An airborne scanning radar forward-looking super-resolution fast imaging method and system

Abstract

This invention belongs to the field of airborne scanning radar imaging technology and discloses a method and system for rapid forward-looking super-resolution imaging of airborne scanning radar. The proposed method first establishes an echo model of the airborne forward-looking imaging radar. Then, based on the echo model, a unified two-dimensional Doppler convolution model is established using the Doppler centroid compensation method, and this model is converted from a two-dimensional convolution form to a one-dimensional vector form. Finally, based on the one-dimensional vector Doppler convolution model, a sparse Bayesian learning method based on alternating direction multipliers is used to reconstruct the target scattering information, thereby achieving forward-looking super-resolution imaging of the airborne scanning radar. This method can obtain high-resolution imaging in a short time, thus realizing forward-looking super-resolution imaging of airborne scanning radar. Furthermore, the process design is simple and clear, making it suitable for widespread application in environments where airborne scanning radar performs forward-looking imaging of sparse targets.

Description

Technical Field

[0001] This invention belongs to the field of airborne scanning radar imaging technology, specifically relating to an airborne scanning radar forward-looking super-resolution rapid imaging method and system. Background Technology

[0002] Airborne scanning radar systems have become a key technology for forward-looking imaging, offering significant advantages in continuous, wide-area observation under arbitrary imaging geometries. Unlike traditional synthetic aperture radar (SAR), scanning forward-looking radar can image the area in front of the platform in real time, which is crucial for applications such as autonomous navigation and disaster monitoring. However, achieving high-resolution forward-looking imaging remains a challenge due to the limitation of radar antenna size on azimuth resolution.

[0003] To obtain high-resolution forward-looking images, super-resolution imaging methods have been extensively studied to improve azimuth resolution. Related research mainly focuses on three aspects: optimization of the forward-looking imaging model, improvement of parameter estimation accuracy, and optimization of computational efficiency. In the super-resolution framework, when the platform is stationary or moving at low speed, the radar echo is modeled as a convolution of the antenna pattern and the target scattering distribution, in which case the Doppler phase is ignored. However, when the platform speed is high, ignoring the Doppler phase leads to convolution model mismatch, thereby reducing imaging performance.

[0004] Due to the low-pass characteristics of antenna patterns, deconvolution can be considered an ill-posed process. Parameter estimation methods aim to alleviate this ill-posedness problem, and existing methods are mainly divided into three categories: regularization methods, array signal methods, and Bayesian methods. Among them, Bayesian methods alleviate the ill-posedness problem by introducing prior information about the target and noise. However, most super-resolution imaging methods require calculating the matrix inverse, resulting in a computational complexity of O(n log n). ,in Let be the azimuth length of the echo signal. When the echo signal size is large, matrix inversion requires more computational resources. Furthermore, existing methods typically require reconstructing the forward-looking scene by processing each distance cell, which further increases computational costs. Summary of the Invention

[0005] The purpose of this invention is to propose a forward-looking super-resolution fast imaging method for airborne scanning radar. This method is based on sparse Bayesian learning and alternating direction multiplier method, and can obtain high-resolution imaging in a short time.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: A forward-looking super-resolution rapid imaging method for airborne scanning radar includes the following steps: Step 1. Establish the echo model of the airborne forward-looking imaging radar; Step 2. Based on the echo model of the airborne forward-looking imaging radar established in Step 1, a unified two-dimensional Doppler convolution model is established through the Doppler centroid compensation method, and it is converted from two-dimensional convolution form to one-dimensional vector form. Step 3. Based on the one-dimensional vector form of the Doppler convolution model obtained in Step 2, the target scattering information is reconstructed using the sparse Bayesian learning (SBL) method based on the alternating direction multiplier method, so as to realize forward-looking super-resolution imaging of airborne scanning radar. The noise is modeled using a complex Gaussian distribution that follows a gamma distribution, and the target scattering coefficient is modeled using a layered Laplace distribution. The imaging parameters are estimated using the expectation-maximization method, and the computational complexity is reduced by a one-dimensional inverse-free estimation method based on split Bregman and singular value decomposition. The two-dimensional form of the SBL method is derived through matrix transformation to reduce memory usage.

[0007] Furthermore, based on the forward-looking super-resolution rapid imaging method for airborne scanning radar, this invention also proposes a corresponding forward-looking super-resolution rapid imaging system for airborne scanning radar, the technical solution of which is as follows: An airborne scanning radar forward-looking super-resolution rapid imaging system includes: The echo model building module is used to build the echo model of the airborne forward-looking imaging radar. The Doppler convolution model building module is used to establish a unified two-dimensional Doppler convolution model based on the echo model of airborne forward-looking imaging radar, and convert it from two-dimensional convolution form to one-dimensional vector form through the Doppler centroid compensation method. And a target scattering information reconstruction module, which is used to reconstruct target scattering information based on a one-dimensional vector form of Doppler convolution model and adopts the sparse Bayesian learning SBL method based on the alternating direction multiplier method to realize forward-looking super-resolution imaging of airborne scanning radar. The noise is modeled using a complex Gaussian distribution that follows a gamma distribution, and the target scattering coefficient is modeled using a layered Laplace distribution. The imaging parameters are estimated using the expectation-maximization method, and the computational complexity is reduced by a one-dimensional inverse-free estimation method based on split Bregman and singular value decomposition. The two-dimensional form of the SBL method is derived through matrix transformation to reduce memory usage.

[0008] Furthermore, based on the aforementioned forward-looking super-resolution rapid imaging method for airborne scanning radar, this invention also proposes a computer device, which includes a memory and one or more processors. The memory stores executable code, and when the processor executes the executable code, it implements the steps of the airborne scanning radar forward-looking super-resolution rapid imaging method described above.

[0009] Furthermore, based on the aforementioned forward-looking super-resolution rapid imaging method for airborne scanning radar, this invention also proposes a computer-readable storage medium storing a program thereon; when executed by a processor, this program is used to implement the steps of the aforementioned forward-looking super-resolution rapid imaging method for airborne scanning radar.

[0010] The present invention has the following advantages: As described above, this invention discloses a fast forward-looking super-resolution imaging method for airborne scanning radar. This method addresses the problems of high computational complexity, large memory footprint, and poor real-time performance in traditional airborne scanning radar forward-looking imaging methods. It employs a sparse Bayesian learning method based on the alternating direction multiplier method to reconstruct target scattering information, achieving efficient forward-looking super-resolution imaging for airborne scanning radar. This method boasts extremely high real-time performance, eliminates the need for inversion during iteration, and implements a two-dimensional method, significantly saving storage space. The process design is simple and clear, making it suitable for widespread application in environments where airborne scanning radar performs forward-looking imaging of sparse targets.

[0011] Specifically, traditional methods based on sparse Bayesian imaging require direct inversion of high-dimensional matrices to solve for target scattering coefficients, resulting in high computational complexity and difficulty in meeting the real-time requirements of airborne platforms. The method of this invention decomposes the high-dimensional convex optimization problem into multiple low-dimensional subproblems using the split Bregman iterative algorithm and combines it with singular value decomposition to achieve inversion-free estimation. The iterative process avoids large-scale matrix inversion operations, significantly reducing computational complexity and improving algorithm efficiency, thereby enhancing the real-time performance of forward-looking imaging by airborne scanning radar.

[0012] Furthermore, the method of this invention derives and implements the SBL algorithm in two-dimensional matrix form through matrix transformation, which can avoid repeated conversion between vectorization and matrixization, thereby reducing the memory overhead caused by vectorization, saving storage space, and better adapting to the hardware resource limitations of airborne platforms.

[0013] Furthermore, the method of this invention models noise using a complex Gaussian distribution with precision following a gamma distribution, models the target scattering coefficient using a layered Laplace distribution, and estimates imaging parameters using the expectation-maximization method. This allows for robust recovery of the target's super-resolution scattering information while significantly reducing computational complexity.

[0014] The method steps of this invention are clearly designed, and the entire process from echo modeling, Doppler compensation to sparse reconstruction is logically coherent. It does not require complicated additional processing steps, making it easy to implement on existing airborne radar platforms and suitable for promotion in environments where airborne scanning radar performs forward-looking imaging of sparse targets. Attached Figure Description

[0015] Figure 1 This is a flowchart of the forward-looking super-resolution fast imaging method for airborne scanning radar in an embodiment of the present invention.

[0016] Figure 2 This is a scene distribution diagram for point target simulation in an embodiment of the present invention.

[0017] Figure 3 This is a schematic diagram of the simulated signal echo when the signal-to-noise ratio (SNR) is 15dB in an embodiment of the present invention.

[0018] Figure 4 This is a schematic diagram of the imaging results obtained by using the method of the present invention in an embodiment of the present invention. Detailed Implementation

[0019] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments: Example 1 This embodiment specifically proposes a two-dimensional forward-looking super-resolution imaging method for airborne scanning radar based on the Sparse Bayesian Learning (SBL) method and the Alternating Direction Multiplier Method (ADMM). This method can obtain high-resolution imaging in a short time.

[0020] This invention establishes a unified two-dimensional Doppler convolution model for airborne radar using a Doppler centroid compensation method. To improve azimuth resolution, the invention further transforms the Doppler convolution model from a two-dimensional form to a one-dimensional form and employs the SBL method to reconstruct target scattering information. In the SBL method, noise is modeled using a complex Gamma-Gaussian distribution, meaning the noise is modeled using a complex Gaussian distribution with gamma accuracy, and the target is modeled using a layered Laplace distribution. Simultaneously, this invention uses the Expectation-Maximization (EM) method to estimate imaging parameters. Direct differentiation methods require calculating matrix inverses, which increases computational complexity. To reduce computational complexity, this invention also employs a 1D inverse-free estimation method based on split Bregman and singular value decomposition, i.e., a one-dimensional inverse-free estimation method based on split Bregman and singular value decomposition to reduce computational complexity. However, the one-dimensional form of parameter estimation also consumes more storage space. To avoid memory waste, this invention also derives the two-dimensional form of the SBL method through matrix transformation.

[0021] The method of the present invention will be described in detail below.

[0022] like Figure 1 As shown, a forward-looking super-resolution fast imaging method for airborne scanning radar specifically includes the following steps: Step 1. Establish the echo model of the airborne forward-looking imaging radar.

[0023] In this embodiment, step 1 specifically includes: Radar platform, i.e., airborne scanning radar along Direction at a constant speed It flies, emitting electromagnetic waves to scan the area in front, with a pitch angle of... The azimuth angle, i.e., the frontal angle, is The aircraft's altitude is ,Target Slant range with radar Approximately: (1) in, Time is the horizontal distance dimension. For radar and targets The initial distance between them.

[0024] Assume the radar transmits a linear frequency modulated (LFM) signal. : (2) in, For window functions, For time in the distance dimension, The duration of the pulse. The imaginary unit, To adjust the frequency, For carrier frequency.

[0025] Subsequently, pulse compression technology was introduced to achieve high resolution in the range dimension. Due to the slant range... and , Correlation can cause distance migration, thus requiring distance migration correction. After pulse compression and distance migration correction, the echo signal... for: (3) in, The target scattering coefficient, This is the antenna pattern modulation function. Indicates the start time. At the speed of light, For signal bandwidth, The wavelength is given; the exponential term on the right side of formula (3) is the Doppler frequency shift: (4) Among them, the Doppler centroid for: (5) In terms of orientation, the first The echo signal of each distance cell is written as: (6) The echo model of the airborne forward-looking imaging radar is established as shown in formula (6).

[0026] in, For the first The echo vector of each distance cell, , The length of the echo signal; For the target vector, ; For imaging noise, ; For the first Doppler convolution matrix of distance units, , This represents the Shul-Adama product.

[0027] and The antenna pattern vector and Doppler phase vector are transformed respectively, as shown in formulas (7) and (8): (7) (8) in, This is the antenna pattern vector. , to They represent The The first to the second One portion, Indicates the dimension of the convolution kernel; Indicates rounding up; These represent the frontal viewpoints of sampling points in different orientations. For the first The pitch angle of each distance unit. and Indicates Doppler frequency shift, This represents the pulse repetition frequency. As can be seen from formula (8), due to the pitch angle of different distance units... Unlike others, the Doppler center of mass exhibits spatial variation.

[0028] Step 2. Based on the echo model of the airborne forward-looking imaging radar established in Step 1, a unified two-dimensional Doppler convolution model is established through the Doppler centroid compensation method, and it is converted from two-dimensional convolution form to one-dimensional vector form.

[0029] The Doppler centroid frequency varies along the distance unit, exhibiting significant spatial variation, which leads to the Doppler convolution matrix... Differences exist across different range cells. During forward-looking imaging, the Doppler convolution matrix needs to be calculated in each range cell, which increases computational complexity. To address this issue, this invention proposes a Doppler centroid compensation method, taking into account the characteristics of forward-looking radar imaging.

[0030] In this embodiment, step 2 specifically includes: In the forward-looking imaging Doppler convolution model, the Doppler compensation matrix is ​​defined. : (9) in, , This indicates the distance to the sampling points of the echo.

[0031] (10) (11) (12) in, This indicates the compensated front view angle. This is the maximum forward field of view.

[0032] For echo signal matrix With Doppler compensation matrix Perform the Shur-Adama product operation, i.e. .

[0033] in, , to They represent the first The first to the second The echo vector of each distance cell.

[0034] In the In a distance cell, the Shul-Adama product is represented as: (13) in, Indicates the first The compensated echo vector in each range cell, ; This indicates the imaging noise after Doppler compensation. Let represent the compensated Doppler convolution matrix, and ,in .

[0035] Since the forward field of view is usually no greater than Right now The cosine value of the front viewpoint changes slowly. Therefore, it is assumed that... Approximately equal to ,in For the first The frontal view angle in each azimuth unit. .

[0036] Compensated Doppler matrix elements Represented as: (14) in Approximately: (15) The compensated convolution matrix remains unchanged in each range cell and is no longer related to the pitch angle, thus avoiding the need to calculate the Doppler convolution matrix in each range cell.

[0037] The two-dimensional Doppler convolution model established by the Doppler centroid compensation method is represented as follows: (16) in, Represents the convolution matrix. The two-dimensional target scattering coefficient matrix, It is noise.

[0038] To simplify the description of the statistical distribution of the target and noise, formula (16) is converted from a two-dimensional convolution form to a one-dimensional vector form, resulting in: (17) in, to These represent the echo signal matrices respectively. Each item; to These represent the two-dimensional target scattering coefficient matrices respectively. Each item; to Representing noise respectively Each item.

[0039] Formula (17) can be simplified as follows: (18) in, Indicates the echo signal. That is, a two-dimensional echo signal matrix The one-dimensional echo signal vector obtained after vectorization; The vectorized convolution matrix is ​​the compensated Doppler convolution matrix. , for The identity matrix.

[0040] Step 3. Based on the one-dimensional vector form of the Doppler convolution model obtained in Step 2, the target scattering information is reconstructed using the sparse Bayesian learning (SBL) method based on the alternating direction multiplier method, so as to realize forward-looking super-resolution imaging of airborne scanning radar.

[0041] The noise is modeled using a complex Gaussian distribution that follows a gamma distribution, and the target scattering coefficient is modeled using a layered Laplace distribution. The imaging parameters are estimated using the expectation-maximization method, and the computational complexity is reduced by a one-dimensional inverse-free estimation method based on split Bregman and singular value decomposition. The two-dimensional form of the SBL method is derived through matrix transformation to reduce memory usage.

[0042] Specifically, forward-looking imaging requires the scattering coefficient of the target. An estimation is performed. The method of this invention compensates for the spatially varying characteristics of the Doppler centroid and compresses the data dimension of the compensated echo signal to a lower level before using a sparse Bayesian learning method to obtain the front view image. In step 3 of this embodiment, the problem is further divided into sub-problems based on the split Bregman method, followed by matrix transformation based on singular value decomposition.

[0043] The following section introduces the Bayesian prior.

[0044] In step 3 of this embodiment, the process of modeling noise using a complex Gaussian distribution with precision following a gamma distribution includes: Assuming imaging noise It is complex Gaussian noise, which has the characteristics of independence, zero mean, and follows a complex Gaussian distribution. : (19) in, Indicates a complex Gaussian distribution. The precision of representing imaging noise, Represents the identity matrix; Indicates imaging noise Follow the mean The covariance matrix is The complex Gaussian distribution.

[0045] Assumption accuracy Follows a gamma distribution : (20) in, Indicates the gamma distribution; and The scaling parameter is... and Usually a smaller value is chosen so that... The prior is a no-information prior; Precision Subject to scaling parameter and The gamma distribution.

[0046] The echo signal is obtained from formula (18). Distribution : (twenty one) in, Represents the identity matrix; Indicates echo signal Follow the mean The covariance matrix is The complex Gaussian distribution.

[0047] The following section introduces the EM solution.

[0048] In step 3 of this embodiment, the process of estimating imaging parameters using the expectation-maximization method, i.e., the EM-MAP algorithm, includes:

[0049] Expected E-step: The function expression is as follows: (twenty two) in, Represents the target scattering coefficient The function, This indicates the operation of calculating the mean. Describes the 2-norm. Indicates the scale parameter. , Represents the target scattering coefficient The item, Indicates a constant.

[0050] Maximize the M-step: Estimate the target scattering coefficient using formula (23) : (twenty three) in, This represents the solution that maximizes the function. The value of , Represents the target scattering coefficient In the The estimated value in the next iteration Indicates the first In the next iteration function.

[0051] Solve ,in Representing the gradient operator, we obtain: (twenty four) in, express The conjugate transpose of .

[0052] (25) in, It is a constant.

[0053] The following section introduces the split Bregman method.

[0054] Target scattering coefficient In the estimation, it is necessary to calculate the matrix. The inverse matrix. If Doppler center compensation and dimensionality reduction are not performed, and sparse Bayesian learning methods are directly applied (i.e., high-dimensional sparse Bayesian learning, HDSBL), then the matrix needs to be calculated. The inverse of the matrix. Although the signal dimension is compressed, the computational complexity of calculating the inverse matrix remains unchanged, still O(n log n). Therefore, this invention introduces the split Bregman method, namely the split Bregman iterative algorithm, to avoid calculating the inverse matrix.

[0055] In step 3 of this embodiment, the process of reducing computational complexity using the one-dimensional inverse-free estimation method based on split Bregman and singular value decomposition is as follows: Solve the following convex optimization problem using the split Bregman iterative algorithm: (26) in, Represents the target scattering coefficient The estimated value; and Let f(x) denote a convex function defined on a Hilbert space, and possibly not differentiable. Indicates in Find the maximum value of the function among all possible values.

[0056] make , .

[0057] Since both terms of the formula depend on variables Therefore, variables are introduced. : (27) Formula (27) is subject to ,in Representing variables The item.

[0058] The constrained problem shown in equation (27) is transformed into the unconstrained problem shown in equation (28): (28) in, It is a positive parameter.

[0059] The split Bregman iterative algorithm simplifies to the following form: (29) (30) in, Indicates the number of iterations. Represents the target scattering coefficient No. The result of the iteration Representing variables No. The result of the iteration Indicates in and Find the maximum value of the function among all possible values. Representing variables No. The result of the iteration Indicates imaging noise No. The result of the iteration Indicates imaging noise No. The result of the next iteration.

[0060] The split Bregman iterative algorithm shown in formulas (29) and (30) contains , , Three variables, which can be broken down into three subproblems. question, question, The problem is to be solved.

[0061] Solve The problem is fixed. , The objective function of the problem is: (31) in, The solution is Represented as: (32) in, Represents the identity matrix.

[0062] Solve The problem is fixed. , The objective function is: (33) in, Indicates in Find the maximum value of the function among all possible values.

[0063] The solution is Represented as: (34) in: (35) Solve question, The solution is As shown in formula (30).

[0064] A one-dimensional inverse-free estimation method using singular value decomposition is employed to estimate the compensated Doppler convolution matrix. Decomposed into: (36) in, and It is a unitary matrix. express The conjugate transpose of; , to Represent matrices respectively The To the item; , to Represent matrices respectively The To the item; for The singular value matrix, , to Represent matrices respectively The To the item.

[0065] Therefore, formula (32) can be written as: (37) The following section describes the conversion to a two-dimensional method.

[0066] In step 3 of this embodiment, the process of deriving the two-dimensional form of the SBL method through matrix transformation includes: To reduce memory usage, the present invention also proposes a two-dimensional parameter estimation algorithm, namely the two-dimensional method. For matrix The vectorized form of formula (37) is rewritten as: (38) in, Indicates by A matrix consisting of diagonal elements whose size and position are similar to... Correspondingly; The dimension is A matrix in which all elements are 1; and They are variables and imaging noise In matrix form, Representation matrix No. The result of the iteration Representation matrix No. The result of the iteration; the target scattering coefficient is estimated in the form of a two-dimensional matrix.

[0067] The other parameters are represented in two-dimensional matrix form as follows: (39) (40) (41) (42) in, .

[0068] Representation matrix No. The result of the iteration Representation matrix No. The result of the iteration scale parameter In matrix form, Representation matrix No. The result of the iteration and Hyperparameters representing scale parameters.

[0069] Furthermore, to verify the effectiveness of the method proposed in this invention, the following specific experiments are presented. The system parameters for the point target simulation are shown in Table 1.

[0070] Table 1 System parameters for point target simulation

[0071] like Figure 2As shown, 11 point scattering targets with the same amplitude are arranged in the scene. The nearest adjacent target in the azimuth dimension is located at 3050 meters, with an angular interval of 0.8°. Table 1 lists the simulation parameters, and the size of the imaging scene is 300 (range direction) × 400 (azimuth direction).

[0072] Simulated signal echo at a signal-to-noise ratio (SNR) of 15 dB is as follows Figure 3 As shown, after antenna pattern modulation, the target appears to broaden in different azimuth cells; and due to the limitation of the main lobe beamwidth, some adjacent targets cannot be directly distinguished. Figure 4 Imaging results using the LDSBL-DC method (the method of this invention) are shown, demonstrating good reconstruction results. This indicates that the LDSBL-DC method is an effective super-resolution imaging method.

[0073] Example 2 This embodiment 2 describes an airborne scanning radar forward-looking super-resolution fast imaging system, which is based on the same inventive concept as the airborne scanning radar forward-looking super-resolution fast imaging method in embodiment 1.

[0074] Specifically, this airborne scanning radar forward-looking super-resolution rapid imaging system includes the following modules: The echo model building module is used to build the echo model of the airborne forward-looking imaging radar.

[0075] The Doppler convolution model building module is used to establish a unified two-dimensional Doppler convolution model based on the echo model of airborne forward-looking imaging radar, and convert it from two-dimensional convolution form to one-dimensional vector form.

[0076] And a target scattering information reconstruction module, which is used to reconstruct target scattering information based on a one-dimensional vector form of Doppler convolution model and adopts the sparse Bayesian learning (SBL) method based on alternating direction multipliers to achieve forward-looking super-resolution imaging of airborne scanning radar.

[0077] The noise is modeled using a complex Gaussian distribution that follows a gamma distribution, and the target scattering coefficient is modeled using a layered Laplace distribution. The imaging parameters are estimated using the expectation-maximization method, and the computational complexity is reduced by a one-dimensional inverse-free estimation method based on split Bregman and singular value decomposition. The two-dimensional form of the SBL method is derived through matrix transformation to reduce memory usage.

[0078] It should be noted that the implementation process of the functions and roles of each functional module in the forward-looking super-resolution rapid imaging system of airborne scanning radar is detailed in the implementation process of the corresponding steps in the method of Example 1, and will not be repeated here.

[0079] Example 3 This embodiment 3 describes a computer device that includes a memory and one or more processors.

[0080] The memory stores executable code, which, when executed by the processor, is used to implement the steps of the airborne scanning radar forward-looking super-resolution rapid imaging method in Embodiment 1 above.

[0081] In this embodiment, the computer device can be any device or apparatus with data processing capabilities, and will not be described in detail here.

[0082] Example 4 This embodiment 4 describes a computer-readable storage medium storing a program that, when executed by a processor, implements the steps of a forward-looking super-resolution fast imaging method for airborne scanning radar.

[0083] The computer-readable storage medium can be an internal storage unit of any device or apparatus with data processing capabilities, such as a hard disk or memory, or an external storage device of any device with data processing capabilities, such as a plug-in hard disk, smart media card (SMC), SD card, flash card, etc.

[0084] Of course, the above description is only a preferred embodiment of the present invention. The present invention is not limited to the above-described embodiments. It should be noted that any equivalent substitutions or obvious modifications made by those skilled in the art under the guidance of this specification fall within the scope of this specification and should be protected by the present invention.

Claims

1. A forward-looking super-resolution rapid imaging method for airborne scanning radar, characterized in that, Includes the following steps: Step 1. Establish the echo model of the airborne forward-looking imaging radar; Step 2. Based on the echo model of the airborne forward-looking imaging radar established in Step 1, a unified two-dimensional Doppler convolution model is established through the Doppler centroid compensation method, and it is converted from two-dimensional convolution form to one-dimensional vector form. Step 3. Based on the one-dimensional vector form of the Doppler convolution model obtained in Step 2, the target scattering information is reconstructed using the sparse Bayesian learning (SBL) method based on the alternating direction multiplier method, so as to realize forward-looking super-resolution imaging of airborne scanning radar. The noise is modeled using a complex Gaussian distribution that follows a gamma distribution, and the target scattering coefficient is modeled using a layered Laplace distribution. The imaging parameters are estimated using the expectation-maximization method, and the computational complexity is reduced by a one-dimensional inverse-free estimation method based on split Bregman and singular value decomposition. The two-dimensional form of the SBL method is derived through matrix transformation to reduce memory usage.

2. The airborne scanning radar forward-looking super-resolution rapid imaging method according to claim 1, characterized in that, Step 1 specifically involves: Radar platform, i.e., airborne scanning radar along Direction at a constant speed It flies, emitting electromagnetic waves to scan the area in front, with a pitch angle of... The azimuth angle, i.e., the frontal angle, is The aircraft's altitude is ,Target Slant range with radar Approximately: (1) in, Time is the horizontal distance dimension. For radar and targets The initial distance between them; Assume the radar transmits a linear frequency modulated signal. : (2) in, For window functions, For time in the distance dimension, The duration of the pulse. The imaginary unit, To adjust the frequency, For carrier frequency; After pulse compression and range migration correction, the echo signal for: (3) in, The target scattering coefficient, This is the antenna pattern modulation function. Indicates the start time. At the speed of light, For signal bandwidth, The wavelength is given; the exponential term on the right side of formula (3) is the Doppler frequency shift: (4) Among them, the Doppler centroid for: (5) In terms of orientation, the first The echo signal of each distance cell is written as: (6) That is, the echo model of the airborne forward-looking imaging radar is established as shown in formula (6); in, For the first The echo vector of each distance cell, , The length of the echo signal; For the target vector, ; For imaging noise, ; For the first Doppler convolution matrix of distance units, , This represents the Shul-Adama product; and The antenna pattern vector and Doppler phase vector are transformed respectively, as shown in formulas (7) and (8): (7) (8) in, This is the antenna pattern vector. , to They represent The The first to the second One portion, Indicates the dimension of the convolution kernel; Indicates rounding up; These represent the frontal viewpoints of sampling points in different orientations. For the first The pitch angle of each distance unit. and Indicates Doppler frequency shift, This indicates the pulse repetition frequency.

3. The airborne scanning radar forward-looking super-resolution rapid imaging method according to claim 2, characterized in that, Step 2 specifically involves: In the forward-looking imaging Doppler convolution model, the Doppler compensation matrix is ​​defined. : (9) in, , Indicates the number of distance sampling points for the echo; (10) (11) (12) in, This indicates the compensated front view angle. Maximum forward field of view; For echo signal matrix With Doppler compensation matrix Perform the Shur-Adama product operation, i.e. ; in, , to They represent the first The first to the second echo vector of each distance cell; In the In a distance cell, the Shul-Adama product is represented as: (13) in, Indicates the first The compensated echo vector in each range cell, ; This indicates the imaging noise after Doppler compensation. Let represent the compensated Doppler convolution matrix, and ,in ; Assumption Approximately equal to ,in For the first The frontal view angle in each azimuth unit. ; Compensated Doppler matrix elements Represented as: (14) in Approximately: (15) The two-dimensional Doppler convolution model established by the Doppler centroid compensation method is represented as follows: (16) in, Represents the convolution matrix. The two-dimensional target scattering coefficient matrix, For noise; Converting formula (16) from a two-dimensional convolution form to a one-dimensional vector form, we get: (17) in, to These represent the echo signal matrices respectively. Each item; to These represent the two-dimensional target scattering coefficient matrices respectively. Each item; to Representing noise respectively Each item; Formula (17) can be simplified as follows: (18) in, Indicates the echo signal. That is, a two-dimensional echo signal matrix The one-dimensional echo signal vector obtained after vectorization; The vectorized convolution matrix is ​​the compensated Doppler convolution matrix. , for The identity matrix.

4. The airborne scanning radar forward-looking super-resolution rapid imaging method according to claim 3, characterized in that, Step 3, the process of modeling noise using a complex Gaussian distribution with precision following a gamma distribution, includes: Assuming imaging noise It is complex Gaussian noise and follows a complex Gaussian distribution. : (19) in, Indicates a complex Gaussian distribution. The precision of representing imaging noise, Represents the identity matrix; Indicates imaging noise Follow the mean The covariance matrix is The complex Gaussian distribution; Assumption accuracy Follows a gamma distribution : (20) in, Represents the gamma distribution. and For scaling parameters; Precision Subject to scaling parameter and The gamma distribution; The echo signal is obtained from formula (18). Distribution : (21) in, Represents the identity matrix; Indicates echo signal Follow the mean The covariance matrix is The complex Gaussian distribution.

5. The airborne scanning radar forward-looking super-resolution rapid imaging method according to claim 4, characterized in that, Step 3, the process of estimating imaging parameters using the expectation-maximization method, i.e., the EM-MAP algorithm, includes: The function expression is as follows: (22) in, Represents the target scattering coefficient The function, This indicates the operation of calculating the mean. Describes the 2-norm. Indicates the scale parameter. , Represents the target scattering coefficient The item, Indicates a constant; Estimate the target scattering coefficient using formula (23) : (23) in, This represents the solution that maximizes the function. The value of , Represents the target scattering coefficient In the The estimated value in the next iteration Indicates the first In the next iteration function; Solve ,in Representing the gradient operator, we obtain: (24) in, express The conjugate transpose of; (25) in, It is a constant.

6. The airborne scanning radar forward-looking super-resolution rapid imaging method according to claim 5, characterized in that, In step 3, the process of reducing computational complexity using the one-dimensional inverse-free estimation method based on split Bregman and singular value decomposition is as follows: Solve the following convex optimization problem using the split Bregman iterative algorithm: (26) in, Represents the target scattering coefficient The estimated value; and Denotes a convex function defined on a Hilbert space; Indicates in Find the maximum value of the function among all possible values. make , ; Introducing variables : (27) Formula (27) is subject to ,in Representing variables The item; The constrained problem shown in equation (27) is transformed into the unconstrained problem shown in equation (28): (28) in, It is a positive parameter; The split Bregman iterative algorithm simplifies to the following form: (29) (30) in, Indicates the number of iterations. Represents the target scattering coefficient No. The result of the iteration Representing variables No. The result of the iteration Indicates in and Find the maximum value of the function among all possible values. Representing variables No. The result of the iteration Indicates imaging noise No. The result of the iteration Indicates imaging noise No. The result of the next iteration; The split Bregman iterative algorithm shown in formulas (29) and (30) contains , , Three variables, which can be broken down into three subproblems. question, question, Solve the problem; Solve The problem is fixed. , The objective function of the problem is: (31) in, The solution is Represented as: (32) in, Represents the identity matrix; Solve The problem is fixed. , The objective function is: (33) in, Indicates in Find the maximum value of the function among all possible values. The solution is Represented as: (34) in: (35) Solve question, The solution is As shown in formula (30); A one-dimensional inverse-free estimation method using singular value decomposition is employed to estimate the compensated Doppler convolution matrix. Decomposed into: (36) in, and It is a unitary matrix. express The conjugate transpose of; , to Represent matrices respectively The To the item; , to Represent matrices respectively The To the item; for The singular value matrix, , to Represent matrices respectively The To the item; Therefore, formula (32) can be written as: (37)。 7. The airborne scanning radar forward-looking super-resolution fast imaging method according to claim 6, characterized in that, Step 3, the process of deriving the two-dimensional form of the SBL method through matrix transformation, includes: For matrix The vectorized form of formula (37) is rewritten as: (38) in, Indicates by A matrix consisting of the diagonal elements; The dimension is A matrix in which all elements are 1; and They are variables and imaging noise In matrix form, Representation matrix No. The result of the iteration Representation matrix No. The results of this iteration; the target scattering coefficient is estimated in two-dimensional matrix form; The other parameters are represented in two-dimensional matrix form as follows: (39) (40) (41) (42) in, ; Representation matrix No. The result of the iteration Representation matrix No. The result of the iteration scale parameter In matrix form, Representation matrix No. The result of the iteration and Hyperparameters representing scale parameters.

8. A forward-looking super-resolution rapid imaging system for airborne scanning radar, characterized in that, include: The echo model building module is used to build the echo model of the airborne forward-looking imaging radar. The Doppler convolution model building module is used to establish a unified two-dimensional Doppler convolution model based on the echo model of airborne forward-looking imaging radar, and convert it from two-dimensional convolution form to one-dimensional vector form through the Doppler centroid compensation method. And a target scattering information reconstruction module, which is used to reconstruct target scattering information based on a one-dimensional vector form of Doppler convolution model and adopts the sparse Bayesian learning SBL method based on the alternating direction multiplier method to realize forward-looking super-resolution imaging of airborne scanning radar. The noise is modeled using a complex Gaussian distribution that follows a gamma distribution, and the target scattering coefficient is modeled using a layered Laplace distribution. The imaging parameters are estimated using the expectation-maximization method, and the computational complexity is reduced by a one-dimensional inverse-free estimation method based on split Bregman and singular value decomposition. The two-dimensional form of the SBL method is derived through matrix transformation to reduce memory usage.

9. A computer device comprising a memory and one or more processors, wherein the memory stores executable code, characterized in that, When the processor executes the executable code, it implements the steps of the airborne scanning radar forward-looking super-resolution fast imaging method as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps of the forward-looking super-resolution fast imaging method for airborne scanning radar as described in any one of claims 1 to 7.

Images