A real-aperture scanning radar forward-looking super-resolution imaging method

Abstract

This invention discloses a super-resolution forward-looking imaging method for real-aperture scanning radar, relating to the field of radar technology. The method includes: establishing an echo signal model in the forward-looking imaging mode of a real-aperture scanning radar; performing pulse compression and range migration correction on the target echo signal obtained from the echo signal model to obtain a pre-processed target echo signal; correcting beam distortion of the existing echo signal model using a block approximation method, transforming the spatially variable convolution matrix into a locally spatially invariant Toeplitz matrix; and combining the beam-distortion-corrected echo signal model with the Gaussian maximum a posteriori algorithm and introducing the Barzilai-Borwein method to adaptively adjust the iteration step size to achieve super-resolution imaging. This solves the problem of insufficient azimuth resolution in forward-looking imaging and improves the signal-to-noise ratio of the imaging results.

Description

Technical Field

[0001] This invention belongs to the field of radar technology, and specifically relates to a forward-looking super-resolution imaging method for real aperture scanning radar. Background Technology

[0002] With its core advantages of all-weather, all-day operation, long detection range, and high imaging resolution, radar plays an irreplaceable role in military and civilian fields such as terrain mapping, battlefield surveillance, and civilian remote sensing. In the development of radar imaging technology, improving the two-dimensional resolution in both the range and azimuth directions is key to optimizing image quality, with azimuth resolution directly determining the ability to identify target details.

[0003] Traditional synthetic aperture radar (SAR) and Doppler beam sharpening (DBS) technologies are classic solutions for improving azimuth resolution, but both have inherent defects when imaging the forward-looking area directly in front of the aircraft: SAR technology has a very small angle between the radar's motion direction and the antenna direction, resulting in a near-zero Doppler bandwidth, and the Doppler trajectories of targets on both sides are similar, which produces a "left-right blurring" phenomenon, making it impossible to accurately distinguish the left and right positions of targets; DBS technology is affected by the slow change of the Doppler frequency gradient and the complete overlap of the Doppler frequency characteristics of targets in symmetrical positions, which destroys the imaging foundation, makes it difficult to achieve target resolution in the forward-looking area, and creates an imaging "blind zone".

[0004] Real aperture scanning radar (RASR) has become an important technical path to solve the problem of forward-looking high-resolution imaging due to its simple structure, strong adaptability, and ability to achieve multi-azimuth imaging without relying on the platform Doppler effect. However, RASR forward-looking imaging still faces key technical bottlenecks: First, the azimuth resolution of the long-distance forward-looking area is inherently insufficient due to the limitation of the antenna aperture; second, the movement of the radar platform causes the beam dwell time to "shrink on the right and lengthen on the left", which leads to antenna pattern distortion and affects imaging accuracy; third, the ill-conditioning of the deconvolution process has not been effectively solved. The paper "A Sparse Bayesian Approach for Forward-Looking Super-resolution Radar Imaging" (Yin Z, Yongchao Z, Yulin H, et al. Sensors, 2017, 17(6):1353.) explores the possibility of realizing RASR forward-looking imaging using the deconvolution method. Through an adaptive iterative imaging algorithm based on weighted least squares, it achieves a beam sharpening effect of more than ten times. The paper "Super-resolution of Radar Forward-Looking Imaging Based on Accelerated TV-Sparse Method" (Zhang Y, Zhang Q, Zhang Y, et al. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2020, PP(99):1-1.) achieves high-resolution forward-looking imaging of RASR through multi-constrained deconvolution using a total variation sparse method based on Gobberg-Semencul decomposition. This method significantly reduces computational complexity and improves computational efficiency. However, the super-resolution performance of the above method degrades significantly under low signal-to-noise ratio conditions. Therefore, to address the "blind zone" problem in forward-looking imaging, it is necessary to overcome the limitations of existing technologies in terms of resolution, noise resistance, and engineering feasibility, and develop a RASR forward-looking super-resolution imaging technology that combines high resolution, strong robustness, and low complexity. This has significant academic research value and practical application significance. Summary of the Invention

[0005] The purpose of this invention is to provide a forward-looking super-resolution imaging method for real aperture scanning radar.

[0006] The technical solution to achieve the objective of this invention is: a forward-looking super-resolution imaging method for real aperture scanning radar, comprising the following steps:

[0007] Based on the relative motion between the radar platform and the target, a spatial geometric model and an echo signal model are established. Range-matched pulse compression and range migration correction are applied to the echo signal to obtain convolution-like data. Discrete convolution models of the target scattering intensity distribution and antenna pattern are also developed. , where s is the radar echo vector, The antenna pattern convolution matrix, Let be the target scattering intensity vector. For noise;

[0008] The mechanism of antenna pattern distortion caused by platform motion is analyzed, and a block approximation method is used to correct the spatially invariant convolution matrix. By quantifying the influence of platform motion velocity and target distance on beam dwell time, the global azimuth domain is divided into multiple non-overlapping sub-regions. The characteristics within each sub-region are approximated by the distortion convolution kernel at the center of the sub-region, and the spatially invariant Toeplitz cyclic matrix of each sub-region is constructed. A diagonal weighted matrix is ​​introduced to weight and synthesize the Toeplitz matrices of each sub-region to obtain the globally distorted convolution matrix. ,in For weighted matrices, For the first Toeplitz matrix of each subregion;

[0009] Gaussian Maximum A posteriori (MAP) super-resolution imaging: Based on a Bayesian estimation framework, super-resolution imaging is achieved by fusing likelihood functions and prior constraints. Assuming echo noise follows a Gaussian distribution, a joint likelihood probability function is constructed, and a generalized Gaussian distribution is used to describe the prior characteristics of target scattering intensity. The nonlinear regularization problem is transformed into a series of weighted least squares problems using an iterative weighted least squares method. The Barzilai-Borwein method is introduced to adaptively adjust the iteration step size, using the residual signal power as the iteration termination criterion. When the residual satisfies… The iteration terminates at time 1, and the estimated target scattering intensity is output. The corresponding super-resolution imaging results, among which As a confidence factor, The variance corresponding to the noise, This represents the number of sampling points in the azimuth direction.

[0010] Furthermore, antenna pattern distortion correction specifically includes:

[0011] Quantification of distortion mechanism: The platform motion speed and beam dwell time have a nonlinear relationship. When the platform speed... At that time, the rate of change of beam dwell time of targets at the edge of the track The pattern distortion rate of close-range targets is higher than that of distant targets. Furthermore, the degree of distortion increases linearly with the increase of the target's deviation from the azimuth angle of the flight path;

[0012] Sub-region division: The global azimuth domain is divided uniformly according to azimuth angle. Each non-overlapping sub-region has a width that satisfies the condition of "vacuum variation error of convolution kernel within the region". , The value is dynamically adjusted based on the platform speed;

[0013] Local matrix construction: Using the target slant range and azimuth at the center of each sub-region as a reference, calculate the distorted antenna pattern of that region and construct a space-invariant Toeplitz cyclic matrix. Matrix dimension and number of azimuth sampling points Consistent;

[0014] Weighted composition correction: weighted matrix It is a diagonal matrix, where its diagonal elements are 1 only in their corresponding subregions and 0 elsewhere. Achieve approximate reconstruction of the global spatially variable matrix.

[0015] Furthermore, Gaussian maximum a posteriori superresolution specifically includes:

[0016] Likelihood function construction: The noise follows a Gaussian distribution, and the joint likelihood probability function is used. ;

[0017] in, For variance, Let it be the conditional likelihood function;

[0018] Prior constraint introduction: target scattering intensity The prior distribution adopts the generalized Gaussian distribution. Where C is the normalization constant of the univariate generalized Gaussian distribution, and the shape parameter is... Scale parameters The target scattering distribution is adaptively adjusted based on its scattering characteristics, and the smoothness of the target scattering distribution is enhanced by the Lp regularization term.

[0019] Iterative optimization solution: The iterative weighted least squares method is used, and the iterative formula is as follows: ,in The weight matrix represents the estimated value of the target scattering intensity vector obtained after iterative updates. diagonal element p is the shape parameter of the generalized Gaussian distribution in the iterative process, with a value of 2. The iteration step size is dynamically updated using the Barzilai-Borwein method, with an initial step size of... .

[0020] Iterative optimization solution: The iterative weighted least squares method is used, and the iterative formula is as follows: The weight matrix diagonal element The iteration step size is dynamically updated using the Barzilai-Borwein method, with an initial step size... .

[0021] An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the method described above.

[0022] A computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the steps of the above-described method.

[0023] A computer program product includes a computer program that, when executed by a processor, implements the steps of the above-described method.

[0024] This invention achieves spatially variable correction of antenna beam distortion through sub-region division and weighted synthesis, and introduces a BB step size acceleration factor to control the iteration speed of the Gaussian maximum a posteriori algorithm, and uses an iterative threshold shrinkage algorithm for optimization.

[0025] The beneficial effects of this invention are that it avoids forward-looking scene imaging distortion by correcting antenna beam distortion, and improves the algorithm's convergence characteristics by combining the Barzilai-Borwein step size acceleration factor iteration, effectively suppressing noise amplification and improving super-resolution performance. Attached Figure Description

[0026] Figure 1 This is a flowchart of the method provided by the present invention.

[0027] Figure 2 This is a schematic diagram of the simulated imaging scene used in the embodiments of the present invention.

[0028] Figure 3 This is the original echo signal in the embodiments of the present invention.

[0029] Figure 4 It is the preprocessed echo signal in the embodiments of the present invention.

[0030] Figure 5 yes Figure 4 The distance cell profile of the target in the middle.

[0031] Figure 6 (a) is the super-resolution result of the method of the present invention and existing methods; (b) is the adaptive iterative imaging algorithm of weighted least squares; and (c) is the total variation sparse regularization method. Detailed Implementation

[0032] Combination Figure 1 This invention provides a radar forward-looking super-resolution imaging method, comprising the following steps:

[0033] Step A. Based on the relative motion state between the radar platform and the target, establish a spatial geometric model and an echo signal model;

[0034] Step B. Perform range matching processing on the echo signal, and perform range-to-Fourier transform on the acquired two-dimensional echo data of the transmitted signal; then construct a frequency domain matching function and perform range-to-range pulse compression to obtain range-compressed frequency domain data. Next, eliminate the interference of platform motion on the echo energy distribution through range east-west correction to obtain convolution-like data; and use a discrete convolution model of the target scattering intensity distribution and antenna pattern. , where s is the radar echo vector, The antenna pattern convolution matrix, Let be the target scattering intensity vector. For noise;

[0035] Step C. Analyze the antenna pattern distortion mechanism caused by platform motion, and use a block approximation method to correct the spatially invariant convolution matrix; by quantifying the influence of platform motion speed and target distance on beam dwell time, divide the global azimuth domain into multiple non-overlapping sub-regions, and use the distortion convolution kernel at the center of the sub-region to approximate the characteristics within the region, and construct the spatially invariant Toeplitz cyclic matrix of each sub-region.

[0036] Step D. Introduce a diagonal weighted matrix to weight and synthesize the Toeplitz matrices of each sub-region, obtaining the convolution matrix after global distortion correction. ,in For weighted matrices, For the first The Toeplitz matrix of each sub-region is used to achieve an approximate reconstruction of the spatially variable convolution matrix;

[0037] Step E. Based on the Bayesian estimation framework, assuming that the echo noise follows a Gaussian distribution, construct the joint likelihood probability function. The prior characteristics of target scattering intensity are described using a generalized Gaussian distribution, and the smoothness of the target scattering distribution is enhanced by an Lp regularization term. The nonlinear regularization problem is transformed into a series of weighted least squares problems, resulting in iterative formulas. The weight matrix diagonal element The iteration step size is dynamically updated using the Barzilai-Borwein method, and the gradient increment is defined as: Step size is Define the initial step size ;

[0038] Step F. Use the residual signal power as the iteration termination criterion; when the residual satisfies... The iteration terminates at time 1, and the estimated target scattering intensity is output. The corresponding super-resolution imaging results, among which As a confidence factor, The variance corresponding to the noise, This represents the number of sampling points in the azimuth direction. The super-resolution imaging result corresponding to the estimated target scattering intensity is output, achieving improved azimuth resolution and optimized signal-to-noise ratio.

[0039] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0040] Example 1

[0041] A forward-looking super-resolution imaging method for real aperture scanning radar includes the following steps:

[0042] Step 1: Spatial geometry and signal echo model construction

[0043] This embodiment uses an X-band airborne forward-looking radar, and its core parameters are:

[0044] carrier frequency Transmitted signal bandwidth Pulse repetition frequency ;

[0045] Radar platform height Platform speed t is time;

[0046] The target slant range history expression is: .

[0047] The slant range between the radar platform and the target is r0, the antenna elevation angle is φ, and the initial azimuth angle of the radar line of sight is θ0. The target point is set as follows: Figure 2 As shown.

[0048] Echo signal model: Transmitting a linear frequency modulated signal , The carrier frequency is , and the pulse width of the transmitted signal is . , To obtain the point target echo signal model using linear frequency modulation. , For rectangular window functions, For antenna pattern modulation terms, such as Figure 3 As shown;

[0049] Step 2: Echo signal preprocessing

[0050] Pulse compression with matched filtering: Perform a 1024-point FFT on time t to obtain the distance-frequency domain signal. Let f be the frequency variable and θ be the azimuth index. Construct a frequency domain matching function. and After multiplying and canceling the quadratic phase terms, an IFFT is performed to obtain the range compression result, achieving a range resolution of 3m.

[0051] Distance migration correction: through linear variable substitution By focusing the target echo energy onto the same range cell and eliminating range offset caused by platform motion, the echo signal is obtained in a convolutional form as the convolution of phase-weighted scattering coefficients and the system objective function. ;

[0052] Step 3: Antenna pattern distortion correction

[0053] The distorted antenna pattern is as follows: ,in This is the initial slope distance. The initial azimuth angle, This represents the antenna scanning speed.

[0054] Preprocessed radar echoes, such as Figure 4 As shown, after pulse compression via matched filtering, the original radar echo signal can be compressed from a wide pulse to a narrow pulse, thereby overcoming the limitation of the transmitted signal bandwidth on range resolution and achieving high-resolution range measurement. After range migration correction, the energy of the target scattered echo can be effectively focused within the same range cell, providing a basis for subsequent deconvolution algorithm processing.

[0055] To balance correction accuracy and computational complexity, the global azimuth domain is uniformly divided according to azimuth angle. Based on the target slant range and distortion characteristics corresponding to the center azimuth angle of each sub-region, a space-invariant Toeplitz cyclic matrix is ​​constructed: for the i-th sub-region, based on the center azimuth angle... Substituting the distortion pattern formula, we obtain the equivalent antenna pattern for this region. ,extract The discrete values ​​are used to construct a Toeplitz cyclic matrix centered at the location of the maximum value, ensuring that the matrix satisfies the Fourier diagonalization property.

[0056] Introducing a diagonal weighted matrix To achieve the splicing and reconstruction of the matrices of each sub-region: for A diagonal matrix, if the first... The sampling point in the azimuth direction belongs to the first Each sub-region, then The The diagonal element is 1; otherwise, it is 0.

[0057] The approximate reconstruction of the spatially variable convolution matrix is ​​achieved through weighted summation, as shown in the formula:

[0058] Step 4: Implementation of Gaussian Maximum A posteriori super-resolution algorithm

[0059] The azimuth profile of the signal before the super-resolution algorithm at the 0-meter distance cell is as follows Figure 5 As shown, due to the superposition of target echo signals and the influence of noise energy, the boundary features of adjacent targets are weakened, making it impossible to accurately locate and distinguish individual targets. Assuming the echo noise follows a Gaussian distribution, the joint likelihood probability function is: ;

[0060] The prior distribution of the standard scattering intensity adopts a generalized Gaussian distribution, and an L2 regularization term is introduced to enhance smoothness. The prior probability is: ;

[0061] Iteration formula: Among them, the weight matrix Diagonal elements are dependent on The element The function, with the expression: L2 regularization terms correspond to It is 2;

[0062] Barzilai-Borwein (BB) step size dynamic update: initial step size The subsequent step size is calculated using gradient increments:

[0063] Iteration termination criterion: Using the residual signal power as a threshold, when the following conditions are met... Termination time, confidence factor The azimuth profile of the signal before the super-resolution algorithm at the 0-meter distance cell is as follows: Figure 6 As shown, compared with the weighted least squares adaptive iterative imaging algorithm and the total variation sparse regularization method, this algorithm effectively suppresses the amplification of noise during the iteration process and can achieve 11.6 times azimuth super-resolution.

[0064] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A forward-looking super-resolution imaging method for real aperture scanning radar, characterized in that, Includes the following steps: Based on the relative motion between the radar platform and the target, a spatial geometric model and an echo signal model are established. Range-matched pulse compression and range migration correction are applied to the echo signal to obtain convolution-like data. Discrete convolution models of the target scattering intensity distribution and antenna pattern are also developed. , where s is the radar echo vector, The antenna pattern convolution matrix, Let be the target scattering intensity vector. For noise; The mechanism of antenna pattern distortion caused by platform motion is analyzed, and a block approximation method is used to correct the spatially invariant convolution matrix. By quantifying the influence of platform motion velocity and target distance on beam dwell time, the global azimuth domain is divided into multiple non-overlapping sub-regions. The characteristics within each sub-region are approximated by the distortion convolution kernel at the center of the sub-region, and the spatially invariant Toeplitz cyclic matrix of each sub-region is constructed. A diagonal weighted matrix is ​​introduced to weight and synthesize the Toeplitz matrices of each sub-region to obtain the globally distorted convolution matrix. ,in For weighted matrices, For the first Toeplitz matrix of each subregion; Gaussian Maximum A posteriori (MAP) super-resolution imaging: Based on a Bayesian estimation framework, super-resolution imaging is achieved by fusing likelihood functions and prior constraints. Assuming echo noise follows a Gaussian distribution, a joint likelihood probability function is constructed, and a generalized Gaussian distribution is used to describe the prior characteristics of target scattering intensity. The nonlinear regularization problem is transformed into a series of weighted least squares problems using an iterative weighted least squares method. The Barzilai-Borwein method is introduced to adaptively adjust the iteration step size, using the residual signal power as the iteration termination criterion. When the residual satisfies… The iteration terminates at time 1, and the estimated target scattering intensity is output. The corresponding super-resolution imaging results, among which As a confidence factor, The variance corresponding to the noise, This represents the number of sampling points in the azimuth direction.

2. The method as described in claim 1, characterized in that, Antenna pattern distortion correction specifically includes: Quantification of distortion mechanism: The platform motion speed and beam dwell time have a nonlinear relationship. When the platform speed... At that time, the rate of change of beam dwell time of targets at the edge of the track The pattern distortion rate of close-range targets is higher than that of distant targets. above; Sub-region division: The global azimuth domain is divided uniformly according to azimuth angle. Each non-overlapping sub-region has a width that satisfies the spatial variation error of the convolution kernel within the region. , The value is dynamically adjusted based on the platform speed; Local matrix construction: Using the target slant range and azimuth at the center of each sub-region as a reference, calculate the distorted antenna pattern of that region and construct a space-invariant Toeplitz cyclic matrix. Matrix dimension and number of azimuth sampling points Consistent; Weighted composition correction: weighted matrix It is a diagonal matrix, where its diagonal elements are 1 only in their corresponding subregions and 0 elsewhere. Achieve approximate reconstruction of the global spatially variable matrix.

3. The method as described in claim 1, characterized in that, Gaussian maximum a posteriori superresolution specifically includes: Likelihood function construction: The noise follows a Gaussian distribution, and the joint likelihood probability function is used. ; in, For variance, Let it be the conditional likelihood function; Prior constraint introduction: target scattering intensity The prior distribution adopts the generalized Gaussian distribution. Where C is the normalization constant of the univariate generalized Gaussian distribution, and the shape parameter is... Scale parameters The target scattering distribution is adaptively adjusted based on its scattering characteristics, and the smoothness of the target scattering distribution is enhanced by the Lp regularization term. Iterative optimization solution: The iterative weighted least squares method is used, and the iterative formula is as follows: ,in The weight matrix represents the estimated value of the target scattering intensity vector obtained after iterative updates. diagonal element p is the shape parameter of the generalized Gaussian distribution in the iterative process. The iteration step size is dynamically updated using the Barzilai-Borwein method, and the initial step size is... .

4. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the method described in any one of claims 1-3.

5. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the program implements the steps of the method described in any of claims 1-3.

6. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the steps of the method described in any one of claims 1-3.

Images