectrum of the STFT, u represents the time integration variable, g(ut) represents the window function, and s(u) represents the signal to be analyzed. Compared to a conventional STFT, the above equation has a modulation factor e jωt This facilitates redistribution in the frequency direction. The uncertainty principle indicates that there is an irreconcilable contradiction between instantaneous frequency resolution and time resolution. Therefore, the traditional STFT method cannot achieve high-resolution ISAR imaging of moving targets.

[0092] Assume |A'(t)| < ε and If the value is small enough, we can use Taylor expansion at different time points t to expand the signal s(u) into:

[0093]

[0094] The expanded signal is replaced with an STFT, and then we obtain:

[0095]

[0096] In the formula Let g() be the Fourier transform of the window. In the frequency direction, we can see the magnitude of the TF coefficients change with the IF trajectory. It decreases as the value increases. Based on the above expression, we can calculate the derivative of G(t,ω) with respect to time as follows:

[0097]

[0098] Among them, O(A'(t)) and The item was ignored. This represents the ideal frequency trajectory. Therefore, based on this expression, we can obtain the two-dimensional IF estimate of the STFT result:

[0099]

[0100] Re() represents the operation of taking a real number, since the frequency value is a real number. To obtain time-varying TF features more accurately, synchronous compression transform is used to improve the resolution of the TF representation. The expression for SST is written as:

[0101]

[0102] Where δ is the Dirac distribution and ω0(t,ω) is the redistribution operator. Furthermore, for signals in the general case, rather than single-component signals, it is possible to calculate...

[0103]

[0104] Among them G g' (t,ω) is the STFT of the window g' after differentiation. g' represents the derivative of g(t) with respect to time. Thus, ω0(t,ω) can be expressed as:

[0105]

[0106] From a geometric perspective, SST redistributes the STFT spectrum |G(t,ω)| from the (t,ω) point to the newly calculated TF position (t,ω0(t,ω)).

[0107] In the context of RID ISAR imaging, we reconsider the frequency coefficient redistribution process. The radar signal in the azimuth direction of a certain range cell τ is represented as s. τ (t). We use STFT to obtain the time-frequency representation of the signal G(t,ω). In a typical frequency redistribution process, assume the instantaneous frequency distribution at t = t0 is G(t = t0,ω), which is plotted on the plane ω-G(t = t0,ω). The ideal instantaneous frequency is... Due to the uncertainty principle, the frequency coefficients are dispersed between 4 and 6. As can be seen from the plane ω-ω0(t=t0,ω), the frequency redistribution operator maps the frequency coefficients in the 4-6 range to a frequency position of 5. Then, frequency redistribution is performed according to the equation, and the SST result for t=t0 is plotted in the plane ω-SST(t=t0,ω). In the processed result, the frequency coefficients in the 4-6 frequency range are redistributed to the ideal instantaneous frequency. Through this operation, the fuzzy energy of the STFT result can be concentrated in a compact region around the ideal frequency trajectory of each component.

[0108] From the expression, we know that the SST method is used to reassign the original TF representation to the newly calculated TF locations to obtain a sharper result. Ideally, all blurred TF coefficients should be reassigned to the IF trajectory along the frequency direction, and it is clear that the SST reassignment can provide a more focused TF representation than the original STFT spectrum. However, when imaging a target using the RID algorithm, both the background and the target are present in the scene. We only want to reassign the frequency components belonging to the target, not the frequency components of the background clutter.

[0109] We can observe that if the redistribution operator satisfies ω0(t,ω)=ω, then after the redistribution function, the coefficient at (t,ω) remains at (t,ω), equivalent to not being redistributed. This can be expressed as:

[0110]

[0111] Based on this characteristic, in ISAR imaging, we aim to enhance the target's frequency response (TF) representation while suppressing clutter TF representation through frequency coefficient redistribution. During the SST process of redistributing the radar signal's frequency coefficients at the azimuth of a range cell, all instantaneous frequency components are redistributed regardless of whether they belong to the target, meaning that both the target and clutter are focused equally. The SST procedure and corresponding redistribution operators are as follows: Figure 2 As shown, Figure 2 In the middle (a), it indicates the SST redistribution method. Figure 2 In Figure (b), the SST frequency allocation operator is shown at time t = t0. Therefore, a feasible method to improve the signal-to-clutter ratio of ISAR images is to first separate the target region from the clutter region, and then only reallocate the time-frequency representation components belonging to the target. We perform frequency coefficient reallocation on the target region, i.e., if ω0(t,ω) belongs to the target region of the range cell, then... If ω0(t,ω) belongs to the clutter region of the range cell, it will not be redistributed, i.e., ω1(t,ω) = ω. Based on this idea, we propose a novel target-oriented frequency redistribution operator, which is written as:

[0112]

[0113]

[0114] Where R τ,target R represents the target region at the τ-th distance unit. τ,background This represents the clutter region of the i-th distance cell. The process is as follows: Figure 3 As shown, Figure 3 In the middle (a), it indicates the redistribution method of TSST. Figure 3In diagram (b), the TSST frequency assignment operator is shown at time t = t0. To extract the target region and clutter region of the τ-th distance cell, the sparse low-rank decomposition method we previously introduced is used to obtain the target region R. τ,target and clutter background region R τ,background The division.

[0115] We set up a simulation experiment to demonstrate our TSST method. Assume a radar signal s in the azimuth direction at a certain range cell τ. τ (t) consists of three signal components, represented as:

[0116] s(n)=exp[j2π(20nT s +2sin(nT s ))]+2exp[j2π(40nT s +2sin(1.5nT s ))]+exp[j2π(70nT s +2sin(2nT s ))]

[0117] Where Ts = 0.01s represents the sampling interval, the first and third terms represent clutter in the range cell, and the second term represents the target signal. The phase term consists of a linear term and a sinusoidal component representing the three-dimensional oscillation of the target. Instantaneous frequency estimation is performed using an STFT with a Hamming window of length 64. Based on the designed allocation operator, we reallocate the frequency coefficients of the target by utilizing the target frequency region as prior information. To illustrate the frequency coefficient reallocation operator, we respectively... Figure 4 and Figure 5 The frequency coefficient redistribution operators for SST and TSST are plotted. For the SST frequency coefficient redistribution operator, the regions near (t, 50), (t, 110), and (t, 200) are relatively flat, with mapping values ​​of 50, 110, and 200, respectively. This means that the frequency coefficients near the frequency index values ​​(t, 50), (t, 110), and (t, 200) will be redistributed to (t, 50), (t, 110), and (t, 200) after processing according to the equation. For the TSST frequency redistribution operator obtained through the equation, the frequency redistribution operator is flat only at the target time-frequency component (t, 110) and varies linearly with the frequency axis at other frequencies. This means that only the target time-frequency component will be redistributed, while the clutter component will not be redistributed. The SST processing results show that all three time-frequency components are redistributed, and both the target and clutter components are enhanced. The results obtained through TSST processing show that, for the TSST method, only the target component undergoes frequency coefficient redistribution, thereby improving frequency resolution and signal-to-noise ratio.

[0118] S3.2. Based on the frequency coefficient rearrangement operator, the target frequency fraction coefficients are iteratively redistributed to obtain the final image.

[0119] It can be observed that by using a single TSST operation, we can obtain a clearer target TF representation than the STFT result. In this case, there is an incentive to perform another TSST operation on the already obtained TSST result. Therefore, without enhancing the clutter time-frequency representation, we can obtain a clearer target TF result than the TSST result. Then, by iteratively applying multiple TSST operations in a stepwise manner, we can gradually focus on the target time-frequency representation while keeping the clutter time-frequency representation unchanged. Based on this idea, we propose a method called TMSST, which is formulated as follows:

[0120]

[0121] The first TSST result is represented as Ts [1] (t,ω), where N is the number of iterations. The equation above is a simple and direct implementation, requiring only multiple iterations of the TSST operation. However, the frequency coefficient redistribution process represented by the equation will be repeated many times. Based on the properties of the above equation, we can obtain a more efficient implementation method. We will use Ts [1] Substitute (t,ξ) into Ts [2] (t,η), and the result can then be expressed as:

[0122]

[0123] Where ζ represents the frequency integral variable.

[0124] From the above equation, it can be seen that TMSST (N=2) constructs the IF estimate through a quadratic composite of a single TSST frequency redistribution operator to redistribute the fuzzy STFT results. If we further consider Ts... [N-1] Substitute (t,η) into Ts [N] (t,η), we can use more iterations to compute the estimated IF of TMSST. For example, ω1(t,ω1(t,ω1(t,ω1(t,ω))) is the TF estimate of TMSST (N=3), ω1(ω1(t,ω1(t,ω1(t,ω)))) is the IF estimate of TMSST (N=4), and so on. Therefore, we can use the frequency assignment operator of TMSST. (Nth order) is represented as It is composed of N composites of a single TSST frequency redistribution operator, and then the expression for TMSST can be rewritten as:

[0125]

[0126] Therefore, to achieve TMSST, we can construct the IF estimate of TMSST by composing a single TSST frequency redistribution operator, and the frequency coefficient redistribution process is performed only once. Through multiple iterations, the IF estimate of TMSST will get closer and closer to the true IF of the signal. Thus, the energy of the target TF representation can be gradually concentrated, while the clutter time-frequency representation will not be concentrated.

[0127] This invention also provides a simulation experiment:

[0128] A geometric model of a multi-scatterer destroyer was considered to validate the proposed ISAR imagery method. The simulated ship target contained 301 scatterers, with a coordinate spacing of 5 m between each scatterer. The parameters set in Table 1 simulate the motion of the radar carrier and the destroyer itself in sea state 5.

[0129] Table 1 ISAR simulation parameters for ship targets

[0130]

[0131] The target imaging performance of the proposed ISAR-TMSST imaging method was compared with that of the STFT, WVD, SST, and SET methods. In several methods, the sliding window length and FFT points were set to 64 and 256, respectively. Figure 6 The instantaneous frequency estimates for the 120th range cell signal are shown by STFT, WVD, SST, SET, and TMSST. We can see that the STFT results have low TF resolution, while WVD has interference with cross terms. Figure 6 This demonstrates that the proposed method can achieve better frequency focusing performance without interference from cross terms, verifying its effectiveness and superiority.

[0132] An ISAR imaging system based on sparse low-rank decomposition and frequency coefficient redistribution includes:

[0133] The image acquisition module is used to acquire the initial image;

[0134] The sparse low-rank decomposition module constructs the target region based on the initial image using a sparse low-rank decomposition method.

[0135] The frequency coefficient redistribution module, based on the RID imaging framework, uses the target region as prior knowledge to redistribute the target frequency coefficients to obtain the final image.

[0136] The content of the above method embodiments is applicable to this system embodiment. The specific functions implemented in this system embodiment are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved in the above method embodiments.

[0137] An ISAR imaging device based on sparse low-rank decomposition and frequency coefficient redistribution:

[0138] At least one processor;

[0139] At least one memory for storing at least one program;

[0140] When the at least one program is executed by the at least one processor, the at least one processor implements an ISAR imaging method based on sparse low-rank decomposition and frequency coefficient redistribution as described above.

[0141] The content of the above method embodiments is applicable to the device embodiments. The specific functions implemented by the device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0142] A storage medium storing processor-executable instructions, which, when executed by a processor, are used to implement an ISAR imaging method based on sparse low-rank decomposition and frequency coefficient redistribution as described above.

[0143] The content of the above method embodiments is applicable to this storage medium embodiment. The specific functions implemented in this storage medium embodiment are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved in the above method embodiments.

[0144] The above is a detailed description of the preferred embodiments of the present invention. However, the present invention is not limited to the embodiments described. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of the present invention. All such equivalent modifications or substitutions are included within the scope defined by the claims of this application.

Claims

1. An ISAR imaging method based on sparse low-rank decomposition and frequency coefficient redistribution, characterized in that, Includes the following steps: Get the initial image; Based on the initial image, the target region is constructed using a sparse low-rank decomposition method; Based on the RID imaging framework, the target region is used as prior knowledge, and the target frequency fraction coefficients are reallocated to obtain the final image. The step of reassigning target frequency fraction coefficients to obtain the final image based on the RID imaging framework, using the target region as prior knowledge, specifically includes: Based on the RID imaging framework, a frequency coefficient rearrangement operator is constructed using the target region as prior knowledge. Based on the frequency coefficient rearrangement operator, the target frequency fraction coefficients are iteratively redistributed to obtain the final image; The frequency coefficient rearrangement operator is expressed as follows: in, Indicates the first The target area of ​​a distance unit. Indicates the first Clutter region of one range cell, This represents the frequency coefficient rearrangement operator. Let t represent the original redistribution operator, ω represent the time point, ω represent the preset frequency variable, and G represent the result of the STFT. This represents the first derivative of the window function g; The process of iteratively reallocating the target frequency fraction coefficients based on the frequency coefficient rearrangement operator is expressed by the following formula: Where N represents the number of iterations, and η represents the transformed frequency variable.

2. The ISAR imaging based on sparse low-rank decomposition and frequency coefficient redistribution according to claim 1, characterized in that, The step of constructing the target region based on the initial image using a sparse low-rank decomposition method specifically includes: The problem of separating the defocused component and the strong scattering point region of the target in the initial image is formulated as a low-rank sparse problem. Norm minimization problem; For the low-rank sparse The norm minimization problem is solved, and the target region is separated by utilizing the low rank of the defocus component and the sparsity of the target, thus obtaining the target region and the clutter region.

3. The ISAR imaging method based on sparse low-rank decomposition and frequency coefficient redistribution according to claim 2, wherein the low-rank sparse The norm minimization problem is expressed by the following formula: in, X represents the initial image, L represents the low-rank portion, and S represents the sparse portion. This represents the weight between the low-rank and sparse components. express Norm.

4. An ISAR imaging system, characterized in that, A method for performing the sparse low-rank decomposition and frequency coefficient redistribution method as described in claim 1 includes: The image acquisition module is used to acquire the initial image; The sparse low-rank decomposition module constructs the target region based on the initial image using a sparse low-rank decomposition method. The frequency coefficient redistribution module, based on the RID imaging framework, uses the target region as prior knowledge to redistribute the target frequency coefficients to obtain the final image.

5. An ISAR imaging device based on sparse low-rank decomposition and frequency coefficient redistribution, characterized in that, include: At least one processor; At least one memory for storing at least one program; When the at least one program is executed by the at least one processor, the at least one processor implements an ISAR imaging method based on sparse low-rank decomposition and frequency coefficient redistribution as described in any one of claims 1-3.

Images