Near-field three-dimensional radar sparse imaging method based on structured regular function
By introducing structured regular functions and ADMM methods in three-dimensional SAR imaging, the problem of insufficient deviation effect and feature characterization in L1 regular functions in three-dimensional SAR imaging is solved, the image quality and edge structure retention are improved, and more efficient sparse imaging is achieved .
Patent Information
- Application Number
- CN202411009383.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-26
- Publication Date
- 2024-11-08
AI Technical Summary
In the existing three-dimensional SAR imaging method, sparse imaging based on L1 regular function has a bias effect, making it difficult to effectively characterize the diverse features of the three-dimensional imaging scene, resulting in low image quality.
A near-field three-dimensional radar sparse imaging method based on structured regular functions is adopted, and a structured regular function is constructed by combining non-convex punishment functions L1/2 and 3D full variations, and a parallel decoupling optimization method and alternating direction multiplier method (ADMM) is used. Solve optimization problems, reduce deviation effects and preserve target edge structure characteristics.
It effectively reduces the deviation effect of the L1 regular function, enhances the retention of the point feature and edge structure information of the target, and improves the quality and imaging accuracy of the three-dimensional sparse images of SAR in the near-field array.
Smart Images

Figure CN118915066A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of radar imaging technology, and particularly relates to the field of near-field three-dimensional imaging radar sparse imaging technology. Background Art
[0002] Synthetic aperture radar is an active microwave radar imaging technology that has the ability to work all day, all weather, and with high resolution. It has great application value in natural resource exploration, target identification and classification, radar cross-sectional area measurement, and other fields. With the development of SAR systems, three-dimensional SAR imaging technology with the ability to obtain complete spatial information of targets has attracted widespread attention from researchers. Array three-dimensional SAR forms a virtual two-dimensional array in the horizontal and vertical directions, thereby obtaining high resolution in the two-dimensional azimuth-altitude plane, while the range resolution is guaranteed by transmitting broadband signals. Compared with traditional two-dimensional SAR, array three-dimensional SAR can not only obtain complete spatial information of the target, but also has high-precision imaging capabilities.
[0003] Most traditional SAR imaging methods are based on matched filtering to achieve three-dimensional imaging. However, the imaging methods based on matched filtering are usually constrained by the resolution Rayleigh criterion and the Nyquist sampling theorem. The resulting SAR images have low resolution and are easily affected by sidelobes and noise interference, which leads to reduced quality of three-dimensional SAR images and cannot meet application requirements. For three-dimensional imaging, the observed target usually only occupies a small part of the imaging space and has strong sparsity. Therefore, sparse signal processing is widely used in SAR imaging to obtain high-quality three-dimensional SAR images. Sparse SAR imaging is essentially a typical ill-conditioned inverse problem. It is necessary to introduce prior terms to alleviate the ill-conditioning, so as to solve the sparse optimization problem. In sparse imaging, the regularization function is usually used as a prior term to characterize the target prior information. By combining the regularization function with the optimization algorithm framework, the sparse optimization problem can be effectively solved.
[0004] L 1 As a sparse penalty function commonly used in compressed sensing theory, the regularization function can restore sparse signals relatively quickly. However, its convex relaxation characteristics usually cause bias effects, which will underestimate the target amplitude and reduce the reconstruction accuracy when applied to SAR imaging. In addition, L 1 The regularization function only enhances the point features of the target and does not consider the edge structure features of the target. 1 Compared with regular functions, non-convex penalty functions can alleviate the bias effect and achieve better imaging effects. A few studies have introduced non-convex penalty functions such as minimax concave penalty function terms into array SAR imaging to improve image quality, and used the imaging results for target scattering information diagnosis. However, non-convex penalty functions only represent the point features of the target and it is difficult to represent the diverse features of three-dimensional imaging scenes.
[0005] Therefore, for L 1 The bias effect introduced by the regularization function and the difficulty of a single regularization function in representing the diversified features in the three-dimensional SAR imaging scene are solved. This paper proposes a near-field three-dimensional radar sparse imaging method based on a structured regularization function, which effectively reduces the L 1 The deviation effect of the regularization function not only enhances the point features of the target, but also preserves the target edge structure information more completely, resulting in a higher quality near-field array SAR three-dimensional sparse image. Summary of the invention
[0006] The present invention proposes a near-field three-dimensional radar sparse imaging method based on structured regularization function. First, the method combines the non-convex penalty function L 1 / 2 The structured regularization function is constructed by 3D total variation. Then, the parallel decoupled optimization method is used to solve the optimization problem of the structured regularization function to achieve 3D sparse imaging, which can effectively reduce the bias effect and retain the edge structure characteristics of the target, improving the accuracy of sparse imaging. Afterwards, in order to reduce the complexity of the algorithm, the sub-optimization problem containing the forward imaging model is linearized by combining the linearization idea and the alternating direction multiplier method to improve the imaging efficiency. 1 Compared with the array synthetic aperture radar three-dimensional sparse imaging method realized by the regular function and the MCP regular function, the method of the present invention effectively reduces L 1 The deviation effect of the regularization function not only enhances the point features of the target, but also preserves the target edge structure information more completely, resulting in a higher quality near-field array SAR three-dimensional sparse image.
[0007] In order to facilitate the description of the present invention, the following terms are first defined:
[0008] Definition 1. Synthetic Aperture Radar
[0009] Synthetic Aperture Radar (SAR) is an observation system that uses the principle of synthetic aperture to achieve high-resolution microwave imaging. It has many characteristics such as all-day, all-weather, and high resolution. It has unique advantages in disaster monitoring, environmental monitoring, ocean monitoring, resource exploration, mapping, and military applications.
[0010] Definition 2. Azimuth, altitude, and distance
[0011] The direction of the lateral movement of the radar platform is called the azimuth, the direction of the longitudinal movement is called the altitude, and the direction perpendicular to the first two is called the range.
[0012] Definition 3.L 1 Regular function term
[0013] L1 The regularization function term is a target prior information constraint term, which is composed of L 1 The norm composition can be expressed as L 1 The norm is a convex function, which is calculated by i The absolute value of is accumulated and calculated. This regularization function term is a common constraint term in existing 3D SAR sparse imaging methods, which is used to enhance the sparsity of the obtained target scene image. For details, see "Wang Y, Zhang X, Zhan X, et al. An RCS Measurement Method Using Sparse Imaging Based 3D SAR Complex Image [J]. IEEE Antennas and Wireless Propagation Letters, 2021".
[0014] Definition 4. MCP regular function term
[0015] The MCP regularization term (minimax-concave penalty, MCP) is a non-convex constraint term of target prior information, expressed as Where min means taking the minimum value, and |v| means the absolute value of vector v. The MCP non-convex penalty function can alleviate L 1 For details, see "Selesnick I. Sparse regularization via convex analysis[J]. IEEE Transactions on Signal Processing, 2017, 65(17): 4481-4494".
[0016] Definition 5.L 1 / 2 Regular function term
[0017] L 1 / 2 The regularization function term is used in radar imaging to alleviate the L 1 The deviation effect brought by the regularization function improves the image quality. 1 / 2 The regularization function is a non-convex constraint term of target prior information, which is composed of L 1 / 2 The norm composition can be expressed as L 1 / 2 Norm compared to L 1The norm has a sparser solution and higher recovery accuracy. For details, see "A.Wu, Y.Wu, Y.Jin, et al. L(1 / 2) regularization for ISAR imaging and target enhancement of complex image[J]. IEEE Transactions on Geoscience and Remote Sensing, 2022, 60: 1-10".
[0018] Definition 6.L 1 / 2 Proximal operator corresponding to the regular function
[0019] The proximal operator is used in proximal gradient methods, which are often used in optimization algorithms related to non-differentiable optimization problems (such as total variation denoising). 1 / 2 The proximal operator corresponding to the regular function is In the formula, g(x i ; The expression of μλ) is In the formula, θ(x i ) is expressed as Where λ represents the regularization factor, generally λ>0, and μ is a fixed positive parameter to ensure convergence. For details, see "A.Wu, Y.Wu, Y.Jin, Y.Wang and Z.Guo.L(1 / 2)Regularization for ISAR Imaging and Target Enhancementof Complex Image[J].IEEE Transactions on Geoscience and Remote Sensing,vol.60,pp.1-10,2022."
[0020] Definition 7.3D Total Variation Regularization Function
[0021] The 3D Total-Variation (3DTV) regularization term is a constraint term that characterizes the edge structure characteristics and can be expressed as
[0022]
[0023] Among them, X represents the target 3D scene image, M x ,M y ,M zRespectively represent the number of sampling points in the range, azimuth and altitude directions, and i, j, and l represent the sampling index values in these three directions. Compared with the traditional TV norm, the 3DTV penalty function is more suitable for three-dimensional SAR imaging scenes with diverse features to enhance the edge structure characteristics of the target. For details, see "Kong, X.; Zhao, Y.; Chan, JC-W.; Xue, J. Hyperspectral Image Restoration via Spatial-Spectral Residual Total Variation Regularized Low-Rank Tensor Decomposition[J]. Remote Sens. 2022, 14, 511."
[0024] Definition 8. Proximal operator corresponding to 3D total variation regularization function
[0025] The proximal operator of 3DTV regularization is a mathematical method used to solve problems such as image restoration and compressed sensing. It usually uses prior information to regularize the image to improve the quality of image restoration. The proximal operator algorithm of 3DTV regularization usually includes the following steps: first, the gradient of the image is calculated to obtain the gradient information of each pixel; then, the image is iteratively processed to continuously update the pixel values of the image so that the pixels with larger gradient magnitudes gradually become sparse; finally, the image is continuously updated through iteration until the convergence condition is reached. For details, see "J.Peng, Q.Xie, Q.Zhao, Y.Wang, L.Yee and D.Meng, Enhanced 3DTV Regularization and ItsApplications on HSIDenoising and Compressed Sensing, IEEE Transactions on Image Processing, vol.29, pp.7889-7903, 2020.".
[0026] Definition 9. Standard ADMM Method
[0027] The alternating direction method of multipliers (ADMM) is an iterative algorithm for solving constrained optimization problems. It uses the block coordinate descent method to optimize the unconstrained sub-problems separately. The basic idea of the ADMM algorithm is to decompose a large optimization problem into several small sub-problems, then alternately find the maximum value of a certain variable, and finally combine them to get the solution to the original problem. For details, see "Stephen Boyd; Neal Parikh; et al. Distributed Optimization and Statistical Learning via the Alternating Direction Method of Multipliers[J], 2011.".
[0028] Definition 10. Traditional gradient calculation method
[0029] For a scalar function f(X), X is a matrix, and the traditional gradient calculation method is:
[0030]
[0031] in, Represents the function f for the element X in the i-th row and j-th column of X ij For details, see "Fu Yingding et al. Optimization Theory and Methods [M]. University of Electronic Science and Technology Press, 1996".
[0032] Definition 11. Traditional variable segmentation method
[0033] The traditional variable segmentation method is to decouple multiple functions by establishing auxiliary variables, so that the data fidelity term and regularization term in the optimization problem model can be processed separately. For details, see "Fu Yingding et al. Optimization Theory and Methods [M]. University of Electronic Science and Technology Press, 1996".
[0034] Definition 12. Standard pulse compression method
[0035] Pulse compression is a modern radar signal processing method. Simply put, the radar transmits a wide pulse, which is then "compressed" into a narrow pulse at the receiving end, thereby improving two radar performances: effective range and range resolution.
[0036] For details of the standard pulse compression method, see "Pi Yiming, Yang Jianyu, Fu Yusheng, Yang Xiaobo. Principles of Synthetic Aperture Radar Imaging [M]. University of Electronic Science and Technology Press. 2007".
[0037] Definition 13. Measurement matrix of 3D radar system
[0038] In near-field array SAR three-dimensional imaging, the measurement matrix is a key tool for mapping the physical characteristics of the target scene into a data space that can be processed by the imaging algorithm. It is usually necessary to define the geometric configuration and beamforming mode of the antenna array, transmit a known signal and record the echo returned from the target, and use the echo data and array configuration information to construct the measurement matrix. For details on the construction method of the measurement matrix of the three-dimensional radar system, please refer to "Wei Shunjun. Research on Sparse Imaging Technology of Linear Array Three-Dimensional Synthetic Aperture Radar [D]. University of Electronic Science and Technology of China, 2013."
[0039] Definition 14. Traditional imaging method based on matched filtering
[0040] The traditional imaging method based on matched filtering is a signal processing technology widely used in the fields of radar, communication and image processing. The implementation of matched filtering usually includes two methods: time domain and frequency domain. In the time domain, matched filtering is realized by directly calculating the convolution of the input signal and the reference signal. In the frequency domain, the signal is usually converted to the frequency domain for processing using algorithms such as fast Fourier transform (FFT), and then converted back to the time domain using inverse Fourier transform (IFFT). For details of the traditional imaging method based on matched filtering, please refer to "Lan G.Cumming Frank H.Wong. Synthetic Aperture Radar Imaging: Algorithm and Implementation [M]. Publishing House of Electronic Industry, 2012".
[0041] Definition 15. Standard distance stacking operation method
[0042] The standard stacking operation method along the range direction refers to a common operation method in the SAR frequency domain three-dimensional imaging algorithm. During the imaging process, two-dimensional imaging is performed on the azimuth-altitude direction corresponding to each range unit to obtain a two-dimensional imaging result. The stacking arranges the two-dimensional imaging results of each range unit in the order of the range unit to form the final three-dimensional imaging result. For details of the standard stacking operation method along the range direction, see "Wang M, Wei S, Liang J, et al. RMIST-net: Joint range migration and sparse reconstruction network for 3-D mmW imaging[J]. IEEE Transactions on Geoscience and Remote Sensing, 2021, 60: 1-17".
[0043] The present invention provides a near-field three-dimensional radar sparse imaging method based on a structured regular function, which is characterized in that it comprises the following steps:
[0044] Step 1. Initialize relevant parameters
[0045] The azimuth sampling number is recorded as l = 1, 2, ..., L, where L represents the total number of azimuth sampling; the altitude sampling number is recorded as m = 1, 2, ..., M, where M represents the total number of altitude sampling; the range sampling number is recorded as i = 1, 2, ..., N, where N represents the total number of range sampling; the original echo of the target is recorded as S L×M×N ; L 1 / 2 The regularization parameter of the penalty function is denoted by λ 1 ; Regularization parameter of 3DTV penalty function, denoted as λ 2 .
[0046] Step 2. Perform range pulse compression on the target original echo to obtain the range pulse compression result
[0047] Take the target original echo S in step 1 L×M×N As input, the pulse compression method of definition 12 is used to generate S L×M×N The distance signal in the range is compressed to obtain the result Y after the distance pulse compression L×M×N ;
[0048] Y L×M×N Extract each range unit one by one to obtain the pulse compression result Y(i) of the i-th range unit; where i is the range sampling number initialized in step 1, and L, M, and N are the total number of azimuth sampling, total number of altitude sampling, and total number of range sampling initialized in step 1.
[0049] Step 3. Use the following formula to construct an imaging optimization model based on structured regularization function:
[0050]
[0051] subject to QX i +BD i =0
[0052] Among them, X i , is the solution of the imaging equation corresponding to the constructed i-th distance unit; i is the range sampling number initialized in step 1; Indicates The minimum X i and Y(i) is the pulse compression processing result corresponding to the i-th range unit obtained in step 2; A is the measurement matrix of the array SAR system described in Definition 13; represents the square of the matrix 2 norm; As described in Definition 5 L 1 / 2 norm regularization term; is the scene image described in Definition 7 , |·| represents the absolute value; λ 1 and λ 2 are L initialized in step 1 1 / 2 Regularization parameter of penalty function and regularization parameter of 3DTV penalty function; Q = (I, I), B = (-I, -I), I is the unit matrix; is the auxiliary variable of the variable partitioning method described in Definition 11.
[0053] Step 4. Solve the imaging optimization model based on the structured regularization function
[0054] The standard ADMM method in Definition 9 is used to iteratively solve the imaging solution equation corresponding to each range unit obtained in step 3 to obtain the azimuth-altitude imaging result corresponding to each range unit.
[0055] Step 4.1. Initialize the azimuth-altitude imaging results corresponding to the range unit to be solved 0 L×M represents an all-zero matrix of dimension L×M, i is the distance sampling number initialized in step 1; initialize the error parameter ε=0.001; initialize the auxiliary variable Initialize the step size factor μ; initialize the regularization parameter T.
[0056] Step 4.2. Use the traditional matched filter-based imaging method in Definition 14 to process the pulse compression result Y(i) obtained in step 2 to obtain the imaging result Y(i) of the i-th distance unit. MF .
[0057] Step 4.3. In the first iteration, perform the following steps:
[0058] (1) Use the following formula to calculate the target scene image X for each distance unit i 1 :
[0059]
[0060] in, is the target scene imaging result of the i-th range unit solved in the first iteration, where i is the range sampling number initialized in step 1; is the azimuth-altitude imaging result corresponding to the range unit to be solved initialized in step 4.1; μ and T are the step size factor and regularization parameter initialized in step 4.1 respectively; Y(i) MF The imaging result based on matched filtering obtained in step 4.2; is the auxiliary variable for the ith distance unit initialized in step 4.1.
[0061] (2) Use the following formula to calculate the auxiliary variable for each distance unit:
[0062]
[0063] in, is the auxiliary variable of the i-th distance unit obtained by the first iteration, i is the distance sampling number initialized in step 1; L in Definition 6 1 / 2 The proximal operator corresponding to the regular function; is the azimuth-altitude imaging result calculated in step (1); is the auxiliary variable initialized in step 4.1.
[0064] (3) Stack along the distance direction to obtain the first iteration of the three-dimensional sparse imaging result X of the target scene:
[0065] For the azimuth-altitude imaging result corresponding to each range unit of the first iteration obtained in step (1), i is the range sampling number initialized in step 1. The standard range stacking operation method described in Definition 16 is used to perform the range stacking operation to obtain the first iteration of the three-dimensional sparse imaging result X of the target scene. 1 .
[0066] (4) Use the following formula to calculate the auxiliary variable
[0067]
[0068] in, is the auxiliary variable obtained by solving the first iteration, is the proximal operator corresponding to the 3D total variation regularization function described in Definition 8; X 1 is the three-dimensional sparse imaging result X obtained in step (3) 1 ;(U 1 ) 0 is the auxiliary variable initialized in step 4.1.
[0069] Pick The slice matrix of each distance unit of , and the auxiliary variable of the i-th distance unit is obtained
[0070] (5) Use the following formula to calculate the auxiliary variable for each distance unit:
[0071]
[0072] in, and is the auxiliary variable of the i-th distance unit obtained by the first iteration; i is the distance sampling number initialized in step 1; and is the auxiliary variable initialized in step 4.1; is the imaging result of the i-th distance unit solved in step (1); and is the auxiliary variable of the i-th distance unit solved in steps (2) and (4).
[0073] (6) Stack along the distance direction to obtain the three-dimensional auxiliary variable (U 1 ) 1 :
[0074] For each distance unit corresponding to the first iteration obtained in step (5), i is the distance sampling number initialized in step 1. The standard distance stacking operation method described in Definition 16 is used to perform distance stacking operation to obtain the three-dimensional auxiliary variable (U) of the first iteration. 1 ) 1 .
[0075] (7) The iterative error of each distance unit is calculated using the following formula:
[0076]
[0077] Among them, Resi represents the iterative error of this distance unit; is the imaging result of the i-th distance unit obtained in the first iteration of step (1); is the azimuth-altitude imaging result of the range unit to be solved initialized in step 4.1; ||·|| F represents the matrix Fibonacci norm; i is the distance sampling number initialized in step 1.
[0078] (8) Iteration termination condition judgment: if Resi ≤ ε, the iteration stops. Resi is the iteration error obtained in step (7), ε is the error parameter initialized in step 1, and the azimuth-altitude imaging result X corresponding to the i-th range unit is i This is the result of the first iteration. Otherwise, continue iterating. i is the distance sampling number initialized in step 1.
[0079] Step 4.4. In the second iteration, perform the following steps:
[0080] (1) Use the following formula to calculate the target scene image of each distance unit
[0081]
[0082] in, is the target scene imaging result of the i-th range unit solved in the second iteration, where i is the range sampling number initialized in step 1; is the azimuth-altitude imaging result of the first iteration calculated in step 4.3.(1); μ and T are the step size factor and regularization parameter initialized in step 4.1 respectively; Y(i) MF The imaging result based on matched filtering obtained in step 4.2; is the auxiliary variable of the first iteration calculated in step 4.3.(2); Step 4.3.(4) Calculate the auxiliary variables of the first iteration; is the auxiliary variable for the first iteration of the calculation in step 4.3.(5).
[0083] (2) Use the following formula to calculate the auxiliary variable for each distance unit:
[0084]
[0085] in, is the auxiliary variable of the i-th distance unit obtained by the second iteration, i is the distance sampling number initialized in step 1; L in Definition 6 1 / 2 The proximal operator corresponding to the regular function; is the azimuth-altitude imaging result of the second iteration calculated in step 4.4.(1); is the auxiliary variable for the first iteration of the calculation in step 4.3.(5).
[0086] (3) Stacking along the distance direction obtains the second iteration of the three-dimensional sparse imaging result X of the target scene:
[0087] For each range unit of the second iteration obtained in step 4.4.(1), the azimuth-altitude imaging result corresponding to i is the range sampling number initialized in step 1. The standard range stacking operation method described in Definition 16 is used to perform the range stacking operation to obtain the second iteration of the three-dimensional sparse imaging result X of the target scene. 2 .
[0088] (4) Use the following formula to calculate the auxiliary variable
[0089]
[0090] in, is the auxiliary variable obtained by solving the second iteration, is the proximal operator corresponding to the 3D total variation regularization function described in Definition 8; X 2 is the three-dimensional sparse imaging result X obtained in step 4.4.(3) 2 ;(U 1 ) 1 is the three-dimensional auxiliary variable (U) of the first iteration obtained in step 4.3.(6) 1 ) 1 .
[0091] Pick The slice matrix of each distance unit of , and the auxiliary variable of the i-th distance unit is obtained
[0092] (5) Use the following formula to calculate the auxiliary variable for each distance unit:
[0093]
[0094]
[0095] in, and is the auxiliary variable of the i-th distance unit obtained by the second iteration; i is the distance sampling number initialized in step 1; and is the auxiliary variable of the first iteration calculated in step 4.3.(5); is the imaging result of the second iteration solved in step 4.4.(1); and is the auxiliary variable of the i-th distance unit solved in steps 4.4.(2) and 4.4.(4).
[0096] (6) Stack along the distance direction to obtain the three-dimensional auxiliary variable (U 1 ) 2
[0097] For each distance unit of the second iteration obtained in step (5), the auxiliary variable corresponding to i is the distance sampling number initialized in step 1. The standard distance stacking operation method described in Definition 16 is used to perform distance stacking operation to obtain the three-dimensional auxiliary variable (U) of the second iteration. 1 ) 2 .
[0098] (7) The iterative error of each distance unit is calculated using the following formula:
[0099]
[0100] Among them, Resi represents the iterative error of this distance unit; is the imaging result of the ith distance unit obtained by the second iteration of step 4.4.(1); is the imaging result of the ith distance unit obtained in the first iteration of step 4.3.(1); ||·|| F represents the matrix Fibonacci norm; i is the distance sampling number initialized in step 1.
[0101] (8) Iteration termination condition judgment: if Resi ≤ ε, the iteration stops. Resi is the iterative error obtained in step 4.4. (7), ε is the error parameter initialized in step 1, and the azimuth-altitude imaging result X corresponding to the i-th range unit i This is the result of the second iteration. Otherwise, continue iterating. i is the distance sampling number initialized in step 1.
[0102] Step 4.5. For the kth iteration, perform the following steps:
[0103] (1) Use the following formula to calculate the target scene image of each distance unit
[0104]
[0105] in, is the target scene imaging result of the i-th distance unit solved in the k-th iteration, k is the number of iterations, and i is the range sampling number initialized in step 1; is the azimuth-altitude imaging result obtained in the k-1th iteration; μ and T are the step size factor and regularization parameter initialized in step 4.1 respectively; Y(i) MF The imaging result based on matched filtering obtained in step 4.2; and is the auxiliary variable obtained at the k-1th iteration.
[0106] (2) Use the following formula to calculate the auxiliary variable for each distance unit:
[0107]
[0108] in, is the auxiliary variable of the i-th distance unit obtained by solving the k-th iteration, and i is the distance sampling sequence number initialized in step 1; L in Definition 6 1 / 2 The proximal operator corresponding to the regular function; is the azimuth-altitude imaging result of the kth iteration calculated in step 4.5.(1); is the auxiliary variable calculated at the k-1th iteration.
[0109] (3) Stack along the distance direction to obtain the k-th iteration of the three-dimensional sparse imaging result X of the target scene:
[0110] For the azimuth-altitude imaging result corresponding to each range unit of the kth iteration obtained in step 4.5.(1), i is the range sampling number initialized in step 1. The standard range stacking operation method described in Definition 16 is used to perform the range stacking operation to obtain the k-th iteration of the three-dimensional sparse imaging result X of the target scene. k .
[0111] (4) Use the following formula to calculate the auxiliary variable
[0112]
[0113] in, is the auxiliary variable obtained by solving the kth iteration, is the proximal operator corresponding to the 3D total variation regularization function described in Definition 8; X k is the three-dimensional sparse imaging result X obtained in step 4.5.(3) k ;(U 1 ) k-1 is the three-dimensional auxiliary variable (U 1 ) k-1 .
[0114] Pick The slice matrix of each distance unit of , and the auxiliary variable of the i-th distance unit is obtained
[0115] (5) Use the following formula to calculate the auxiliary variable for each distance unit:
[0116]
[0117] in, and is the auxiliary variable of the i-th distance unit obtained by the k-th iteration; i is the distance sampling number initialized in step 1; and is the auxiliary variable calculated in the k-1th iteration; is the imaging result of the kth iteration solved in step 4.5.(1); and is the auxiliary variable of the i-th distance unit solved in steps 4.5.(2) and 4.5.(4).
[0118] (6) Stack along the distance direction to obtain the three-dimensional auxiliary variable (U 1 ) k
[0119] For each distance unit corresponding to the kth iteration obtained in step (5), i is the distance sampling number initialized in step 1. The standard distance stacking operation method described in Definition 16 is used to perform distance stacking operation to obtain the three-dimensional auxiliary variable (U) of the kth iteration. 1 ) k .
[0120] (7) The iterative error of each distance unit is calculated using the following formula:
[0121]
[0122] Among them, Resi represents the iterative error of this distance unit; is the imaging result of the i-th distance unit obtained in the k-th iteration of step 4.5.(1); is the imaging result of the i-th distance unit calculated in the k-1th iteration; ||·|| F represents the matrix Fibonacci norm; i is the distance sampling number initialized in step 1.
[0123] (8) Iteration termination condition judgment: if Resi ≤ ε, the iteration stops. Resi is the iterative error obtained in step 4.5. (7), ε is the error parameter initialized in step 1, and the azimuth-altitude imaging result X corresponding to the i-th range unit i This is the result of the kth iteration. Otherwise, continue iterating. i is the distance sampling number initialized in step 1.
[0124] Step 5. Stack along the distance to obtain the overall three-dimensional sparse imaging result of the target scene
[0125] For each distance unit obtained in step 4, the azimuth-altitude imaging result X i , i is the range sampling number initialized in step 1, and the standard range stacking operation method described in Definition 15 is used to perform the range stacking operation to obtain the final target overall three-dimensional sparse imaging result X, where X is the final imaging result of the proposed near-field three-dimensional radar sparse imaging method based on structured regular function.
[0126] The innovation and advantage of the present invention are: 1Compared with the array synthetic aperture radar three-dimensional sparse imaging method implemented by the regular function and the MCP regular function, the method of the present invention combines the non-convex penalty function L 1 / 2 and 3DTV to construct a structured regularization function, effectively reducing L 1 The deviation effect of the regularization function not only enhances the point features of the target, but also preserves the target edge structure information more completely, resulting in a higher quality near-field three-dimensional sparse image. The parallel decoupling optimization method is used to solve the optimization problem of the structured regularization function to achieve three-dimensional sparse imaging, which can effectively reduce the deviation effect and preserve the target edge structure characteristics, thereby improving the accuracy of sparse imaging. Combining the linearization idea with ADMM, the sub-optimization problem containing the forward imaging model is linearized to improve the imaging efficiency. BRIEF DESCRIPTION OF THE DRAWINGS
[0127] Figure 1 It is a flow chart of the present invention.
[0128] Figure 2 This is the verification result of the simulation experiment of the present invention. DETAILED DESCRIPTION
[0129] The present invention is mainly verified by simulation experiment method, and all steps and conclusions are verified correct on the mathematical calculation software Matlab2019b. The specific implementation steps are as follows:
[0130] Step 1. Initialize relevant parameters
[0131] The sampling number in azimuth is recorded as l=1,2,…,512; the sampling number in altitude is recorded as m=1,2,…,512; the sampling number in range is recorded as i=1,2,…,1024; the original echo of the target is recorded as S 512×512×1024 ; L 1 / 2 The regularization parameter λ of the penalty function 1 =0.5; regularization parameter λ of 3DTV penalty function 2 =0.5.
[0132] Step 2. Perform range pulse compression on the target original echo to obtain the range pulse compression result
[0133] Take the target original echo S in step 1 512×512×1024 As input, the pulse compression method of definition 12 is used to generate S 512×512×1024 The distance signal in the range is compressed to obtain the result Y after the distance pulse compression 512×512×1024 . 512×512×1024 Perform range unit extraction to obtain the pulse compression result Y(i) of the i-th range unit. i is the range sampling number initialized in step 1.
[0134] Step 3. Use the following formula to construct an imaging optimization model based on structured regularization function:
[0135]
[0136] subject to QX i +BD i =0
[0137] Among them, X i , is the solution of the imaging equation corresponding to the constructed i-th distance unit; i is the range sampling number initialized in step 1; Indicates The minimum X i and Y(i) is the pulse compression processing result corresponding to the i-th range unit obtained in step 2; A is the measurement matrix of the array SAR system described in Definition 13; represents the square of the matrix 2 norm; As described in Definition 5 L 1 / 2 norm regularization term; is the scene image described in Definition 7 , |·| represents the absolute value; Q = (I, I), B = (-I, -I), I is the unit matrix; is the auxiliary variable of the variable partitioning method described in Definition 11.
[0138] Step 4. Solve the imaging optimization model based on the structured regularization function
[0139] The standard ADMM method in Definition 9 is used to iteratively solve the imaging solution equation corresponding to each range unit obtained in step 3 to obtain the azimuth-altitude imaging result corresponding to each range unit.
[0140] Step 4.1. Initialize the azimuth-altitude imaging results corresponding to the range unit to be solved 0 512×512 represents a zero matrix with a dimension of 512×512, i is the distance sampling number initialized in step 1; initialize the error parameter ε=0.001; initialize the auxiliary variable (U 1 ) 0 =0 512×512×1024 , Initialize the step size factor μ=0.1; initialize the regularization parameter T=0.5.
[0141] Step 4.2. Use the traditional matched filter-based imaging method in Definition 14 to process the pulse compression result Y(i) obtained in step 2 to obtain the imaging result Y(i) of the i-th distance unit. MF .
[0142] Step 4.3. In the first iteration, perform the following steps:
[0143] (1) Use the following formula to calculate the target scene image of each distance unit
[0144]
[0145] in, is the target scene imaging result of the i-th range unit solved in the first iteration, where i is the range sampling number initialized in step 1; is the azimuth-altitude imaging result corresponding to the distance unit to be solved obtained by initialization in step 4.1; Y(i) MF The imaging result based on matched filtering obtained in step 4.2; is the auxiliary variable for the ith distance unit initialized in step 4.1.
[0146] (2) Use the following formula to calculate the auxiliary variable for each distance unit:
[0147]
[0148] in, is the auxiliary variable of the i-th distance unit obtained by the first iteration, i is the distance sampling number initialized in step 1; L in Definition 6 1 / 2 The proximal operator corresponding to the regular function; is the azimuth-altitude imaging result calculated in step (1); is the auxiliary variable initialized in step 4.1.
[0149] (3) Stack along the distance direction to obtain the first iteration of the three-dimensional sparse imaging result X of the target scene:
[0150] For the azimuth-altitude imaging result corresponding to each range unit of the first iteration obtained in step (1), i is the range sampling number initialized in step 1. The standard range stacking operation method described in Definition 16 is used to perform the range stacking operation to obtain the first iteration of the three-dimensional sparse imaging result X of the target scene. 1 .
[0151] (4) Use the following formula to calculate the auxiliary variable
[0152]
[0153] in, is the auxiliary variable obtained by solving the first iteration, is the proximal operator corresponding to the 3D total variation regularization function described in Definition 8; X 1 is the three-dimensional sparse imaging result X obtained in step (3) 1 ;(U 1 ) 0 is the auxiliary variable initialized in step 4.1.
[0154] Pick The slice matrix of each distance unit of , and the auxiliary variable of the i-th distance unit is obtained
[0155] (5) Use the following formula to calculate the auxiliary variable for each distance unit:
[0156]
[0157] in, and is the auxiliary variable of the i-th distance unit obtained by the first iteration; i is the distance sampling number initialized in step 1; and is the auxiliary variable initialized in step 4.1; is the imaging result of the i-th distance unit solved in step (1); and is the auxiliary variable of the i-th distance unit solved in steps (2) and (4).
[0158] (6) Stack along the distance direction to obtain the three-dimensional auxiliary variable (U 1 ) 1 :
[0159] For each distance unit corresponding to the first iteration obtained in step (5), i is the distance sampling number initialized in step 1. The standard distance stacking operation method described in Definition 16 is used to perform distance stacking operation to obtain the three-dimensional auxiliary variable (U) of the first iteration. 1 ) 1 .
[0160] (7) The iterative error of each distance unit is calculated using the following formula:
[0161]
[0162] Among them, Resi represents the iterative error of this distance unit; is the imaging result of the i-th distance unit obtained in the first iteration of step (1); is the azimuth-altitude imaging result of the range unit to be solved initialized in step 4.1; ||·|| F represents the matrix Fibonacci norm; i is the distance sampling number initialized in step 1.
[0163] (8) Iteration termination condition judgment: if Resi ≤ 0.001, the iteration stops. Resi is the iteration error obtained in step (7). The azimuth-altitude imaging result X corresponding to the i-th range unit is i This is the result of the first iteration. Otherwise, continue iterating. i is the distance sampling number initialized in step 1.
[0164] Step 4.4. In the second iteration, perform the following steps:
[0165] (1) Use the following formula to calculate the target scene image of each distance unit
[0166]
[0167] in, is the target scene imaging result of the i-th range unit solved in the second iteration, where i is the range sampling number initialized in step 1; is the azimuth-altitude imaging result of the first iteration calculated in step 4.3.(1); Y(i) MF The imaging result based on matched filtering obtained in step 4.2; is the auxiliary variable of the first iteration calculated in step 4.3.(2); Step 4.3.(4) Calculate the auxiliary variables of the first iteration; is the auxiliary variable for the first iteration of the calculation in step 4.3.(5).
[0168] (2) Use the following formula to calculate the auxiliary variable for each distance unit:
[0169]
[0170] in, is the auxiliary variable of the i-th distance unit obtained by the second iteration, i is the distance sampling number initialized in step 1; L in Definition 6 1 / 2 The proximal operator corresponding to the regular function; is the azimuth-altitude imaging result of the second iteration calculated in step 4.4.(1); is the auxiliary variable for the first iteration of the calculation in step 4.3.(5).
[0171] (3) Stacking along the distance direction obtains the second iteration of the three-dimensional sparse imaging result X of the target scene:
[0172] For each range unit of the second iteration obtained in step 4.4.(1), the azimuth-altitude imaging result corresponding to i is the range sampling number initialized in step 1. The standard range stacking operation method described in Definition 16 is used to perform the range stacking operation to obtain the second iteration of the three-dimensional sparse imaging result X of the target scene. 2 .
[0173] (4) Use the following formula to calculate the auxiliary variable
[0174]
[0175] in, is the auxiliary variable obtained by solving the second iteration, is the proximal operator corresponding to the 3D total variation regularization function described in Definition 8; X 2 is the three-dimensional sparse imaging result X obtained in step 4.4.(3) 2 ;(U 1 ) 1 is the three-dimensional auxiliary variable (U) of the first iteration obtained in step 4.3.(6) 1 ) 1 .
[0176] Pick The slice matrix of each distance unit of , and the auxiliary variable of the i-th distance unit is obtained
[0177] (5) Use the following formula to calculate the auxiliary variable for each distance unit:
[0178]
[0179]
[0180] in, and is the auxiliary variable of the i-th distance unit obtained by the second iteration; i is the distance sampling number initialized in step 1; and is the auxiliary variable of the first iteration calculated in step 4.3.(5); is the imaging result of the second iteration solved in step 4.4.(1); and is the auxiliary variable of the i-th distance unit solved in steps 4.4.(2) and 4.4.(4).
[0181] (6) Stack along the distance direction to obtain the three-dimensional auxiliary variable (U 1 ) 2
[0182] For each distance unit of the second iteration obtained in step (5), the auxiliary variable corresponding to i is the distance sampling number initialized in step 1. The standard distance stacking operation method described in Definition 16 is used to perform distance stacking operation to obtain the three-dimensional auxiliary variable (U) of the second iteration. 1 ) 2 .
[0183] (7) The iterative error of each distance unit is calculated using the following formula:
[0184]
[0185] Among them, Resi represents the iterative error of this distance unit; is the imaging result of the ith distance unit obtained by the second iteration of step 4.4.(1); is the imaging result of the ith distance unit obtained in the first iteration of step 4.3.(1); ||·|| F represents the matrix Fibonacci norm; i is the distance sampling number initialized in step 1.
[0186] (8) Iteration termination condition judgment: if Resi≤0.001, the iteration stops. Resi is the iteration error obtained in step 4.4.(7), and the azimuth-altitude imaging result X corresponding to the i-th range unit is i This is the result of the second iteration. Otherwise, continue iterating. i is the distance sampling number initialized in step 1.
[0187] Step 4.5. For the kth iteration, perform the following steps:
[0188] (1) Use the following formula to calculate the target scene image of each distance unit
[0189]
[0190] in, is the target scene imaging result of the i-th distance unit solved in the k-th iteration, k is the number of iterations, and i is the range sampling number initialized in step 1; is the azimuth-altitude imaging result obtained in the k-1th iteration; Y(i) MF The imaging result based on matched filtering obtained in step 4.2; is the auxiliary variable obtained at the k-1th iteration.
[0191] (2) Use the following formula to calculate the auxiliary variable for each distance unit:
[0192]
[0193] in, is the auxiliary variable of the i-th distance unit obtained by solving the k-th iteration, and i is the distance sampling sequence number initialized in step 1; L in Definition 6 1 / 2 The proximal operator corresponding to the regular function; is the azimuth-altitude imaging result of the kth iteration calculated in step 4.5.(1); is the auxiliary variable calculated at the k-1th iteration.
[0194] (3) Stack along the distance direction to obtain the k-th iteration of the three-dimensional sparse imaging result X of the target scene:
[0195] For the azimuth-altitude imaging result corresponding to each range unit of the kth iteration obtained in step 4.5.(1), i is the range sampling number initialized in step 1. The standard range stacking operation method described in Definition 16 is used to perform the range stacking operation to obtain the k-th iteration of the three-dimensional sparse imaging result X of the target scene. k .
[0196] (4) Use the following formula to calculate the auxiliary variable
[0197]
[0198] in, is the auxiliary variable obtained by solving the kth iteration, is the proximal operator corresponding to the 3D total variation regularization function described in Definition 8; X k is the three-dimensional sparse imaging result X obtained in step 4.5.(3) k ;(U 1 ) k-1 is the three-dimensional auxiliary variable (U 1 ) k-1 .
[0199] Pick The slice matrix of each distance unit of , and the auxiliary variable of the i-th distance unit is obtained
[0200] (5) Use the following formula to calculate the auxiliary variable for each distance unit:
[0201]
[0202] in, and is the auxiliary variable of the i-th distance unit obtained by the k-th iteration; i is the distance sampling number initialized in step 1; and is the auxiliary variable calculated in the k-1th iteration; is the imaging result of the kth iteration solved in step 4.5.(1); and is the auxiliary variable of the i-th distance unit solved in steps 4.5.(2) and 4.5.(4).
[0203] (6) Stack along the distance direction to obtain the three-dimensional auxiliary variable (U 1 ) k
[0204] For each distance unit corresponding to the kth iteration obtained in step (5), i is the distance sampling number initialized in step 1. The standard distance stacking operation method described in Definition 16 is used to perform distance stacking operation to obtain the three-dimensional auxiliary variable (U) of the kth iteration. 1 ) k .
[0205] (7) The iterative error of each distance unit is calculated using the following formula:
[0206]
[0207] Among them, Resi represents the iterative error of this distance unit; is the imaging result of the i-th distance unit obtained in the k-th iteration of step 4.5.(1); is the imaging result of the i-th distance unit calculated in the k-1th iteration; ||·|| F represents the matrix Fibonacci norm; i is the distance sampling number initialized in step 1.
[0208] (8) Iteration termination condition judgment: if Resi≤0.001, the iteration stops. Resi is the iteration error obtained in step 4.5. (7), and the azimuth-altitude imaging result X corresponding to the i-th range unit isi This is the result of the kth iteration. Otherwise, continue iterating. i is the distance sampling number initialized in step 1.
[0209] Step 5. Stack along the distance to obtain the overall three-dimensional sparse imaging result of the target scene
[0210] For each distance unit obtained in step 4, the azimuth-altitude imaging result X i , i is the range sampling number initialized in step 1, and the standard range stacking operation method described in Definition 15 is used to perform the range stacking operation to obtain the final target overall three-dimensional sparse imaging result X, where X is the final imaging result of the proposed near-field three-dimensional radar sparse imaging method based on structured regular function.
[0211] After computer simulation, peak signal-to-noise ratio and structural similarity were used as evaluation criteria. The results showed that the 1 Compared with the near-field 3D radar sparse imaging method implemented by the regular function and the MCP regular function, this method shows higher signal-to-noise ratio and structural similarity at sampling rates of 100%, 75% and 50%. For details, see Figure 2 ,The simulation results verify that the proposed method has smaller ,estimation deviation, more complete geometric contour preservation, and higher ,image quality.
Claims
1. A near-field three-dimensional radar sparse imaging method based on structured regularization function, characterized in that it The following steps are involved: Step 1. Initialize relevant parameters The azimuth sampling number is recorded as l = 1, 2, ..., L, where L represents the total number of azimuth sampling; the altitude sampling number is recorded as m = 1, 2, ..., M, where M represents the total number of altitude sampling; the range sampling number is recorded as i = 1, 2, ..., N, where N represents the total number of range sampling; the original echo of the target is recorded as S L×M×N ; L 1 / 2 The regularization parameter of the penalty function is denoted as λ1; the regularization parameter of the 3DTV penalty function is denoted as λ2; Step 2. Perform range pulse compression on the target original echo to obtain the range pulse compression result Take the target original echo S in step 1 L×M×N As input, a standard pulse compression method is used to L×M×N The distance signal in the range direction is compressed to obtain the result Y after the distance pulse compression L×M×N ; Y L×M×N Extract each range unit to obtain the pulse compression result Y(i) of the i-th range unit; where i is the range sampling number initialized in step 1, and L, M, and N are the total number of azimuth sampling, total number of altitude sampling, and total number of range sampling initialized in step 1; Step 3. Use the following formula to construct an imaging optimization model based on structured regularization function: subject to QX i +BD i =0 Among them, X i , is the solution of the imaging equation corresponding to the constructed i-th distance unit; i is the range sampling number initialized in step 1; Indicates The minimum X i and Y(i) is the pulse compression processing result corresponding to the i-th range unit obtained in step 2; A is the measurement matrix of the array SAR system; represents the square of the matrix 2 norm; for L 1 / 2 norm regularization term; For scene images , |·| represents the absolute value; λ1 and λ2 are L initialized in step 1. 1 / 2 Regularization parameter of penalty function and regularization parameter of 3DTV penalty function; Q = (I, I), B = (-I, -I), I is the unit matrix; It is an auxiliary variable for the variable segmentation method; Step 4. Solve the imaging optimization model based on the structured regularization function The imaging solution equation corresponding to each distance unit obtained in step 3 is iteratively solved using the standard ADMM method to obtain the azimuth-altitude imaging result corresponding to each distance unit; Step 4.
1. Initialize the azimuth-altitude imaging results corresponding to the range unit to be solved 0 L×M represents an all-zero matrix of dimension L×M, i is the distance sampling number initialized in step 1; initialize the error parameter ε=0.001; initialize the auxiliary variable Initialize the step size factor μ; Initialize the regularization parameter T; Step 4.
2. Use the traditional imaging method based on matched filtering to process the pulse compression result Y(i) obtained in step 2 to obtain the imaging result Y(i) of the i-th distance unit. MF ; Step 4.
3. In the first iteration, perform the following steps: (1) Use the following formula to calculate the target scene image of each distance unit in, is the target scene imaging result of the i-th range unit solved in the first iteration, where i is the range sampling number initialized in step 1; is the azimuth-altitude imaging result corresponding to the range unit to be solved initialized in step 4.1; μ and T are the step size factor and regularization parameter initialized in step 4.1 respectively; Y(i) MF The imaging result based on matched filtering obtained in step 4.2; is the auxiliary variable of the i-th distance unit initialized in step 4.1; (2) Use the following formula to calculate the auxiliary variable for each distance unit: in, is the auxiliary variable of the i-th distance unit obtained by the first iteration, i is the distance sampling number initialized in step 1; For L 1 / 2 The proximal operator corresponding to the regular function; is the azimuth-altitude imaging result calculated in step (1); is the auxiliary variable initialized in step 4.1; (3) Stack along the distance direction to obtain the first iteration of the three-dimensional sparse imaging result X of the target scene: For the azimuth-altitude imaging result corresponding to each range unit of the first iteration obtained in step (1), i is the range sampling number initialized in step 1. The standard range stacking operation method is used to perform the range stacking operation to obtain the first iteration of the three-dimensional sparse imaging result X of the target scene. 1 ; (4) Use the following formula to calculate the auxiliary variable in, is the auxiliary variable obtained by solving the first iteration, is the proximal operator corresponding to the 3D total variation regularization function; X 1 is the three-dimensional sparse imaging result X obtained in step (3) 1 ; (U1) 0 is the auxiliary variable initialized in step 4.1; Pick The slice matrix of each distance unit of , and the auxiliary variable of the i-th distance unit is obtained (5) Use the following formula to calculate the auxiliary variable for each distance unit: in, and is the auxiliary variable of the i-th distance unit obtained by the first iteration; i is the distance sampling number initialized in step 1; and is the auxiliary variable initialized in step 4.1; is the imaging result of the i-th distance unit solved in step (1); and is the auxiliary variable of the i-th distance unit solved in steps (2) and (4); (6) Stack along the distance direction to obtain the three-dimensional auxiliary variable (U1) of the first iteration 1 : For each distance unit corresponding to the first iteration obtained in step (5), i is the distance sampling number initialized in step 1. The standard distance stacking operation method is used to perform distance stacking operation to obtain the three-dimensional auxiliary variable (U1) of the first iteration. 1 ; (7) The iterative error of each distance unit is calculated using the following formula: Among them, Resi represents the iterative error of this distance unit; is the imaging result of the i-th distance unit obtained in the first iteration of step (1); is the azimuth-altitude imaging result of the range unit to be solved initialized in step 4.1; ||·|| F represents the matrix Fibonacci norm; i is the distance sampling number initialized in step 1; (8) Iteration termination condition judgment: if Resi ≤ ε, the iteration stops. Resi is the iteration error obtained in step (7), ε is the error parameter initialized in step 1, and the azimuth-altitude imaging result X corresponding to the i-th range unit is i This is the result of the first iteration. Otherwise, continue to iterate; i is the distance sampling number initialized in step 1; Step 4.
4. In the second iteration, perform the following steps: (1) Use the following formula to calculate the target scene image of each distance unit in, is the target scene imaging result of the i-th range unit solved in the second iteration, where i is the range sampling number initialized in step 1; is the azimuth-altitude imaging result of the first iteration calculated in step 4.3.(1); μ and T are the step size factor and regularization parameter initialized in step 4.1 respectively; Y(i) MF The imaging result based on matched filtering obtained in step 4.2; is the auxiliary variable of the first iteration calculated in step 4.3.(2); Step 4.3.(4) Calculate the auxiliary variables of the first iteration; is the auxiliary variable of the first iteration calculated in step 4.3.(5); (2) Use the following formula to calculate the auxiliary variable for each distance unit: in, is the auxiliary variable of the i-th distance unit obtained by the second iteration, i is the distance sampling number initialized in step 1; For L 1 / 2 The proximal operator corresponding to the regular function; is the azimuth-altitude imaging result of the second iteration calculated in step 4.4.(1); is the auxiliary variable of the first iteration calculated in step 4.3.(5); (3) Stacking along the distance direction obtains the second iteration of the three-dimensional sparse imaging result X of the target scene: For each range unit of the second iteration obtained in step 4.4.(1), the azimuth-altitude imaging result corresponding to i is the range sampling number initialized in step 1. The standard range stacking operation method is used to perform the range stacking operation to obtain the second iteration of the three-dimensional sparse imaging result X of the target scene. 2 ; (4) Use the following formula to calculate the auxiliary variable in, is the auxiliary variable obtained by solving the second iteration, is the proximal operator corresponding to the 3D total variation regularization function; X 2 is the three-dimensional sparse imaging result X obtained in step 4.4.(3) 2 ; (U1) 1 is the three-dimensional auxiliary variable (U1) of the first iteration obtained in step 4.3.(6) 1 ; Pick The slice matrix of each distance unit of , and the auxiliary variable of the i-th distance unit is obtained (5) Use the following formula to calculate the auxiliary variable for each distance unit: in, and is the auxiliary variable of the i-th distance unit obtained by the second iteration; i is the distance sampling number initialized in step 1; and is the auxiliary variable of the first iteration calculated in step 4.3.(5); is the imaging result of the second iteration solved in step 4.4.(1); and is the auxiliary variable of the i-th distance unit solved in steps 4.4.(2) and 4.4.(4); (6) Stack along the distance direction to obtain the three-dimensional auxiliary variable (U1) of the second iteration 2 For each distance unit of the second iteration obtained in step (5), the auxiliary variable corresponding to i is the distance sampling number initialized in step 1. The standard distance stacking operation method is used to perform distance stacking operation to obtain the three-dimensional auxiliary variable (U1) of the second iteration. 2 ; (7) The iterative error of each distance unit is calculated using the following formula: Among them, Resi represents the iterative error of this distance unit; is the imaging result of the i-th distance unit obtained by the second iteration of step 4.4.(1); is the imaging result of the ith distance unit obtained in the first iteration of step 4.3.(1); ||·|| F represents the matrix Fibonacci norm; i is the distance sampling number initialized in step 1; (8) Iteration termination condition judgment: if Resi ≤ ε, the iteration stops. Resi is the iterative error obtained in step 4.
4. (7), ε is the error parameter initialized in step 1, and the azimuth-altitude imaging result X corresponding to the i-th range unit i This is the result of the second iteration. Otherwise, continue to iterate; i is the distance sampling number initialized in step 1; Step 4.
5. For the kth iteration, perform the following steps: (1) Use the following formula to calculate the target scene image of each distance unit in, is the target scene imaging result of the i-th distance unit solved in the k-th iteration, k is the number of iterations, and i is the range sampling number initialized in step 1; is the azimuth-altitude imaging result obtained in the k-1th iteration; μ and T are the step size factor and regularization parameter initialized in step 4.1 respectively; Y(i) MF The imaging result based on matched filtering obtained in step 4.2; and is the auxiliary variable obtained in the k-1th iteration; (2) Use the following formula to calculate the auxiliary variable for each distance unit: in, is the auxiliary variable of the i-th distance unit obtained by solving the k-th iteration, and i is the distance sampling sequence number initialized in step 1; For L 1 / 2 The proximal operator corresponding to the regular function; is the azimuth-altitude imaging result of the kth iteration calculated in step 4.5.(1); is the auxiliary variable calculated in the k-1th iteration; (3) Stack along the distance direction to obtain the k-th iteration of the three-dimensional sparse imaging result X of the target scene: For the azimuth-altitude imaging result corresponding to each range unit of the kth iteration obtained in step 4.5.(1), i is the range sampling number initialized in step 1. The standard range stacking operation method is used to perform the range stacking operation to obtain the k-th iteration of the three-dimensional sparse imaging result X of the target scene. k ; (4) Use the following formula to calculate the auxiliary variable in, is the auxiliary variable obtained by solving the kth iteration, is the proximal operator corresponding to the 3D total variation regularization function; X k is the three-dimensional sparse imaging result X obtained in step 4.5.(3) k ; (U1) k-1 is the three-dimensional auxiliary variable (U1) calculated at the k-1th iteration k-1 ; Pick The slice matrix of each distance unit of , and the auxiliary variable of the i-th distance unit is obtained (5) Use the following formula to calculate the auxiliary variable for each distance unit: in, and is the auxiliary variable of the i-th distance unit obtained by the k-th iteration; i is the distance sampling number initialized in step 1; and is the auxiliary variable calculated in the k-1th iteration; is the imaging result of the kth iteration solved in step 4.5.(1); and is the auxiliary variable of the i-th distance unit solved in steps 4.5.(2) and 4.5.(4); (6) Stack along the distance direction to obtain the three-dimensional auxiliary variable (U1) of the kth iteration k For each distance unit corresponding to the kth iteration obtained in step (5), i is the distance sampling number initialized in step 1. The standard distance stacking operation method is used to perform distance stacking operation to obtain the three-dimensional auxiliary variable (U1) of the kth iteration. k ; (7) The iterative error of each distance unit is calculated using the following formula: Among them, Resi represents the iterative error of this distance unit; is the imaging result of the i-th distance unit obtained in the k-th iteration of step 4.5.(1); is the imaging result of the i-th distance unit calculated in the k-1th iteration; ||·|| F represents the matrix Fibonacci norm; i is the distance sampling number initialized in step 1; (8) Iteration termination condition judgment: if Resi ≤ ε, the iteration stops. Resi is the iterative error obtained in step 4.
5. (7), ε is the error parameter initialized in step 1, and the azimuth-altitude imaging result X corresponding to the i-th range unit i This is the result of the kth iteration. Otherwise, continue to iterate; i is the distance sampling number initialized in step 1; Step 5. Stack along the distance to obtain the overall three-dimensional sparse imaging result of the target scene For each distance unit obtained in step 4, the azimuth-altitude imaging result X i , i is the range sampling number initialized in step 1, and the standard range stacking operation method is used to perform the range stacking operation to obtain the final target overall three-dimensional sparse imaging result X, where X is the final imaging result of the proposed near-field three-dimensional radar sparse imaging method based on structured regularization function.
Citation Information
Cited By
Sparse near-field SAR (Synthetic Aperture Radar) imaging method based on multi-constraint index high-order total variation
CN121208824A
