A phase information guided real-aperture radar super-resolution imaging method

Abstract

The application discloses a kind of phase information guiding real aperture radar super-resolution imaging method, first establish multiple frames echo model, then construct interframe consistency mask based on multiple frames phase sequence, then construct super-resolution imaging model based on missing data mask matrix, finally based on the super-resolution reconstruction of missing data of generalized matrix norm, realize the multiple frames super-resolution imaging of real aperture radar.The echo layering strategy of interframe phase guidance proposed in the method of the application can effectively separate stable scatterer and moving scatterer and background clutter, the super-resolution imaging model based on missing data mask matrix can maintain the performance of super-resolution imaging under the condition of incomplete data, the optimization strategy of iterative weighted least squares and fast iterative threshold algorithm combination can realize efficient and robust solution, compared with existing method, can suppress the disturbance of background clutter and moving target on super-resolution imaging performance, improve the interframe robustness of stable scattering target.

Description

Technical Field

[0001] This invention belongs to the field of radar detection and imaging technology, specifically relating to a phase information-guided real aperture radar super-resolution imaging method. Background Technology

[0002] Real aperture radar (SAR) relies on antenna scanning to achieve Earth observation, offering unique advantages in applications such as forward-looking and omnidirectional ground building mapping. However, in ground building mapping, the angular resolution of SAR is limited by the physical antenna aperture, making it difficult to meet the high-precision mapping requirements. Furthermore, existing single-frame super-resolution imaging methods are susceptible to background clutter and moving targets, leading to a decrease in imaging resolution and stability. Existing multi-frame super-resolution imaging research struggles to effectively distinguish between stable rigid bodies and unstable scattering bodies.

[0003] To overcome antenna aperture limitations, researchers commonly employ super-resolution inversion imaging techniques to improve angular resolution without increasing hardware complexity. Existing super-resolution methods can be broadly categorized into Bayesian methods and regularization methods. The paper "H. Chen, J. Yu, W. Zhang, Y. Li, J. Li, L. Cai and Y. Lu, 'Probabilitymodel-driven airborne bayesian forward-looking super-resolution imaging for multitarget scenario' Journal of Radars, vol. 12, pp. 1125-1137, 2023" proposes a probability model-driven airborne Bayesian forward-looking super-resolution multitarget imaging method. Based on a Bayesian framework, it achieves sparse forward-looking imaging, which can suppress clutter conforming to a specific statistical distribution to a certain extent. The paper "W. Li, W. Zhang, Q. Zhang, Y. Zhang, Y. Huang and J. Yang, 'Simultaneous super-resolution and target detection of forward-looking scanning radar via low-rank and sparsity constrained method,' IEEE Transactions on Geoscience and Remote Sensing, vol. 58, pp. 7085-7095, 2020" proposes an algorithm for simultaneous super-resolution imaging and target detection. It introduces low-rank and sparsity constraints as regularization criteria into forward-looking scanning radar imaging, transforming the super-resolution imaging and target detection problem into an optimization problem, providing an important means for refined mapping and scene understanding. However, in omnidirectional mapping applications targeting ground buildings, single-frame echoes often contain moving scatterers such as people and vehicles, as well as background clutter information such as vegetation undulations. These scatterings, unrelated to the building structure, not only create a large number of non-interesting targets in the imaging results but also violate the assumptions of sparsity or smoothness in the super-resolution model, leading to a significant decrease in resolution and stability. By jointly modeling multiple frames of echo data, the spatial sparsity can be enhanced or the impact of noise fluctuations can be mitigated to some extent.The paper "K. Tan, S. Zhou, X. Lu, J. Yang, H. Gu and W. Su, 'A multi-frame super-resolution imaging method for forward-looking scanning radar,' in IGARSS 2024-2024 IEEE International Geoscience and Remote Sensing Symposium, 2024, pp. 10554-10557" proposes a multi-frame super-resolution imaging method for forward-looking scanning radar. It constructs a high-resolution image by utilizing complementary information from multiple real-beam images. However, its emphasis on the synergistic utilization of amplitude information fails to fully exploit the phase evolution patterns of different scatterers between frames, thus making it difficult to effectively distinguish between stable rigid bodies and unstable scatterers. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides a phase information-guided real aperture radar super-resolution imaging method, which can significantly improve the inter-frame consistency and imaging resolution of ground building mapping in complex scenarios. It is applicable to multi-frame super-resolution imaging of ground buildings by real aperture radar in complex scenarios.

[0005] The technical solution adopted in this invention is: a phase information-guided real aperture radar super-resolution imaging method, the specific steps of which are as follows:

[0006] Step 1: The real aperture scanning radar is deployed on a fixed platform and performs mechanical scanning in the azimuth direction through a servo antenna. During the scanning process, the radar periodically transmits linear frequency modulated signals to establish a multi-frame echo model of the area to be mapped. After range pulse compression, the echo signal is represented as the convolution of target scattering and the measurement matrix.

[0007] Step 2: Construct an inter-frame consistency mask based on the phase sequence of multiple frames by utilizing the phase consistency of the stable scatterer in multiple frames of images;

[0008] Step 3: Use a consistent mask to separate stable scatterers from background clutter and moving targets, and construct a super-resolution imaging model based on the mask matrix of missing data;

[0009] Step 4: Introduce the generalized mixed matrix norm regularization term to jointly constrain the sparsity and structural continuity of the solution, and perform super-resolution reconstruction of missing data based on the generalized mixed matrix norm. The solution is obtained by combining the optimization strategy of iterative weighted least squares and fast iterative thresholding algorithm.

[0010] Furthermore, step 1 is detailed as follows:

[0011] Assume the radar antenna operates at a constant angular velocity. Perform a scan; antenna elevation angle is... azimuth angle is Then the echo of a target at a certain point in the area to be surveyed The expression is as follows:

[0012]

[0013] in, Represents the target scattering coefficient. Indicates the phase of the target point. Indicates slow time. Indicates a fast time. Represents the antenna pattern function. Indicates two-way echo delay. Indicates the pulse width. Indicates the frequency modulation slope. Indicates the radar carrier frequency. Represents the imaginary unit. , It is a window function in the time domain;

[0014] And if there are multiple scattering units in the scene, then the first Frame echo The expression is as follows:

[0015]

[0016] in, Indicates the scan range. This represents the total number of observation frames.

[0017] Then, after distance pulse pressure, the first... The echo signal of the frame is represented as the convolution of the target scattering and the measurement matrix, as shown in the following expression:

[0018]

[0019] in, , , , They represent the first The frame echo, target scattering amplitude matrix, target scattering phase matrix, and noise matrix. Indicates the first The frame is a measurement matrix determined by the antenna pattern and scanning modulation.

[0020] Furthermore, step 2 is detailed as follows:

[0021] The targets are divided into stable scatterers and unstable scatterers. The echo signal of the frame can be written as the following expression:

[0022]

[0023] in, and They represent the first Scattering coefficient and phase of a frame-stable scatterer and They represent the first The scattering coefficient and phase of an unstable scatterer. This represents the measurement matrix determined by both the antenna pattern and the scanning modulation.

[0024] Then, considering the characteristic of stable scatterers maintaining phase consistency across frames, an inter-frame consistency mask based on multi-frame phase sequences is constructed. For the ... The first frame of echo Each sampling point, the phase change between adjacent frames The expression is defined as follows:

[0025]

[0026] in, Indicates the first The first echo Phase values ​​at each sampling point;

[0027] When the phase change Less than the preset phase threshold At that time, it is assumed that the point corresponds to a stable scatterer, forming a set of stable scatterer points. The corresponding inter-frame consistency mask matrix The expression is defined as follows:

[0028] .

[0029] Furthermore, step 3 is detailed as follows:

[0030] By using a uniform mask and processing targets with different motion amplitudes into layers based on different phase thresholds, the layered echo data expression is obtained as follows:

[0031]

[0032] in, Indicates the first The first frame Layer data, Indicates the first The inter-frame consistency mask matrix of the layer, This represents the Hadamard product, and only considers the distinction between stable and unstable scatterers, i.e., it is divided into only two layers. ;

[0033] Based on the inter-frame consistency mask matrix To construct the defect measurement matrix, firstly, consistent sampling is performed on the echo sampling points with a mask value of 1 along the azimuth direction, as shown in the following expression:

[0034]

[0035] in, Indicates the number of extractions after the first extraction. Layered echoes of each distance cell, Indicates the first The azimuth mask vector of a distance cell, , Indicates the distance from the sampling points. Indicates the number of azimuth sampling points;

[0036] Then, based on the correspondence between the measurement matrix and the echo matrix, Extraction to form a defect measurement matrix The expression is as follows:

[0037]

[0038] in, Indicates the first Layer The defect measurement matrix of each distance cell, Represents a vector consisting entirely of 1s. Represents the Kronecker product;

[0039] The final expression for the separated echo is as follows:

[0040]

[0041] in, Indicates the first Layer Scattering coefficient per unit distance, Indicates the first Layer Noise per distance unit.

[0042] Furthermore, step 4 is detailed as follows:

[0043] First, we construct a generalized mixed matrix norm regularization solution model, expressed as follows:

[0044]

[0045] in, , , Representing the sets respectively The corresponding number Layer echo data, and ensemble The corresponding number Layer measurement matrix, and set The corresponding number Layer target scattering coefficient, Represents the regularization parameter. It represents the 2-norm of a vector. Denotes the norm of a mixed matrix. and These represent the norm order of two different dimensions;

[0046] A solution strategy combining iterative weighted least squares and fast iterative thresholding is adopted. By introducing a weight matrix to approximate the mixture norm twice, the solution model is transformed into a series of weighted least squares subproblems, as shown in the following expression:

[0047]

[0048] in, Let r represent the weighted least squares subproblem in the r-th iteration. Indicates the number of iterations. Indicates the first The weighted values ​​of each scattering point;

[0049] when When there are multiple columns, differentiate the weighted least squares subproblem and set its gradient to 0 to obtain the final result for each column. The iterative update formula is as follows:

[0050]

[0051] in, This indicates the transpose operation. For the first The diagonal weight matrix at the next iteration has diagonal elements that are the weighted values ​​of each distance cell in the current iteration step. , express The The process is repeated until convergence, thus completing the solution.

[0052] Furthermore, the phase threshold is determined based on the maximum allowable micro-motion amplitude of the target and the system wavelength.

[0053] Furthermore, the phase threshold is specifically calculated in the following manner:

[0054] Set the radar carrier frequency to The corresponding wavelength The expression is as follows:

[0055]

[0056] in, Indicates the speed of electromagnetic wave propagation;

[0057] When there are moving targets in the scene, the inter-frame phase change amount The expression is as follows:

[0058]

[0059] in, Indicates the first Frame and the The change in distance between frames; the corresponding phase threshold is calculated by limiting the magnitude of the change in distance.

[0060] Furthermore, the solution model is transformed into a series of weighted least squares subproblems, as follows:

[0061] The specific expression for the norm of a mixed matrix is ​​as follows:

[0062]

[0063] First, define the row norm scale. The expression is as follows:

[0064]

[0065] The regularization term can then be written as the following expression:

[0066]

[0067] in, Represents the regularization term; uses the current iteration value. The weights are constructed using the following expression:

[0068]

[0069]

[0070] in, Indicates the first Row weights Represents a small positive number. Indicates the first The energy of movement, This represents the number of iterations; the following expression is obtained through a first-order approximation:

[0071]

[0072] Then for each element After a second approximation, we obtain the following expression:

[0073]

[0074] The entire regularization term can then be approximated as a weighted square, as shown in the following expression:

[0075]

[0076] in, Indicates the first The weighted value of each scattering point.

[0077] Substituting the entire regularization term into the solution model transforms it into a standard weighted least squares problem.

[0078] The beneficial effects of this invention are as follows: The method of this invention first establishes a multi-frame echo model, then constructs an inter-frame consistency mask based on the multi-frame phase sequence, then constructs a super-resolution imaging model based on the missing data mask matrix, and finally performs super-resolution reconstruction of the missing data based on the generalized matrix norm, thereby realizing multi-frame super-resolution imaging of real aperture radar. The inter-frame phase-guided echo layering strategy proposed in this invention can effectively separate stable scatterers from moving scatterers and background clutter. The super-resolution imaging model based on the missing data mask matrix can maintain the performance of super-resolution imaging under incomplete data conditions. The optimization strategy combining iterative weighted least squares and fast iterative thresholding algorithm can achieve efficient and robust solutions. Compared with existing methods, it can suppress the disturbance of background clutter and moving targets on the performance of super-resolution imaging and improve the inter-frame robustness of stable scattering targets. Attached Figure Description

[0079] Figure 1 This is a flowchart of a phase information-guided real aperture radar super-resolution imaging method according to the present invention.

[0080] Figure 2 This is a schematic diagram of a real aperture radar mapping scenario in an embodiment of the present invention.

[0081] Figure 3 This is a schematic diagram of a multi-frame simulation scene in an embodiment of the present invention.

[0082] Figure 4 This is a schematic diagram of multi-frame simulated echoes in an embodiment of the present invention.

[0083] Figure 5 This is a schematic diagram of the phase and amplitude changes between target frames in an embodiment of the present invention.

[0084] Figure 6 This is a schematic diagram of the simulation results of multi-frame imaging using the L1 method in an embodiment of the present invention.

[0085] Figure 7 This is a schematic diagram of the simulation results of multi-frame imaging using the MAP method in an embodiment of the present invention.

[0086] Figure 8 This is a schematic diagram of the simulation results of multi-frame imaging using the L1-rank method in an embodiment of the present invention.

[0087] Figure 9 This is a schematic diagram of the multi-frame imaging simulation results of the post-detection reconstruction method in an embodiment of the present invention.

[0088] Figure 10 This is a schematic diagram of the multi-frame imaging simulation experiment results of the method of the present invention in an embodiment of the present invention. Detailed Implementation

[0089] The method of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0090] This embodiment uses a real-world experimental method for verification, and all steps and conclusions are presented in [the document / document / etc.]. The above verification is correct, such as Figure 1 The flowchart of a phase information-guided real aperture radar super-resolution imaging method of the present invention is shown below, and the specific steps are as follows:

[0091] Step 1: Establish a multi-frame echo model for ground observation using a real aperture radar with a servo antenna;

[0092] The scanning process of a real aperture radar for Earth observation using a servo antenna is as follows: Figure 2 As shown. The real aperture scanning radar is deployed on a fixed platform and performs mechanical scanning in the azimuth direction via a servo antenna to achieve omnidirectional observation of the ground scene. Assume the radar antenna operates at a constant angular velocity. Perform a scan; antenna elevation angle is... azimuth angle is During the scanning process, the radar periodically transmits linear frequency modulated signals, and the echo of a target at a certain point in the area to be mapped will be detected. The expression is as follows:

[0093] Equation (1);

[0094] in, Represents the target scattering coefficient. Indicates the phase of the target point. Indicates slow time. Indicates a fast time. Represents the antenna pattern function. Indicates two-way echo delay. Indicates the pulse width. Indicates the frequency modulation slope. Indicates the radar carrier frequency. Represents the imaginary unit. , It is a window function in the time domain.

[0095] And if there are multiple scattering units in the scene, then the first Frame echo The expression is as follows:

[0096] Equation (2);

[0097] in, Indicates the scan range. This represents the total number of observation frames.

[0098] Then, after distance pulse pressure, the first... The echo signal of the frame is represented as the convolution of the target scattering and the measurement matrix, as shown in the following expression:

[0099] Equation (3);

[0100] in, , , , They represent the first The frame echo, target scattering amplitude matrix, target scattering phase matrix, and noise matrix. Indicates the first The frame is a measurement matrix determined by the antenna pattern and scanning modulation.

[0101] Step 2: Based on Step 1, utilize the phase consistency of the stable scatterer in multiple frames of images to construct an inter-frame consistency mask based on the phase sequence of multiple frames;

[0102] In multi-frame observations using real aperture radar, the transmitted radar signal remains constant between frames, while the received noise exhibits statistically independent characteristics between frames. For different types of scatterers, the amplitude and phase evolution patterns differ significantly between frames.

[0103] For stable rigid targets such as ground buildings and streetlights, their physical dimensions are much larger than the radar wavelength, and their geometry and scattering center position remain essentially unchanged over short timescales. Therefore, their inter-frame scattering amplitude changes are small, and the phase changes are mainly caused by system noise, which can be approximated as constant. In contrast, flexible scatterers such as leaves and flowers are affected by wind disturbances, causing their scattering centers to undergo small but random displacements between frames; furthermore, their scattering intensity is affected by attitude changes, changes in coherent superposition conditions, and local geometric changes, exhibiting significant fluctuations between frames. Moving targets such as vehicles and pedestrians produce significant displacements between frames. These displacements cause noticeable phase changes in the millimeter-wave or centimeter-wave bands.

[0104] Then equation (3) can be rewritten as the following expression:

[0105] Equation (4);

[0106] in, and They represent the first Scattering coefficient and phase of a frame-stable scatterer and They represent the first The scattering coefficient and phase of an unstable scatterer. This represents the measurement matrix determined by both the antenna pattern and the scanning modulation.

[0107] Then, considering the characteristic of stable scatterers maintaining phase consistency across frames, an inter-frame consistency mask based on multi-frame phase sequences is constructed. Therefore, for the ... The first frame of echo Each sampling point, the phase change between adjacent frames The expression is defined as follows:

[0108] Equation (5);

[0109] in, Indicates the first The first echo Phase value of each sampling point.

[0110] When the phase change is less than the preset phase threshold If a point is considered to correspond to a stable scatterer, then a stable scatterer point set is defined. The expression is as follows:

[0111] Equation (6);

[0112] Corresponding inter-frame consistency mask matrix The expression is defined as follows:

[0113] Equation (7);

[0114] Phase threshold The phase threshold is determined based on the target's maximum permissible micro-motion amplitude and the system wavelength, thus possessing clear physical meaning. To determine the set value of the phase threshold, the radar carrier frequency is set to... The corresponding wavelength The expression is as follows:

[0115] Equation (8);

[0116] in, This indicates the speed of electromagnetic wave propagation.

[0117] When there are moving targets in the scene, the inter-frame phase change amount The expression is as follows:

[0118] Equation (9);

[0119] in, Indicates the first Frame and the The change in distance between frames. For a stable rigid body, since... Approximately zero, the phase remains almost unchanged between frames; for flexible scatterers such as leaves and flowers, although their displacement is usually small, the relatively short wavelength still results in a significant phase change. Meanwhile, for saliently moving targets such as people and vehicles... The phase difference is relatively large, resulting in more dramatic inter-frame phase changes. Therefore, in practical applications, The threshold setting should take into account the micro-motion characteristics of the object to be measured under normal conditions. The corresponding phase threshold can be calculated by limiting the magnitude of the distance change.

[0120] Step 3: Based on Step 2, use a consistent mask to separate stable scatterers from background clutter and moving targets, and then construct a super-resolution imaging model based on the mask matrix of missing data to maintain the performance of super-resolution imaging under the condition of incomplete data.

[0121] Using a consistency mask, the first Frame echoes are layered; that is, in practical applications, targets with different motion amplitudes are layered according to different phase thresholds. The resulting layered echo data expression is as follows:

[0122] Equation (10);

[0123] in, Indicates the first The first frame Layer data, Indicates the first The inter-frame consistency mask matrix of the layer, This represents the Hadamard product. Furthermore, it only considers the distinction between stable rigid bodies and unstable scatterers, i.e., it divides them into only two layers. .

[0124] Because masking operations will set some sampling points to zero, i.e. Echo data The effect is that some echo data points have amplitude and phase values ​​of 0. This results in data loss in both the range and azimuth directions of the layered echo. However, since antenna beam scanning is a continuous modulation process, and In the measurement matrix The corresponding sampling points are not zero. If the original measurement matrix is ​​directly used for super-resolution reconstruction, a measurement model mismatch problem will occur, leading to a decrease in resolution or even imaging failure. Therefore, based on the inter-frame consistency mask matrix... To construct the defect measurement matrix, firstly, consistent sampling is performed on the echo sampling points with a mask value of 1 along the azimuth direction, as shown in the following expression:

[0125] , Equation (11);

[0126] in, Indicates the number of extractions after the first extraction. Layered echoes of each distance cell, Indicates the first An azimuth mask vector with distance units, whose elements are either 0 or 1. , Indicates the distance from the sampling points. This indicates the number of azimuth sampling points.

[0127] Then, based on the correspondence between the measurement matrix and the echo matrix, Extraction to form a defect measurement matrix The expression is as follows:

[0128] Equation (12);

[0129] in, Indicates the first Layer The defect measurement matrix of each distance cell, Represents a vector consisting entirely of 1s. This represents the Kronecker product.

[0130] The final expression for the separated echo is as follows:

[0131] Equation (13);

[0132] in, Indicates the first Layer Scattering coefficient per unit distance, Indicates the first Layer Noise per distance unit.

[0133] Step 4: Based on Step 3, the sparsity and structural continuity of the solution are jointly constrained by the generalized mixed matrix norm regularization term. Super-resolution reconstruction of missing data based on the generalized mixed matrix norm is carried out. An optimization strategy combining iterative weighted least squares and fast iterative thresholding algorithm is used to achieve efficient and robust solution.

[0134] Echo data are separated in both the range and azimuth dimensions, resulting in sparsity in these dimensions. However, some stable building targets exhibit continuity within the layered echoes, causing the layered scattered field to simultaneously exhibit sparsity and structural continuity in both the range and azimuth dimensions. Therefore, this embodiment first constructs a generalized hybrid matrix norm regularization solution model, expressed as follows:

[0135] , Equation (14);

[0136] in, , , , Representing the sets respectively The corresponding number Layer echo data, and ensemble The corresponding number Layer measurement matrix, and set The corresponding number Layer target scattering coefficient, Represents the regularization parameter. It represents the 2-norm of a vector. The norm of a mixture matrix is ​​expressed as follows:

[0137] Equation (15);

[0138] in, and These represent the norm order of two different dimensions.

[0139] Then, regarding the nonconvexity of equation (14), this embodiment adopts a solution strategy combining iterative weighted least squares and fast iterative thresholding algorithm. By introducing a weight matrix to perform a second approximation of the mixed norm, the problem is transformed into a series of weighted least squares subproblems, and each column is solved independently to achieve efficient iterative updates until convergence, as follows:

[0140] First, define the row norm scale. The expression is as follows:

[0141] , Equation (16);

[0142] The regularization term can then be written as the following expression:

[0143] , Equation (17);

[0144] in, This represents the regularization term.

[0145] To facilitate the solution, the current iteration value will be used again. The weights are constructed using the following expression:

[0146] Equation (18);

[0147] , Equation (19);

[0148] in, Indicates the first Row weights Represents a small positive number. Indicates the first The energy of movement, Indicates the number of iterations.

[0149] And through a first-order approximation, the following expression is obtained:

[0150] Equation (20);

[0151] Then for each element After a second approximation, we obtain the following expression:

[0152] , Equation (21);

[0153] The entire regularization term can then be approximated as a weighted square, as shown in the following expression:

[0154] Equation (22);

[0155] in, Indicates the first The weighted value of each scattering point.

[0156] Equation (14) then becomes a standard weighted least squares problem. The expression is as follows:

[0157] Equation (23);

[0158] when When there are multiple columns, differentiate equation (23) and set its gradient to 0 to obtain the final value for each column. The iterative update formula is as follows:

[0159] , formula (24);

[0160] in, This indicates the transpose operation. For the first The diagonal weight matrix at the next iteration, its diagonal elements , , , These represent the weight coefficients of each distance unit in the current iteration step. , express The List.

[0161] Finally, repeat the iterative formula (24) until... The solution is now complete.

[0162] in, This represents a preset small value.

[0163] To verify the effectiveness of the method of this invention, this embodiment also constructs a multi-frame simulation scene including a stable rigid target, a moving target, and complex background clutter, to simulate a ground mapping task in a real urban scene. The scene includes four frames of observation data, including two rows of streetlights as fixed-position, structurally stable rigid building targets; moving vehicles that shift along the road direction in adjacent frames; and two rows of trees whose trunks are stable, but whose leaves undergo random micro-motions between frames due to wind disturbance. The specific details of frames 1-4 are as follows: Figure 3 As shown in (a), (b), (c), and (d).

[0164] Then, based on the set radar system parameters, corresponding multi-frame real aperture scanning radar echoes are generated. The radar system parameters are shown in Table 1.

[0165] Table 1

[0166] carrier frequency 9.6GHz bandwidth 30MHz Scan speed 60° / s Scan range ±10° Main lobe beamwidth 2° Pulse repetition frequency 1000Hz

[0167] Figure 4 (a), (b), (c), and (d) in the figure represent the echoes of frames 1-4, respectively. As can be seen from the echoes in the figure, due to the limitation of the azimuth resolution of the real aperture radar by the antenna aperture, the echoes of streetlights, vehicles, and trees are significantly broadened in the real beam domain, and some targets are aliased in the range-azimuth plane, which increases the difficulty of subsequent super-resolution imaging.

[0168] Figure 5 This shows the phase and amplitude changes of the target between four echo frames, where Figure 5 (a) in the figure represents the phase change between target frames. Figure 5(b) shows the inter-frame amplitude variation of the target. Analysis of the inter-frame amplitude and phase variations in the four frames of echo data reveals that the phase at the streetlight location remains highly stable across frames, exhibiting only weak random perturbations. While its amplitude fluctuates slightly due to scattering from nearby vehicles and trees, the overall structure remains consistent. In contrast, the amplitude and phase in the vehicle and leaf regions show significant changes across frames, with the magnitude of these changes increasing with the target's motion. This phenomenon validates the physical rationale for distinguishing stable rigid bodies from unstable scatterers based on inter-frame phase consistency.

[0169] The proposed method is then compared with L1 sparse regularization, maximum a posteriori probability estimation (MAP) Bayesian method, low-rank sparse decomposition (L1-rank) method, and detection-after-reconstruction (DBR) method. The multi-frame imaging results are shown below. Figure 6 , Figure 7 , Figure 8 , Figure 9 and Figure 10 As shown.

[0170] The imaging results of frames 1-4 of the L1 method are as follows: Figure 6 As shown in (a), (b), (c), and (d), targets such as streetlights, vehicles, and trees can be resolved. However, the echoes from streetlights, trees, and vehicles interfere with each other, resulting in differences in the recovered shapes of these targets between frames. Streetlights, being stable rigid structures, exhibit poor inter-frame stability. The imaging results for frames 1-4 using the MAP method are shown below. Figure 7 As shown in (a), (b), (c), and (d) in the figure, similar to the L1 method, the inter-frame stability of streetlights, a stable rigid body structure, is poor. Figure 8 Images (a), (b), (c), and (d) in the figure show the imaging results of frames 1-4 using the L1-rank method, which can distinguish between low-rank background and sparse targets. Figure 8 As can be seen, some strong sparse targets are well resolved, but the sparse targets in each frame are random and lack stability, making it impossible to obtain stable imaging results for streetlights. The imaging results of frames 1-4 using the Detection-Based Reconstruction (DBR) method are shown below. Figure 9 In examples (a), (b), (c), and (d), because the detection and reconstruction are based on simple amplitude threshold detection, they can only filter out weak noise and some small target echoes, so strong point targets can be distinguished. However, due to the lack of some echo information, the resolution effect deteriorates, and even targets are lost, resulting in poor stability of streetlight imaging between frames. Figure 10 As can be seen from (a), (b), (c), and (d) in the figure, the method proposed in this invention can effectively extract stable rigid targets, namely streetlights. High-resolution imaging of two rows of streetlights is achieved, and the imaging results of the streetlights are very stable between frames.

[0171] Then, this embodiment uses the Normalized Cross-Correlation (NCC) index and the Structural Similarity Index (SSIM) index to quantitatively evaluate the consistency of the scattering structure. Let the first... Frame and the The imaging results of the frames are as follows and And expand it into a vector within the target region. , The NCC definition expression is as follows:

[0172] Equation (25);

[0173] in, Representing vectors The mean.

[0174] The SSIM definition expression is as follows:

[0175] , Equation (26);

[0176] in, yes The average value, yes The average value, yes variance yes variance yes and covariance, and This represents a constant used to maintain stability.

[0177] To demonstrate that the method of this invention can still restore the overall scene after layered processing, this embodiment also evaluates the similarity between the fused imaging result and the original scene. This evaluation uses the two metrics mentioned above, comparing the NCC and SSIM between the resulting imaging frames and the scene frames. The comparison results are shown in Table 2.

[0178] Table 2 Comparison of imaging similarity evaluation results of different methods on each frame.

[0179]

[0180] As shown in Table 2, the MAP method has the lowest NCC (0.4941) in the second frame, while the L1-rank method has the lowest SSIM (0.5850) in the second frame. The proposed method, however, has a higher NCC in every frame than the other comparative methods, reaching a maximum of 0.6592, and its SSIM is also higher in every frame than the other comparative methods, reaching a maximum of 0.7723, demonstrating that the proposed method has the best full-scene mapping capability.

[0181] In summary, the method of this invention effectively separates stable scatterers from moving scatterers and background clutter by mining the phase consistency characteristics of stable rigid structures in multi-frame observations. Furthermore, addressing the data loss problem introduced by layered processing, a super-resolution imaging model based on a missing data mask matrix is ​​established to maintain super-resolution imaging performance under incomplete data conditions. A generalized hybrid matrix norm regularization term is introduced to jointly constrain the sparsity and structural continuity of the solution. Finally, an optimization strategy combining iterative weighted least squares and a fast iterative thresholding algorithm achieves efficient and robust solution. Compared with existing methods, the method of this invention can suppress the perturbation of super-resolution imaging performance by background clutter and moving targets, and improve the inter-frame robustness of stable scattering targets.

[0182] Those skilled in the art will recognize that the embodiments described herein are intended to help the reader understand the principles of the invention, and should be understood that the scope of protection of the invention is not limited to such specific statements and embodiments. Various modifications and variations can be made to the invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the invention should be included within the scope of the claims of the invention.

Claims

1. A phase information-guided real aperture radar super-resolution imaging method, characterized in that, Includes the following steps: Step 1: The real aperture scanning radar is deployed on a fixed platform and performs mechanical scanning in the azimuth direction through a servo antenna. During the scanning process, the radar periodically transmits linear frequency modulated signals to establish a multi-frame echo model of the area to be mapped. After range pulse compression, the echo signal is represented as the convolution of target scattering and the measurement matrix. Step 2: Construct an inter-frame consistency mask based on the phase sequence of multiple frames by utilizing the phase consistency of stable scatterers in multi-frame images; Step 3: Use a consistent mask to separate stable scatterers from background clutter and moving targets, and construct a super-resolution imaging model based on the mask matrix of missing data; Step 4: Introduce the generalized mixed matrix norm regularization term to jointly constrain the sparsity and structural continuity of the solution, and perform super-resolution reconstruction of missing data based on the generalized mixed matrix norm. Solve the problem by combining iterative weighted least squares and fast iterative thresholding algorithm.

2. The phase information-guided real aperture radar super-resolution imaging method according to claim 1, characterized in that, Step 1 is described in detail as follows: Assume the radar antenna operates at a constant angular velocity. Perform a scan; antenna elevation angle is... azimuth angle is Then the echo of a target at a certain point in the area to be surveyed The expression is as follows: in, Represents the target scattering coefficient. Indicates the phase of the target point. Indicates slow time. Indicates a fast time. Represents the antenna pattern function. Indicates two-way echo delay. Indicates the pulse width. Indicates the frequency modulation slope. Indicates the radar carrier frequency. Represents the imaginary unit. , It is a window function in the time domain; And if there are multiple scattering units in the scene, then the first Frame echo The expression is as follows: in, Indicates the scan range. Indicates the total number of observation frames; Then, after distance pulse pressure, the first The echo signal of the frame is represented as the convolution of the target scattering and the measurement matrix, as shown in the following expression: in, , , , They represent the first The frame echo, target scattering amplitude matrix, target scattering phase matrix, and noise matrix; Indicates the first The frame is a measurement matrix determined by the antenna pattern and scanning modulation.

3. The phase information-guided real aperture radar super-resolution imaging method according to claim 2, characterized in that, Step 2 is described in detail below: The targets are divided into stable scatterers and unstable scatterers. The echo signal of the frame can be written as the following expression: in, and They represent the first Scattering coefficient and phase of a frame-stable scatterer and They represent the first The scattering coefficient and phase of an unstable scatterer. This represents the measurement matrix determined by both the antenna pattern and the scanning modulation. Then, considering the characteristic of stable scatterers maintaining phase consistency across frames, an inter-frame consistency mask based on multi-frame phase sequences is constructed. For the ... The first frame of echo Each sampling point, the phase change between adjacent frames The expression is defined as follows: in, Indicates the first The first echo Phase values ​​at each sampling point; When the phase change Less than the preset phase threshold At that time, it is assumed that the point corresponds to a stable scatterer, forming a set of stable scatterer points. The corresponding inter-frame consistency mask matrix The expression is defined as follows: 。 4. The phase information-guided real aperture radar super-resolution imaging method according to claim 3, characterized in that, Step 3 is described in detail below: By using a uniform mask and processing targets with different motion amplitudes into layers based on different phase thresholds, the layered echo data expression is obtained as follows: in, Indicates the first The first frame Layer data, Indicates the first The inter-frame consistency mask matrix of the layer, This represents the Hadamard product, and only considers the distinction between stable and unstable scatterers, i.e., it is divided into only two layers. ; Based on the inter-frame consistency mask matrix To construct the defect measurement matrix, firstly, consistent sampling is performed on the echo sampling points with a mask value of 1 along the azimuth direction, as shown in the following expression: in, Indicates the number of extractions after the first extraction. Layered echoes of each distance cell, Indicates the first The azimuth mask vector of a distance cell, , Indicates the distance from the sampling points. Indicates the number of azimuth sampling points; Then, based on the correspondence between the measurement matrix and the echo matrix, Extraction to form a defect measurement matrix The expression is as follows: in, Indicates the first Layer The defect measurement matrix of each distance cell, Represents a vector consisting entirely of 1s. Represents the Kronecker product; The final expression for the separated echo is as follows: in, Indicates the first Layer Scattering coefficient per unit distance, Indicates the first Layer Noise per distance unit.

5. The phase information-guided real aperture radar super-resolution imaging method according to claim 4, characterized in that, Step 4 is described in detail below: First, we construct a generalized mixed matrix norm regularization solution model, expressed as follows: in, , , Representing the sets respectively The corresponding number Layer echo data, and set The corresponding number Layer measurement matrix, and set The corresponding number Layer target scattering coefficient, Represents the regularization parameter. The 2-norm of a vector; Denotes the norm of a mixed matrix. and These represent the norm orders of two different dimensions; A solution strategy combining iterative weighted least squares and fast iterative thresholding is adopted. By introducing a weight matrix to approximate the mixture norm twice, the solution model is transformed into a series of weighted least squares subproblems, as shown in the following expression: in, Let r represent the weighted least squares subproblem in the r-th iteration. Indicates the number of iterations. Indicates the first The weighted values ​​of each scattering point; when When there are multiple columns, differentiate the weighted least squares subproblem and set its gradient to 0 to obtain the final result for each column. The iterative update formula is as follows: in, This indicates the transpose operation. For the first The diagonal weight matrix at the next iteration has diagonal elements that are the weighted values ​​of each distance cell in the current iteration step. , express The The process is repeated until convergence, thus completing the solution.

6. The phase information-guided real aperture radar super-resolution imaging method according to claim 5, characterized in that, The phase threshold is determined based on the maximum allowable micro-motion amplitude of the target and the system wavelength.

7. The phase information-guided real aperture radar super-resolution imaging method according to claim 6, characterized in that, The phase threshold is specifically calculated in the following manner: Set the radar carrier frequency to The corresponding wavelength The expression is as follows: in, Indicates the speed of electromagnetic wave propagation; When there are moving targets in the scene, the inter-frame phase change amount The expression is as follows: in, Indicates the first Frame and the The change in distance between frames; the corresponding phase threshold is calculated by limiting the magnitude of the change in distance.

8. The phase information-guided real aperture radar super-resolution imaging method according to claim 7, characterized in that, The solution model is transformed into a series of weighted least squares subproblems, as follows: The specific expression for the norm of a mixed matrix is ​​as follows: First, define the row norm scale. The expression is as follows: The regularization term can then be written as the following expression: in, Represents the regularization term; uses the current iteration value. The weights are constructed using the following expression: in, Indicates the first Row weights Represents a small positive number. Indicates the first The energy of movement, This represents the number of iterations; the following expression is obtained through a first-order approximation: Then for each element After a second approximation, we obtain the following expression: The entire regularization term can then be approximated as a weighted square, as shown in the following expression: in, Indicates the first The weighted values ​​of each scattering point; Substituting the entire regularization term into the solution model transforms it into a standard weighted least squares problem.

Images