A radar super-resolution forward-looking imaging method and system

Abstract

The application discloses a radar super-resolution forward-looking imaging method and system, constructs a weight vector and a manifold matrix in the distance and azimuth dimensions according to the waveform parameters and the array structure of the radar, and further constructs a range spread function and an azimuth spread function.On this basis, target features are extracted through mean shift theory, and the constructed range spread function and azimuth spread function are used when calculating a mean vector, so that the diffusion mode of the target in the radar image is more in line with the diffusion mode, an image with a smaller full width at half maximum can be obtained, and super-resolution imaging is realized.The application innovatively considers the super-resolution imaging problem from the perspective of the image domain and in combination with the characteristics of the radar.Compared with the previous super-resolution method based on signal processing, the application is more universal, is not easily limited by factors such as the waveform and the array structure, and has no strict performance upper limit, so that the resolution can be further improved on the basis of the existing super-resolution imaging result.

Description

Technical Field

[0001] This invention belongs to the field of radar imaging algorithms, and in particular relates to a radar super-resolution forward-looking imaging method and system. Background Technology

[0002] For a long time, the exploration of forward-looking imaging algorithms for practical aperture radar has mainly focused on signal processing in the radar echo domain. This exploration relies on array signal processing techniques to improve resolution, enhance signal-to-noise ratio (SNR), or reduce clutter, thereby improving image quality. However, this approach has two limitations. First, the developed methods are usually only applicable to specific or limited radar systems, waveforms, or array configurations, resulting in poor versatility. Second, most techniques have inherent performance limits and lack complementarity, thus preventing them from being combined to achieve further performance breakthroughs.

[0003] In classic radar signal processing, high-quality range-azimuth (RA) thermal imaging largely depends on accurate direction-of-arrival (DOA) estimation. Traditional DOA estimation methods, such as digital beamforming (DBF), are limited by the Rayleigh limit, hindering their ability to achieve super-resolution performance. Subsequent subspace-based super-resolution methods, represented by Multiple Signal Classification (MUSIC) and Rotation Invariant Subspace Algorithm (ESPRIT), have overcome the limitations of antenna physical aperture, achieving significant progress in DOA estimation. Iterative Adaptive Algorithm (IAA) is another classic method widely used in DOA estimation and forward-looking imaging. It derives the spatial energy spectrum by iteratively updating the target energy on the angular grid, achieving high-resolution performance with a minimal number of snapshots.

[0004] In recent DOA estimation research, deep learning-based methods have attracted considerable attention and praise. Studies have shown that even a simple deep neural network (DNN) composed of only partially fully connected (FC) layers can achieve super-resolution DOA estimation for individual targets. Furthermore, by increasing network depth and employing more complex model structures, super-resolution DOA estimation for even more targets can be achieved, and these methods can adapt well to special cases such as low SNR, limited snapshots, and array defects. Although deep learning-based methods offer significant advantages in real-time performance and robustness, and their outputs can even represent energy spectra suitable for range-azimuth (RA) heatmap forward-looking imaging, these methods suffer from insufficient interpretability. Fundamentally, these methods still process each range unit in isolation, failing to perceive the image as a whole.

[0005] Deconvolutional imaging is a forward-looking technique widely used in practical aperture radar systems, especially on airborne and other scanning radar platforms. It models radar echoes as a convolution of the target scattering coefficients and antenna patterns. Depending on the subsequent processing, deconvolutional imaging methods can be broadly categorized into inverse filtering, statistical optimization, and regularization techniques. Deconvolutional imaging addresses the ambiguity caused by the point spread function (PSF) through a novel signal model. However, accurately separating the target using the signal model remains a significant challenge. Furthermore, most existing methods only consider azimuth ambiguity.

[0006] In summary, many scholars have conducted extensive research on radar super-resolution forward-looking imaging, but existing research focuses primarily on signal processing, limiting its versatility and limiting its performance. Furthermore, most methods concentrate on improving resolution in a single azimuth dimension, which is insufficient for scenarios requiring high resolution in both range and azimuth, failing to achieve two-dimensional super-resolution. To address these issues, exploring an imaging method that starts from the image domain and considers the target's diffusion patterns in both dimensions is of great significance. This method could not only achieve super-resolution performance based on non-super-resolution imaging results but also potentially further improve upon the imaging results of traditional super-resolution algorithms. Summary of the Invention

[0007] Purpose of the invention: The purpose of this invention is to provide a radar super-resolution forward-looking imaging method and system to further reduce the target diffusion range and achieve super-resolution imaging effect.

[0008] Technical solution: To achieve the above-mentioned objectives, the present invention adopts the following technical solution:

[0009] In a first aspect, the present invention provides a radar super-resolution forward-looking imaging method, comprising the following steps:

[0010] Obtain the radar's waveform parameters and array structure;

[0011] Receive radar echo data of the forward-looking scene, and based on the radar echo data, obtain preliminary imaging results in the form of range-azimuth RA heatmap and normalize them;

[0012] Based on the waveform parameters and array structure, weight vectors and manifold matrices for the range and azimuth dimensions are constructed, and then the range spread function and azimuth spread function are constructed.

[0013] For each cell in the RA heatmap, the corresponding mean offset vector is calculated. The value of the mean offset vector is inverted and the negative value is discarded to form a mean offset image. The mean offset image is multiplied with the original image to obtain the final super-resolution imaging result. The calculation of the mean offset vector of each cell includes: determining the scale bandwidth in the three dimensions of range, azimuth, and intensity; determining the neighborhood based on the scale bandwidth; calculating the probability density function using mean offset theory; and calculating the gradient of the probability density function to obtain the mean offset vector at the cell. The probability density function is calculated using the constructed range spread function and azimuth spread function.

[0014] Furthermore, for distances of azimuth angle is The reference point, and the weight vectors for distance and orientation dimensions are respectively and ,in Indicates an index. Represents the imaginary unit. The speed at which electromagnetic waves propagate in the air. For wavelength, This represents the frequency corresponding to each distance unit. , For bandwidth, For fast sampling points, This represents the position coordinates corresponding to each array element. The number of array elements. The distance between array elements is denoted by T, which indicates transpose.

[0015] Furthermore, the manifold matrices for the distance and orientation dimensions are respectively and , For length is The distance vector represents the distance unit divided within the detection range; For length is An angular vector represents the angular units divided within the field of view.

[0016] Furthermore, the distance spread function azimuth spread function ,in Distance vector The location of the peak Elements of one unit, Distance vector The location of the peak Elements of one unit, , The superscript H indicates conjugate transpose.

[0017] Furthermore, unit probability density function ,in:

[0018] ,

[0019] This represents the total number of cells in the RA heatmap. Represents the scale bandwidth based on the distance, orientation, and intensity dimensions. Determined unit The neighborhood, , These represent the total dimensions of distance and orientation, respectively. The contour representing the intensity dimension is calculated using a Gaussian kernel. , , These represent the distance, azimuth coordinates, and intensity value of the current element, respectively. , , These represent the distance, azimuth coordinates, and intensity value of each cell within the neighborhood, respectively. This represents the Euclidean norm.

[0020] As a preferred option, the scale bandwidth in the distance and orientation dimensions is half of the full width at half maximum, and the scale bandwidth in the intensity dimension is the maximum intensity value in the distance-orientation neighborhood of the current cell.

[0021] Preferably, in the preliminary imaging stage, the radar data is preprocessed, and a radar heatmap is obtained by range-dimensional fast Fourier transform and angle spectrum estimation. The angle spectrum estimation adopts a digital beamforming method or an iterative adaptive algorithm.

[0022] In a second aspect, the present invention provides a radar super-resolution forward-looking imaging system, comprising:

[0023] The parameter acquisition module is used to acquire the radar's waveform parameters and array structure.

[0024] The initial imaging module is used to receive radar echo data of the forward-looking scene, and based on the radar echo data, obtain and normalize the preliminary imaging results in the form of range-azimuth RA heatmap.

[0025] The diffusion model construction module is used to construct the weight vectors and manifold matrices of the distance and azimuth dimensions based on the waveform parameters and array structure, and then construct the distance diffusion function and the azimuth diffusion function.

[0026] The super-resolution imaging module is used to calculate the corresponding mean offset vector for each cell on the RA heatmap, invert the value of the mean offset vector and discard the negative value to form a mean offset image, and multiply the mean offset image with the original image to obtain the final super-resolution imaging result. The calculation of the mean offset vector for each cell includes: determining the scale bandwidth in the three dimensions of range, azimuth and intensity, determining the neighborhood based on the scale bandwidth, calculating the probability density function using mean offset theory, and calculating the gradient of the probability density function to obtain the mean offset vector at the cell. The probability density function is calculated using the constructed range spread function and azimuth spread function.

[0027] Thirdly, the present invention provides a computer system including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the radar super-resolution forward-looking imaging method.

[0028] Fourthly, the present invention provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the radar super-resolution forward-looking imaging method.

[0029] Beneficial Effects: This invention provides a radar super-resolution forward-looking imaging method that constructs an accurate target diffusion model based on radar parameters and array structure, including range-dimensional diffusion functions and azimuth-dimensional diffusion functions. It utilizes the mean-shift principle to extract target features in the radar image domain, reducing the target's diffusion range and thus achieving super-resolution imaging. Compared to previous radar super-resolution forward-looking imaging methods, this invention is based on the image domain rather than signal processing, allowing for integration with signal processing methods to further improve the quality of imaging results. Furthermore, during image domain processing, this invention accurately models the target's diffusion patterns in both range and azimuth dimensions, rather than considering only the azimuth dimension, thus achieving a two-dimensional super-resolution effect. In addition, the mean-shift theory is introduced. After accurately modeling the target's two-dimensional diffusion model based on radar parameters and array structure, range and azimuth diffusion functions replace the kernel commonly used in the mean-shift method, more accurately representing the target's contour and intensity changes, thereby achieving super-resolution imaging. Attached Figure Description

[0030] Figure 1 This is a flowchart of a method according to an embodiment of the present invention.

[0031] Figure 2 This is a simulated scattering point map from the perspective of the vehicle's front.

[0032] Figure 3The images are comparison diagrams of the imaging results from the frontal view of the vehicle. (a) is the imaging result of the DBF method for the simulated frontal view of the vehicle, and (b) is the imaging result of the super-resolution imaging result based on DBF of the method in the embodiment of the present invention for the simulated frontal view of the vehicle.

[0033] Figure 4 This is a simulated scattering point diagram from the side view of the vehicle.

[0034] Figure 5 The images show a comparison of the imaging results from the front and side views of the vehicle. (a) is the imaging result of the DBF method for the simulated front and side views of the vehicle, and (b) is the imaging result of the super-resolution imaging result based on DBF of the method in the embodiment of the present invention for the simulated front and side views of the vehicle.

[0035] Figure 6 This is a DBF imaging result image of a parking lot scene.

[0036] Figure 7 The image shows the super-resolution imaging results based on DBF in a parking lot scenario according to an embodiment of the present invention.

[0037] Figure 8 This is an IAA imaging result image of a parking lot scene.

[0038] Figure 9 The image shows the super-resolution imaging results based on IAA of the method of this invention in a parking lot scenario. Detailed Implementation

[0039] To better understand the technical solution and effects of the present invention, the present invention will be described in further detail below with reference to the accompanying drawings.

[0040] like Figure 1As shown, this invention discloses a radar super-resolution forward-looking imaging method, comprising: acquiring radar waveform parameters and array structure; receiving radar echo data of the forward-looking scene, obtaining and normalizing a preliminary range-azimuth (RA) heatmap based on the radar echo data; constructing weight vectors and manifold matrices for range and azimuth dimensions according to the waveform parameters and array structure, and then constructing range spread function and azimuth spread function; calculating the corresponding mean offset vector for each cell on the RA heatmap, inverting the value of the mean offset vector and discarding negative values ​​to form a mean offset image, and multiplying the mean offset image with the original image to obtain the final super-resolution imaging result. In this embodiment, the calculation of the mean offset vector for each cell includes: determining the scale bandwidth in the three dimensions of range, azimuth, and intensity; determining the neighborhood according to the scale bandwidth; calculating the probability density function using the mean offset theory; and calculating the gradient of the probability density function to obtain the mean offset vector at the cell; wherein the constructed range spread function and azimuth spread function are used when calculating the probability density function, that is, the kernel commonly used in the mean offset method is replaced by the range and azimuth spread functions.

[0041] In a specific implementation scenario, the radar super-resolution forward-looking imaging method specifically includes the following steps:

[0042] Step 101: Determine the waveform parameters of the frequency modulated continuous wave (FMCW) radar signal and the structure of the antenna array.

[0043] Specifically, in this step, the carrier frequency is determined to be... bandwidth is , wavelength is Fast time sampling points are The number of array elements is The spacing between array elements is .

[0044] Step 102: The radar transmits signals and receives reflected signals from the forward-looking scene. The radar echo ADC sampled signals are preprocessed, and the range-azimuth (RA) heatmap imaging results are obtained through range-dimensional Fast Fourier Transform (FFT) and angle spectrum estimation. And normalize.

[0045] Specifically, the angle spectrum estimation in this step can employ either traditional beamforming methods or super-resolution algorithms, and the size of the imaging result is... ,in , These represent the total dimensions of the distance and orientation dimensions, respectively.

[0046] Step 103: Based on the waveform parameters and array structure in Step 101, construct the weight vectors and manifold matrices for the range and azimuth dimensions, and then construct the range spread function and azimuth spread function.

[0047] Specifically, a distance can be chosen as azimuth angle is The reference point is generally taken as the center of the field of view, i.e. Half of the maximum detection range, The weight vectors for distance and orientation dimensions are respectively and Furthermore, the distance manifold matrix can be determined separately. and orientation manifold matrix ;in, This represents the frequency corresponding to each distance unit. , Represents the position coordinates of each array element; distance vector Length is , representing the distance grid points within the detection range; angle vector Length is , represents the angular grid points within the field of view. The speed at which electromagnetic waves propagate in the air. Indicates an index. It represents the imaginary unit, and the superscript T indicates transpose.

[0048] Step 104: Based on the weight vectors and manifold matrix of the two dimensions in Step 103, a diffusion model of a target at any coordinate on the RA heatmap can be constructed in two dimensions. By calculating the two dimensions separately and normalizing them, the distance diffusion function can be obtained. and azimuth spread function ,in Distance vector The location of the peak Elements of one unit, Distance vector The location of the peak Elements of one unit, , The superscript H indicates conjugate transpose.

[0049] Step 105: Obtain the RA heatmap The unit above Determine the scale bandwidth in the three dimensions of distance, orientation, and intensity. The neighborhood is determined based on the selected cell and scale bandwidth. The probability density function is calculated using mean shift (MS) theory. , can be represented as

[0050]

[0051] in,

[0052]

[0053] This represents the diffusion profile at that cell. This represents the total number of cells in the RA heatmap. The contour representing the intensity dimension is calculated using a Gaussian kernel. , , These represent the distance, azimuth coordinates, and intensity value of the current element, respectively. , , These represent the distance, azimuth coordinates, and intensity value of each cell within the neighborhood, respectively. This represents the Euclidean norm. The gradient of the probability density function is used to obtain the principal unit. The mean offset vector at that point.

[0054] Scale bandwidth in distance and orientation dimensions and It can be adaptively adjusted according to the processing method used when calculating the RA heatmap, usually taking half of the full width at half maximum (FWHM). The value is taken as the maximum intensity value within the distance-azimuth neighborhood of the current element. For an element... The determined neighborhood is represented as

[0055]

[0056] Step 106: Perform the same operation on each cell in the RA heatmap to obtain the corresponding mean offset vector, invert its value and discard the negative value to form a mean offset image, and finally multiply it with the original image to suppress sidelobes and achieve super-resolution imaging effect.

[0057] In a simulation scenario, the radar super-resolution forward-looking imaging method includes the following steps:

[0058] Step 201: Using the determined radar parameters, array structure, and target point position in the forward-looking scene, simulate the radar transmitted signal, target point, and corresponding radar echo ADC data using MATLAB software.

[0059] Step 202: Perform simple preprocessing on the radar ADC data. The range dimension is processed using a simple FFT. Then, for the received data of each channel, calculate the spatial spectrum using DBF or other super-resolution DOA estimation methods to obtain the forward-looking RA thermal imaging results of the radar.

[0060] Step 203: Construct the range and azimuth weight vectors based on the actual radar waveform parameters and array structure, and calculate the range manifold matrix and azimuth manifold matrix according to the formula in step 103. The maximum detection range is considered to be 20m, and the field of view is [-60°, 60°]. Therefore, the reference point is (10m, 0°). The range and azimuth resolutions are consistent with those in the preprocessing process.

[0061] Step 204: Obtain the target range spread function and azimuth spread function based on the weight vector and manifold matrix. On this basis, determine the scale bandwidth of the range and azimuth dimensions. In this embodiment, half of the full width at half maximum is taken. For each cell, the maximum intensity value of its range-azimuth neighborhood is taken as the scale bandwidth of the intensity dimension.

[0062] Step 205: For each cell in the RA heatmap, the neighborhood and contour are determined by scale bandwidth, distance and azimuth spread function. The mean offset vector of each cell in the whole image is obtained by mean offset theory. The value of the mean offset vector is inverted and the value less than 0 is discarded to form the mean offset image. Finally, it is multiplied with the original image to suppress sidelobes and obtain the final super-resolution imaging result.

[0063] It is understood that the image processing in this embodiment is in polar coordinates. The image shown in this embodiment is transformed from a polar coordinate system to a Cartesian coordinate system using grid interpolation.

[0064] The effectiveness of this invention is demonstrated below through simulation and experimental results.

[0065] The algorithm proposed in this invention is implemented using MATLAB R2024b, running on Windows 10 with an Intel Core i7-13650HX CPU. Simulation experiments directly obtain radar echo signals based on pre-set radar parameters and target scattering points, and then perform subsequent processing. In the field experiment, the radar device is placed at the front of the vehicle, and the acquired radar ADC data is read and pre-processed using MATLAB. Finally, the algorithm in this invention is used to obtain the final super-resolution imaging result.

[0066] like Figure 2 The image shown is the first test scenario used to verify the super-resolution imaging performance of this invention. The front of a car is positioned directly in front of the scene, and nine target points are set up for easy observation of the effect. Figure 3This demonstrates a comparison between the traditional DBF method (left) and the super-resolution imaging results based on DBF of the present invention (right) in this scenario. Analysis shows that in the image generated by the DBF method, most scattering points are mixed with neighboring points within the same cell, making them difficult to distinguish; while using the method of the present invention, clear separation and identification of each target point can be achieved.

[0067] like Figure 4 The image shows a second test scenario used to evaluate the super-resolution imaging capability of this invention. A car is placed to the right front of this scenario, and the radar can capture part of the front and side of the car. Therefore, the vehicle presents an almost L-shaped outline within the radar's scanning range, and a total of 9 scattering points are arranged in this scenario. Figure 5 The image presents a comparison of the imaging results of the traditional DBF method (left) and the DBF-based super-resolution imaging results of the present invention (right) in this scenario. The comparison reveals that the DBF method only obtains a blurred outline of the target and cannot effectively identify specific scattering points; most scattering points overlap. In contrast, the super-resolution imaging method of the present invention can clearly distinguish all scattering points and clearly display them on the RA thermal map.

[0068] like Figure 6 The image shown is the imaging result of the DBF method in a real-world parking lot scenario, including six vehicles. The radar's forward-looking field of view is directly facing the vehicles. The final image is obtained by converting multiple frames of imaging results to the Cartesian coordinate system and then performing distance compensation. The results show that the DBF method produces images with significant noise and clutter, and the targets exhibit noticeable sidelobes in the azimuth dimension, resulting in a less clear image. However, as... Figure 7 The image shows the super-resolution imaging results based on DBF using the method of this invention. It can be seen that the diffusion of each target reflection point is significantly reduced, the outline of the vehicle is more clearly displayed, and the surrounding clutter is significantly suppressed.

[0069] like Figure 8 The image shown is the imaging result of the IAA method in a parking lot scene. The final image is obtained by transforming the imaging results of multiple frames into the Cartesian coordinate system and then performing distance compensation. Compared with the DBF method, the resolution is improved and the azimuth spread of each target reflection point is reduced, but it still contains a lot of noise and clutter, resulting in a poor final imaging result. And as... Figure 9 The image shows the super-resolution imaging results based on IAA using the method of the present invention. It can be seen that the method of the present invention can further improve the imaging effect on the basis of traditional super-resolution methods.

[0070] This invention addresses the problem of radar super-resolution imaging by proposing a radar super-resolution forward-looking imaging method. Compared to traditional super-resolution methods, this invention operates in the image domain, avoiding some limitations and performance ceilings faced by traditional signal processing methods. Furthermore, the method comprehensively considers target diffusion in both range and azimuth dimensions. Based on accurately modeling the diffusion model using waveform parameters and array structure, it further extracts target features using mean shift theory, fully utilizing the target characteristics of radar. Additionally, it is noteworthy that this invention's method can further improve imaging performance compared to traditional super-resolution imaging methods.

[0071] Based on the same inventive concept, this invention discloses a radar super-resolution forward-looking imaging system, comprising: a parameter acquisition module for acquiring radar waveform parameters and array structure; a preliminary imaging module for receiving radar echo data of the forward-looking scene, and obtaining and normalizing a preliminary range-azimuth (RA) heatmap image based on the radar echo data; a diffusion model construction module for constructing weight vectors and manifold matrices for range and azimuth dimensions according to the waveform parameters and array structure, and then constructing range diffusion functions and azimuth diffusion functions; and a super-resolution imaging module for calculating the corresponding mean offset vector for each cell on the RA heatmap, inverting the value of the mean offset vector and discarding negative values ​​to form a mean offset image, and multiplying the mean offset image with the original image to obtain the final super-resolution imaging result.

[0072] This invention also discloses a computer system, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, it implements the steps of the radar super-resolution forward-looking imaging method.

[0073] This invention also discloses a computer program product, including a computer program that, when executed by a processor, implements the steps of the radar super-resolution forward-looking imaging method.

[0074] It is understood that the present invention has been described through some embodiments, and those skilled in the art will recognize that various changes or equivalent substitutions can be made to these features and embodiments without departing from the spirit and scope of the invention. Furthermore, under the teachings of the present invention, these features and embodiments can be modified to adapt to specific situations and materials without departing from the spirit and scope of the invention. Therefore, the present invention is not limited to the specific embodiments disclosed herein, and all embodiments falling within the scope of the claims of this application are within the protection scope of the present invention.

Claims

1. A radar super-resolution forward-looking imaging method, characterized in that, Includes the following steps: Obtain the radar's waveform parameters and array structure; Receive radar echo data of the forward-looking scene, and based on the radar echo data, obtain preliminary imaging results in the form of range-azimuth RA thermal map and normalize them; Based on the waveform parameters and array structure, weight vectors and manifold matrices for the range and azimuth dimensions are constructed, and then the range spread function and azimuth spread function are constructed. For each cell in the RA heatmap, the corresponding mean offset vector is calculated. The value of the mean offset vector is inverted and the negative value is discarded to form a mean offset image. The mean offset image is multiplied with the original image to obtain the final super-resolution imaging result. The calculation of the mean offset vector of each cell includes: determining the scale bandwidth in the three dimensions of range, azimuth, and intensity; determining the neighborhood based on the scale bandwidth; calculating the probability density function using mean offset theory; and calculating the gradient of the probability density function to obtain the mean offset vector at the cell. The probability density function is calculated using the constructed range spread function and azimuth spread function.

2. The radar super-resolution forward-looking imaging method according to claim 1, characterized in that, For distance is azimuth angle is The reference point, and the weight vectors for distance and orientation dimensions are respectively and ,in Indicates an index. Represents the imaginary unit. The speed at which electromagnetic waves propagate in the air. For wavelength, This represents the frequency corresponding to each distance unit. , For bandwidth, For fast sampling points, This represents the position coordinates corresponding to each array element. The number of array elements. The distance between array elements is denoted by T, which indicates transpose.

3. The radar super-resolution forward-looking imaging method according to claim 2, characterized in that, The manifold matrices for the distance and orientation dimensions are respectively and , For length is The distance vector represents the distance unit divided within the detection range; For length is An angular vector represents the angular units divided within the field of view.

4. The radar super-resolution forward-looking imaging method according to claim 3, characterized in that, Distance spread function azimuth spread function ,in Distance vector The location of the peak Elements of one unit, Distance vector The location of the peak Elements of one unit, , The superscript H indicates conjugate transpose.

5. The radar super-resolution forward-looking imaging method according to claim 1, characterized in that, unit probability density function ,in: ， This represents the total number of cells in the RA heatmap. Represents the scale bandwidth based on the distance, orientation, and intensity dimensions. Determined unit The neighborhood, , These represent the total dimensions of distance and orientation, respectively. , Let these represent the range spread function and the azimuth spread function, respectively. The contour representing the intensity dimension is calculated using a Gaussian kernel. , , These represent the distance, azimuth coordinates, and intensity value of the current element, respectively. , , These represent the distance, azimuth coordinates, and intensity value of each cell within the neighborhood, respectively. This represents the Euclidean norm.

6. The radar super-resolution forward-looking imaging method according to claim 1, characterized in that, The scale bandwidth in the distance and orientation dimensions is half of the full width at half the peak, and the scale bandwidth in the intensity dimension is the maximum intensity value in the distance-orientation neighborhood of the current cell.

7. The radar super-resolution forward-looking imaging method according to claim 1, characterized in that, In the initial imaging stage, radar data is preprocessed, and a radar heatmap is obtained by range-dimensional fast Fourier transform and angle spectrum estimation. The angle spectrum estimation adopts a digital beamforming method or an iterative adaptive algorithm.

8. A radar super-resolution forward-looking imaging system, characterized in that, include: The parameter acquisition module is used to acquire the radar's waveform parameters and array structure. The initial imaging module is used to receive radar echo data of the forward-looking scene, and based on the radar echo data, obtain and normalize the preliminary imaging results in the form of range-azimuth RA heatmap. The diffusion model construction module is used to construct the weight vectors and manifold matrices of the distance and azimuth dimensions based on the waveform parameters and array structure, and then construct the distance diffusion function and the azimuth diffusion function. The super-resolution imaging module is used to calculate the corresponding mean offset vector for each cell on the RA heatmap, invert the value of the mean offset vector and discard the negative value to form a mean offset image, and multiply the mean offset image with the original image to obtain the final super-resolution imaging result. The calculation of the mean offset vector for each cell includes: determining the scale bandwidth in the three dimensions of range, azimuth and intensity, determining the neighborhood based on the scale bandwidth, calculating the probability density function using mean offset theory, and calculating the gradient of the probability density function to obtain the mean offset vector at the cell. The probability density function is calculated using the constructed range spread function and azimuth spread function.

9. A computer system comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the computer program is executed by the processor, it implements the steps of a radar super-resolution forward-looking imaging method according to any one of claims 1-7.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of a radar super-resolution forward-looking imaging method according to any one of claims 1-7.

Images