SAR (Synthetic Aperture Radar) self-focusing method for multi-stage sub

Abstract

The invention provides an SAR (Synthetic Aperture Radar) self-focusing method for multi-stage sub-aperture segmentation, and relates to the technical field of radar signal processing. The method comprises the following steps: preprocessing original SAR echo data to obtain a coarse focusing SAR image; secondly, for a coarse focusing SAR image, reducing the image resolution by dividing frequency domain sub-apertures, and reducing the sensitivity of the image to a motion error; then, a focusing algorithm based on a model or a phase gradient is adopted in each stage of sub-aperture, and the image quality of each stage of sub-aperture with a good focusing effect is obtained; and finally realizing global focusing of the SAR image through step-by-step sub-aperture focusing. Compared with direct division of a plurality of sub-apertures, the method greatly improves the focusing effect of the SAR image while improving the efficiency, and is effectively verified in simulation, actual airborne flight and satellite-borne data.

Description

Technical Field

[0001] The present invention relates to the technical field of radar signal processing, and particularly to a SAR autofocus method with multi-level sub-aperture segmentation. Background Art

[0002] Synthetic Aperture Radar (SAR), as a powerful remote sensing technology, provides high-resolution imaging capabilities of the earth's surface under all-weather and all-time conditions, and is widely used in fields such as geographical mapping, disaster monitoring, ocean observation, and military reconnaissance. However, due to factors such as the imperfections of the radar system itself, the instability of the aircraft movement, and the atmospheric effects, the collected original SAR data often contains phase errors, which will seriously affect the focusing quality and spatial resolution of the final image, resulting in blurred and distorted images and affecting the accurate extraction and analysis of information. Therefore, motion compensation is an important link in obtaining high-quality SAR images.

[0003] In related technologies, common motion compensation methods include two major categories: sensor measurement data-based compensation strategies and echo data-based autofocus strategies. In most cases, the motion error measurements obtained only by inertial navigation systems cannot meet the requirements of high-quality SAR imaging, especially for light and small platforms that are vulnerable to atmospheric disturbances. In addition, compared with low-resolution SAR systems, high-resolution SAR systems are more sensitive to motion errors and have higher requirements for error measurement and estimation accuracy. In actual application scenarios, whether in civilian or military fields, users' requirements for SAR resolution are also gradually increasing. Therefore, it is particularly urgent to conduct research on high-precision autofocus algorithms based on echo data. Autofocus algorithms based on echo data are mainly divided into two major categories: non-parametric estimation algorithms and model-based parametric estimation algorithms. Non-parametric estimation algorithms have high estimation accuracy but strong dependence on the scene, such as requiring strong scatterers in the scene. Model-based parametric estimation algorithms have no special requirements for the scene and high estimation efficiency, but the error form and error order depend on the model.

[0004] In addition, the above two major categories of autofocus algorithms are often used in cooperation with sub-apertures, and previous sub-aperture algorithms directly divide the data into multiple sub-apertures for processing. However, since the original data collected by high-resolution SAR systems is severely contaminated by motion errors, the estimation effect of a single-level sub-aperture is not necessarily good. Summary of the Invention

[0005] (1) Technical Problems to be Solved

[0006] Aiming at the deficiencies of the prior art, the present invention provides a SAR autofocus method with multi-level sub-aperture segmentation, which solves the technical problem that the potential of sub-aperture estimation needs to be further explored.

[0007] (2) Technical solution

[0008] To achieve the above objectives, the present invention is implemented through the following technical solutions:

[0009] A SAR autofocus method based on multi-level sub-aperture segmentation, comprising:

[0010] Obtain and preprocess the original SAR echo data to obtain a coarsely focused SAR image;

[0011] Divide the coarsely focused SAR image into several first-level sub-apertures in the azimuth frequency domain;

[0012] Shift the Doppler center of each of the first-level sub-apertures to zero frequency;

[0013] Evenly divide each first-level sub-aperture after Doppler center adjustment into several second-level sub-apertures;

[0014] Based on a model or a phase gradient-based focusing algorithm, perform autofocus processing on each of the second-level sub-apertures separately;

[0015] Re-synthesize each autofocused second-level sub-aperture into the corresponding first-level sub-aperture;

[0016] Based on a model or a phase gradient-based focusing algorithm, perform autofocus processing on each synthesized first-level sub-aperture separately;

[0017] Synthesize each fully focused first-level sub-aperture into a single full-aperture SAR image.

[0018] Preferably, dividing the coarsely focused SAR image into two first-level sub-apertures in the azimuth frequency domain includes:

[0019] Multiply the azimuth frequency domain signal of the coarsely focused SAR image by an initialized matched filtering function to obtain a range-Doppler signal;

[0020] Process the range-Doppler signal using an inverse fast Fourier transform and take its absolute value to obtain left and right sub-aperture images respectively;

[0021] Calculate the relative displacement between the left and right sub-aperture images;

[0022] Based on the relative displacement and the effective Doppler bandwidth, calculate the azimuth frequency domain linear chirp rate error;

[0023] Based on the azimuth frequency domain linear chirp rate error, correct the matched filtering function and repeat the above operations until the azimuth frequency domain linear chirp rate error meets the preset accuracy;

[0024] Output the finally obtained left and right sub-aperture images, and use them as the two first-level sub-apertures.

[0025] Preferably, the calculation process of the azimuth frequency-domain linear chirp rate error includes:

[0026] Let the azimuth Doppler spectrum model of the coarsely focused SAR image be shown as the following formula:

[0027]

[0028] where S is the azimuth spectrum after azimuth ideal quadratic phase matching, f a is the azimuth frequency, B d is the Doppler bandwidth, j is the imaginary unit, Δk af is the azimuth chirp rate error, rect(·) is the rectangular window function, and exp(·) is the exponential function with the natural constant as the base;

[0029] Divide the azimuth into left and right sub-aperture images in the frequency domain:

[0030]

[0031] where S1 and S2 are the azimuth spectra of the left and right sub-aperture images after dividing the azimuth spectrum S;

[0032] The corresponding time-domain expression is:

[0033]

[0034] where s1(t a ) and s2(t a ) respectively represent the time-domain images corresponding to the left and right sub-aperture images. The first rect(·) in formulas (4) and (5) is the envelope of the time-domain signal, and the relative displacement between the left and right sub-aperture images is:

[0035]

[0036] Based on the relationship between the azimuth linear chirp rate, relative displacement, and Doppler bandwidth, obtain the azimuth frequency-domain linear chirp rate error as:

[0037]

[0038] where Δk af is the azimuth frequency-domain linear chirp rate error, and Δt is the relative displacement.

[0039] Preferably, the step of at least evenly dividing each first-level sub-aperture after Doppler center adjustment into two second-level sub-apertures includes:

[0040] Perform azimuth Fourier transform on each first-level sub-aperture adjusted by the Doppler center to evenly divide the effective Doppler bandwidth of the signal into N sub-bands in the azimuth frequency domain and use them as N second-level sub-apertures; where N is a positive integer greater than or equal to 2.

[0041] Preferably, the self-focusing process for each of the second-level sub-apertures includes:

[0042] Perform inverse Fourier transform on the N second-level sub-apertures in the azimuth direction to obtain N sub-aperture time-domain images;

[0043] Calculate the cross-correlation function between any pair of sub-aperture time-domain images to obtain cross-correlation functions;

[0044] Search for the peak point positions corresponding to the cross-correlation functions to obtain the offsets between the pairs of sub-aperture time-domain images;

[0045] Based on the offsets, construct an offset matrix;

[0046] Construct a position offset matrix between any pair of sub-aperture images introduced by the second to Nth order phase errors;

[0047] Based on the offset matrix and the position offset matrix, solve the polynomial coefficients of each order corresponding to the Nth order phase error;

[0048] Based on the polynomial coefficients of each order, perform a compensation operation to ensure that each of the second-level sub-apertures reaches the optimal focusing state.

[0049] Preferably, the calculation process of the polynomial coefficients of each order corresponding to the Nth order phase error includes:

[0050] Assume that the error phase contained in any sub-aperture in the frequency domain of the first-level sub-aperture is in the form of an Nth order polynomial, excluding the linear error; the phase error model is:

[0051]

[0052] where, φ e is the phase error, B sd is the Doppler bandwidth corresponding to the first-level sub-aperture a k is the kth order phase error coefficient in the polynomial phase error model, is the kth power of the azimuth frequency f a i.e., the kth order phase error in the phase error model;

[0053] Divide any sub-aperture in the first-level sub-aperture into N sub-apertures, and the width of the sub-aperture is The phase error within the i-th sub-aperture range is:

[0054]

[0055] Among them, φ i represents the phase error within the i-th sub-aperture, and f ai is the center of the i-th sub-aperture, and the expression is:

[0056]

[0057] In the phase error model, through binomial decomposition, it can be known that the model of the linear phase error component is:

[0058]

[0059] Among them, φ lin,i is the linear error component of φ i in the phase error within the i-th sub-aperture, is the (k - 1)th power of the center f ai of the i-th sub-aperture;

[0060] The offset Δ i,j between any two sub-aperture images i and j is:

[0061]

[0062] Expressing the above linear equations in matrix form, the position offset matrix Δ:

[0063] Δ = δa(13)

[0064] Among them

[0065] Δ = [Δ 1,2 … Δ 1,N Δ 2,3 … Δ 2,N Δ 3,4 … Δ N-1,N T (14)

[0066] a = [a2 a3 … a N T (15)

[0067]

[0068] Among them, the superscript T represents transpose; represents the position offset introduced by the k-th order phase error between the i-th and j-th sub-aperture images, and is defined as follows:

[0069]

[0070] ​​Then the solution of the polynomial error coefficient is approximately:

[0071] a = δ -1 Δ (18)

[0072] where δ -1 is the pseudo-inverse matrix of the position offset matrix δ.

[0073] A multi-level sub-aperture segmentation SAR autofocus system, comprising:

[0074] An acquisition module, configured to acquire and preprocess the original SAR echo data and obtain a coarsely focused SAR image;

[0075] A first-level division module, configured to divide the coarsely focused SAR image into a plurality of first-level sub-apertures in the azimuth frequency domain;

[0076] An adjustment module, configured to shift the Doppler center of each of the first-level sub-apertures to zero frequency;

[0077] A second-level division module, configured to uniformly divide each first-level sub-aperture after Doppler center adjustment into a plurality of second-level sub-apertures;

[0078] A first autofocus module, configured to perform autofocus processing on each of the second-level sub-apertures respectively based on a model or a phase gradient-based focusing algorithm;

[0079] A first synthesis module, configured to re-synthesize each autofocused second-level sub-aperture into a corresponding first-level sub-aperture;

[0080] A second autofocus module, configured to perform autofocus processing on each synthesized first-level sub-aperture respectively based on a model or a phase gradient-based focusing algorithm;

[0081] A second synthesis module, configured to synthesize each fully focused first-level sub-aperture into a single full-aperture SAR image.

[0082] A storage medium stores a computer program for SAR autofocus with multi-level sub-aperture segmentation, wherein the computer program causes a computer to execute the SAR autofocus method as described above.

[0083] An electronic device, comprising:

[0084] One or more processors; a memory; and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, and the programs include those for executing the SAR autofocus method as described above.

[0085] (III) Beneficial effects

[0086] The present invention provides a SAR autofocus method based on multi-level sub-aperture segmentation. Compared with the prior art, it has the following beneficial effects:

[0087] In the present invention, the original SAR echo data is preprocessed to obtain a roughly focused SAR image; then, for the roughly focused SAR image, the image resolution is reduced by dividing the frequency domain sub-apertures, and the sensitivity of the image to motion errors is reduced; then, in each level of sub-aperture, a focusing algorithm based on a model or a phase gradient is adopted to obtain the image quality of each level of sub-aperture with better focusing effect; through step-by-step sub-aperture focusing, the global focusing of the SAR image is finally realized. Compared with directly dividing multiple sub-apertures, while improving the efficiency, the focusing effect of the SAR image is greatly improved, and it has been effectively verified in simulations, actual airborne flights, and spaceborne data. BRIEF DESCRIPTION OF THE DRAWINGS

[0088] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0089] Figure 1 It is a block diagram of a SAR autofocus method based on multi-level sub-aperture segmentation provided by an embodiment of the present invention;

[0090] Figure 2 It is a flowchart of a SAR autofocus algorithm provided by an embodiment of the present invention;

[0091] Figure 3 It is a conceptual diagram of an autofocus algorithm based on multi-level aperture segmentation provided by an embodiment of the present invention (taking two-level sub-apertures as an example);

[0092] Figures 4(a) to 4(b) They are respectively the imaging results of a simulation scenario without compensation and the azimuthal imaging results of a single target provided by an embodiment of the present invention;

[0093] Figures 5(a) to 5(b) They are respectively the imaging results and the azimuthal imaging results of a single target after multi-level sub-aperture compensation provided by an embodiment of the present invention;

[0094] Figures 6(a) to 6(b) They are respectively the imaging results and the azimuthal imaging results of a single target after traditional single-level sub-aperture compensation provided by an embodiment of the present invention;

[0095] Figure 7 、 Figure 8 They are respectively schematic diagrams of the azimuth slices before and after focusing of the first-level sub-aperture 1 and the first-level sub-aperture 2 provided by an embodiment of the present invention;

[0096] Figure 9 Schematic diagram of the azimuth slice after the original image provided by the embodiment of the present invention is synthesized with sub-apertures;

[0097] Figure 10 Schematic diagram of the azimuth slice after focusing after sub-aperture synthesis provided by the embodiment of the present invention;

[0098] Figure 11 An SAR original image provided by the embodiment of the present invention;

[0099] Figure 12 An SAR image after focusing using the proposed algorithm provided by the embodiment of the present invention;

[0100] Figure 13 、 Figure 14 Schematic diagrams of the azimuth slices before and after focusing of the first-level sub-aperture 1 and the first-level sub-aperture 2 provided by the embodiment of the present invention respectively;

[0101] Figure 15 Schematic diagram of the azimuth slice after the original image provided by the embodiment of the present invention is synthesized with sub-apertures;

[0102] Figure 16 Schematic diagram of the azimuth slices of the direct multi-aperture and the multi-level sub-apertures provided by the embodiment of the present invention;

[0103] Figure 17 An "Sentinel-1" SAR original image provided by the embodiment of the present invention;

[0104] Figure 18 An SAR image after focusing of the direct multi-sub-apertures provided by the embodiment of the present invention;

[0105] Figure 19 An SAR image after focusing using the proposed algorithm provided by the embodiment of the present invention;

[0106] Figures 20(a) to 20(b) Point target performance diagrams of the sub-aperture 1 before focusing of the second-level sub-aperture and after focusing of the multi-level sub-apertures provided by the embodiment of the present invention respectively;

[0107] Figure 21 Schematic diagram of the azimuth slice before and after focusing of the first-level sub-aperture 1 provided by the embodiment of the present invention;

[0108] Figures 22(a) to 22(b) Point target performance diagrams of the sub-aperture 3 before and after focusing of the second-level sub-aperture provided by the embodiment of the present invention respectively;

[0109] Figure 23 Schematic diagram of the azimuth slice before and after focusing of the first-level sub-aperture 2 provided by the embodiment of the present invention;

[0110] Figures 24(a) to 24(b) The point target qualities before and after multi - level sub - aperture processing, i.e., before and after multi - level sub - aperture focusing, provided by the embodiments of the present invention respectively.

[0111] Figure 25 A schematic diagram of the original image and the azimuth slice after sub - aperture synthesis provided by the embodiments of the present invention.

[0112] Figures 26(a) to 26(b) Schematic diagrams of point targets of direct multi - sub - aperture and multi - level sub - aperture provided by the embodiments of the present invention respectively.

[0113] Figure 27 A schematic diagram of the azimuth slices of direct multi - aperture and multi - level sub - aperture provided by the embodiments of the present invention. Detailed implementation manners

[0114] To make the objectives, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention are described clearly and completely. Apparently, the described embodiments are some but not all of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0115] The embodiments of the present application provide a SAR autofocus method based on multi - level sub - aperture segmentation, which solves the technical problem that the potential of sub - aperture estimation needs to be further explored.

[0116] The overall idea of the technical solutions in the embodiments of the present application to solve the above - mentioned technical problems is as follows:

[0117] Considering that existing sub - aperture algorithms directly divide data into multiple sub - apertures for processing without further secondary sub - aperture division of the obtained sub - apertures, and the raw data collected by high - resolution SAR systems is seriously contaminated by motion errors, it is necessary to further explore the potential of sub - aperture estimation. The embodiments of the present invention propose a SAR autofocus algorithm based on frequency - domain multi - level sub - aperture segmentation for high - resolution SAR imaging systems. The basic idea is to reduce the image resolution by dividing frequency - domain sub - apertures, reduce the sensitivity of the image to motion errors, and then adopt a model - based or phase - gradient - based focusing algorithm for each level of sub - aperture to obtain sub - aperture images with better focusing effects at each level. Through gradual sub - aperture focusing, global focusing of the SAR image is finally achieved.

[0118] It should be noted that the present application creatively proposes the concept of "multi - level sub - aperture", which refers to the result obtained by dividing the SAR image into sub - apertures level by level.

[0119] To better understand the above technical solution, the above technical solution will be described in detail below in conjunction with the accompanying drawings of the specification and specific implementation manners.

[0120] Embodiment 1:

[0121] As Figure 1 shown, an embodiment of the present invention provides a multi-level sub-aperture segmentation SAR autofocus method, including:

[0122] S1. Obtain and preprocess the original SAR echo data to obtain a coarsely focused SAR image;

[0123] S2. Divide the coarsely focused SAR image into several first-level sub-apertures in the azimuth frequency domain;

[0124] S3. Move the Doppler center of each first-level sub-aperture to zero frequency;

[0125] S4. Uniformly divide each first-level sub-aperture after Doppler center adjustment into several second-level sub-apertures;

[0126] S5. Based on a model or a phase gradient-based focusing algorithm, perform autofocus processing on each second-level sub-aperture respectively;

[0127] S6. Re-synthesize each autofocused second-level sub-aperture into the corresponding first-level sub-aperture;

[0128] S7. Based on a model or a phase gradient-based focusing algorithm, perform autofocus processing on each synthesized first-level sub-aperture respectively;

[0129] S8. Synthesize each fully focused first-level sub-aperture into a single full-aperture SAR image.

[0130] Compared with directly dividing multiple sub-apertures, the example of the present invention improves the efficiency while greatly improving the focusing effect of the SAR image, and has been effectively verified in simulations, actual airborne flights, and spaceborne data.

[0131] As Figure 2 shown, Figure 2 a SAR autofocus algorithm flow chart is disclosed. Among them: First, divide the original data into multiple time-domain sub-apertures, and perform imaging processing on the echoes corresponding to the time-domain apertures respectively. Then, divide each sub-image into multiple levels of sub-apertures, and perform focusing processing on each level of sub-apertures step by step. Finally, estimate the high-order motion error and complete the focusing of the image. The focusing order starts from the sub-aperture with a deeper level. After completing the secondary focusing, it is synthesized to the upper level, and then processed level by level upwards until finally returning to the original image domain.

[0132] Taking the two-level sub-aperture as an example to elaborate the principle of the algorithm, as Figure 3As shown below. First, the frequency domain is divided into two sub-apertures, and then multi-aperture processing is performed on the two sub-apertures respectively; the two sub-aperture images obtained by the processing are synthesized into one image and subjected to autofocus processing to obtain a focused SAR image. Taking the secondary sub-aperture division as an example, the effectiveness of the idea is illustrated by combining the MD (Map Drift) algorithm and the MAM (Multiple Aperture Map Drift) algorithm:

[0133] In step S1, the original SAR echo data is acquired and preprocessed to obtain a coarsely focused SAR image.

[0134] In this step, the original SAR echo data is subjected to preliminary processing such as motion compensation and two-dimensional focusing to obtain a coarsely focused SAR image.

[0135] It can be understood that although the resolution of the coarsely focused SAR image in the range direction and azimuth direction is relatively low and the detailed features of the target cannot be clearly distinguished. However, compared with the original SAR data, it can already present the basic outline and general structure of the target scene.

[0136] In addition, the embodiments of the present invention do not limit the source of the original SAR echo data. Whether it is from actual airborne flight data or spaceborne SAR satellite data, or data collected by other devices equipped with high-resolution SAR, it can be used as the processing object of this step.

[0137] In step S2, the coarsely focused SAR image is divided into several primary sub-apertures in the azimuth frequency domain.

[0138] For the obtained coarsely focused SAR image, this step divides it into multiple primary sub-apertures in the azimuth frequency domain. This process is the basis of the entire algorithm. It divides the original SAR image into multiple sub-aperture images with relatively low azimuth resolution by means of Doppler spectrum segmentation. In this way, the obtained sub-images are less sensitive to errors, easier to focus, and the processing efficiency can be improved by parallel computing. At the same time, this division also lays the foundation for subsequent focusing and synthesis operations, enabling the algorithm to gradually refine the details of the image and finally achieve high-resolution imaging.

[0139] Specifically, this step includes:

[0140] S21. Multiply the azimuth frequency domain signal of the coarsely focused SAR image by the initialized matched filtering function to obtain the range-Doppler signal.

[0141] S22. Process the range-Doppler signal by using the inverse fast Fourier transform and take its absolute value to obtain the left and right sub-aperture images respectively.

[0142] S23. Calculate the relative displacement between the left and right sub-aperture images.

[0143] S24. Calculate the azimuth frequency domain linear chirp rate error based on the relative displacement and the effective Doppler bandwidth.

[0144] S25. Correct the matched filtering function based on the azimuth frequency domain linear chirp rate error, and repeat the operations of S21 - S24 above until the azimuth frequency domain linear chirp rate error meets the preset accuracy.

[0145] S26. Output the finally obtained left and right sub-aperture images and use them as the two first-level sub-apertures.

[0146] In an optional implementation manner, the calculation process of the azimuth frequency domain linear chirp rate error includes:

[0147] Assume that the azimuth Doppler spectrum model of the rough focused SAR image is shown as the following formula:

[0148]

[0149] where S is the azimuth spectrum after azimuth ideal quadratic phase matching, f a is the azimuth frequency, B d is the Doppler bandwidth, j is the imaginary unit, Δk af is the azimuth chirp rate error, rect(·) is the rectangular window function, and exp(·) is the exponential function with the natural constant as the base.

[0150] Divide the azimuth direction into left and right sub-aperture images in the frequency domain:

[0151]

[0152] where S1 and S2 are the azimuth spectra of the left and right sub-aperture images after the division of the azimuth spectrum S respectively.

[0153] The corresponding time domain expression is:

[0154]

[0155] where s1(t a ) and s2(t a ) represent the time domain images corresponding to the left and right sub-aperture images respectively. The first term rect(·) in formulas (4) and (5) is the envelope of the time domain signal. The relative displacement between the left and right sub-aperture images is:

[0156]

[0157] Based on the relationship between the azimuth linear chirp rate and the relative displacement and Doppler bandwidth, the azimuth frequency domain linear chirp rate error is obtained as follows:

[0158]

[0159] where, Δk af is the azimuth frequency domain linear chirp rate error, and Δt is the relative displacement.

[0160] It should be noted that when obtaining the azimuth frequency domain linear chirp rate error through the above operations, the following points need to be noted:

[0161] 1) Block calculation: To reduce the influence of noise and improve the parameter estimation accuracy, it is usually necessary to perform block calculation along the range term. For example, calculate one parameter for every 50 range cells. This method of block calculation can effectively reduce the influence of noise on signal processing and improve the accuracy of parameter estimation.

[0162] 2) Selection of the number of range cells: The number of range cells should not be too large. Within the selected range, the change in the Doppler chirp rate should be negligible. This is because too large a number of range cells may cause the change in the Doppler chirp rate to be too significant, thus affecting the accuracy of signal processing. Generally speaking, it is necessary to select an appropriate number of range cells according to the specific application scenario and signal characteristics.

[0163] 3) Polynomial fitting: Finally, perform a second-order polynomial fitting on the multiple estimation results obtained along the range direction. This step can further eliminate the influence of noise and improve the stability and reliability of signal processing. Polynomial fitting is a commonly used mathematical method that smooths the data by fitting a curve to remove the interference caused by noise.

[0164] In step S3, the Doppler center of each of the first-level sub-apertures is shifted to zero frequency.

[0165] After obtaining the first-level sub-apertures, this step needs to shift the Doppler center of each sub-aperture to zero frequency. This adjustment step is crucial as it standardizes the process of dividing each sub-aperture into second-level sub-apertures. The division of the first-level sub-apertures causes the equivalent Doppler center of the SAR image of each sub-aperture to shift. By shifting the Doppler center frequency to zero, the Doppler center of the first-level sub-aperture is made consistent with the Doppler center of the original SAR image, thus providing a more accurate signal basis for subsequent sub-aperture division and focusing processes. In addition, zero Doppler alignment also ensures the alignment of sub-apertures in the frequency domain, which paves the way for the effective synthesis at subsequent levels, enabling the algorithm to smoothly transfer and process signals between different levels.

[0166] In step S4, each of the first-level sub-apertures whose Doppler center has been adjusted is evenly divided into a number of second-level sub-apertures.

[0167] After the Doppler center of each first-level sub-aperture is adjusted, it has similar properties to the original SAR image. Therefore, in this step, in a form similar to the division of the first-level sub-apertures, each first-level sub-aperture is evenly divided into multiple smaller second-level sub-apertures, and this secondary division allows the algorithm to perform more detailed and localized processing.

[0168] Specifically, this step includes:

[0169] Perform azimuth Fourier transform on each first-level sub-aperture after Doppler center adjustment to evenly divide the effective Doppler bandwidth of the signal into N sub-bands in the azimuth frequency domain and use them as N second-level sub-apertures; where N is a positive integer greater than or equal to 2.

[0170] In step S5, based on the model or phase gradient-based focusing algorithm, perform autofocus processing on each of the second-level sub-apertures respectively.

[0171] At this level, this step applies autofocus technology to each second-level sub-aperture to correct phase errors and enhance image sharpness. Autofocus processing is one of the key technologies in improving SAR imaging quality, such as PGA and MAM. It estimates and compensates for phase errors caused by motion, etc., to ensure that each sub-aperture reaches the best focused state. Operating at this finer level, the algorithm can solve local distortion problems that may be ignored in coarser divisions. In some complex terrain or target scenarios, there may be local motion errors or signal distortions, which may be difficult to detect in coarse divisions but can be effectively detected and corrected in the processing of second-level sub-apertures. Through this hierarchical and refined focusing method, the algorithm can gradually optimize the details of the image and improve the imaging quality. The focusing of the second-level sub-apertures can use the MAM algorithm because the MAM algorithm shows good performance and robustness when processing certain types of SAR data.

[0172] Specifically, this step includes:

[0173] S51. Perform inverse Fourier transform on the N second-level sub-apertures along the azimuth direction to obtain N sub-aperture time-domain images;

[0174] S52. Calculate the cross-correlation function between any two sub-aperture time-domain images to obtain cross-correlation functions;

[0175] S53. Search for the peak point position corresponding to the cross-correlation function to obtain the offset between the two sub-aperture time-domain images;

[0176] S54. Based on the offset, construct an offset matrix;

[0177] S55. Construct a position offset matrix between any pair of sub-aperture images introduced by 2nd to Nth order phase errors;

[0178] S56. Based on the offset matrix and the position offset matrix, solve the polynomial coefficients of each order corresponding to the Nth order phase error;

[0179] S57. Based on the polynomial coefficients of each order, perform a compensation operation to ensure that each of the secondary sub-apertures reaches the optimal focusing state.

[0180] In an optional implementation, the calculation process of the polynomial coefficients of each order corresponding to the Nth order phase error includes:

[0181] Assume that the error phase included in the frequency domain of any sub-aperture in the first-level sub-aperture is in the form of an Nth order polynomial, excluding the linear error; the phase error model is:

[0182]

[0183] where, φ e is the phase error, B sd is the Doppler bandwidth corresponding to the first-level sub-aperture a k is the kth order phase error coefficient in the polynomial phase error model, is the kth power of the azimuth frequency f a i.e., the kth order phase error in the phase error model.

[0184] Divide any sub-aperture in the first-level sub-aperture into N sub-apertures, and the width of the sub-aperture is The phase error within the i-th sub-aperture range is:

[0185]

[0186] where, φ i represents the phase error within the i-th sub-aperture, f ai is the center of the i-th sub-aperture, and the expression is:

[0187]

[0188] In the phase error model, through binomial decomposition, it can be known that the model of the linear phase error component is:

[0189]

[0190] where, φ lin,i is the linear error component of φ i in the phase error within the i-th sub-aperture, is the center f ai of the i-th sub-apertureto the power of k - 1.

[0191] The offset Δ between any two sub - aperture images i and j i,j is:

[0192]

[0193] Expressing the above - mentioned linear equations in matrix form, the position offset matrix Δ:

[0194] Δ = δa (13)

[0195] where

[0196] Δ = [Δ 1,2 … Δ 1,N Δ 2,3 … Δ 2,N Δ 3,4 Δ N-1,N T (14)

[0197] a = [a2 a3 … a N T (15)

[0198]

[0199] where the superscript T represents transpose; represents the position offset introduced by the k - th order phase error between the i - th and j - th sub - aperture images, and is defined as follows:

[0200]

[0201] Then the solution of the polynomial error coefficients is approximately:

[0202] a = δ -1 Δ (18)

[0203] where δ -1 is the pseudo - inverse matrix of the position offset matrix δ.

[0204] In step S6, each self - focused secondary sub - aperture is recombined into the corresponding primary sub - aperture.

[0205] After separately focusing the secondary sub - apertures, this step recombines these focused secondary sub - apertures into their respective primary sub - apertures. This recombination step is an important part of the algorithm, which combines the locally optimized sub - apertures into larger and more coherent units. During the recombination process, the algorithm needs to ensure the phase continuity and amplitude consistency between the recombined sub - apertures. This is because if there are phase discontinuities or amplitude inconsistencies during the recombination process, it may lead to artifacts or distortions in the image, thus affecting the final imaging effect.​​

[0206] Through a carefully designed synthesis algorithm, the focused paragraphs can be seamlessly combined to form a first-level sub-aperture with a larger effective Doppler bandwidth, while retaining the improvements made by autofocus. This step is crucial for maintaining the integrity of the SAR image when it returns to a coarser level, providing high-quality intermediate results for subsequent further processing.

[0207] In step S7, based on a model or a phase gradient-based focusing algorithm, each synthesized first-level sub-aperture is separately subjected to autofocus processing.

[0208] With the synthesized first-level sub-aperture, this step applies a focusing process to it to further improve its quality. This step utilizes the corrections made in the previous levels to ensure accurate focusing of the first-level sub-aperture. At this stage, various autofocus techniques can also be used for processing, such as phase gradient autofocus (PGA), MD, or MAM, etc. These techniques can automatically adjust the focusing parameters according to the characteristics and error conditions of the image, thereby enhancing the sharpness and resolution of the sub-aperture. The focusing of the first-level sub-aperture can be processed using the MD algorithm. This step is the final preparation before reconstructing the full-aperture SAR image. Through the focusing processing at this stage, it can be ensured that the first-level sub-aperture reaches the best imaging state before entering the final synthesis stage.

[0209] It should be noted that the focusing algorithm adopted here can be the same as or different from that in step S5. Of course, in an optional implementation manner, the calculation processes of the two focusings before and after can be limited to be consistent. Correspondingly, the autofocus processing process in step S5 can be referred to, and details are not described here again.

[0210] In step S8, each fully focused first-level sub-aperture is synthesized into a single full-aperture SAR image.

[0211] In this step, the fully focused first-level sub-apertures are synthesized into a single full-aperture SAR image. This synthesis step combines all the corrected and focused sub-apertures into a coherent high-resolution image.

[0212] Specifically, this step includes: First, restore the Doppler center of each first-level sub-aperture to the original true Doppler center. Then, reconstruct the complete aperture Doppler spectrum after focusing in the Doppler domain. Finally, perform an inverse Fourier transform on the complete aperture Doppler spectrum to obtain the reconstructed and focused full-aperture SAR image.

[0213] So far, the embodiment of the present invention has completed all the processes of the SAR autofocus method for multi-level sub-aperture segmentation.

[0214] By dividing into multiple levels of sub-apertures, not only can high-order motion errors be effectively estimated, but also the computational load is improved compared to directly dividing into multiple sub-apertures. Whether directly dividing into multiple sub-apertures or dividing into multiple levels of sub-apertures for polynomial error estimation, the most time-consuming operation in the core solution process is the cross-correlation operation.

[0215] Therefore, to illustrate that the multi-level sub-aperture autofocus method can reduce the computational load, the number of cross-correlation operations for the sub-aperture division methods of the two ways is statistically analyzed:

[0216] Directly dividing into 2N sub-apertures requires L1 times of cross-correlation operations to solve the polynomial coefficients. The expression of L1 is:

[0217]

[0218] First, divide into 2 sub-apertures at the first-level sub-aperture, and then divide into N sub-apertures respectively at the second-level sub-aperture. It requires L2 times of cross-correlation operations to solve the polynomial coefficients at each level. The expression is:

[0219]

[0220] L1 - L2 = N - 1 (21)

[0221] Generally, the number of second-level sub-apertures satisfies N ≥ 2. Then, the number of cross-correlation operations required for the multi-level sub-apertures is less than that of directly dividing the sub-apertures.

[0222] In addition, to verify the effectiveness of the proposed algorithm, a simulation experiment analysis is carried out. The parameters of the simulation system are shown in Table 1.

[0223] Table 1 Simulation System Parameters

[0224] Parameter Value Center frequency 14.25 GHz Bandwidth 500 MHz Platform speed 100 m / s Pulse repetition frequency 2000 Hz Pulse width 500 μs

[0225] The imaging result without compensation is as Figures 4(a) to 4(b) shown. The azimuth slice diagram of a single target is shown in Figure 4(b). The azimuth resolution is 0.4219 m, and the peak sidelobe ratio is -3.2 dB. As Figures 5(a) to 5(b) shown, after multi-level sub-aperture error estimation and compensation, the azimuth resolution is 0.1422 m, the peak sidelobe ratio is -12.56 dB, and the integrated sidelobe ratio is: -11.05 dB. As Figures 6(a) to 6(b) shown, the azimuth resolution of the traditional single-level sub-aperture is 0.1452, the peak sidelobe ratio is -11.18 dB, and the integrated sidelobe ratio is: -9.93 dB.

[0226] As Figure 7As shown in the figure, the MAM processing result analysis of sub-aperture 1 is carried out using the secondary sub-aperture image. Before focusing, the azimuth resolution is 0.9375 m, and the peak sidelobe ratio is -5.37 dB; after MAM processing: the azimuth resolution is 0.2672 m, the peak sidelobe ratio is -12.69 dB, and the integrated sidelobe ratio is -11.63 dB.

[0227] As Figure 8 shown in the figure, the MAM processing result analysis of sub-aperture 2 is carried out: without MAM processing, the azimuth resolution is 0.7469 m, and the peak sidelobe ratio is -2.46 dB; after MAM processing: the azimuth resolution is 0.2781 m, the peak sidelobe ratio is -11.00 dB, and the integrated sidelobe ratio is -7.66 dB.

[0228] As Figure 9 shown in the figure, for the full aperture after the synthesis of the two sub-apertures, before synthesis: the azimuth resolution is 1.98 m; after synthesis: the azimuth resolution is 0.1359 m, the peak sidelobe ratio is -7.15 dB, and the integrated sidelobe ratio is -6.16 dB.

[0229] As Figure 10 shown in the figure, for the processed result after synthesis, the polynomial coefficients are solved by using the relative relationship between the already focused first-level sub-apertures for focusing processing. The processing effect is as follows. The green line is the image slice after re-focusing, and the resolution, integrated sidelobe ratio, and peak sidelobe ratio are optimized.

[0230] Furthermore, in order to verify the effectiveness of the proposed algorithm, focusing processing is respectively carried out on the actual airborne flight L-band data and the Sentinel-1 spaceborne SAR satellite data:

[0231] First, the proposed algorithm is used to perform focusing processing on the actual flight L-band SAR data, and the system parameters are shown in Table 2.

[0232] Table 2 Simulation system parameters

[0233] Parameter Value Center frequency 1.3 GHz Bandwidth 200 MHz Platform speed 70 m / s Pulse repetition frequency 1000 Hz Pulse width 10 μs Flying height 5.1 km

[0234] After the motion compensation of the inertial navigation data for the original data, the imaging result is as Figure 11 shown. It can be seen that although the inertial navigation data has been used for compensation, there are still residual errors affecting the image focusing.

[0235] The proposed frequency-domain multi-level sub-aperture autofocus algorithm is used to process the SAR original image after the above processing, and the processing result is as Figure 12 shown. Compared with Figure 11 , the image focusing effect has been greatly improved. In order to evaluate the performance of the algorithm, point targets in the scene are selected, and the focusing effects at each level of sub-aperture are quantitatively analyzed.

[0236] Sub-aperture 1 is divided into 3 sub-apertures in the secondary sub-apertures. After performing MAM focusing processing on it, the SAR image corresponding to the improved sub-aperture 1 is obtained. For the point target slices before and after processing, such as Figure 13 shown, where the red and green curves represent the situations before and after focusing respectively. Azimuth slice before focusing: azimuth resolution is 2.707 m, peak sidelobe ratio -5 dB, integrated sidelobe ratio -4.68 dB; azimuth slice after focusing: azimuth resolution 1.8165 m, peak sidelobe ratio -10.19 dB, integrated sidelobe ratio: -6.2918 dB. From the intuitive and specific indicators, the focusing effect of the first-level sub-aperture 1 has been greatly improved.

[0237] The same operation as that for sub-aperture 1 is performed on sub-aperture 2. For the point target slices before and after processing, such as Figure 14 shown, where the red and green curves represent the situations before and after focusing respectively. Azimuth slice before focusing: azimuth resolution is 2.903 m, peak sidelobe ratio -6.27 dB, integrated sidelobe ratio -8.298 dB; azimuth slice after focusing: azimuth resolution 1.6384 m, peak sidelobe ratio -12.65 dB, integrated sidelobe ratio: -9.44 dB. From the intuitive and specific indicators, the focusing effect of the first-level sub-aperture 2 has been greatly improved.

[0238] The directly synthesized first-level sub-apertures after focusing are directly synthesized. The comparison of the azimuth slices of the original SAR image and the synthesized image is as Figure 15 shown, where the red and green curves represent the situations before and after focusing respectively. Azimuth slice before focusing: azimuth resolution is 3.2412 m, peak sidelobe ratio -4.25 dB, integrated sidelobe ratio -4.69 dB; azimuth slice after focusing: azimuth resolution 1.2644 m, peak sidelobe ratio -7.5152 dB, integrated sidelobe ratio: -9.49 dB. From the intuitive and specific indicators, the focusing of the first-level sub-apertures has greatly improved the quality of the original SAR image

[0239] Since there are still offsets and errors between the two sub-apertures, continue to extract the focusing parameters according to the relationship between these two amplitudes. The target azimuth slice is as Figure 16 shown, where the red and green curves represent the situations of directly dividing multiple sub-apertures and dividing multi-level sub-apertures respectively. Azimuth slice before focusing: azimuth resolution is 0.9973 m, peak sidelobe ratio -16.42 dB, integrated sidelobe ratio -13.57 dB; azimuth slice after focusing: azimuth resolution 0.9261 m, peak sidelobe ratio -17.74 dB, integrated sidelobe ratio: -17.88 dB. From the intuitive and specific indicators, the multi-level sub-aperture method has greatly improved the quality of the original SAR image.

[0240] Next, the proposed algorithm is used to focus the Sentinel-1 spaceborne SAR data, and the system parameters are shown in Table 3.

[0241] Table 3 SAR system parameters

[0242] Parameter Value Center frequency 5.4 GHz Bandwidth 42.2 MHz Sampling rate 46.9 MHz Platform speed 7000 m / s Pulse repetition frequency 1663.5 Hz Pulse width 51.1 μs

[0243] After the motion compensation of the inertial navigation data for the original data, the imaging results are as Figure 17 shown. It can be seen that although the compensation has been carried out using inertial navigation data, there are still residual errors affecting the focusing of the image.

[0244] The SAR original image processed above is directly divided into 5 sub-apertures for focusing processing, and the polynomial coefficients of the estimation error are estimated in the frequency domain. The processing results are as Figure 18 shown, and the image focusing effect has been greatly improved compared with Figure 17 .

[0245] The SAR original image processed above is processed using the proposed frequency-domain multi-level sub-aperture autofocus algorithm. The processing results are as Figure 19 shown. Compared with Figure 17 and Figure 18 , the image focusing effect has been greatly improved. To evaluate the performance of the algorithm, point targets in the scene are selected, and the focusing effects in each level of sub-apertures are quantitatively analyzed.

[0246] For sub-aperture 1, 3 sub-apertures are divided in the second-level sub-aperture. After performing MAM focusing processing on it, the SAR image corresponding to the improved sub-aperture 1 is obtained, and the point targets in the scene are extracted. As shown in Figure 20, after the second-level sub-aperture focusing, the imaging effect of the point targets in the scene has been greatly improved. The sliced analysis of the point targets before and after processing is as Figure 21 shown, where the red and green curves are the situations before and after focusing respectively. Azimuth slice before focusing: azimuth resolution is 34.9794m, peak sidelobe ratio -5.2301dB, integrated sidelobe ratio -6.3191dB; azimuth slice after focusing: azimuth resolution 15.1227m, peak sidelobe ratio -14.0826dB, integrated sidelobe ratio: -8.8524dB. From the intuitive and specific indicators, the focusing effect of the first-level sub-aperture 1 has been greatly improved.

[0247] The same operations as those for sub-aperture 1 are performed on sub-aperture 2, and the point targets in the scene are extracted. As shown in Figure 22, after the second-level sub-aperture focusing, the imaging effect of the point targets in the scene has been greatly improved. The sliced analysis of the point targets before and after processing is as Figure 23As shown, where the red and green curves represent the situations before and after focusing respectively. Azimuth slice before focusing: azimuth resolution is 33.1384 m, peak sidelobe ratio is -2.7446 dB, and integrated sidelobe ratio is -2.9529 dB; azimuth slice resolution after focusing is 11.1776 m, peak sidelobe ratio is -14.7018 B, and integrated sidelobe ratio is -9.6112 dB. From the intuitive and specific indicators, the focusing effect of the first-level sub-aperture 2 has been greatly improved.

[0248] The first-level sub-apertures after focusing are directly synthesized. The comparison of the point target patterns and their azimuth slices in the original SAR image and the synthesized image is shown in Figures 24 and Figure 25 respectively. In Figure 24, (a) and (b) are the imaging results of point targets before and after multi-level sub-aperture focusing respectively. It can be clearly seen that the focusing effect of the point targets after the proposed algorithm is processed has been greatly improved. Perform azimuth slice analysis on the point targets in Figure 24 to obtain Figure 25 , where the red and green curves represent the situations before and after focusing respectively. Azimuth slice before focusing: azimuth resolution is 110.9873 m, the target defocus phenomenon is serious, there is a splitting phenomenon, and it is meaningless to calculate the peak sidelobe ratio and the integrated sidelobe ratio; azimuth slice azimuth resolution after focusing is 8.4161 m, peak sidelobe ratio is -17.3265 dB, and integrated sidelobe ratio is -14.4211 dB. From the intuitive and specific indicators, the focusing of the first-level sub-aperture has greatly improved the quality of the original SAR image. (Imaging azimuth weighting - Taylor window).

[0249] Since there are still offsets and errors between the two sub-apertures, continue to extract the focusing parameters according to the relationship between these two amplitudes. Compare the finally obtained multi-level sub-aperture focused SAR image with the method of directly dividing multiple sub-apertures for focusing, as shown in Figures 26 and Figure 27 respectively. In Figure 26, (a) and (b) are the imaging results of point targets before and after using multiple sub-apertures and multi-level sub-apertures for focusing respectively. It can be clearly seen that the focusing effect of the point targets after the proposed algorithm is processed has been greatly improved. Perform azimuth slice analysis on the point targets in Figure 26 to obtain Figure 27 , where the red and green curves represent the situations of directly dividing multiple sub-apertures and dividing multi-level sub-apertures respectively. Azimuth slice of directly multiple sub-apertures: azimuth resolution is 8.0216 m, peak sidelobe ratio is -17.439 dB, and integrated sidelobe ratio is -15.05391 dB; azimuth slice azimuth resolution after focusing is 7.1011 m, peak sidelobe ratio is -21.0164 dB, and integrated sidelobe ratio is -17.8426 dB. From the intuitive and specific indicators, the multi-level sub-aperture method has greatly improved the quality of the original SAR image.

[0250] Embodiment 2:

[0251] An embodiment of the present invention provides a multi-level sub-aperture segmentation SAR autofocus system, including:

[0252] An acquisition module, configured to acquire and preprocess original SAR echo data and obtain a coarsely focused SAR image;

[0253] A first-level division module, configured to divide the coarsely focused SAR image into a plurality of first-level sub-apertures in the azimuth frequency domain;

[0254] An adjustment module, configured to shift the Doppler center of each of the first-level sub-apertures to zero frequency;

[0255] A second-level division module, configured to uniformly divide each first-level sub-aperture after Doppler center adjustment into a plurality of second-level sub-apertures;

[0256] A first autofocus module, configured to perform autofocus processing on each of the second-level sub-apertures respectively based on a model or a phase gradient-based focusing algorithm;

[0257] A first synthesis module, configured to re-synthesize each autofocused second-level sub-aperture into a corresponding first-level sub-aperture;

[0258] A second autofocus module, configured to perform autofocus processing on each synthesized first-level sub-aperture respectively based on a model or a phase gradient-based focusing algorithm;

[0259] A second synthesis module, configured to synthesize each fully focused first-level sub-aperture into a single full-aperture SAR image.

[0260] Embodiment 3:

[0261] An embodiment of the present invention provides a storage medium, which stores a computer program for SAR autofocus of multi-level sub-aperture segmentation, wherein the computer program enables a computer to execute the SAR autofocus method as described in Embodiment 1.

[0262] Embodiment 4:

[0263] An embodiment of the present invention provides an electronic device, including:

[0264] One or more processors; a memory; and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, and the programs include those for executing the SAR autofocus method as described in Embodiment 1.

[0265] It can be understood that the SAR self-focusing system, storage medium and electronic device with multi-level sub-aperture segmentation provided in the embodiments of the present invention correspond to the SAR self-focusing method with multi-level sub-aperture segmentation provided in the embodiments of the present invention, and the explanations, examples and beneficial effects of the relevant contents can refer to the corresponding parts in the synthetic SAR self-focusing method, which will not be repeated here.

[0266] In summary, compared with the prior art, the present invention has the following beneficial effects:

[0267] In the embodiment of the present invention, the original SAR echo data is preprocessed to obtain a coarsely focused SAR image; then, for the coarsely focused SAR image, the image resolution is reduced by dividing the frequency domain sub-aperture, thereby reducing the sensitivity of the image to motion errors; then, a model-based or phase gradient-based focusing algorithm is used at each level of sub-aperture to obtain sub-aperture image quality at each level with good focusing effect; and by focusing the sub-apertures step by step, the global focusing of the SAR image is finally achieved. Compared with directly dividing multiple sub-apertures, the efficiency is improved while the focusing effect of the SAR image is greatly improved, and it has been effectively verified in simulation, actual airborne flight, and satellite-borne data.

[0268] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0269] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A SAR self-focusing method with multi-level sub-aperture segmentation, characterized in that: include: Acquire and preprocess raw SAR echo data to obtain coarse-focused SAR images; Dividing the coarsely focused SAR image into a plurality of first-order sub-apertures in the azimuth frequency domain; Moving the Doppler center of each of the first-order sub-apertures to zero frequency; Each first-order sub-aperture adjusted by the Doppler center is evenly divided into a number of second-order sub-apertures; Based on a model or a phase gradient-based focusing algorithm, each of the secondary sub-apertures is subjected to self-focusing processing; Resynthesize each secondary sub-aperture after self-focusing into a corresponding primary sub-aperture; Based on the model or phase gradient based focusing algorithm, each synthesized first-order sub-aperture is subjected to self-focusing treatment respectively; Each fully focused primary sub-aperture is combined into a single full-aperture SAR image.

2. The SAR self-focusing method according to claim 1, characterized in that: The coarsely focused SAR image is divided into two first-order sub-apertures in the azimuth frequency domain, including: Multiplying the azimuth frequency domain signal of the coarsely focused SAR image with the initialized matched filter function to obtain a range Doppler signal; The range Doppler signal is processed by inverse fast Fourier transform, and its absolute value is taken to obtain left and right sub-aperture images respectively; Calculating the relative displacement between the left and right sub-aperture images; Calculating the azimuth frequency domain chirp rate error based on the relative displacement and the effective Doppler bandwidth; Based on the azimuth frequency domain linear modulation rate error, the matched filter function is corrected, and the above operation is repeated until the azimuth frequency domain linear modulation rate error meets the preset accuracy; The left and right sub-aperture images finally obtained are output as the two first-level sub-apertures.

3. The SAR self-focusing method according to claim 1, characterized in that: The calculation process of the azimuth frequency domain linear modulation rate error includes: Assume that the azimuth Doppler spectrum model of the coarsely focused SAR image is as follows: Where S is the azimuth spectrum after the azimuth ideal secondary phase matching, f a is the azimuth frequency, B d is the Doppler bandwidth, j is the imaginary unit, Δk af is the azimuth frequency modulation error, rect(·) is the rectangular window function, and exp(·) is an exponential function with a natural constant as the base; In the frequency domain, the azimuth is divided into left and right sub-aperture images: Among them, S1 and S2 are the azimuth spectra of the left and right sub-aperture images after the azimuth spectrum represents S division; The corresponding time domain expression is: Among them, s1(t a ) and s2(t a ) represent the time domain images corresponding to the left and right sub-aperture images respectively. The first term rect(·) in equations (4) and (5) is the envelope of the time domain signal. The relative displacement between the left and right sub-aperture images is: Based on the relationship between the azimuth linear modulation rate, relative displacement, and Doppler bandwidth, the linear modulation rate error in the azimuth frequency domain is obtained as: Among them, Δk af is the azimuth frequency domain linear modulation rate error, and Δt is the relative displacement.

4. The SAR self-focusing method according to claim 3, characterized in that: The step of evenly dividing each Doppler center-adjusted primary subaperture into at least two secondary subapertures comprises: Each first-order subaperture adjusted by the Doppler center is subjected to azimuth Fourier transform to evenly divide the effective Doppler bandwidth of the signal into N subbands in the azimuth frequency domain, and serve as N second-order subapertures; wherein N is a positive integer greater than or equal to 2.

5. The SAR self-focusing method according to claim 4, characterized in that: The self-focusing process is performed on each of the secondary sub-apertures, comprising: Performing inverse Fourier transform on the N secondary sub-apertures along the azimuth direction to obtain N sub-aperture time domain images; Calculate the cross-correlation function between any pair of sub-aperture time-domain images and obtain mutual functions; Searching for the peak point position corresponding to the cross-correlation function to obtain the offset between the sub-aperture time domain image pair; Based on the offset, construct an offset matrix; Constructing a position offset matrix between any pair of sub-aperture images introduced by 2nd to Nth order phase errors; Based on the offset matrix and the position offset matrix, solving the polynomial coefficients of each order corresponding to the N-order phase error; Based on the polynomial coefficients of each order, a compensation operation is performed to ensure that each of the secondary sub-apertures reaches an optimal focusing state.

6. The SAR self-focusing method according to claim 5, characterized in that: The calculation process of the polynomial coefficients of each order corresponding to the N-order phase error includes: Assume that the error phase contained in the frequency domain of any sub-aperture in the first-order sub-aperture is in the form of an N-order polynomial, excluding linear errors; the phase error model is: Among them, φ e is the phase error, B sd is the Doppler bandwidth corresponding to the first-order sub-aperture a k is the k-th order phase error coefficient in the polynomial phase error model, is the azimuth frequency f a The kth power, that is, the kth order phase error in the phase error model; Divide any sub-aperture in the first-level sub-aperture into N sub-apertures, and the width of the sub-aperture is The phase error within the i-th sub-aperture is: Among them, φ i represents the phase error within the i-th sub-aperture, f ai is the center of the i-th sub-aperture, and the expression is: In the phase error model, through binomial decomposition, it can be seen that the model of the linear phase error component is: Among them, φ lin,i is the phase error φ in the i-th sub-aperture i The linear error component of is the center of the i-th subaperture f ai to the k-1th power; The offset Δ between any two subaperture images i and j is i,j for: The above linear equations are expressed in matrix form, and the position offset matrix Δ is: Δ=δa (13) in D=[D 1,2 …D 1,N D 2,3 …D 2,N D 3,4 …D N-1,N ] T (14) <h2 style=";text-align:left;direction:ltr">a=[a2 a3 … a<h2 style=";text-align:left;direction:ltr"> N <h2 style=";text-align:left;direction:ltr"> ]<h2 style=";text-align:left;direction:ltr"> T <h2 style=";text-align:left;direction:ltr"> (15) Among them, the superscript T represents transposition; represents the position offset of the i-th and j-th sub-aperture images introduced by the k-th order phase error and is defined as follows: Then the solution of the polynomial error coefficient is approximately: a=δ -1 D (18) Among them, δ -1 is the pseudo-inverse matrix of the position offset matrix δ.

7. A SAR self-focusing system with multi-level sub-aperture segmentation, characterized in that: include: An acquisition module is used to acquire and preprocess raw SAR echo data to obtain a coarsely focused SAR image; A primary division module, used for dividing the coarsely focused SAR image into a plurality of primary sub-apertures in the azimuth frequency domain; An adjustment module, used for moving the Doppler center of each of the first-order sub-apertures to zero frequency; A secondary division module, used for evenly dividing each primary sub-aperture adjusted by the Doppler center into a plurality of secondary sub-apertures; A first self-focusing module is used to perform self-focusing processing on each of the secondary sub-apertures based on a model or a phase gradient-based focusing algorithm; A first synthesis module, used for resynthesizing each secondary sub-aperture after self-focusing into a corresponding primary sub-aperture; A second self-focusing module is used to perform self-focusing processing on each synthesized primary sub-aperture based on a model or a phase gradient-based focusing algorithm; The second synthesis module is used to synthesize each fully focused primary sub-aperture into a single full-aperture SAR image.

8. A storage medium, characterized in that: The computer program for SAR self-focusing with multi-level sub-aperture segmentation is stored therein, wherein the computer program enables the computer to execute the SAR self-focusing method according to any one of claims 1 to 6.

9. An electronic device, characterized in that: include: one or more processors; Memory; and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the programs comprising commands for executing the SAR self-focusing method according to any one of claims 1 to 6.

Images