Synthetic aperture radar spectrum error estimation compensation method

Abstract

The invention provides a synthetic aperture radar spectrum error estimation compensation method, and particularly relates to the technical field of synthetic aperture radar signal processing, and the method comprises the steps: optimizing the image quality through multi-step processing, obtaining a coarse compensation signal through conjugate multiplication and inertial navigation coarse compensation, and generating a coarse focusing sub-image through sub-aperture division and BP imaging technologies. Image details are improved through up-sampling, spectrum decompressing and coherent fusion, high-order and linear phase errors are accurately estimated, a phase gradient is calculated based on a PGA algorithm and a least square method so as to correct full-aperture image spectrum errors, coherent fusion is carried out on compensated sub-images again, and the phase error estimation and compensation process is repeated until the optimal phase error is achieved. And finally performing fusion processing to generate a high-precision radar image. The technical problems of low accuracy of spectrum error estimation and low imaging efficiency of a high-precision image in the existing motion compensation technology are solved, so that the accuracy of spectrum error estimation and the imaging efficiency of the high-precision image are effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of synthetic aperture radar signal processing, and in particular to a synthetic aperture radar spectrum error estimation and compensation method. Background Art

[0002] In recent years, with the rapid development of remote sensing technology, synthetic aperture radar (SAR), as a high-resolution earth observation method, has played an irreplaceable role in military reconnaissance, environmental monitoring, disaster assessment and other fields. However, in practical applications, SAR imaging systems, especially airborne and vehicle-mounted platforms, face complex and changing environmental challenges, among which airflow disturbances and insufficient accuracy of inertial navigation systems are particularly prominent. These factors make it difficult for the platform to maintain an ideal motion trajectory, and then introduce significant motion errors in SAR images, which manifest as image defocusing, geometric distortion and other phenomena, seriously reducing image quality and interpretation accuracy.

[0003] In order to solve the problem of motion error in SAR imaging, traditional motion compensation methods mainly rely on frequency domain imaging algorithms to compensate for motion error by estimating and correcting phase error. These methods can improve image quality to a certain extent, but they have many limitations. For example, the traditional method based on phase gradient autofocusing (PGA) gradually approaches the real motion trajectory by iteratively optimizing the phase error estimation, thereby achieving image focusing. However, such methods usually require a long computing time and high computing resources, and are highly dependent on the initial phase error estimation, and are prone to fall into local optimal solutions. Most of the existing motion compensation methods are based on frequency domain imaging algorithms, but time domain imaging algorithms have better versatility because they have no restrictions on motion trajectory, radar observation angle, etc. Based on this, many scholars at home and abroad have developed motion compensation methods based on the back projection algorithm of the ground rectangular coordinate system. Chen Leping proposed a motion compensation method based on back-projection to deal with the phase error in the data under the maximum image clarity criterion. However, this method has the following problems: (1) This method is based on the traditional back-projection algorithm and has high computational complexity; (2) This method requires cyclic iteration to estimate the optimal parameters and has low computational efficiency; (3) This method is based on multiple sub-aperture data images, and the phase error between the sub-aperture data is discontinuous, making it impossible to estimate the full image spectrum error.

[0004] In summary, the technical problems of low accuracy of spectrum error estimation and low imaging efficiency of high-precision images in existing motion compensation technologies are difficult problems that need to be overcome urgently in this field. Summary of the invention

[0005] The main purpose of the present invention is to provide a synthetic aperture radar spectrum error estimation compensation method to overcome the technical problems of low accuracy of spectrum error estimation and low imaging efficiency of high-precision images in existing motion compensation technology, thereby effectively improving the accuracy of spectrum error estimation and the imaging efficiency of high-precision images.

[0006] In order to achieve the above object, the present invention provides a synthetic aperture radar spectrum error estimation compensation method, thereby effectively improving the accuracy of spectrum error estimation and the imaging efficiency of high-precision images.

[0007] The present invention provides a synthetic aperture radar spectrum error estimation and compensation method, the method comprising the following steps:

[0008] Step 1: obtaining an original callback signal and a reference signal, and obtaining a coarse compensation signal by conjugate multiplication, removing a residual video phase and coarse compensation processing of an inertial navigation system according to the original callback signal and the reference signal;

[0009] Step 2: According to the coarse compensation signal, a coarse focusing sub-image of N sub-aperture coarse compensation signals is obtained by sub-aperture division, BP imaging and spectrum compression processing;

[0010] Step 3: performing upsampling, spectrum decompression and coherent fusion processing on each of the coarsely focused sub-images to obtain K-1 sub-images;

[0011] Step 4: According to the K-1 sub-images, obtain K-1 high-order phase errors and linear phase errors through phase error estimation;

[0012] Step 5: According to the K-1 high-order phase errors and the linear phase errors, K-1 high-order phase gradients and K-1 linear phase gradients are obtained by PGA algorithm and least square method, and according to the K-1 high-order phase gradients and the K-1 linear phase gradients, a full-aperture image spectrum error is obtained;

[0013] Step 6: According to the full-aperture image spectrum error, obtain K-1 compensated sub-images through error compensation, and coherently fuse the K-1 compensated sub-images to obtain M sub-images;

[0014] Step 7: According to the M sub-images, phase error estimation and error compensation processing are performed on the M sub-images, or coherent fusion processing is performed to obtain a final radar image.

[0015] Optionally, the step 1 includes:

[0016] Obtaining a conjugate multiplied signal by conjugate multiplying the original callback signal and the reference signal;

[0017] After performing Fourier transformation on the conjugate multiplied signal in terms of distance, a residual video phase is removed to obtain a residual video phase-removed signal;

[0018] The signal after removing the residual video phase is subjected to Fourier transformation to obtain pulse pressure data, and the pulse pressure data is coarsely compensated using an inertial navigation system to obtain a coarse compensation signal.

[0019] Optionally, the step 2 includes:

[0020] Dividing the coarse compensation signal into N sub-aperture coarse compensation signals, and establishing a rectangular coordinate system imaging coarse grid on a ground plane based on the sub-aperture coarse compensation signals;

[0021] The BP imaging and spectrum compression processing of the sub-aperture coarse compensation signal are completed coarsely grid by coarse grid to obtain coarse focusing sub-images of N sub-aperture coarse compensation signals.

[0022] Optionally, the step 3 includes:

[0023] Constructing an overlapping sub-image architecture, and performing upsampling and spectrum decompression processing on each of the coarsely focused sub-images at a first layer of the overlapping sub-image architecture to obtain an upsampled coarsely focused sub-image;

[0024] Then, adjacent up-sampled coarsely focused sub-images are coherently fused to obtain K-1 sub-images.

[0025] Optionally, the step 4 includes:

[0026] Performing spectrum compression and upsampling on the K-1 sub-images, and obtaining the K-1 high-order phase errors by high-order phase error estimation;

[0027] The linear phase error is obtained by linear phase error estimation according to the K-1 sub-images.

[0028] Optionally, the performing spectrum compression and upsampling on the K-1 sub-images and obtaining the K-1 high-order phase errors by high-order phase error estimation includes:

[0029] The K-1 sub-images are subjected to spectrum compression and up-sampling, and the high-order phase errors of the K-1 sub-images are estimated using a phase gradient autofocusing algorithm to obtain the K-1 high-order phase errors.

[0030] Optionally, obtaining the linear phase error by linear phase error estimation according to the K-1 sub-images includes:

[0031] Cross-correlation is performed on the adjacent K-1 sub-images to obtain the offset distance of the same-name point targets between adjacent sub-images as △x img ;

[0032] According to the offset distance △x of the same-name point targets between adjacent sub-images img , the linear phase error is calculated by the following formula:

[0033]

[0034] Among them, △F′ x is the azimuth frequency of the sub-image after spectrum compression.

[0035] Optionally, the step 5 includes:

[0036] Estimate the error gradients of the K-1 high-order phase errors by a phase gradient algorithm to obtain K-1 high-order phase gradients;

[0037] According to the linear phase error, the continuous K-1 linear phase gradients are obtained by least square method:

[0038] l=[l1,l2,L,l K-1 ] 1×(K-1) ;

[0039] The K-1 high-order phase gradients and the K-1 linear phase gradients are added to obtain a combined full-aperture image phase error gradient, and the full-aperture image phase error gradient is integrated to obtain a full-aperture image spectrum error.

[0040] Optionally, the step 6 includes:

[0041] The full-aperture image spectrum error is equally divided according to the spectrum width and compensated into the corresponding K-1 sub-images to obtain K-1 compensated sub-images;

[0042] The K-1 compensated sub-images are coherently fused to obtain M sub-images.

[0043] Optionally, the step 7 includes:

[0044] Determining the number of the M sub-images;

[0045] If the number of M sub-images is greater than 2, then execute step 4 to step 7;

[0046] If the number of the M sub-images is equal to 2, the M sub-images are directly subjected to coherent fusion processing after being upsampled to obtain a final radar image;

[0047] If the number of M sub-images is less than 2, the final radar image is directly obtained.

[0048] The synthetic aperture radar spectrum error estimation and compensation method provided by the present application first extracts a coarse compensation signal from the original callback signal through conjugate multiplication, video phase removal and inertial navigation coarse compensation. Subsequently, multiple coarsely focused sub-images are generated by sub-aperture division and BP imaging technology, and the sub-image quality is optimized by upsampling, spectrum decompression and coherent fusion. In the key step, high-order and linear phase errors are accurately identified and separated through phase error estimation, and then the PGA algorithm and the least squares method are used to calculate the phase gradient and comprehensively correct the spectrum error of the full-aperture image. This process effectively eliminates the phase distortion introduced by factors such as platform motion and atmospheric disturbance. Finally, error compensation is performed based on the calculated phase gradient to generate multiple high-quality compensated sub-images, and further coherent fusion is used to integrate them into a more refined radar image. This method overcomes the technical problems of low accuracy of spectrum error estimation and low imaging efficiency of high-precision images in existing motion compensation technology, thereby effectively improving the accuracy of spectrum error estimation and the imaging efficiency of high-precision images. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] The drawings constituting a part of the present application are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0050] Figure 1 A schematic flow chart of a synthetic aperture radar spectrum error estimation and compensation method provided in this application.

[0051] The above drawings have shown clear embodiments of the present application, which will be described in more detail later. These drawings and text descriptions are not intended to limit the scope of the present application in any way, but to illustrate the concept of the present application to those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0052] In order to make the purpose, technical solutions and advantages of this application clearer, the technical solutions in this application will be clearly and completely described below in conjunction with the drawings in this application. Obviously, the described embodiments are part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0053] The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in sequences other than those illustrated or described herein.

[0054] In the present invention, words such as "exemplary" or "for example" are used to indicate examples, illustrations or descriptions. Any embodiment or design described as "exemplary" or "for example" in this application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary" or "for example" is intended to present related concepts in a specific way.

[0055] In recent years, with the rapid development of remote sensing technology, synthetic aperture radar (SAR), as a high-resolution earth observation method, has played an irreplaceable role in military reconnaissance, environmental monitoring, disaster assessment and other fields. However, in practical applications, SAR imaging systems, especially airborne and vehicle-mounted platforms, face complex and changing environmental challenges, among which airflow disturbances and insufficient accuracy of inertial navigation systems are particularly prominent. These factors make it difficult for the platform to maintain an ideal motion trajectory, and then introduce significant motion errors in SAR images, which manifest as image defocusing, geometric distortion and other phenomena, seriously reducing image quality and interpretation accuracy.

[0056] In order to solve the problem of motion error in SAR imaging, traditional motion compensation methods mainly rely on frequency domain imaging algorithms to compensate for motion error by estimating and correcting phase error. These methods can improve image quality to a certain extent, but they have many limitations. For example, the traditional method based on phase gradient autofocusing (PGA) gradually approaches the real motion trajectory by iteratively optimizing the phase error estimation, thereby achieving image focusing. However, such methods usually require a long computing time and high computing resources, and are highly dependent on the initial phase error estimation, and are prone to fall into local optimal solutions. Most of the existing motion compensation methods are based on frequency domain imaging algorithms, but time domain imaging algorithms have better versatility because they have no restrictions on motion trajectory, radar observation angle, etc. Based on this, many scholars at home and abroad have developed motion compensation methods based on the back projection algorithm of the ground rectangular coordinate system. Chen Leping proposed a motion compensation method based on back-projection to deal with the phase error in the data under the maximum image clarity criterion. However, this method has the following problems: (1) This method is based on the traditional back-projection algorithm and has high computational complexity; (2) This method requires cyclic iteration to estimate the optimal parameters and has low computational efficiency; (3) This method is based on multiple sub-aperture data images, and the phase error between the sub-aperture data is discontinuous, making it impossible to estimate the full image spectrum error.

[0057] In summary, how to overcome the technical problems of low accuracy of spectrum error estimation and low imaging efficiency of high-precision images in existing motion compensation technologies is a difficult problem that needs to be solved urgently in this field.

[0058] The synthetic aperture radar spectrum error estimation and compensation method provided in the present application enables the phase error estimation method to be combined with the imaging process of the ground rectangular coordinate system fast back-projection algorithm through the architecture of overlapping sub-images, thereby improving the high-precision imaging efficiency. By analyzing the spectrum error of the rectangular coordinate system back-projection image, the relationship between the linear error and the image offset is constructed, so that the spectrum error between the sub-images is continuous, thereby effectively improving the accuracy of the spectrum error estimation and the imaging efficiency of the high-precision image.

[0059] The technical solution of the present application and how the technical solution of the present application solves the above-mentioned technical problems are described in detail below with specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of the present application will be described below in conjunction with the accompanying drawings.

[0060] Figure 1 The present invention provides a flowchart of a method for compensating a spectrum error of a synthetic aperture radar, and describes the method for compensating a spectrum error of a synthetic aperture radar in detail. Figure 1As shown, the synthetic aperture radar spectrum error estimation and compensation method provided in this embodiment includes:

[0061] S101: obtaining an original callback signal and a reference signal, and obtaining a coarse compensation signal through conjugate multiplication, removal of residual video phase and coarse compensation processing of an inertial navigation system according to the original callback signal and the reference signal.

[0062] Further, the obtaining of the original callback signal and the reference signal, and obtaining the coarse compensation signal by conjugate multiplication, removing the residual video phase and coarse compensation processing of the inertial navigation system according to the original callback signal and the reference signal include:

[0063] Obtaining a conjugate multiplied signal by conjugate multiplying the original callback signal and the reference signal;

[0064] After performing Fourier transformation on the conjugate multiplied signal in terms of distance, a residual video phase is removed to obtain a residual video phase-removed signal;

[0065] The signal after removing the residual video phase is subjected to Fourier transformation to obtain pulse pressure data, and the pulse pressure data is coarsely compensated using an inertial navigation system to obtain a coarse compensation signal.

[0066] In specific implementation, it can be achieved through the following implementation methods:

[0067] Step 1: Conjugate multiplication

[0068] First, obtain the original callback signal received by the radar or sensor (which can be expressed as: Among them, t r =[-n / 2+1:n / 2] / F s , n is the number of distance sampling points, F s is the distance sampling rate, j is an imaginary number, γ is the modulation frequency, T p is the pulse width. The above parameters are preset parameters of the radar and are known before the radar data is recorded. τ=2R / c is the fast time, R is the distance from the radar to the target, and c is the speed of light. is the distance window function, a(t m ) is the azimuth window function. The range window function and the azimuth window function can be calculated by the distance between the radar and the target and the known parameters. ) and the known reference signal (can be expressed as: Among them, τ ref =2R s / c is the reference fast time, R s The reference signal is usually preset or generated according to system characteristics, and has a corresponding relationship with the original echo signal in terms of frequency, waveform, etc. The original echo signal is conjugate multiplied with the reference signal to obtain the conjugate multiplied signal:

[0069] This step aims to preliminarily eliminate the phase difference between signals by means of phase conjugation.

[0070] Step 2: Remove residual video phase

[0071] Next, the conjugate multiplied signal is subjected to a fast Fourier transform (FFT) in the distance dimension (i.e., the time or delay axis) to obtain its frequency domain representation. Since the signal may be affected by the residual video phase during propagation, the frequency domain signal has a phase offset. By designing an appropriate filter or phase correction algorithm, these residual video phases are removed to obtain the signal after the residual video phase is removed:

[0072]

[0073] Among them, f r ' is the new frequency sampling. Then, the signal after removing the residual video phase is subjected to an inverse Fourier transform (IFFT) to restore its time domain representation.

[0074] Step 3: Rough compensation of inertial navigation system

[0075] The signal after removing the residual video phase is further processed to obtain pulse compression data (i.e., the signal after pulse compression processing):

[0076]

[0077] in, F s ′=nγ / F s is the new frequency sampling rate, B = γT p . Pulse compression processing is a common technique in radar signal processing, which is used to improve the resolution and signal-to-noise ratio of the signal. Then, the pulse compression data is roughly compensated using the platform motion parameters (such as position, velocity, acceleration, etc.) provided by the inertial navigation system (INS). The compensation process is mainly based on the prediction and correction of the signal propagation path to eliminate or reduce the signal distortion caused by the platform motion. Finally, a signal roughly compensated by the inertial navigation system is obtained, that is, a coarse compensation signal.

[0078] It can be understood that through the above steps, the phase distortion and platform motion error in the original callback signal can be effectively eliminated, the accuracy and reliability of signal processing can be improved, and high-quality signal input can be provided for subsequent target detection, tracking and recognition tasks.

[0079] S102: According to the coarse compensation signal, coarse focusing sub-images of N sub-aperture coarse compensation signals are obtained through sub-aperture division, BP imaging and spectrum compression processing.

[0080] Furthermore, the method of acquiring coarse focusing sub-images of N sub-aperture coarse compensation signals through sub-aperture division, BP imaging and spectrum compression processing according to the coarse compensation signal includes:

[0081] Dividing the coarse compensation signal into N sub-aperture coarse compensation signals, and establishing a rectangular coordinate system imaging coarse grid on a ground plane based on the sub-aperture coarse compensation signals;

[0082] The BP imaging and spectrum compression processing of the sub-aperture coarse compensation signal are completed coarsely grid by coarse grid to obtain coarse focusing sub-images of N sub-aperture coarse compensation signals.

[0083] The specific implementation process is as follows:

[0084] The specific implementation includes the following steps:

[0085] Step 1: Acquisition and processing of coarse compensation signal

[0086] First, the system receives the reflected or scattered signals from the target area through a preset sensor array, and after preliminary signal preprocessing (such as filtering, gain control, etc.), obtains a coarse compensation signal containing the target position and morphological information. The purpose of the coarse compensation signal is to preliminarily correct the signal distortion caused by factors such as the transmission medium and sensor characteristics, laying the foundation for subsequent precise imaging.

[0087] Step 2: Sub-aperture division

[0088] Determine the number and layout of sub-apertures: Divide the full-aperture coarse compensation signal into N sub-apertures evenly or unevenly according to actual requirements (such as imaging resolution, computing resources, etc.). Each sub-aperture covers a part of the sensor array, and the sub-apertures may or may not overlap, depending on the application scenario and performance requirements.

[0089] Signal segmentation: The overall coarse compensation signal is segmented into N sub-aperture coarse compensation signals according to the sub-aperture division results. Each sub-aperture coarse compensation signal contains the signal information received by all sensors in the corresponding sub-aperture.

[0090] Step 3: BP imaging and spectrum compression based on sub-aperture coarse compensation signal

[0091] Establish a rectangular coordinate system and imaging coarse grid: On the ground plane (or the plane where the target is located), establish a rectangular coordinate system according to the imaging range and resolution requirements, and divide the imaging coarse grid into uniform grids. Each grid point represents a potential target position for subsequent imaging algorithm processing.

[0092] BP imaging processing:

[0093] For each sub-aperture coarse compensation signal, BP imaging processing is performed on each grid point. The BP algorithm reversely projects (i.e., replays the possible emission source position) the signal received by each sub-aperture onto the imaging plane, and forms the brightness or intensity of the image by accumulating the projection values ​​of all sub-apertures.

[0094] During the projection process, physical effects such as signal propagation delay and attenuation are taken into account and the signal is compensated accordingly.

[0095] Spectrum compression processing:

[0096] In order to further suppress noise and enhance image details, the sub-aperture image after BP imaging is subjected to spectral compression. Spectral compression usually involves converting the image from the spatial domain to the frequency domain (such as through Fourier transform), then adjusting the high-frequency and low-frequency components to varying degrees (such as low-pass filtering to reduce noise, high-pass filtering to enhance edges), and finally converting it back to the spatial domain.

[0097] Through spectrum compression, the contrast and clarity of the image can be effectively improved while maintaining the overall image quality.

[0098] Step 4: Generate a coarsely focused sub-image

[0099] After completing the above BP imaging and spectrum compression processing, coarse focusing sub-images corresponding to N sub-apertures are obtained. These sub-images each reflect the target information observed from different perspectives (i.e., different sub-apertures). In subsequent processing, these sub-images can be fused to generate the final high-resolution, high signal-to-noise ratio imaging results.

[0100] It can be understood that this step describes in detail how to generate N sub-aperture coarse focusing sub-images from the coarse compensation signal through sub-aperture division, BP imaging and spectrum compression processing. This step not only improves the accuracy and efficiency of imaging, but also provides a reliable data basis for subsequent advanced image processing and analysis.

[0101] S103: Perform upsampling, spectrum decompression and coherent fusion processing on each of the coarsely focused sub-images to obtain K-1 sub-images.

[0102] Further, the upsampling, spectrum decompression and coherent fusion processing of each of the coarsely focused sub-images to obtain K-1 sub-images includes:

[0103] Constructing an overlapping sub-image architecture, and performing upsampling and spectrum decompression processing on each of the coarsely focused sub-images at a first layer of the overlapping sub-image architecture to obtain an upsampled coarsely focused sub-image;

[0104] Then, adjacent up-sampled coarsely focused sub-images are coherently fused to obtain K-1 sub-images.

[0105] The specific implementation includes the following steps:

[0106] Step 1: Construct overlapping sub-image architecture

[0107] First, a multi-layer overlapping sub-image architecture is designed according to imaging requirements and computing resources. The architecture aims to gradually improve the resolution and fusion effect of sub-images through layer-by-layer processing. In the steps, we focus on the first layer of the architecture, which directly operates on the coarsely focused sub-image.

[0108] Step 2: Upsampling and spectrum decompression

[0109] Upsampling: For each coarsely focused sub-image, an appropriate upsampling technique (such as bilinear interpolation, bicubic interpolation or a more advanced image super-resolution reconstruction algorithm) is used to increase the resolution. The purpose of upsampling is to increase the number of pixels in the image, thereby making it possible to restore more image details.

[0110] During the upsampling process, it is necessary to select appropriate interpolation methods and interpolation factors according to the characteristics of the image and the expected effect.

[0111] Decompression: After upsampling, the sub-image is decompressed to restore the high-frequency information that was suppressed during the spectral compression process. This step is usually achieved by inverse transform (such as inverse Fourier transform) combined with high-frequency enhancement filter.

[0112] After spectrum decompression, the sub-image has richer details and clearer edges, but some noise may also be introduced. Therefore, in subsequent processing, attention should be paid to the balance between noise suppression and detail enhancement.

[0113] Step 3: Coherent fusion processing

[0114] Determine the fusion strategy: There are often overlapping areas between adjacent upsampled coarsely focused sub-images, which provide the basis for coherent fusion. According to the actual application scenario, select a suitable fusion strategy, such as weighted average, maximum fusion, minimum fusion, or a more complex adaptive fusion algorithm based on image content.

[0115] Performing a fusion operation: For adjacent upsampled coarsely focused sub-images, a fusion operation is performed in their overlapping regions. During the fusion process, the pixel values ​​in the overlapping region are calculated according to the selected fusion strategy to generate fused pixel values.

[0116] After the fusion is completed, K-1 new sub-images are obtained (assuming that the number of original coarsely focused sub-images is K). These sub-images have improved resolution, signal-to-noise ratio and detail richness, and the transition between adjacent sub-images is smoother.

[0117] It can be understood that this step describes in detail how to optimize the coarse focus sub-image by constructing an overlapping sub-image architecture, upsampling processing, spectrum decompression processing, and coherent fusion processing. This series of processing steps not only improves the resolution and signal-to-noise ratio of the sub-image, but also achieves a smooth transition between adjacent sub-images through coherent fusion, laying a solid foundation for subsequent advanced image processing and analysis.

[0118] S104: Obtain K-1 high-order phase errors and linear phase errors through phase error estimation according to the K-1 sub-images.

[0119] Further, the obtaining of K-1 high-order phase errors and linear phase errors through phase error estimation according to the K-1 sub-images includes:

[0120] The K-1 sub-images are spectrally compressed and up-sampled, and the K-1 high-order phase errors are obtained by high-order phase error estimation. Specifically, the K-1 sub-images are spectrally compressed and up-sampled, and the high-order phase errors of the K-1 sub-images are estimated using a phase gradient autofocusing algorithm to obtain the K-1 high-order phase errors.

[0121] According to the K-1 sub-images, the linear phase error is obtained by linear phase error estimation, specifically including: performing cross correlation on the adjacent K-1 sub-images to obtain an offset distance of the same-name point target between adjacent sub-images as △x img ; According to the offset distance △x of the same-name point targets between adjacent sub-images img , the linear phase error is calculated by the following formula: Among them, △F′ x is the azimuth frequency of the sub-image after spectrum compression.

[0122] The specific implementation includes the following steps:

[0123] Step 1: High-order phase error estimation

[0124] Spectral compression and upsampling: Spectral compression is performed on the obtained K-1 sub-images to reduce the amount of data and retain key information. The specific method of spectrum compression can be selected according to the actual application, such as retaining low-frequency components after Fourier transform.

[0125] Subsequently, the compressed sub-image is upsampled to improve the resolution and detail of the image. The upsampling method may use bilinear interpolation, bicubic interpolation or a more advanced image super-resolution reconstruction algorithm.

[0126] High-order phase error estimation: The Phase Gradient Autofocus (PGA) algorithm is used to estimate the high-order phase error of the upsampled sub-image. The PGA algorithm gradually adjusts the phase of the image through an iterative optimization process to maximize the contrast or sharpness of the image, thereby estimating the high-order phase error.

[0127] In specific implementation, the initial phase error may be set to zero or an initial value may be set based on prior knowledge, and then the phase gradient is iteratively calculated, and the phase error estimate is updated accordingly until a convergence condition is met or a preset number of iterations is reached.

[0128] Step 2: Linear Phase Error Estimation

[0129] Cross-correlation calculation: Cross-correlation calculation is performed on adjacent K-1 sub-images to find the offset distance between the same-name point targets (i.e., targets corresponding to the same physical point in different sub-images). Cross-correlation calculation can be implemented by the sliding window method, calculating the similarity between the sub-images in each window, and finding the position with the highest similarity as the offset distance.

[0130] Linear phase error calculation: According to the offset distance of the same-name point targets between adjacent sub-images, the linear phase error is calculated using the formula. The azimuth frequency (usually expressed as f) of the sub-image after spectrum compression in the formula is known, which depends on the specific implementation of spectrum compression and the geometric characteristics of the sub-image.

[0131] Since the linear phase error only affects the position of the same target in K-1 sub-images, the image migration method is used to estimate the linear phase error. The image migration method first calculates the offset distance of the same-name points between adjacent images, and cross-correlates the adjacent sub-images to obtain the offset distance of the same-name point targets between adjacent sub-images as △x img , the motion error analysis of the K-1 sub-image shows that the linear phase error in the image range compression phase history domain is:

[0132]

[0133] in, ξ(△F′ x / △F yc ) is the structure of the image spectrum with respect to the azimuth time, △F′ x and ∆F yc Corresponding to the only azimuth time, this structure is used to represent △F′ x is the azimuth frequency of the image after spectrum compression, △F yc is the center distance frequency of the image, t m is the azimuth time, △R0 represents the constant phase error, β represents the radar viewing angle, R brepresents the projected distance from the radar to the target on the ground, and v is the speed of the radar flight platform. From the Fourier transform relationship, it can be seen that the offset in the image can be expressed as a linear phase in the frequency domain, so the existence of linear error will affect the focus position of the target in the image domain, so the calculation formula for the linear phase error can be obtained as: Among them, △F′ x is the azimuth frequency of the sub-image after spectrum compression.

[0134] It can be understood that this step describes in detail how to estimate the high-order phase error and linear phase error of the K-1 sub-images through spectrum compression, upsampling, phase gradient autofocusing algorithm, and cross-correlation calculation. This process not only improves the focusing effect and geometric position accuracy of the sub-images, but also provides important phase correction information for subsequent image reconstruction and fusion.

[0135] S105: According to the K-1 high-order phase errors and the linear phase errors, K-1 high-order phase gradients and K-1 linear phase gradients are obtained by PGA algorithm and least squares method, and according to the K-1 high-order phase gradients and the K-1 linear phase gradients, a full-aperture image spectrum error is obtained.

[0136] Further, according to the K-1 high-order phase errors and the linear phase errors, K-1 high-order phase gradients and K-1 linear phase gradients are obtained by PGA algorithm and least squares method, and according to the K-1 high-order phase gradients and the K-1 linear phase gradients, a full-aperture image spectrum error is obtained, which includes:

[0137] Estimate the error gradients of the K-1 high-order phase errors by using a PGA algorithm to obtain K-1 high-order phase gradients;

[0138] According to the linear phase error, the continuous K-1 linear phase gradients are obtained by least square method:

[0139] l=[l1,l2,L,l K-1 ] 1×(K-1) ;

[0140] The K-1 high-order phase gradients and the K-1 linear phase gradients are added to obtain a combined full-aperture image phase error gradient, and the full-aperture image phase error gradient is integrated to obtain a full-aperture image spectrum error.

[0141] The specific implementation includes the following steps:

[0142] Step 1: Data preparation and preprocessing

[0143] First, the phase error information corresponding to each sub-aperture extracted according to the above steps includes K-1 high-order phase errors (excluding the reference sub-aperture to simplify the calculation) and the global linear phase error. These phase error information are usually stored in the form of arrays or matrices for subsequent processing.

[0144] Step 2: PGA algorithm processing of high-order phase error

[0145] PGA algorithm application: For each high-order phase error (set as φ i (n) , where i represents the sub-aperture number and n represents the order of the phase error, such as 3 represents the cubic phase error), the PGA algorithm is used to estimate the corresponding error gradient. The PGA algorithm compares the phase differences between adjacent sub-aperture images and iteratively calculates the gradient of the phase error based on these differences, that is, The specific iterative process depends on the standard implementation of the PGA algorithm, including but not limited to constructing a cost function, selecting an optimization algorithm, etc.

[0146] Obtaining high-order phase gradients: After multiple rounds of iterations, the PGA algorithm converges and outputs the gradient value corresponding to each high-order phase error, recorded as A set of K-1 high-order phase gradients is formed.

[0147] Step 3: Least Squares Processing of Linear Phase Error

[0148] Establishing a model: Based on the linear relationship between the linear phase error and the distance (or time), it is assumed that the linear phase error can be expressed in the form of ax+b, where a is the slope, b is the intercept, and x is the distance or time variable related to a certain reference point.

[0149] Apply the least squares method: Use the known linear phase error data points (usually the linear phase error values ​​corresponding to each of the K sub-apertures) to solve the optimal slope a and intercept b through the least squares method. After obtaining these two parameters, the continuous linear phase gradient, i.e. a, can be calculated over the entire aperture range (because the linear phase gradient is the slope).

[0150] Step 4: Combine phase error gradient and integral

[0151] Combined full-aperture image phase error gradient: Add the high-order phase gradient obtained in step 2 to the linear phase gradient obtained in step 3 (regarded as a constant gradient over the entire aperture) to obtain the comprehensive gradient of the full-aperture image phase error.

[0152] Integrate to obtain the full-aperture image spectrum error: Integrate the integrated phase error gradient to obtain the phase error distribution of the full-aperture image in the spectrum domain (i.e., the full-aperture image spectrum error). This step is completed by numerically integrating the gradient function, and numerical integration methods such as the trapezoidal integration method and the Simpson integration method can be used.

[0153] Through the above steps, we have demonstrated how to systematically combine the PGA algorithm with the least squares method to accurately estimate and correct high-order and linear phase errors in SAR images, thereby improving the focus quality and resolution of the image.

[0154] S106: Obtain K-1 compensated sub-images through error compensation according to the full-aperture image spectrum error, and coherently fuse the K-1 compensated sub-images to obtain M sub-images.

[0155] Further, obtaining K-1 compensated sub-images by error compensation according to the full-aperture image spectrum error, and coherently fusing the K-1 compensated sub-images to obtain M sub-images includes:

[0156] The full-aperture image spectrum error is equally divided according to the spectrum width and compensated into the corresponding K-1 sub-images to obtain K-1 compensated sub-images;

[0157] The K-1 compensated sub-images are coherently fused to obtain M sub-images.

[0158] The specific implementation includes the following steps:

[0159] Step 1: Compensation of full-aperture image spectrum error

[0160] Error division and compensation: First, the previously calculated full-aperture image spectrum error (which includes the combined effects of high-order and linear phase errors) is divided equally according to the spectrum width. This means that if the full-aperture spectrum is divided into multiple equal-width intervals, the spectrum error in each interval will be considered to be uniformly distributed within the interval.

[0161] Sub-image compensation: Next, the spectral error of each equally divided image is compensated into the corresponding K-1 sub-images. This step usually involves converting the spectral error from the frequency domain to the spatial domain (if it is not already in the spatial domain), and then performing phase correction at the corresponding spatial position for each sub-image. Phase correction can be achieved by directly adjusting the phase value of each pixel in the complex domain of the image to ensure that it cancels out the spectral error.

[0162] Obtain compensated sub-images: After the above compensation process, we obtain K-1 sub-images that have been corrected for phase errors. These images are more accurate in phase, so in theory they should have higher focus quality and less blur.

[0163] Step 2: Coherent fusion to obtain M sub-images

[0164] Select fusion strategy: Before fusing K-1 compensated sub-images into M higher quality sub-images, a suitable fusion strategy needs to be determined. Coherent fusion is a fusion method based on inter-image coherence (i.e., a measure of similarity or consistency between images), which aims to retain useful information in each sub-image while reducing redundancy and noise.

[0165] Perform coherent fusion: According to the selected fusion strategy, the K-1 compensated sub-images are coherently fused. This step may involve multiple sub-steps, such as image registration (ensuring that all sub-images are spatially aligned), coherence calculation (assessing the similarity between sub-images), weight assignment (assigning fusion weights to each sub-image based on coherence), and finally image fusion (merging the weighted sub-images into M new sub-images).

[0166] Optimization and Adjustment: During the fusion process, optimization and adjustment may also be required to ensure the quality of the final result. For example, the image quality can be further improved by adjusting the fusion weights, applying image enhancement techniques, or performing post-processing steps such as denoising, sharpening, etc.

[0167] Obtaining M sub-images: After the above coherent fusion and optimization process, we obtain M fused sub-images. These images maintain the original sub-image information, eliminate phase errors and fuse redundant information, achieving higher resolution, clearer details and less noise.

[0168] In summary, this step describes in detail how to perform phase compensation on sub-images based on the full-aperture image spectrum error and obtain M sub-images of higher quality through the coherent fusion method. This process not only improves the focus quality and clarity of the image, but also provides a more reliable data basis for subsequent image processing and analysis.

[0169] S107: Based on the M sub-images, determine whether the number of the M sub-images is greater than 2. If so, return to execute the steps including and after step S104 to perform phase error estimation and error compensation processing on the M sub-images. If not, execute step S108.

[0170] The specific implementation process has been introduced before and will not be repeated here.

[0171] S108: Determine whether the number of the M sub-images is equal to 2, if so, execute S109, if not, execute S110.

[0172] S109: directly performing coherent fusion processing on the M sub-images after upsampling to obtain a final radar image.

[0173] Among them, the coherent fusion process has been introduced before and will not be repeated here;

[0174] S110: directly obtain the final radar image.

[0175] Among them, M is a positive integer in the specific implementation. After completing the steps of phase error estimation, error compensation processing and coherent fusion, the number of M sub-images gradually decreases until they are fused into the final image, which is the final radar image.

[0176] Those skilled in the art will appreciate that all or some of the steps in the methods disclosed above, and the functional modules / units in the systems and devices may be implemented as software, firmware, hardware, or a suitable combination thereof.

[0177] Those skilled in the art will readily appreciate other embodiments of the present application after considering the specification and practicing the invention disclosed herein. The present application is intended to cover any modification, use or adaptation of the present application, which follows the general principles of the present application and includes common knowledge or customary techniques in the art that are not disclosed in the present application. The specification and examples are intended to be exemplary only, and the true scope and spirit of the present application are indicated by the following claims.

[0178] It should be understood that the present application is not limited to the precise structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present application is limited only by the appended claims.

Claims

1. A synthetic aperture radar spectrum error estimation and compensation method, characterized in that: The method comprises the following steps: Step 1: obtaining an original callback signal and a reference signal, and obtaining a coarse compensation signal by conjugate multiplication, removing a residual video phase and coarse compensation processing of an inertial navigation system according to the original callback signal and the reference signal; Step 2: According to the coarse compensation signal, a coarse focusing sub-image of N sub-aperture coarse compensation signals is obtained by sub-aperture division, BP imaging and spectrum compression processing; Step 3: performing upsampling, spectrum decompression and coherent fusion processing on each of the coarsely focused sub-images to obtain K-1 sub-images; Step 4: According to the K-1 sub-images, obtain K-1 high-order phase errors and linear phase errors through phase error estimation; Step 5: According to the K-1 high-order phase errors and the linear phase errors, K-1 high-order phase gradients and K-1 linear phase gradients are obtained by PGA algorithm and least square method, and according to the K-1 high-order phase gradients and the K-1 linear phase gradients, a full-aperture image spectrum error is obtained; Step 6: According to the full-aperture image spectrum error, obtain K-1 compensated sub-images through error compensation, and coherently fuse the K-1 compensated sub-images to obtain M sub-images; Step 7: According to the M sub-images, phase error estimation and error compensation processing are performed on the M sub-images, or coherent fusion processing is performed to obtain a final radar image.

2. The method according to claim 1, characterized in that The step 1 comprises: Obtaining a conjugate multiplied signal by conjugate multiplying the original callback signal and the reference signal; After performing Fourier transformation on the conjugate multiplied signal in terms of distance, a residual video phase is removed to obtain a residual video phase-removed signal; The signal after removing the residual video phase is subjected to Fourier transformation to obtain pulse pressure data, and the pulse pressure data is coarsely compensated using an inertial navigation system to obtain a coarse compensation signal.

3. The method according to claim 1, characterized in that The step 2 comprises: Dividing the coarse compensation signal into N sub-aperture coarse compensation signals, and establishing a rectangular coordinate system imaging coarse grid on a ground plane based on the sub-aperture coarse compensation signals; The BP imaging and spectrum compression processing of the sub-aperture coarse compensation signal are completed coarsely grid by coarse grid to obtain coarse focusing sub-images of N sub-aperture coarse compensation signals.

4. The method according to claim 1, characterized in that: The step 3 comprises: Constructing an overlapping sub-image architecture, and performing upsampling and spectrum decompression processing on each of the coarsely focused sub-images at a first layer of the overlapping sub-image architecture to obtain an upsampled coarsely focused sub-image; Then, adjacent up-sampled coarsely focused sub-images are coherently fused to obtain K-1 sub-images.

5. The method according to claim 1, characterized in that The step 4 comprises: Performing spectrum compression and upsampling on the K-1 sub-images, and obtaining the K-1 high-order phase errors by high-order phase error estimation; The linear phase error is obtained by linear phase error estimation according to the K-1 sub-images.

6. The method according to claim 5, characterized in that The performing spectrum compression and upsampling on the K-1 sub-images and obtaining the K-1 high-order phase errors by high-order phase error estimation includes: The K-1 sub-images are subjected to spectrum compression and up-sampling, and the high-order phase errors of the K-1 sub-images are estimated using a phase gradient autofocusing algorithm to obtain the K-1 high-order phase errors.

7. The method according to claim 5, characterized in that The step of estimating the linear phase error according to the K-1 sub-images by linear phase error includes: Cross-correlation is performed on the adjacent K-1 sub-images to obtain the offset distance of the same-name point targets between adjacent sub-images as △x img ; According to the offset distance △x of the same-name point targets between adjacent sub-images img , the linear phase error is calculated by the following formula: Among them, △F x ′ is the azimuth frequency of the sub-image after spectrum compression.

8. The method according to claim 1, characterized in that The step 5 comprises: Estimate the error gradients of the K-1 high-order phase errors by using a PGA algorithm to obtain K-1 high-order phase gradients; According to the linear phase error, the continuous K-1 linear phase gradients are obtained by least square method: l＝[l1,l2,L,l K-1 ] 1×(K-1) ； The K-1 high-order phase gradients and the K-1 linear phase gradients are added to obtain a combined full-aperture image phase error gradient, and the full-aperture image phase error gradient is integrated to obtain a full-aperture image spectrum error.

9. The method according to claim 1, characterized in that: The step 6 comprises: The full-aperture image spectrum error is equally divided according to the spectrum width and compensated into the corresponding K-1 sub-images to obtain K-1 compensated sub-images; The K-1 compensated sub-images are coherently fused to obtain M sub-images.

10. The method according to claim 1, characterized in that The step 7 comprises: Determining the number of the M sub-images; If the number of M sub-images is greater than 2, then execute step 4 to step 7; If the number of the M sub-images is equal to 2, the M sub-images are directly subjected to coherent fusion processing after being upsampled to obtain a final radar image; If the number of M sub-images is less than 2, the final radar image is directly obtained.

Images