Space-time fast foresight imaging method for motion platform foresight array radar
By employing a space-time sparsity adaptive forward-looking imaging method, and utilizing Doppler domain dimensionality reduction and space-time rank reduction processing, the problem of high computational complexity in airborne radar forward-looking imaging is solved, achieving efficient real-time imaging.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- NANJING UNIV OF AERONAUTICS & ASTRONAUTICS
- Filing Date
- 2025-12-25
- Publication Date
- 2026-05-12
AI Technical Summary
Existing technologies for forward-looking imaging with airborne radar suffer from high computational complexity, low imaging resolution, and low computational efficiency, making it difficult to achieve real-time imaging, especially in dynamic environments.
A spatiotemporal sparsity adaptive forward-looking imaging method is adopted. By using Doppler domain dimensionality reduction and spatiotemporal rank reduction processing, combined with an adaptive iterative algorithm, the radar data processing flow is optimized, reducing computational complexity while maintaining imaging resolution.
It significantly improves the computational efficiency of airborne array radar while maintaining high-resolution imaging, thus enhancing real-time imaging capabilities.
Smart Images
Figure CN122017826A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of radar imaging technology, and relates to airborne radar forward-looking imaging signal processing technology, specifically to a space-time fast forward-looking imaging method for a moving platform forward-looking array radar. Background Technology
[0002] Current forward-looking radar imaging technology faces significant bottlenecks in achieving high resolution across the flight path. This challenge stems from the extremely small Doppler frequency shift observed under forward-looking geometry, making it difficult to effectively distinguish targets in the frequency domain. Furthermore, symmetrical terrain features produce nearly identical Doppler characteristics, further exacerbating the difficulty of target separation.
[0003] Traditional real-aperture radar generates low-resolution images through mechanical scanning, and its azimuth resolution is inherently limited by the physical characteristics of the antenna. To improve the forward-looking area observation capability of moving platforms, researchers have proposed two main technical approaches over the past decade: temporal enhancement methods based on deconvolution theory and spatial optimization methods using super-resolution spectral estimation. The deconvolution method reconstructs images by establishing a convolution model of the antenna pattern and the target scattering characteristics, but as a typical inverse problem, it is extremely sensitive to measurement noise and systematic errors.
[0004] Another type of solution employs a tangential multi-antenna configuration, directly integrating super-resolution processing into the imaging link. This approach can overcome traditional beamwidth limitations, achieving a resolution improvement of several times. Several mature algorithms have been developed in the field of array signal processing, including adaptive beamforming and subspace decomposition methods. It is worth noting that although Doppler variations in the forward-looking scene are weak, they still have potential value for resolution improvement. Currently, space-time joint super-resolution forward-looking imaging technology, as a new technical means for radar forward-looking imaging, has gradually been recognized and valued by relevant research institutions, and is one of the effective means to solve the forward-looking imaging challenges of moving platform radar in the future.
[0005] However, in forward-looking imaging applications, computational efficiency is just as important as resolution improvement. While existing super-resolution methods can effectively improve azimuth resolution, they often face the challenge of excessively high computational complexity. How to optimize algorithms to significantly improve computational efficiency while maintaining imaging accuracy, and ultimately achieve real-time imaging, is an urgent problem to be solved in the field of forward-looking imaging. Summary of the Invention
[0006] Purpose of the invention: This invention proposes a space-time sparsity adaptive forward-looking imaging method for airborne array radar systems operating in dynamic environments, which significantly improves computational efficiency without sacrificing imaging resolution.
[0007] Summary of the Invention: The present invention provides a space-time sparsity adaptive forward-looking imaging method for airborne array radar, comprising the following steps:
[0008] (1) Data preprocessing: The multi-channel radar system with tangential trajectory is used to collect echoes in the forward-looking area to obtain three-dimensional range-pulse-channel data; then, high-resolution range processing is achieved through pulse compression and range migration correction.
[0009] (2) Doppler domain dimensionality reduction: Traverse each distance gate and perform azimuth-to-Fourier transform on its corresponding spatiotemporal two-dimensional data matrix; Utilize the characteristics of the Doppler spectrum set of the forward-looking scene, only extract the dominant Doppler units in the Doppler spectrum that are greater than the energy threshold and perform inverse Fourier transform.
[0010] (3) Space-time rank reduction processing: Perform eigenvalue decomposition on the sampling covariance matrix, arrange it in descending order of eigenvalues, retain the energy of the signal, further compress the data size and suppress noise;
[0011] (4) Super-resolution imaging: For the data after dimensionality reduction and rank reduction, an adaptive iterative algorithm is used to estimate the azimuth position and scattering intensity of the target. The estimation results are then projected onto a two-dimensional plane to form a high-resolution forward-looking image.
[0012] Furthermore, the implementation process of step (1) is as follows:
[0013] A full beam scan covers a single point target throughout the entire imaging process. The specific form of the obtained echo is:
[0014]
[0015] in, The scattering coefficient is given by coordinates. , For radar array antenna reference elements and target The slant distance between them The azimuth and angle of the target. The pitch angle of the target. Represents a rectangular window function. For distance-oriented fast time domain variables, The carrier frequency for radar transmission signals. For the range-direction linear frequency modulation, The azimuth-direction pattern of a two-way antenna varies with slow time variables. The change represents the modulation effect of azimuth. This indicates that the center of the antenna beam has swept over the point target. Time; To transmit signals to the target After the target Reflected to the first The propagation delay of each channel, Indicates the channel spacing, then the delay for:
[0016]
[0017] in, At the speed of light, The instantaneous slant range between the radar and the target is given. The echo is down-converted, followed by range-direction pulse compression and migration correction, i.e., multiplying the range frequency domain by the pulse compression reference function and the migration correction phase factor to achieve high-resolution range-direction imaging. The echo after inverse Fourier transform of the range is represented as:
[0018] .
[0019] Furthermore, the process of multiplying the pulse compression reference function and the migration correction phase factor in the distance frequency domain is as follows:
[0020] After performing a range FFT on the echo of each channel after downconversion, the following results were obtained:
[0021]
[0022] in, For signal bandwidth, For the range-oriented frequency domain variable; multiply by the following pulse compression reference function in the frequency domain:
[0023]
[0024] get:
[0025]
[0026] Multiply by the following migration correction phase factor in the distance frequency domain to eliminate the effects of platform motion:
[0027]
[0028] That is, multiply by the following function in the distance frequency domain:
[0029]
[0030] This enables high-resolution imaging in the range direction.
[0031] Furthermore, the implementation process of step (2) is as follows:
[0032] For distance is The Azimuth echo of each channel, point target The echo is represented as:
[0033]
[0034] The forward-looking imaging region is divided into grids. A distance door, The azimuth angles of a potential target are expressed as follows: For the first A coherent pulse interval of CPI, at which point the beam center points Spacetime two-dimensional signal Represented as:
[0035]
[0036] in, For azimuth angle The scattering coefficient of the target at that location. For spacetime guiding vector, For noise; the guidance vector is represented by the Kronecker product of the spatial guidance vector and the temporal guidance vector:
[0037] ;
[0038] For two-dimensional signals in spacetime Perform an orientation Fourier transform to the Doppler domain.
[0039]
[0040]
[0041] in, ;
[0042] The core dimensionality reduction process identifies dominant Doppler units through energy threshold screening. Energy-based methods are used to sort unit energies and select index sets that meet certain criteria. :
[0043]
[0044] in, For the energy retention threshold, The number of selected Doppler units, and satisfying the following conditions: ;
[0045] Dimensionality reduction is performed using the basis vectors corresponding to the selected units. The constructed projection matrix To achieve this, the data is projected onto a reduced-dimensional subspace; the reduced-dimensional data in the Doppler domain is... Therefore, by performing an inverse Fourier transform on the Doppler domain compressed data, the dimension-reduced spatiotemporal data matrix is obtained: ; Through orthogonal projection, the original data cube Transformed into a reduced-dimensional space .
[0046] Furthermore, the spatial guidance vector and time-domain guidance vector The results are obtained by calculating the following formulas respectively:
[0047]
[0048] .
[0049] Furthermore, the energy retention threshold The value ranges from 0.95 to 1.
[0050] Furthermore, the implementation process of step (3) is as follows:
[0051] In the space-time covariance matrix composed of space-time snapshots In this context, eigenvalues characterize the distribution of signal energy along the direction of the corresponding eigenvector; large eigenvalues correspond to the dominant signal component, whose eigenvectors capture the spatiotemporal coupling structure of the target or strong clutter; while small eigenvalues span a noise subspace, which is orthogonal to the signal subspace.
[0052] Set an energy retention threshold, by... Perform principal component analysis and retain the previous results. eigenvectors; spatiotemporal covariance matrix The singular value decomposition (SVD) is represented as:
[0053]
[0054] in, Including the signal subspace corresponding to One dominant feature vector, Including the corresponding eigenvalues arranged in descending order, These are the eigenvectors spanning the noisy subspace. Indicates noise components;
[0055] Before keeping The compressed covariance matrix is obtained from the eigenvectors. And then through Construct a dimension-reduced space-time guidance matrix. Construct a spacetime snapshot after dimensionality reduction and rank reduction.
[0056] Furthermore, the constant energy retention threshold is 95% of the total energy.
[0057] Furthermore, the implementation process of step (4) is as follows:
[0058] Using dimensionality reduction and rank reduction of the covariance matrix Using iterative algorithms to reconstruct the azimuth spectrum of a single snapshot, based on the least squares criterion, we obtain:
[0059]
[0060] Introducing an iterative algorithm, initially:
[0061]
[0062] The autocorrelation matrix of the signal is calculated based on this. , , Represented as a diagonal matrix, where each diagonal element The autocorrelation matrix of the echo is calculated as follows:
[0063]
[0064] in, For regularization parameters, The identity matrix is used; the autocorrelation matrix is reduced in rank and then substituted into the iterative process to repeatedly estimate the azimuth spectrum. The azimuth spectrum estimated in each iteration is used to update the signal autocorrelation matrix, which in turn updates the echo autocorrelation matrix. This process is repeated until convergence.
[0065] Finally, the azimuth spectrum corresponding to each distance-CPI unit is estimated, projected, transformed into spatial coordinates, and displayed to obtain the final forward-looking imaging result.
[0066] Beneficial Effects: Compared with existing technologies, the beneficial effects of this invention are as follows: This invention deeply explores the unique spectral characteristics of forward-looking radar scenarios. The inherent narrow Doppler bandwidth in this configuration creates significant information redundancy in the frequency domain. This invention develops an innovative two-stage acceleration architecture: First, it utilizes the extremely strong Doppler block sparsity of the forward-looking radar echo to achieve data reduction through spectral compression. Subsequently, it performs optimized rank reduction processing on the covariance matrix. This breakthrough processing reduces computational complexity from... Down to —After dimensional and rank compression At the same time, it fully maintains the image fidelity. Attached Figure Description
[0067] Figure 1 This is a flowchart of the present invention;
[0068] Figure 2 This is a schematic diagram of the forward-looking imaging geometric model of an array radar.
[0069] Figure 3 This is a schematic diagram of a simulation scene for point targets.
[0070] Figure 4This is a traditional full-dimensional spatiotemporal imaging result image;
[0071] Figure 5 Spatiotemporal forward-looking imaging results after dimensionality reduction of Doppler;
[0072] Figure 6 The results are fast spatiotemporal imaging based on the dimensionality reduction-rank reduction framework;
[0073] Figure 7 This is an azimuth profile of a traditional full-dimensional spatiotemporal imaging result;
[0074] Figure 8 Azimuth profile of the spatiotemporal forward-looking imaging results after dimensionality reduction of Doppler;
[0075] Figure 9 This is an azimuth profile of the fast spatiotemporal imaging results based on the dimensionality reduction-rank reduction framework;
[0076] Figure 10 For generating SAR images of forward-looking echoes;
[0077] Figure 11 This is a traditional full-dimensional spatiotemporal surface target imaging result image;
[0078] Figure 12 Forward-looking imaging results of a spatiotemporal surface target after Doppler dimensionality reduction;
[0079] Figure 13 The results are for fast spatiotemporal surface target imaging based on the dimensionality reduction-rank reduction framework. Detailed Implementation
[0080] The invention will now be further described with reference to the accompanying drawings.
[0081] like Figure 1 As shown, this invention provides a space-time fast forward-looking imaging method for a forward-looking array radar of a moving platform, specifically including the following steps:
[0082] Step 1: The missile-borne radar operates in scanning mode and uses an array antenna oriented along the flight path to receive echoes from the forward-looking region. It samples and obtains range-pulse-array three-dimensional echo data. Through pulse compression and range migration correction, high-resolution processing in the range direction is achieved.
[0083] like Figure 2 As shown, the missile-borne radar uses a horizontal array antenna along the tangent to the flight path, and synthesizes data through subarrays. Each receiving channel receives the echo from the forward-looking area of the missile's onboard radar. The forward-looking area generally considers the area in front of the missile's flight path. Area; When the radar operates in downward-looking scanning mode, it transmits a linear frequency modulated pulse (LFM) at each pulse repetition interval (PRI), the time-domain expression of which is:
[0084]
[0085] in, For distance-oriented fast time domain variables, The carrier frequency for radar transmission signals. For the range-direction linear frequency modulation, The pulse width of the signal. Represents a rectangular window function, namely:
[0086]
[0087] Assumption This is a slow-time domain variable in the azimuth direction, which is related to radar motion and antenna scanning, for a point target within the imaging area. For example, assuming Its position coordinates at time are The scattering coefficient is , Let be the downward angle of the radar beam, and be a constant. In any... At any given moment, the radar array antenna reference element and the target Instantaneous slant distance between Approximately expressed as:
[0088]
[0089] in, This refers to the speed of the missile-borne radar. For multi-channel radars... For each receiving channel After one cycle of beam scanning, the specific form of the echo obtained for a single point target during the entire imaging scanning process is as follows:
[0090]
[0091] in, The azimuth-direction pattern of a two-way antenna varies with slow time variables. The change represents the modulation effect of azimuth. This indicates that the center of the antenna beam has swept over the point target. Time; To transmit signals to the target After the target Reflected to the first The propagation delay of each channel, Indicates the first The distance between each receiving channel and the reference array element, then the delay... for:
[0092]
[0093] in, It is the speed of light.
[0094] After down-conversion processing, the first The range-azimuth two-dimensional echo of each receiving channel is represented as follows:
[0095]
[0096] By sampling the echoes, three-dimensional range-pulse-channel echo data of the forward-looking area of the missile-borne radar can be obtained. Range-axis pulse compression and migration correction are performed on the array radar's three-dimensional echo data. Multiplying the data in the frequency domain by a pulse compression reference function and a migration correction phase factor achieves high range resolution. Then, the received data from multiple channels of each range-pulse unit are selected and arranged sequentially to form the corresponding spatial snapshot for that unit.
[0097] After performing a range FFT on the echo of each channel after downconversion, the following results were obtained:
[0098]
[0099] in, For signal bandwidth, For the range-oriented frequency domain variable, multiply by the following pulse compression reference function in the frequency domain:
[0100]
[0101] get:
[0102]
[0103] Due to its high-speed movement, the missile-borne radar experiences range migration with the target. This migration introduces Doppler frequency shift and range-azimuth coupling. To decouple the radar, range migration correction is needed for the echo. Therefore, the following migration correction phase factor is multiplied in the range frequency domain to eliminate the effects of platform motion:
[0104]
[0105] That is, multiply by the following function in the distance frequency domain:
[0106]
[0107] After pulse compression and migration correction in the range direction, high-resolution imaging in the range direction has been achieved. The echo after inverse Fourier transform of the range is represented as:
[0108]
[0109] Step 2: Traverse each range gate and perform an azimuth-to-Fourier transform on its corresponding spatiotemporal two-dimensional data matrix. Utilizing the characteristics of the forward-looking scene Doppler spectrum set, only the dominant Doppler units above the energy threshold are truncated for inverse Fourier transform, thereby significantly reducing the data dimensionality while preserving the main signal components.
[0110] For distance is The Azimuth echo of each channel, point target The echo is represented as:
[0111]
[0112] The forward-looking imaging region is divided into grids, assuming the grid division... A distance door, The azimuth angles of a potential target are expressed as follows: For the first One coherent pulse interval (CPI), at which point the beam center points to Spacetime two-dimensional signal Represented as:
[0113]
[0114] in, For azimuth angle The scattering coefficient of the target at that location. For spacetime guiding vector, For noise. The guidance vector can be represented by the Kronecker product of the spatial guidance vector and the temporal guidance vector:
[0115]
[0116] Among them, the spatial guidance vector and time-domain guidance vector The results are obtained by calculating the following formulas respectively:
[0117]
[0118] .
[0119] For two-dimensional signals in spacetime Perform azimuth Fourier transform to the Doppler domain:
[0120]
[0121]
[0122] in, .
[0123] The core dimensionality reduction process identifies dominant Doppler units through energy threshold screening. An energy-based method is used to sort unit energies and select an index set that meets certain criteria. :
[0124]
[0125] in, The energy retention threshold is typically set between 0.95 and 1. The number of selected Doppler units, and satisfying the following conditions: .
[0126] Dimensionality reduction is performed using the basis vectors corresponding to the selected units. The constructed projection matrix This involves projecting the data onto a reduced-dimensional subspace. The reduced-dimensional data in the Doppler domain is... Therefore, by performing an inverse Fourier transform on the Doppler domain compressed data, the dimension-reduced spatiotemporal data matrix can be obtained: .
[0127] The original data cube is obtained through orthogonal projection. Transformed into a reduced-dimensional space .
[0128] Step 3: Based on the low-rank characteristic of the signal covariance matrix, principal component analysis and other methods are used to extract the main signal components, retain most of the energy, further compress the data size and suppress noise.
[0129] Based on the established spatial-temporal echo data Doppler domain dimensionality reduction framework, an enhanced acceleration method is further developed, which organically combines optimized subspace projection with adaptive processing techniques.
[0130] In the space-time covariance matrix composed of space-time snapshots In this context, eigenvalues characterize the distribution of signal energy along the direction of the corresponding eigenvector. Specifically, larger eigenvalues correspond to the dominant signal component, whose eigenvector captures the spatiotemporal coupling structure of the target or strong clutter; while smaller eigenvalues span a noise subspace, which is orthogonal to the signal subspace.
[0131] Set the energy retention threshold to 95% of the total energy, and then... Perform principal component analysis (PCA) and retain the results before... Using eigenvectors (signal subspaces) can significantly reduce computational complexity. Spatiotemporal covariance matrix The singular value decomposition (SVD) is represented as:
[0132]
[0133] in, Includes the signal subspace corresponding to One dominant feature vector, It contains the corresponding feature values arranged in descending order. These are the eigenvectors spanning the noisy subspace. This represents the noise component, whose mean is approximately equal to the receiver noise power.
[0134] Before keeping The compressed covariance matrix can be obtained from the eigenvectors. And then through Construct a dimension-reduced space-time guidance matrix. Construct a spacetime snapshot after dimensionality reduction and rank reduction.
[0135] Step 4: For the dimensionality-reduced and rank-reduced data, an adaptive iterative algorithm is used to estimate the target's azimuth and scattering intensity with high accuracy; finally, the estimation results are projected onto a two-dimensional plane to form a high-resolution forward-looking image.
[0136] Using dimensionality reduction and rank reduction of the covariance matrix Using iterative algorithms to reconstruct the azimuth spectrum of a single snapshot, the least squares criterion yields the following:
[0137]
[0138] Introducing an iterative algorithm, initially:
[0139]
[0140] The autocorrelation matrix of the signal is calculated based on this. , , Represented as a diagonal matrix, where each diagonal element Therefore, the autocorrelation matrix of the echo can be calculated as follows: ( For regularization parameters, (The identity matrix is used). The autocorrelation matrix is reduced in rank and then substituted into the iterative process to repeatedly estimate the azimuth spectrum. The azimuth spectrum estimated in each iteration is used to update the signal autocorrelation matrix, which in turn updates the echo autocorrelation matrix. This process is repeated until convergence.
[0141] Finally, the azimuth spectrum corresponding to each distance-CPI unit is estimated, projected, transformed into spatial coordinates, and displayed to obtain the final forward-looking imaging result.
[0142] The following section presents point target simulation and area target simulation verification, comparing the traditional spatiotemporal super-resolution imaging method with the spatiotemporal fast forward-looking imaging proposed in this invention. Figure 3This is a schematic diagram of a point target simulation scenario. The beamwidth is 4.2 degrees. Nine point targets are arranged 2 kilometers in front of the motion platform. Every three targets are located in the same distance cell, with azimuth angles of -3°, -1° and 2° respectively. Figures 4 to 6 The image shows the results of the imaging using the adaptive iterative algorithm. Figure 4 This is a result of traditional spatiotemporal imaging. Figure 5 The spatiotemporal imaging results after Doppler dimension reduction are shown. Figure 6 The results of spatiotemporal fast imaging after dimensionality reduction and rank reduction show that the Doppler-based dimensionality reduction-rank reduction method can effectively maintain resolution in spatiotemporal super-resolution forward-looking imaging. Figures 7 to 9 The image is a azimuth profile comparison diagram of a certain distance gate in the imaging results. The profile comparison diagram more clearly shows that the method proposed in this invention can effectively maintain the azimuth super-resolution performance, and clear and steep peaks are formed in the three azimuths of -3°, -1° and 2°.
[0143] To compare the computational efficiency before and after dimensionality reduction or rank reduction, processing time was measured on an Intel(R) Xeon(R) Platinum 8168 CPU, and the results are summarized in Table 1. The reported times are the average of 10 independent trials using simulated point target data of size 2048×7680×8 (range gate × pulse number × channel number). The results demonstrate that traditional imaging methods have the highest computational complexity and longest processing time due to the need to invert large-scale spatiotemporal covariance matrices. Doppler-based dimensionality reduction significantly improves this problem, reducing the average computation time by several times. Further acceleration is achieved through the joint dimensionality reduction-rank reduction framework proposed in this invention, ultimately reducing the processing time to the second level.
[0144] Table 1 Comparison of Simulation Calculation Time for Point Targets
[0145]
[0146] Next, we will conduct simulation verification of the surface target. Figure 10 The original SAR images used to generate forward-looking radar echoes are presented. Figure 11 The processing results using the traditional spatiotemporal forward-looking imaging method are shown; Figure 12 The processing effect based on Doppler dimensionality reduction (compressing the pulse number from 64 to 32) is shown; Figure 13 Imaging results were demonstrated using a fast algorithm combining Doppler dimensionality reduction and rank reduction. All dimensionality reduction and rank reduction schemes achieved significant improvements in computational efficiency while maintaining imaging resolution and quality comparable to the full-dimensional method. The computation times are shown in Table 2.
[0147] Table 2 Comparison of simulation computation time for surface targets
[0148]
[0149] This strong robustness confirms that the proposed fast spatiotemporal imaging method based on the dimensionality reduction-rank reduction framework can effectively preserve key signal features while significantly reducing computational complexity.
[0150] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.
Claims
1. A space-time fast forward-looking imaging method for a forward-looking array radar for a moving platform, characterized in that, Includes the following steps: (1) Data preprocessing: The multi-channel radar system with tangential trajectory is used to collect echoes in the forward-looking area to obtain three-dimensional range-pulse-channel data; then, high-resolution range processing is achieved through pulse compression and range migration correction. (2) Doppler domain dimensionality reduction: Traverse each distance gate and perform azimuth-to-Fourier transform on its corresponding spatiotemporal two-dimensional data matrix; Utilize the characteristics of the Doppler spectrum set of the forward-looking scene, only extract the dominant Doppler units in the Doppler spectrum that are greater than the energy threshold and perform inverse Fourier transform. (3) Space-time rank reduction processing: Perform eigenvalue decomposition on the sampling covariance matrix, arrange it in descending order of eigenvalues, retain the energy of the signal, further compress the data size and suppress noise; (4) Super-resolution imaging: For the data after dimensionality reduction and rank reduction, an adaptive iterative algorithm is used to estimate the azimuth position and scattering intensity of the target. The estimation results are then projected onto a two-dimensional plane to form a high-resolution forward-looking image.
2. The spatiotemporal fast forward-looking imaging method for a forward-looking array radar for a moving platform according to claim 1, characterized in that, The implementation process of step (1) is as follows: A full beam scan covers a single point target throughout the entire imaging process. The specific form of the obtained echo is: in, The scattering coefficient is given by coordinates. , For radar array antenna reference elements and target The slant distance between them The azimuth and angle of the target. The pitch angle of the target. Represents a rectangular window function. For distance-oriented fast time domain variables, The carrier frequency for radar transmission signals. For the range-direction linear frequency modulation, The azimuth-direction pattern of a two-way antenna varies with slow time variables. The change represents the modulation effect of orientation. This indicates that the center of the antenna beam has swept over the point target. Time; To transmit signals to the target After the target Reflected to the first The propagation delay of each channel, Indicates the channel spacing, then the delay for: in, At the speed of light, The instantaneous slant range between the radar and the target is given. The echo is down-converted, followed by range-direction pulse compression and migration correction, i.e., multiplying the range frequency domain by the pulse compression reference function and the migration correction phase factor to achieve high-resolution range-direction imaging. The echo after inverse Fourier transform of the range is represented as: 。 3. The space-time fast forward-looking imaging method for a forward-looking array radar for a moving platform according to claim 2, characterized in that, The process of multiplying the pulse compression reference function and the migration correction phase factor in the distance frequency domain is as follows: After performing a range FFT on the echo of each channel after downconversion, the following results were obtained: in, For signal bandwidth, For the range-oriented frequency domain variable; multiply by the following pulse compression reference function in the frequency domain: get: Multiply by the following migration correction phase factor in the distance frequency domain to eliminate the effects of platform motion: That is, multiply by the following function in the distance frequency domain: This enables high-resolution imaging in the range direction.
4. The space-time fast forward-looking imaging method for a forward-looking array radar for a moving platform according to claim 1, characterized in that, The implementation process of step (2) is as follows: For distance is The Azimuth echo of each channel, point target The echo is represented as: The forward-looking imaging region is divided into grids. A distance door, The azimuth angles of a potential target are expressed as follows: For the first A coherent pulse interval of CPI, at which point the beam center points Spacetime two-dimensional signal Represented as: in, For azimuth angle The scattering coefficient of the target at that location. For spacetime guiding vector, For noise; the guidance vector is represented by the Kronecker product of the spatial guidance vector and the temporal guidance vector: ； For two-dimensional signals in spacetime Perform an orientation Fourier transform to the Doppler domain. in, ; The core dimensionality reduction process identifies dominant Doppler units through energy threshold screening. Energy-based methods are used to sort unit energies and select index sets that meet certain criteria. : in, For the energy retention threshold, The number of selected Doppler units, and satisfying the following conditions: ; Dimensionality reduction is performed using the basis vectors corresponding to the selected units. The constructed projection matrix To achieve this, the data is projected onto a reduced-dimensional subspace; the reduced-dimensional data in the Doppler domain is... Therefore, by performing an inverse Fourier transform on the Doppler domain compressed data, the dimension-reduced spatiotemporal data matrix is obtained: ; Through orthogonal projection, the original data cube Transformed into a reduced-dimensional space .
5. A space-time fast forward-looking imaging method for a forward-looking array radar for a moving platform according to claim 4, characterized in that, The spatial guidance vector and time-domain guidance vector They are calculated using the following formulas respectively: 。 6. A space-time fast forward-looking imaging method for a forward-looking array radar for a moving platform according to claim 4, characterized in that, The energy retention threshold The value ranges from 0.95 to 1.
7. The space-time fast forward-looking imaging method for a forward-looking array radar for a moving platform according to claim 1, characterized in that, The implementation process of step (3) is as follows: In the space-time covariance matrix composed of space-time snapshots In this context, eigenvalues characterize the distribution of signal energy along the direction of the corresponding eigenvector; large eigenvalues correspond to the dominant signal components. Its eigenvectors capture the spatiotemporal coupling structure of the target or strong clutter; while small eigenvalues span the noise subspace, which is orthogonal to the signal subspace; Set an energy retention threshold, by... Perform principal component analysis and retain the previous results. eigenvectors; spatiotemporal covariance matrix The singular value decomposition (SVD) is represented as: in, Including the signal subspace corresponding to One dominant feature vector, Including the corresponding eigenvalues arranged in descending order, These are the eigenvectors spanning the noisy subspace. Indicates noise components; Before keeping The compressed covariance matrix is obtained from the eigenvectors. And then through Construct a dimension-reduced space-time guidance matrix. Construct a spacetime snapshot after dimensionality reduction and rank reduction.
8. A space-time fast forward-looking imaging method for a forward-looking array radar for a moving platform according to claim 7, characterized in that, The constant energy retention threshold is 95% of the total energy.
9. A space-time fast forward-looking imaging method for a forward-looking array radar for a moving platform according to claim 1, characterized in that, The implementation process of step (4) is as follows: Using dimensionality reduction and rank reduction of the covariance matrix Using iterative algorithms to reconstruct the azimuth spectrum of a single snapshot, based on the least squares criterion, we obtain: Introducing an iterative algorithm, initially: The autocorrelation matrix of the signal is calculated based on this. , , Represented as a diagonal matrix, where each diagonal element The autocorrelation matrix of the echo is calculated as follows: in, For regularization parameters, The identity matrix is used; the autocorrelation matrix is reduced in rank and then substituted into the iterative process to repeatedly estimate the azimuth spectrum. The azimuth spectrum estimated in each iteration is used to update the signal autocorrelation matrix, which in turn updates the echo autocorrelation matrix. This process is repeated until convergence. Finally, the azimuth spectrum corresponding to each distance-CPI unit is estimated, projected, transformed into spatial coordinates, and displayed to obtain the final forward-looking imaging result.