Satellite-borne multi-frequency band synthetic aperture radar signal processing optimization method and system

Abstract

The application provides a kind of spaceborne multi-band synthetic aperture radar signal processing optimization method and system, it is related to signal processing technical field, including obtaining the original echo data of spaceborne multi-band synthetic aperture radar, carries out Doppler center frequency estimation, carries out two-dimensional matrix multiplication operation, obtains compensation echo data, based on the distance direction matching filter algorithm of sub-band decomposition, obtains distance direction position information and scattering intensity information;Determine spatial distribution density, carry out adaptive block processing, in each block, carry out phase error modeling, generate continuous phase correction function;Carry out wavelet domain convolution operation, obtain correction echo data, obtain multi-band two-dimensional image, extract multi-dimensional feature vector through deep neural network, and carry out weight learning, carry out nonlinear weighted fusion, through adaptive histogram equalization enhancement, determine synthetic aperture radar imaging result.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of signal processing, and in particular to a spaceborne multi-band synthetic aperture radar signal processing optimization method and system. BACKGROUND

[0002] The spaceborne multi-band synthetic aperture radar is a kind of efficient remote sensing imaging technology, which is widely used in ground monitoring, resource exploration and disaster assessment fields. By transmitting microwave signals and receiving reflected echoes, combined with a moving platform (such as a satellite) to synthesize a virtual long aperture, the imaging resolution is improved. The multi-band synthetic aperture radar system can simultaneously acquire data at different frequency bands, and utilize the different characteristics of each frequency band to provide more accurate and reliable imaging results in complex terrain and variable environment.

[0003] Although the synthetic aperture radar technology has made significant progress in imaging quality, the existing signal processing methods still face various challenges. First, the phase error caused by Doppler frequency variation will affect the image quality, especially in the case of target uniform acceleration or variable acceleration motion, the modeling and compensation of this error is very complex. Second, the current phase compensation technology is often targeted at a single frequency band or specific conditions, lacking comprehensive processing capability for multi-band data. In addition, the existing feature extraction and data fusion methods still need to be improved in terms of computational efficiency and image quality improvement.

[0004] In summary, there is an urgent need for an adaptive piecewise polynomial fitting algorithm to achieve accurate Doppler center frequency estimation, combined with spline interpolation-based phase error modeling, complex optimization objective function of image entropy and contrast, and deep learning method for adaptive learning of multi-dimensional features, which significantly improves the clarity, accuracy and processing efficiency of imaging. The application can solve the problems in the prior art. SUMMARY

[0005] The embodiment of the application provides a spaceborne multi-band synthetic aperture radar signal processing optimization method and system, which can solve the problems in the prior art.

[0006] The first aspect of the embodiment of the application is,

[0007] A spaceborne multi-band synthetic aperture radar signal processing optimization method is provided, which comprises:

[0008] The original echo data of a spaceborne multi-frequency synthetic aperture radar is acquired, a Doppler center frequency of the original echo data is estimated by using an adaptive segmented polynomial fitting algorithm, a frequency domain adaptive compensation factor matrix is constructed, a two-dimensional matrix multiplication operation is performed between the frequency domain adaptive compensation factor matrix and the original echo data, compensated echo data is obtained, a range compression is performed on the compensated echo data based on a sub-band decomposition range matching filtering algorithm, a range image is generated, and range position information and scattering intensity information of a target scattering point are calculated based on the range image by using a maximum likelihood estimation method;

[0009] Based on the range position information and the scattering intensity information of the target scattering point, a spatial distribution density of the target scattering point is determined, the compensated echo data is adaptively processed based on the spatial distribution density, a block size is dynamically adjusted, a generalized polynomial function containing a quadratic phase error term, a cubic phase error term and a high-order cross term is used to model a phase error of the compensated echo data in each block, a composite optimization objective function fusing an image entropy and a contrast is constructed, a conjugate gradient algorithm is used to quickly iterate the composite optimization objective function, a phase error estimation value is solved, and a continuous phase correction function is generated by using a piecewise spline interpolation based on the phase error estimation value.

[0010] The phase correction function and the compensated echo data are subjected to a wavelet domain convolution operation to obtain corrected echo data, the corrected echo data is processed based on a variable scale kernel function azimuth compression algorithm to obtain a multi-frequency two-dimensional image, local correlation coefficients, edge intensities and texture features are extracted from each frequency band in the multi-frequency two-dimensional image by using a deep neural network to construct a multi-dimensional feature vector, weight learning is performed on the multi-dimensional feature vector, adaptive weight coefficients are learned based on the adaptive weight coefficients, nonlinear weighted fusion is performed on the two-dimensional images of different frequency bands, adaptive histogram equalization enhancement is performed to determine a synthetic aperture radar imaging result.

[0011] In an alternative embodiment,

[0012] The adaptive segmented polynomial fitting algorithm is used to estimate the Doppler center frequency of the original echo data, which includes:

[0013] The original echo data is segmented and processed, the original echo data is divided into a plurality of data segments according to a preset window length, adjacent data segments have an overlap rate of 50%, and a to-be-processed data segment is obtained;

[0014] The to-be-processed data segment is subjected to residual error evaluation, normalized residual error values of different order polynomial fitting are calculated, a ratio of the normalized residual error value to a normalized residual error value of a previous order polynomial fitting is compared with a preset threshold value, and when the ratio is less than the preset threshold value, an optimal polynomial order is determined;

[0015] A Vandermonde matrix based on an optimal polynomial order is constructed, elements of the Vandermonde matrix are composed of different order power terms of a time variable, a Hanning weighting matrix is constructed, diagonal elements of the Hanning weighting matrix are calculated by a cosine function, polynomial coefficients determined by the Vandermonde matrix and the Hanning weighting matrix are solved by a least square method, and a fitting polynomial function is constructed according to the polynomial coefficients;

[0016] A data signal-to-noise ratio of a to-be-processed sub-data segment is calculated, a preset window length is adaptively adjusted according to the data signal-to-noise ratio, when the data signal-to-noise ratio is greater than a first preset data signal-to-noise ratio threshold, the preset window length is reduced to one half of the original, when the data signal-to-noise ratio is less than a second preset data signal-to-noise ratio threshold, the preset window length is increased to twice the original, data fitting is performed again according to the adjusted preset window length and the fitting polynomial function, and an optimized fitting polynomial function is obtained;

[0017] An error function between the to-be-processed sub-data segment and the optimized fitting polynomial function is calculated, a mean and a variance of the error function are obtained, a compensation term based on a Gaussian function is constructed, the compensation term is added to an initial frequency estimation value calculated according to the fitting polynomial function, and a final Doppler center frequency estimation result is obtained.

[0018] In an alternative embodiment,

[0019] A sub-band decomposition-based range direction matching filter algorithm is used to compress the compensation echo data in the range direction, generate a range image, and calculate the range direction position information and scattering intensity information of a target scattering point by using a maximum likelihood estimation method according to the range image, including:

[0020] The spectrum of the compensation echo data is divided into a plurality of frequency sub-bands, adjacent frequency sub-bands have an overlap region, the bandwidth of the overlap region is 20% of the bandwidth of a sub-band, and a Hamming window function is used to weight the spectrum of each frequency sub-band;

[0021] A reference signal is generated based on the transmission signal parameters, the spectrum of the reference signal contains the influence of the antenna pattern and the propagation attenuation, and the reference signal is multiplied with the signals of each frequency sub-band in the frequency domain to obtain a filtering result;

[0022] The phase difference between the frequency sub-bands is calculated, a linear phase compensation method is used to correct the phase difference, the signal signal-to-noise ratio of the signals corresponding to each frequency sub-band is calculated, and the signal signal-to-noise ratio is used as a weight coefficient to weight and sum the filtering results to obtain a synthesized range image;

[0023] performing energy normalization on the synthetic range image, dividing range cells according to system resolution, establishing a mapping relationship between amplitude and range, searching for a peak position of the synthetic range image, and taking the peak position as a coarse positioning result of the scattering point;

[0024] performing ten-fold subdivision on the range cells in a neighborhood of the coarse positioning result of the scattering point by using a parabolic interpolation method, performing peak searching on the subdivided range cells, and obtaining an accurate position of the scattering point;

[0025] constructing a likelihood function based on a Gaussian noise model, wherein the likelihood function contains a range direction position parameter and a scattering intensity parameter, and performing iterative optimization and solving on the likelihood function, and taking parameters corresponding to a maximum value of the likelihood function as an initial estimation result;

[0026] calculating a Cramer-Rao bound of the initial estimation result, judging whether the initial estimation result reaches a pre-determined theoretical precision limit, if not, constructing a new search space in a neighborhood of the initial estimation result, and continuing parameter iterative optimization, and finally obtaining position information and scattering intensity information of a target scattering point.

[0027] In an alternative embodiment,

[0028] In each sub-block, a generalized polynomial function containing a quadratic phase error term, a cubic phase error term and a high-order cross term is used to model phase error of the compensated echo data, and a composite optimization objective function fusing image entropy and contrast is constructed, including:

[0029] In each sub-block, a quadratic term taking range direction position information of a target scattering point as a variable is established as a quadratic phase error term to compensate for phase error caused by target uniform acceleration motion, a cubic term taking range direction position information of the target scattering point as a variable is established as a cubic phase error term to compensate for phase error caused by target variable acceleration motion, and a product term of range direction position information and azimuth direction position information of the target scattering point is established as a cross phase error term to determine a corresponding relationship of phase error caused by range direction and azimuth direction coupling; and the quadratic phase error term, the cubic phase error term and the high-order cross phase error term are combined to construct a generalized polynomial function;

[0030] The generalized polynomial function is applied to the compensation echo data to perform phase compensation to obtain a compensation image; an intensity distribution of the compensation image is calculated, the intensity distribution is normalized to obtain a probability distribution, and an image entropy is calculated based on the probability distribution; an intensity maximum value and an intensity minimum value of the compensation image are extracted, and an image contrast is calculated; a difference between the image contrast and 1 is calculated to obtain an image contrast loss; a first weight coefficient is determined based on a spatial distribution density of the target region, 1 is subtracted from the first weight coefficient to obtain a second weight coefficient; the image entropy is multiplied by the first weight coefficient to obtain a first weighted term, and the image contrast loss is multiplied by the second weight coefficient to obtain a second weighted term; and the first weighted term and the second weighted term are added to construct a compound optimization objective function.

[0031] In an optional embodiment,

[0032] The compound optimization objective function is iterated quickly by a conjugate gradient algorithm to obtain a phase error estimation value, and a continuous phase correction function is generated by piecewise spline interpolation according to the phase error estimation value, including:

[0033] A numerical differentiation method is used to calculate a gradient of the compound optimization objective function with respect to each coefficient in the generalized polynomial function, a negative direction of the gradient is taken as an initial search direction, an optimal step size is determined by one-dimensional search in the initial search direction, each coefficient in the generalized polynomial function is updated based on the optimal step size to obtain updated coefficients, an updated gradient of the compound optimization objective function is calculated based on the updated coefficients, a conjugate coefficient is calculated by calculating an inner product ratio of the updated gradient and a previous gradient, and a new search direction is obtained by combining a negative direction of the updated gradient and a conjugate of the initial search direction; the iteration is repeated until a difference between two adjacent iterations of the compound optimization objective function is less than a preset convergence threshold to obtain a convergence coefficient, and the convergence coefficient is substituted into the generalized polynomial function to obtain the phase error estimation value.

[0034] The center positions of each block are determined as reference interpolation nodes, and supplementary interpolation nodes are additionally arranged in the overlapping regions of adjacent blocks; a cubic spline basis function is selected as an interpolation basis function, and a spline coefficient equation set is constructed based on the reference interpolation nodes and the supplementary interpolation nodes; a boundary condition of zero second-order derivative at the end points is set for the spline coefficient equation set, the spline coefficient equation set is solved to obtain spline coefficients, the spline coefficients are combined with the cubic spline basis function, and the phase error estimation value is subjected to piecewise interpolation to generate a continuous phase correction function.

[0035] In an optional embodiment,

[0036] The phase correction function is subjected to wavelet domain convolution operation with the compensation echo data to obtain corrected echo data, the corrected echo data is processed based on an azimuth direction compression algorithm of a variable scale kernel function to obtain multi-band two-dimensional images, based on each frequency band in the multi-band two-dimensional images, local correlation coefficients, edge intensities and texture features are extracted through a deep neural network, a multi-dimensional feature vector is constructed, weight learning is performed on the multi-dimensional feature vector, nonlinear weighted fusion is performed on the two-dimensional images of different frequency bands based on adaptive weight coefficients learned, and adaptive histogram equalization enhancement is performed to determine a synthetic aperture radar imaging result, including:

[0037] The phase correction function is subjected to wavelet domain convolution operation with the compensation echo data to obtain corrected echo data, the corrected echo data is processed based on an azimuth direction compression algorithm of a variable scale kernel function to obtain multi-band two-dimensional images, based on each frequency band in the multi-band two-dimensional images, local correlation coefficients, edge intensities and texture features are extracted through a deep neural network, a multi-dimensional feature vector is constructed, weight learning is performed on the multi-dimensional feature vector, nonlinear weighted fusion is performed on the two-dimensional images of different frequency bands based on adaptive weight coefficients learned, and adaptive histogram equalization enhancement is performed to determine a synthetic aperture radar imaging result, including:

[0038] The phase correction function is subjected to wavelet domain convolution operation with the compensation echo data to obtain corrected echo data, the corrected echo data is processed based on an azimuth direction compression algorithm of a variable scale kernel function to obtain multi-band two-dimensional images, based on each frequency band in the multi-band two-dimensional images, local correlation coefficients, edge intensities and texture features are extracted through a deep neural network, a multi-dimensional feature vector is constructed, weight learning is performed on the multi-dimensional feature vector, nonlinear weighted fusion is performed on the two-dimensional images of different frequency bands based on adaptive weight coefficients learned, and adaptive histogram equalization enhancement is performed to determine a synthetic aperture radar imaging result, including:

[0039] The phase correction function is subjected to wavelet domain convolution operation with the compensation echo data to obtain corrected echo data, the corrected echo data is processed based on an azimuth direction compression algorithm of a variable scale kernel function to obtain multi-band two-dimensional images, based on each frequency band in the multi-band two-dimensional images, local correlation coefficients, edge intensities and texture features are extracted through a deep neural network, a multi-dimensional feature vector is constructed, weight learning is performed on the multi-dimensional feature vector, nonlinear weighted fusion is performed on the two-dimensional images of different frequency bands based on adaptive weight coefficients learned, and adaptive histogram equalization enhancement is performed to determine a synthetic aperture radar imaging result, including:

[0040] The phase correction function is subjected to wavelet domain convolution operation with the compensation echo data to obtain corrected echo data, the corrected echo data is processed based on an azimuth direction compression algorithm of a variable scale kernel function to obtain multi-band two-dimensional images, based on each frequency band in the multi-band two-dimensional images, local correlation coefficients, edge intensities and texture features are extracted through a deep neural network, a multi-dimensional feature vector is constructed, weight learning is performed on the multi-dimensional feature vector, nonlinear weighted fusion is performed on the two-dimensional images of different frequency bands based on adaptive weight coefficients learned, and adaptive histogram equalization enhancement is performed to determine a synthetic aperture radar imaging result, including:

[0041] The phase correction function is subjected to wavelet domain convolution operation with the compensation echo data to obtain corrected echo data, the corrected echo data is processed based on an azimuth direction compression algorithm of a variable scale kernel function to obtain multi-band two-dimensional images, based on each frequency band in the multi-band two-dimensional images, local correlation coefficients, edge intensities and texture features are extracted through a deep neural network, a multi-dimensional feature vector is constructed, weight learning is performed on the multi-dimensional feature vector, nonlinear weighted fusion is performed on the two-dimensional images of different frequency bands based on adaptive weight coefficients learned, and adaptive histogram equalization enhancement is performed to determine a synthetic aperture radar imaging result, including:

[0042] The second aspect of the embodiment of the application,

[0043] The satellite-borne multi-band synthetic aperture radar signal processing optimization system comprises:

[0044] The first unit is configured to acquire original echo data of a spaceborne multi-band synthetic aperture radar, estimate a Doppler center frequency of the original echo data by using an adaptive segmented polynomial fitting algorithm, construct a frequency domain adaptive compensation factor matrix, perform two-dimensional matrix multiplication operation on the frequency domain adaptive compensation factor matrix and the original echo data, obtain compensated echo data, perform range compression on the compensated echo data based on a sub-band decomposition range matching filtering algorithm, generate a range image, and calculate range position information and scattering intensity information of a target scattering point based on the range image by using a maximum likelihood estimation method.

[0045] The second unit is configured to determine a spatial distribution density of the target scattering point based on the range position information and the scattering intensity information of the target scattering point, perform adaptive block processing on the compensated echo data based on the spatial distribution density, dynamically adjust a block size, model the compensated echo data by using a generalized polynomial function containing a quadratic phase error term, a cubic phase error term and a high-order cross term in each block, construct a composite optimization objective function fusing an image entropy and a contrast, quickly iterate the composite optimization objective function by using a conjugate gradient algorithm, and obtain a phase error estimation value.

[0046] The third unit is configured to perform wavelet domain convolution operation on the phase correction function and the compensated echo data to obtain corrected echo data, perform processing on the corrected echo data based on a variable scale kernel function azimuth compression algorithm to obtain multi-band two-dimensional images, extract local correlation coefficients, edge intensities and texture features by using a deep neural network based on each frequency band in the multi-band two-dimensional images, construct a multi-dimensional feature vector, perform weight learning on the multi-dimensional feature vector, perform nonlinear weighted fusion on the two-dimensional images of different frequency bands based on adaptive weight coefficients learned, and determine a synthetic aperture radar imaging result by using adaptive histogram equalization enhancement.

[0047] A third aspect of the embodiment of the application,

[0048] An electronic device is provided, comprising:

[0049] A processor;

[0050] A memory for storing processor-executable instructions;

[0051] The processor is configured to invoke the instructions stored in the memory to perform the method described above.

[0052] A fourth aspect of the embodiment of the application,

[0053] A computer readable storage medium is provided, which stores computer program instructions, and the computer program instructions are executed by a processor to implement the method.

[0054] In the embodiment of the application, the error is reduced by an adaptive piecewise polynomial fitting algorithm, and the accuracy of frequency estimation is improved; a frequency domain adaptive compensation factor matrix is constructed to realize stability in the signal frequency domain; an effective range compression is performed by using a range matching filter algorithm based on sub-band decomposition; a maximum likelihood estimation method can accurately calculate the range position information and scattering intensity information of the target scattering point; the data block size is dynamically adjusted according to the spatial distribution density of the target scattering point, and the phase error modeling efficiency is improved; a generalized polynomial function containing a polynomial term is used to improve the accuracy of phase error estimation; a composite optimization objective function is constructed by combining image entropy and contrast to improve imaging clarity and contrast; a conjugate gradient algorithm and piecewise spline interpolation are used to generate a continuous phase correction function; a range compression algorithm based on a variable scale kernel function ensures the retention of detail and feature information of multi-frequency band data; a deep neural network is used to extract features to realize efficient fusion of multi-frequency band images; and nonlinear weighted fusion and adaptive histogram equalization technology improve image contrast and detail highlighting. BRIEF DESCRIPTION OF DRAWINGS

[0055] Figure 1 A flowchart of a spaceborne multi-frequency band synthetic aperture radar signal processing optimization method is shown in FIG.

[0056] Figure 2 A structural diagram of a spaceborne multi-frequency band synthetic aperture radar signal processing optimization system is shown in FIG. DETAILED DESCRIPTION

[0057] To make the objectives, technical solutions, and advantages of the embodiments of the application clearer, the technical solutions in the embodiments of the application will be described below in connection with the drawings of the embodiments of the application. Obviously, the described embodiments are only some of the embodiments of the application, rather than all the embodiments of the application. Based on the embodiments in the application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the protection scope of the application.

[0058] The technical solutions of the application will be 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 can not be described in some embodiments.

[0059] Figure 1 A flowchart of a spaceborne multi-frequency band synthetic aperture radar signal processing optimization method is shown in FIG. Figure 1 As shown in FIG.

[0060] S101. Obtain the original echo data of the spaceborne multi-band synthetic aperture radar, estimate the Doppler center frequency of the original echo data by using the adaptive piecewise polynomial fitting algorithm, construct a frequency domain adaptive compensation factor matrix, perform two-dimensional matrix multiplication operation on the frequency domain adaptive compensation factor matrix and the original echo data to obtain compensated echo data, perform range compression on the compensated echo data based on the sub-band decomposition range matching filtering algorithm to generate a range image, and calculate the range position information and scattering intensity information of the target scattering point based on the range image by using the maximum likelihood estimation method;

[0061] S102. Determine the spatial distribution density of the target scattering point based on the range position information and scattering intensity information of the target scattering point, perform adaptive block processing on the compensated echo data based on the spatial distribution density, dynamically adjust the block size, and in each block, model the compensated echo data by using a generalized polynomial function containing a quadratic phase error term, a cubic phase error term and a high-order cross term, construct a composite optimization objective function fusing image entropy and contrast, and solve the phase error estimation value by quickly iterating the composite optimization objective function through the conjugate gradient algorithm; and generate a continuous phase correction function through piecewise spline interpolation according to the phase error estimation value.

[0062] S103. Perform wavelet domain convolution operation on the phase correction function and the compensated echo data to obtain corrected echo data, process the corrected echo data based on the variable scale kernel function azimuth compression algorithm to obtain a multi-band two-dimensional image, extract local correlation coefficients, edge intensity and texture features from each frequency band in the multi-band two-dimensional image by using a deep neural network, construct a multi-dimensional feature vector, and perform weight learning on the multi-dimensional feature vector; based on the adaptive weight coefficients learned, perform nonlinear weighted fusion on the two-dimensional images of different frequency bands, and determine the synthetic aperture radar imaging result through adaptive histogram equalization enhancement.

[0063] First, the original echo data of the spaceborne multi-band synthetic aperture radar is obtained. These data usually contain echo information of multiple frequency bands, such as X, C, L bands, etc. For each frequency band, the original echo data is usually a two-dimensional complex matrix, where one dimension represents range sampling and the other dimension represents azimuth sampling.

[0064] Next, the adaptive piecewise polynomial fitting algorithm is used to estimate the Doppler center frequency of the original echo data. In the implementation, the echo data can be divided into multiple sub-blocks in the azimuth direction, and the Fourier transform is performed on each sub-block to find the frequency corresponding to the spectral peak as the Doppler center frequency estimate of the sub-block. Then, the polynomial function is used to fit these estimates to obtain the Doppler center frequency variation curve of the entire data in the azimuth direction. Taking the X-band as an example, the estimated Doppler center frequency result can vary from -320 Hz to 320 Hz.

[0065] Based on the estimated Doppler center frequency, a frequency domain adaptive compensation factor matrix is constructed. Each element of the matrix corresponds to a sampling point of the original echo data. The two-dimensional matrix multiplication operation is performed between the compensation factor matrix and the original echo data to obtain the compensated echo data. This step can correct the phase error caused by the Doppler frequency variation.

[0066] The distance compression is performed on the compensated echo data, and the sub-band decomposition based distance matching filtering algorithm is used. First, the echo data is Fourier transformed in the distance direction, then the spectrum is divided into several sub-bands, and each sub-band is matched filtered respectively, and finally the results of each sub-band are synthesized to obtain the range image. This method can effectively suppress the range ambiguity. Taking the X-band 150 MHz bandwidth signal as an example, the spectrum can be equally divided into 10 sub-bands of 15 MHz for processing.

[0067] According to the obtained range image, the maximum likelihood estimation method is used to calculate the distance position information and scattering intensity information of the target scattering points. Specifically, the scattering point position can be determined by finding the local peak value in the range image, and the peak amplitude is the scattering intensity. For example, it can be obtained that a target has 3 main scattering points with distances of 1000 m, 1002 m and 1005 m, and relative scattering intensities of 1, 0.8 and 0.6.

[0068] Based on the spatial distribution of the target scattering points, the adaptive block processing is performed on the compensated echo data. Smaller block size is used in the area where the target scattering points are dense, and larger block size is used in the sparse area. For example, the data can be divided into sub-blocks of different sizes from 64x64 to 256x256. In each block, a generalized polynomial function is used to model the phase error, which includes a quadratic phase error term, a cubic phase error term and a high-order cross term, which can accurately describe the complex phase error.

[0069] A composite optimization objective function is constructed by combining image entropy and contrast. Image entropy reflects the focusing degree of the image, and contrast reflects the sharpness of the image. Combining the two can get a more comprehensive image quality evaluation. The conjugate gradient algorithm is used to quickly iterate and optimize the objective function, and the phase error estimate value is obtained. Usually, 10-20 iterations can converge.

[0070] According to the obtained discrete phase error estimate value, a continuous phase correction function is generated by cubic spline interpolation. Convolution operation of the function with the compensated echo data in the wavelet domain can obtain the corrected echo data. Wavelet domain processing can suppress noise while preserving image details.

[0071] The azimuth compression is performed on the corrected echo data, and the algorithm based on the variable scale kernel function is used. This algorithm can adapt to the Doppler frequency variation characteristics of different regions by adjusting the scale of the kernel function, so as to obtain better focusing effect. The processing result is a two-dimensional image of multiple frequency bands.

[0072] For the obtained multi-frequency band image, a deep convolutional neural network is constructed to extract local correlation coefficients, edge strength and texture features, etc. The network contains multiple convolutional layers, pooling layers and fully connected layers, the input is a multi-frequency band image, and the output is a feature vector extracted. For example, a 256-dimensional feature vector can be extracted to represent various properties of the image.

[0073] The extracted multi-dimensional feature vector is subjected to weight learning, and algorithms such as stochastic gradient descent can be used to optimize the weight coefficients. The learned weights reflect the importance of different features in image fusion. For example, the weight of the edge feature may be larger, 0.4, and the weight of the texture feature may be smaller, 0.1.

[0074] Based on the learned weight coefficients, the images of different frequency bands are nonlinearly weighted and fused. When fusing, the pixel value of each frequency band image is multiplied by the corresponding weight, and then the results are added to obtain the fused image. Nonlinear fusion can better preserve the detail information of each frequency band image.

[0075] Finally, the fused image is subjected to adaptive histogram equalization enhancement processing. This method can effectively improve the contrast of the image and highlight the target features. When processing, the image is divided into multiple sub-blocks, histogram equalization is performed in each sub-block, then the results of each sub-block are smoothly fused by bilinear interpolation method to obtain the final enhanced image. This is the final imaging result of the spaceborne multi-frequency band synthetic aperture radar.

[0076] Through the above series of processing steps, the imaging quality of the spaceborne multi-band synthetic aperture radar can be effectively improved, and a radar image with high contrast can be obtained. The method comprehensively considers key technologies such as Doppler frequency estimation, adaptive segmentation, phase error correction, multi-band feature extraction and fusion, and can better adapt to complex imaging environments.

[0077] In the embodiment, by using the adaptive segmented polynomial fitting algorithm, accurate Doppler center frequency estimation of the original echo data can be performed, errors can be reduced, and the accuracy of frequency estimation can be improved; a frequency domain adaptive compensation factor matrix is constructed, and matrix multiplication operation is performed with the original echo data, efficient frequency compensation is realized, and the stability of the signal in the frequency domain is ensured; the distance direction matching filter algorithm based on sub-band decomposition is used to effectively compress the distance direction and generate a high-resolution range image; the maximum likelihood estimation method is used to accurately calculate the distance direction position information and scattering intensity information of the target scattering point, and the reliability of the positioning and intensity information is improved; according to the spatial distribution density of the target scattering point, the data block size is dynamically adjusted to ensure more flexible data processing, and the efficiency of phase error modeling is improved; the generalized polynomial function containing quadratic and cubic phase error terms and high-order cross terms is used to realize more accurate phase error modeling and improve the accuracy of phase error estimation; the image entropy and contrast are combined to construct a composite optimization objective function, which can comprehensively improve the definition and contrast of the image; the conjugate gradient algorithm is used for fast iterative optimization to obtain the phase error estimation value, and a continuous phase correction function is generated by using piecewise spline interpolation to realize high-precision phase correction of the compensated echo data; the azimuth direction compression algorithm based on the variable scale kernel function makes the compression and fusion of multi-band data more accurate, and ensures that the details and feature information in the two-dimensional image are retained; the local correlation coefficient, edge strength and texture features are extracted by using a deep neural network, and weight learning is performed on the multi-dimensional feature vector to automatically adapt to the differences between different frequency bands, and efficient fusion of multi-band images is realized; nonlinear weighted fusion is used, combined with adaptive weight coefficients, to ensure that the best target features are retained in the multi-band image, and the definition and resolution of the image are further improved; the adaptive histogram equalization enhancement technology can improve the image contrast, highlight the details of the key area, and make the synthetic aperture radar imaging result clearer and more interpretable.

[0078] In an alternative embodiment, the adaptive segmented polynomial fitting algorithm is used to estimate the Doppler center frequency of the original echo data, which includes:

[0079] The original echo data is segmented and processed, and the original echo data is divided into a plurality of sub-data segments according to a preset window length, adjacent sub-data segments have an overlap rate of 50%, and a to-be-processed sub-data segment is obtained;

[0080] The residual evaluation is performed on the to-be-processed sub-data segment, the normalized residual values of polynomial fitting of different orders are calculated, the ratio of the normalized residual value to the normalized residual value of the last-order polynomial fitting is compared with a preset threshold, and when the ratio is less than the preset threshold, the optimal polynomial order is determined.

[0081] A Vandermonde matrix based on the optimal polynomial order is constructed, elements of the Vandermonde matrix are composed of different order powers of time variables, a Hanning weighting matrix is constructed, diagonal elements of the Hanning weighting matrix are calculated by a cosine function, polynomial coefficients determined by the Vandermonde matrix and the Hanning weighting matrix are solved by a least square method, and a fitting polynomial function is constructed according to the polynomial coefficients.

[0082] The data signal-to-noise ratio of the to-be-processed sub-data segment is calculated, the preset window length is adaptively adjusted according to the data signal-to-noise ratio, when the data signal-to-noise ratio is greater than a first preset data signal-to-noise ratio threshold, the preset window length is reduced to one half of the original, when the data signal-to-noise ratio is less than a second preset data signal-to-noise ratio threshold, the preset window length is doubled, data fitting is performed again according to the adjusted preset window length and the fitting polynomial function, and an optimized fitting polynomial function is obtained.

[0083] An error function between the to-be-processed sub-data segment and the optimized fitting polynomial function is calculated, a mean value and a variance of the error function are obtained, a compensation term based on a Gaussian function is constructed, the compensation term is added to an initial frequency estimation value calculated according to the fitting polynomial function, and a final Doppler center frequency estimation result is obtained.

[0084] The Vandermonde matrix specifically refers to a matrix used to represent the power relationship between polynomial coefficients and input variables. In the Vandermonde matrix, the elements of each row are arranged in order of different powers of the same variable. Such a matrix is commonly used in polynomial fitting problems. By arranging the power terms of different orders into a matrix form, the coefficients of the polynomial can be solved by matrix operations. The structure of the Vandermonde matrix is very regular, and it is widely used in the fields of data fitting and signal processing, especially in solving linear equation groups and least square problems, which can help to construct the expression of the polynomial.

[0085] The Hann window matrix is a diagonal matrix used for weighting signals in signal processing, typically in window functions. The diagonal elements of the matrix are calculated from the Hann window function, which can smooth the signal boundaries and reduce the impact of spectral leakage. By using a cosine function for weighting, the Hann window function effectively reduces the discontinuity of data at the boundaries, ensuring smoother and more stable data processing. Hann window matrices are widely used in signal processing and filtering tasks, especially in scenarios requiring spectral analysis or frequency estimation, where weighting can reduce calculation errors and improve the reliability of the results.

[0086] Specifically, the method of estimating the Doppler center frequency of the original echo data by using the adaptive segmented polynomial fitting algorithm includes the following steps:

[0087] First, the original echo data is segmented. The original echo data is divided into multiple data segments according to a preset window length (e.g., 1024 sampling points), with a 50% overlap between adjacent data segments. For example, for original echo data with a length of 10000 sampling points, 18 data segments can be divided, each containing 1024 sampling points, with an overlap of 512 sampling points between adjacent data segments. This way, a sequence of data segments to be processed is obtained.

[0088] Next, the residual error of each data segment to be processed is evaluated to determine the optimal polynomial fitting order. Starting from a first-order polynomial, the order is gradually increased, and the normalized residual error values of different order polynomials are calculated. The normalized residual error value of the current order is compared with that of the previous order, and the ratio of the two is calculated. When the ratio is less than a preset threshold (e.g., 0.95), it is considered that the increase in order does not significantly improve the fitting accuracy, and the order at this time is the optimal polynomial order. For example, for a certain data segment, the normalized residual error values of 1st to 6th order polynomials are 0.8, 0.6, 0.4, 0.35, 0.33, and 0.32, respectively, and the 4th order polynomial is the optimal order.

[0089] Then, according to the determined optimal polynomial order, a Vandermonde matrix and a Hann weighting matrix are constructed. The elements of the Vandermonde matrix are composed of different order powers of the time variable, and the diagonal elements of the Hann weighting matrix are calculated from the cosine function. The least squares method is used to solve the polynomial coefficients determined by the two matrices, thereby constructing a fitting polynomial function. For example, for a 4th order polynomial fitting, a fitting function of the form a0+a1t+a2t 2 +a3t 3 +a4t 4 can be obtained.

[0090] Afterwards, the data signal-to-noise ratio of the to-be-processed sub-data segment is calculated, and the preset window length is adaptively adjusted according to the data signal-to-noise ratio. When the data signal-to-noise ratio is greater than a first preset threshold (for example, 20 dB), the preset window length is reduced to half of the original, that is, 512 sampling points. When the data signal-to-noise ratio is less than a second preset threshold (for example, 5 dB), the preset window length is increased to twice the original, that is, 2048 sampling points. According to the adjusted window length and the fitting polynomial function, data fitting is performed again to obtain an optimized fitting polynomial function.

[0091] Finally, an error function between the to-be-processed sub-data segment and the optimized fitting polynomial function is calculated to obtain the mean and variance of the error function. A compensation term is constructed based on a Gaussian function, and the compensation term is added to an initial frequency estimation value calculated according to the optimized fitting polynomial function to obtain a final Doppler center frequency estimation result. For example, the initial frequency estimation value of a certain sub-data segment is 1000 Hz, the mean of the error function is 5 Hz, and the variance is 2 Hz 2 , and the final frequency estimation result is about 1005 Hz.

[0092] Through the above steps, adaptive segmented polynomial fitting of the original echo data can be realized to obtain an accurate Doppler center frequency estimation result. The method can adaptively adjust the processing parameters according to the data characteristics, effectively improving the accuracy and robustness of the frequency estimation.

[0093] In the embodiment, the optimal polynomial order is dynamically selected through residual evaluation to ensure that the accuracy of polynomial fitting is optimal, thereby improving the estimation accuracy of the Doppler center frequency. The adaptive adjustment mechanism based on the data signal-to-noise ratio flexibly adjusts the window length of data processing, so that the best fitting effect can be achieved in high-noise or low-noise environments, further optimizing the data processing efficiency and accuracy. By constructing a compensation term based on a Gaussian function and adding it to the initial frequency estimation value, the error in data fitting is significantly reduced, and the reliability of the frequency estimation result is improved. The use of the Vandermonde matrix and the Hanning weighting matrix ensures that the data weighting processing during polynomial fitting is smoother, further improving the stability and robustness of the fitting result. The least squares method is combined with the adaptive window adjustment and compensation mechanism to dynamically optimize the polynomial fitting process, so that the data processing can remain efficient and accurate under various noise conditions.

[0094] In an optional implementation, based on a sub-band decomposition distance matching filter algorithm, the compensation echo data is compressed in the distance direction to generate a range image, and the range image is used to calculate the distance position information and scattering intensity information of the target scattering point by using a maximum likelihood estimation method.

[0095] The spectrum of the compensation echo data is divided into a plurality of frequency subbands, adjacent frequency subbands have an overlap region, the bandwidth of the overlap region is 20% of the bandwidth of the subband, and a Hamming window function is used to weight the spectrum of each frequency subband;

[0096] A reference signal is generated based on the transmission signal parameters, the spectrum of the reference signal contains the antenna pattern and the propagation attenuation influence, the reference signal is multiplied with the signal of each frequency subband in the frequency domain to obtain a filtering result;

[0097] The phase difference between the frequency subbands is calculated, a linear phase compensation method is used to correct the phase difference, the signal signal-to-noise ratio of the signal corresponding to each frequency subband is calculated, and the signal signal-to-noise ratio is used as a weight coefficient to weight and sum the filtering result to obtain a combined range image;

[0098] The combined range image is energy-normalized, distance units are divided according to the system resolution, a mapping relationship between amplitude and distance is established, and the peak position of the combined range image is searched to obtain a coarse positioning result of the scattering point;

[0099] In the neighborhood of the coarse positioning result of the scattering point, the distance units are ten times subdivided using a parabolic interpolation method, and the peak of the subdivided distance units is searched to obtain an accurate position of the scattering point;

[0100] A likelihood function based on a Gaussian noise model is constructed, the likelihood function includes a distance position parameter and a scattering intensity parameter, and the likelihood function is optimized and solved in an iterative manner, and the maximum value of the likelihood function is used as an initial estimation result;

[0101] The Cramer-Rao bound of the initial estimation result is calculated, and it is judged whether the initial estimation result reaches a predetermined theoretical precision limit, if not, a new search space is constructed in the neighborhood of the initial estimation result, and parameter iterative optimization is continued, and finally the target scattering point position information and scattering intensity information are obtained.

[0102] The Hamming window function is a specific weighting function commonly used in spectral analysis in signal processing. Its main function is to reduce spectral leakage, especially when the signal is processed in segments, it can smooth the edges of the signal, thereby reducing the distortion in the frequency domain. Compared with other window functions, the Hamming window introduces specific weighting coefficients to smooth the transition of the signal at the boundary, making the spectral leakage smaller and better highlighting the main frequency components in the signal. The weighting value of the Hamming window function is larger in the middle of the window function and gradually decreases at the edges, showing an approximate parabolic shape.

[0103] The Cramer-Rao bound is a concept in parameter estimation theory, which represents the theoretically minimum variance of parameter estimation under given noise conditions. In other words, it defines a limit that the variance of any unbiased estimation cannot be lower than. The CRB can provide a reference standard for evaluating the precision of the estimation, and by comparing the actual variance of the estimation with the CRB, the efficiency of the estimation method or the potential for improvement can be judged. In signal processing and statistics, the Cramer-Rao bound is often used to measure whether the parameter estimation result has reached the optimal theoretical precision.

[0104] The specific implementation of the range-matching filter algorithm based on sub-band decomposition is as follows:

[0105] First, the spectrum of the compensated echo data is divided into multiple frequency sub-bands. Specifically, the entire spectrum can be equally divided into 8 sub-bands, with a 20% overlap region between adjacent sub-bands. For example, for an echo signal with a bandwidth of 500MHz, it can be divided into 8 sub-bands, each with a bandwidth of 62.5MHz, and an overlap bandwidth of 12.5MHz between adjacent sub-bands. Then, a Hamming window function is used to weight the spectrum of each frequency sub-band to reduce spectral leakage.

[0106] Next, a reference signal is generated based on the transmitted signal parameters. The spectrum of the reference signal needs to include the effects of the antenna pattern and the propagation attenuation. For example, for a system operating in the X-band with a center frequency of 10GHz, the frequency domain expression of the reference signal can be generated according to the antenna gain model and the free space propagation loss model. The reference signal is multiplied by the signals of each frequency sub-band in the frequency domain to obtain the filtering result.

[0107] Then, the phase difference between the frequency sub-bands is calculated. Since the sub-band division will introduce phase discontinuity, the phase difference needs to be corrected. A linear phase compensation method can be used to calculate the average phase difference in the overlapping region of adjacent sub-bands and perform linear interpolation within the entire sub-band to achieve phase correction. At the same time, the signal-to-noise ratio of the signal corresponding to each frequency sub-band is calculated, and the signal-to-noise ratio is used as a weight coefficient to weight and sum the filtering results to obtain a composite range image.

[0108] The composite range image is energy-normalized to normalize the maximum amplitude value to 1. According to the system distance resolution, the distance units are divided, and the mapping relationship between amplitude and distance is established. For example, for a system with a distance resolution of 0.3m, the distance axis can be divided into multiple distance units with a spacing of 0.3m. The peak position of the composite range image is searched, and the peak position is taken as the coarse positioning result of the scattering point.

[0109] In the neighborhood of the coarse positioning result of the scattering point, a parabolic interpolation method is used to subdivide the distance unit by ten times. For example, if the coarse positioning result is the 100th distance unit, 100 new distance points are inserted between the 99th and 101st distance units. After subdivision, the distance unit is searched for a peak value to obtain the accurate position of the scattering point.

[0110] A likelihood function based on a Gaussian noise model is constructed, which contains the distance direction position parameter and the scattering intensity parameter. The likelihood function is optimized and solved in an iterative manner, and optimization algorithms such as gradient descent or Newton method can be used. The maximum value of the likelihood function corresponds to the initial estimation result.

[0111] The Cramer-Rao bound of the initial estimation result is calculated to determine whether the initial estimation result meets the pre-determined theoretical precision limit. For example, the distance direction position estimation accuracy can be set to no less than 1 / 10 of the system resolution, and the scattering intensity estimation accuracy can be set to no less than 3dB. If the accuracy requirement is not met, a new search space is constructed in the neighborhood of the initial estimation result, and the parameter iterative optimization is continued. The search range can be reduced to ±5% of the initial estimation result, and the step size can be reduced to 1 / 10 of the original step size. The optimization process is repeated until the accuracy requirement is met or the maximum number of iterations is reached, and finally the position information and scattering intensity information of the target scattering point are obtained.

[0112] Through the above steps, the distance direction matched filtering and maximum likelihood estimation based on sub-band decomposition can be realized, and high-precision target scattering characteristic information can be obtained. This method fully utilizes the frequency domain information, improves the anti-interference ability through sub-band processing, and further improves the parameter estimation accuracy through maximum likelihood estimation, which can be effectively applied to imaging radar systems.

[0113] In the embodiment, by using Hamming window function to weight the spectrum of each frequency sub-band, the spectrum leakage can be effectively reduced, the resolution of the sub-band frequency and the signal fidelity can be improved, and the filtering effect can be enhanced; the linear phase compensation is performed on the phase difference between the frequency sub-bands, the phase difference of the frequency sub-bands is corrected, the phase error is reduced, and the accuracy of the synthesized range image is ensured; the weighted sum of the filtering results is performed based on the signal-to-noise ratio as the weight coefficient, so that the weight of different frequency bands is adaptively adjusted in the signal processing process, and the signal-to-noise ratio and the image quality of the synthesized range image are improved; the distance unit is subdivided through parabolic interpolation, and the precise positioning of the scattering point is realized in combination with peak value search, and the spatial resolution of the radar imaging is improved; the likelihood function is constructed by using the Gaussian noise model, and the position information and the scattering intensity information of the scattering point are optimized and solved by using the iterative optimization method, so that the parameter estimation is more accurate; the Cramer-Rao bound of the initial estimation result is calculated to determine whether the estimation reaches the theoretical precision limit, and the estimation accuracy of the position information and the scattering intensity of the scattering point is further ensured. If the precision requirement is not met, the accuracy and the robustness of the final result are ensured through further iterative optimization.

[0114] In an alternative embodiment, within each patch, the phase error of the compensation echo data is modeled by a generalized polynomial function including a quadratic phase error term, a cubic phase error term and a high-order cross term, and a composite optimization objective function f integrating the entropy and the contrast of the fused image is constructed, including:

[0115] Within each patch, a quadratic term taking the distance position information of the target scattering point as a variable is established as a quadratic phase error term to compensate the phase error caused by the target uniform acceleration motion; a cubic term taking the distance position information of the target scattering point as a variable is established as a cubic phase error term to compensate the phase error caused by the target variable acceleration motion; a product term of the distance position information and the azimuth position information of the target scattering point is established as a cross phase error term to determine the corresponding relationship of the phase error caused by the distance-azimuth coupling; the quadratic phase error term, the cubic phase error term and the high-order cross phase error term are combined to construct a generalized polynomial function;

[0116] The generalized polynomial function is applied to the compensation echo data to perform phase compensation to obtain a compensation image; an intensity distribution of the compensation image is calculated, the intensity distribution is normalized to obtain a probability distribution, and an image entropy is calculated based on the probability distribution; an intensity maximum value and an intensity minimum value of the compensation image are extracted, and an image contrast is calculated; a difference between the image contrast and 1 is calculated to obtain an image contrast loss; a first weight coefficient is determined based on a spatial distribution density of the target region, 1 is subtracted from the first weight coefficient to obtain a second weight coefficient; the image entropy is multiplied by the first weight coefficient to obtain a first weighted term, the image contrast loss is multiplied by the second weight coefficient to obtain a second weighted term, and the first weighted term and the second weighted term are added to construct a compound optimization target function.

[0117] Specifically, first, a phase error model is established in each sub-block. Specifically, a quadratic term is constructed as a quadratic phase error term by taking the distance-wise position information of the target scattering point as a variable, which is used to compensate for the phase error caused by the target uniform acceleration motion. For example, a quadratic term of the form (r-r0)2may be used, where r represents the distance-wise position of the target scattering point, and r0represents the reference position. At the same time, a cubic term (r-r0)3is constructed as a cubic phase error term, which is used to compensate for the phase error caused by the target variable acceleration motion. In addition, a product term (r-r0)(a-a0) of the distance-wise position information and the azimuth-wise position information of the target scattering point is also needed to be established as a cross-phase error term, where a represents the azimuth-wise position of the target scattering point, and a0represents the reference position. This term is used to determine the corresponding relationship of the phase error caused by the distance-wise and azimuth-wise coupling. 2 3 Specifically, first, a phase error model is established in each sub-block. Specifically, a quadratic term is constructed as a quadratic phase error term by taking the distance-wise position information of the target scattering point as a variable, which is used to compensate for the phase error caused by the target uniform acceleration motion. For example, a quadratic term of the form (r-r0)2may be used, where r represents the distance-wise position of the target scattering point, and r0represents the reference position. At the same time, a cubic term (r-r0)3is constructed as a cubic phase error term, which is used to compensate for the phase error caused by the target variable acceleration motion. In addition, a product term (r-r0)(a-a0) of the distance-wise position information and the azimuth-wise position information of the target scattering point is also needed to be established as a cross-phase error term, where a represents the azimuth-wise position of the target scattering point, and a0represents the reference position. This term is used to determine the corresponding relationship of the phase error caused by the distance-wise and azimuth-wise coupling.

[0118] Next, the above-mentioned quadratic phase error term, cubic phase error term and high-order cross-phase error term are combined to construct a generalized polynomial function. For example, a polynomial function can be obtained as follows:

[0119] where α, β, γ are coefficients to be estimated.

[0120] Then, the constructed generalized polynomial function is applied to the compensation echo data to perform phase compensation to obtain a compensation image. Specifically, the original echo data can be multiplied by , where j is an imaginary unit, so as to realize phase compensation.

[0121] ​Then, the intensity distribution of the compensated image is calculated. The intensity image can be obtained by taking the modulus of the compensated complex image, and the probability distribution can be obtained by normalizing the intensity image. Based on the probability distribution, the image entropy can be calculated. For example, assuming that the normalized intensity values are 0.1, 0.2, 0.3, and 0.4, the image entropy can be calculated as -(0.1log0.1+0.2log0.2+0.3log0.3+0.4log0.4).

[0122] At the same time, the maximum intensity value and the minimum intensity value of the compensated image are extracted, and the image contrast is calculated. For example, if the maximum intensity value is 100 and the minimum intensity value is 10, the contrast can be calculated as (100-10) / (100+10). Then, the difference between the image contrast and 1 is calculated to obtain the image contrast loss. In this example, the contrast loss is 1-(100-10) / (100+10).

[0123] In addition, the first weight coefficient needs to be determined based on the spatial distribution density of the target region. For example, the proportion of non-zero pixel points in the target region can be counted, and if the proportion is 60%, the first weight coefficient can be set to 0.6. The second weight coefficient 0.4 is obtained by subtracting the first weight coefficient from 1.

[0124] Finally, the image entropy is multiplied by the first weight coefficient to obtain the first weighted term, and the image contrast loss is multiplied by the second weight coefficient to obtain the second weighted term. The first weighted term and the second weighted term are added to construct the composite optimization objective function. For example, assuming that the calculated image entropy is 2.5 and the contrast loss is 0.2, the composite optimization objective function can be expressed as 0.6x2.5+0.4x0.2=1.58.

[0125] Through the above steps, the phase error compensation method based on the generalized polynomial function can be realized, and the composite optimization objective function integrating the image entropy and the contrast can be constructed. In practical applications, the coefficients of the polynomial function can be continuously adjusted through iterative optimization, so that the composite optimization objective function reaches the optimal value, thereby obtaining the best phase error compensation effect.

[0126] In the embodiment, by constructing quadratic, cubic phase error terms and cross phase error terms, the phase errors caused by target uniform acceleration and variable acceleration motion and distance-azimuth coupling are effectively compensated, and the imaging precision and quality are greatly improved; using a generalized polynomial function to model and compensate different types of phase errors provides a flexible error correction method, which is suitable for complex target motion models and ensures accurate compensation for various errors; by combining image entropy and image contrast loss to construct a composite optimization objective function, the imaging results have both information richness (high entropy) and good visual clarity (high contrast), thereby improving the overall quality of the image; based on the spatial distribution density of the target area, the weight coefficients of the image entropy and contrast loss are dynamically adjusted to achieve adaptive weighted optimization, ensuring that the balance and adaptability of image quality optimization are stronger in different imaging scenarios; through intensity distribution normalization processing and contrast optimization after phase compensation, the details and edge features in the image can be better presented, and the resolution and fineness of radar imaging are improved.

[0127] In an optional implementation, the conjugate gradient algorithm is used to iteratively solve the composite optimization objective function to obtain a phase error estimation value, and based on the phase error estimation value, a continuous phase correction function is generated through piecewise spline interpolation, including:

[0128] The numerical differentiation method is used to calculate the gradient of the composite optimization objective function with respect to each coefficient in the generalized polynomial function, the negative direction of the gradient is taken as the initial search direction, the optimal step size is determined through one-dimensional search in the initial search direction, each coefficient in the generalized polynomial function is updated based on the optimal step size to obtain updated coefficients, the updated gradient of the composite optimization objective function is calculated based on the updated coefficients, the conjugate coefficient is obtained by calculating the inner product ratio of the updated gradient and the previous gradient, and the negative direction of the updated gradient and the conjugate combination of the initial search direction are taken as the new search direction; the iteration is repeated until the difference between the adjacent two iterations of the composite optimization objective function is less than a preset convergence threshold, and the convergence coefficient is obtained; the convergence coefficient is substituted into the generalized polynomial function to obtain the phase error estimation value.

[0129] The center positions of each block are determined as reference interpolation nodes, and supplementary interpolation nodes are added in the overlapping area of adjacent blocks; a cubic spline basis function is selected as the interpolation basis function, and a spline coefficient equation set is constructed based on the reference interpolation nodes and the supplementary interpolation nodes; the boundary condition of zero second derivative at the end point is set for the spline coefficient equation set, and the spline coefficient equation set is solved to obtain the spline coefficient; the spline coefficient is combined with the cubic spline basis function to perform piecewise interpolation on the phase error estimation value, and a continuous phase correction function is generated.

[0130] Specifically, first, the conjugate gradient algorithm is used to iteratively solve the composite optimization objective function to obtain the phase error estimate. The specific steps are as follows:

[0131] The numerical differentiation method is used to calculate the gradient of the composite optimization objective function with respect to each coefficient in the generalized polynomial function. Taking a 5th order generalized polynomial as an example, the function includes 6 coefficients a0-a5. For each coefficient, the central difference method is used to calculate the partial derivative. For example, the partial derivative calculation formula for a0 is: (f(a0+h)-f(a0-h)) / (2h), where h is a small perturbation, generally taken as 1e-6. The partial derivatives of the 6 coefficients are calculated in turn to form the gradient vector.

[0132] The negative direction of the gradient is taken as the initial search direction. For example, if the gradient vector is [1.2, -0.5, 0.8, -0.3, 1.5, -0.6], the initial search direction is [-1.2, 0.5, -0.8, 0.3, -1.5, 0.6].

[0133] The optimal step size is determined by one-dimensional search in the initial search direction. The golden section method is used to search for the step size that makes the objective function minimum in the [0, 1] interval. For example, the optimal step size obtained by the final search is 0.35.

[0134] Based on the optimal step size, the coefficients in the generalized polynomial function are updated to obtain the updated coefficients. The update formula is: new coefficient = old coefficient + step size * search direction. For example, the update result of a0 is: a0_new = a0_old + 0.35*(-1.2).

[0135] The updated gradient of the composite optimization objective function is calculated based on the updated coefficients. The same numerical differentiation method as the first step is used.

[0136] The inner product ratio of the updated gradient and the previous gradient is calculated to obtain the conjugate coefficient. The conjugate coefficient calculation formula is: β = (g new T *g new ) / (g old T *g old ), where g new and g old are the new and old gradient vectors, respectively.

[0137] The negative direction of the updated gradient is combined with the conjugate of the initial search direction to obtain the new search direction. The new search direction calculation formula is: d new =-g new +β*d old , where d old is the old search direction.

[0138] The above steps are repeated until the difference between two adjacent iterations of the composite optimization objective function is less than a preset convergence threshold, and a convergence coefficient is obtained. For example, the convergence threshold is set to 1e-6, and after 50 iterations, the target function value changes less than 1e-6, and the coefficient obtained at this time is the convergence coefficient.

[0139] The convergence coefficient is substituted into the generalized polynomial function to obtain the phase error estimation value. For example, if the convergence coefficient is [1.5, -0.8, 0.6, -0.2, 0.1, -0.05], the phase error estimation value is:

[0140] Next, according to the phase error estimation value, a continuous phase correction function is generated by piecewise spline interpolation. The specific steps are as follows:

[0141] The center positions of each block are determined as the reference interpolation nodes. For example, for 5 blocks, the center positions are [-0.8, -0.4, 0, 0.4, 0.8], and these 5 points are used as reference interpolation nodes.

[0142] Supplementary interpolation nodes are added in the overlapping area of adjacent blocks. For example, one supplementary interpolation node is added at each of the positions [-0.6, -0.2, 0.2, 0.6], and a total of 9 interpolation nodes are formed.

[0143] A cubic spline basis function is selected as the interpolation basis function. The expression of the cubic spline basis function is: S(x) = a(x-xi) 3 +b(x-xi) 2 +c(x-xi)+d, where xi is the interpolation node.

[0144] Based on the reference interpolation nodes and the supplementary interpolation nodes, a spline coefficient equation set is constructed. The equation set includes interpolation conditions and smooth connection conditions. The interpolation condition requires that the function value of the spline function at the node is equal to the given phase error estimation value. The smooth connection condition requires that the first and second derivatives of adjacent spline segments at the connection are continuous.

[0145] The spline coefficient equation set is set with a boundary condition that the second derivative at the end point is zero. This boundary condition is called natural boundary condition, which can ensure the smoothness of the interpolation function at the end point.

[0146] The spline coefficient equation set is solved to obtain the spline coefficient. The chasing method is used to solve the three-diagonal linear equation set, and four coefficients a, b, c, d of each spline segment are obtained.

[0147] The spline coefficients are combined with the cubic spline basis functions to segmentally interpolate the phase error estimation value to generate a continuous phase correction function. In each segment interval, the corresponding spline function is used for interpolation calculation. The final obtained phase correction function is continuous and smooth in the entire definition domain, and accurately passes all interpolation nodes.

[0148] Through the above steps, the estimation and correction of the phase error are completed, and a continuous phase correction function is obtained. The function can be used for subsequent phase correction processing to improve the radar imaging quality.

[0149] In the embodiment, the gradient is calculated by numerical differentiation, and the conjugate gradient method is used for optimization iteration, which can quickly find the optimal estimation value of the phase error, accelerate the convergence process, reduce the calculation complexity, and improve the processing efficiency; the optimal step size is determined by one-dimensional search in the initial search direction, and the coefficients of the generalized polynomial function are dynamically updated based on the step size, so that each iteration is closer to the optimal solution, ensuring that the iteration process converges quickly; the conjugate direction of the updated gradient is used to avoid the "zigzag phenomenon" in the traditional gradient descent method, improving the efficiency and accuracy of the optimization path and ensuring faster global optimization; the phase correction function is generated by the cubic spline interpolation method, and supplementary interpolation nodes are added in the overlapping area of adjacent blocks to ensure the continuity and smoothness of the phase correction function in the entire area, reducing the phase jump or discontinuity; the boundary condition of the second derivative being zero at the end point is set in the spline coefficient equation system, further improving the smoothness and accuracy of the phase correction function in the edge area, and ensuring the stability of the overall imaging quality; the adaptive interpolation processing based on the block center and the nodes in the overlapping area can more accurately correct the phase error and ensure more accurate image processing results,

[0150] In an optional implementation, the phase correction function is subjected to wavelet domain convolution operation with the compensation echo data to obtain corrected echo data, the corrected echo data is processed based on an azimuth compression algorithm of a variable scale kernel function to obtain a multi-frequency band two-dimensional image, local correlation coefficients, edge intensities and texture features are extracted from each frequency band in the multi-frequency band two-dimensional image through a deep neural network to construct a multi-dimensional feature vector, weight learning is performed on the multi-dimensional feature vector, adaptive weight coefficients learned based on the weight learning are used to perform nonlinear weighted fusion on the two-dimensional images of different frequency bands, and adaptive histogram equalization enhancement is performed to determine the synthetic aperture radar imaging result, including:

[0151] The phase correction function is subjected to discrete wavelet transform to obtain first wavelet coefficients, the compensation echo data is subjected to discrete wavelet transform to obtain second wavelet coefficients, the first wavelet coefficients are multiplied with the second wavelet coefficients in the wavelet domain, and inverse wavelet transform is performed on the multiplication result to obtain corrected echo data;

[0152] The kernel function parameters are determined according to the scattering characteristics of the corrected echo data, a variable-scale kernel function is constructed based on the kernel function parameters, and the variable-scale kernel function is used for azimuth compression processing of the corrected echo data to obtain a multi-frequency band two-dimensional image.

[0153] A normalized cross-correlation operator is used to perform local correlation calculation on the multi-frequency band two-dimensional image to obtain a correlation coefficient matrix, a Canny edge detection operator is used to perform edge extraction on the multi-frequency band two-dimensional image to obtain an edge intensity matrix, and a Gabor filter bank is used to filter the multi-frequency band two-dimensional image to obtain a texture feature matrix; the correlation coefficient matrix, the edge intensity matrix, and the texture feature matrix are concatenated in the feature dimension to obtain a multi-dimensional feature vector.

[0154] A deep neural network including an input layer, a hidden layer, and an output layer is constructed, the number of neurons of the input layer is the same as the dimension of the multi-dimensional feature vector, the number of neurons of the output layer is the same as the number of frequency bands, the multi-dimensional feature vector is input into the deep neural network for training to obtain adaptive weight coefficients, and the multi-frequency band two-dimensional image is subjected to nonlinear weighted calculation according to the adaptive weight coefficients to obtain a fusion image.

[0155] The gray scale statistical distribution of the fusion image is calculated, histogram equalization parameters are determined, and the fusion image is subjected to enhancement processing by using the histogram equalization parameters to obtain a final multi-frequency band synthetic aperture radar image.

[0156] First, the phase correction function is subjected to discrete wavelet transform to obtain first wavelet coefficients. Specifically, the Haar wavelet basis function can be used to perform 3-layer wavelet decomposition on the phase correction function to obtain low-frequency approximation coefficients and high-frequency detail coefficients, which constitute the first wavelet coefficients.

[0157] Then, the compensation echo data is also subjected to the same discrete wavelet transform to obtain second wavelet coefficients. For compensation echo data with a size of 128x128, the Haar wavelet basis function is also used to perform 3-layer decomposition to obtain 7 sub-band wavelet coefficients.

[0158] Next, in the wavelet domain, the first wavelet coefficients are multiplied by the second wavelet coefficients. After the coefficients at corresponding positions are multiplied, new wavelet coefficients are obtained.

[0159] Finally, the multiplied wavelet coefficients are subjected to inverse wavelet transform to reconstruct the corrected echo data. The inverse transform process uses the same Haar basis function as the forward transform, and through 3 times of inverse transform, the wavelet coefficients are reconstructed into a time domain signal, which is the corrected echo data.

[0160] After obtaining the corrected echo data, it is necessary to perform azimuth compression processing to obtain a multi-band two-dimensional image. First, the kernel function parameters are determined according to the scattering characteristics of the corrected echo data. Specifically, by analyzing the amplitude and phase characteristics of the corrected echo data, the scattering center distribution is calculated, and according to the spatial distribution characteristics of the scattering center, the scale parameter and shape parameter of the kernel function are determined.

[0161] Then, a variable-scale kernel function is constructed based on the above parameters. A Gaussian kernel function can be used as the basis, and the scale and shape of the kernel function are adjusted according to the determined parameters to obtain a variable-scale kernel function that adapts to the current scene. For example, for a point target scene, the kernel function scale is small; for a surface target scene, the kernel function scale is relatively large.

[0162] The variable-scale kernel function is used to perform azimuth compression processing on the corrected echo data. Specifically, the corrected echo data is two-dimensionally convolved with the variable-scale kernel function to realize azimuth matching filtering and obtain a compressed two-dimensional image. For echo data of different frequency bands, the above processing is performed respectively, and finally multi-band two-dimensional images are obtained.

[0163] In order to extract features from the multi-band two-dimensional image, first, a normalized cross-correlation operator is used to calculate the local correlation. Specifically, for each frequency band two-dimensional image, a 5x5 sliding window is selected, and the normalized cross-correlation coefficient of the pixels in the window and the center pixel is calculated to obtain a correlation coefficient matrix. For a 128x128 image, a 124x124 correlation coefficient matrix can be obtained.

[0164] Next, the Canny edge detection operator is used to extract edge features. For each frequency band two-dimensional image, first, a Gaussian filter is used to smooth the image, then the image gradient amplitude and direction are calculated, and the edge intensity matrix is obtained through non-maximum suppression and double threshold detection. Similarly, a 124x124 edge intensity matrix can be obtained.

[0165] Then, the Gabor filter bank is used to extract texture features. A Gabor filter bank with 4 directions and 3 scales is constructed, and the two-dimensional image of each frequency band is filtered to obtain 12 filter responses. These responses are spliced at each pixel position to obtain a 124x124x12 texture feature matrix.

[0166] The three kinds of feature matrices described above are concatenated in the feature dimension to obtain a multi-dimensional feature vector. For each pixel position, the feature vector dimension is 1+1+12=14. Therefore, a 124x124x14 multi-dimensional feature vector is finally obtained.

[0167] In order to learn adaptive weight coefficients, a deep neural network including an input layer, a hidden layer and an output layer is constructed. The number of input layer neurons is 14, which is the same as the dimension of the feature vector. The hidden layer adopts a 3-layer structure, and the number of neurons is 64, 32 and 16 respectively. The number of output layer neurons is the same as the number of frequency bands, assuming 4 frequency bands.

[0168] The multi-dimensional feature vector is input into the deep neural network for training. The back propagation algorithm is adopted, the mean square error is used as the loss function, and the network parameters are optimized through multiple iterations. After training, the output values of the 4 neurons of the output layer are the adaptive weight coefficients corresponding to the 4 frequency bands.

[0169] According to the learned adaptive weight coefficients, nonlinear weighted calculation is performed on the multi-frequency band two-dimensional images. Specifically, the two-dimensional image of each frequency band is multiplied by the corresponding weight coefficient, and then the weighted images are added to obtain the fused image.

[0170] Finally, the fused image needs to be histogram equalized for enhancement. First, the gray scale statistical distribution of the fused image is calculated to obtain the gray scale histogram. Then, the equalization parameters such as the truncation percentage and the contrast limit are determined according to the histogram distribution. Using these parameters, the fused image is adaptively histogram equalized to adjust the pixel gray scale distribution and enhance the image contrast, and the final multi-frequency band synthetic aperture radar image is obtained.

[0171] Through the above steps, the fusion and enhancement of multi-frequency band SAR images are realized, and the image quality and information richness are effectively improved. This method comprehensively utilizes wavelet transform, variable scale kernel function, multi-feature extraction, deep learning and other technologies, and can adaptively process SAR images of different scenes, with strong robustness and universality.

[0172] In this embodiment, the phase correction operation is performed in the wavelet domain through discrete wavelet transform, which effectively improves the correction accuracy and reduces the influence of phase error on echo data, thereby improving the quality of subsequent imaging; the kernel function scale is adaptively adjusted according to the scattering characteristics to compress the azimuth of the corrected echo data, which can significantly improve the resolution of the multi-frequency band two-dimensional image and enhance the performance of target details; the multi-dimensional feature vector is trained using a deep neural network to obtain adaptive weight coefficients, realizing nonlinear weighted fusion of multi-frequency band two-dimensional images, effectively integrating information of different frequency bands, and improving the accuracy and clarity of the final imaging; the histogram equalization is used to optimize the gray scale distribution of the fused image, enhancing the contrast and detail performance of the image, so that the finally generated multi-frequency band synthetic aperture radar image has better visual effect and interpretability.

[0173] Figure 2 The structure diagram of the satellite-borne multi-frequency band synthetic aperture radar signal processing optimization system of the embodiment of the present application is as follows:Figure 2 The system comprises:

[0174] The first unit is configured to acquire original echo data of a space-borne multi-band synthetic aperture radar, estimate a Doppler center frequency of the original echo data by using an adaptive segmented polynomial fitting algorithm, construct a frequency domain adaptive compensation factor matrix, perform a two-dimensional matrix multiplication operation on the frequency domain adaptive compensation factor matrix and the original echo data to obtain compensated echo data, perform distance compression on the compensated echo data based on a sub-band decomposition distance matching filtering algorithm to generate a range image, and calculate distance position information and scattering intensity information of a target scattering point by using a maximum likelihood estimation method based on the range image.

[0175] The second unit is configured to determine a spatial distribution density of the target scattering point based on the distance position information and the scattering intensity information of the target scattering point, perform adaptive block processing on the compensated echo data based on the spatial distribution density, dynamically adjust a block size, model the compensated echo data by using a generalized polynomial function containing a quadratic phase error term, a cubic phase error term and a high-order cross term in each block, construct a composite optimization objective function fusing an image entropy and a contrast, quickly iterate the composite optimization objective function by using a conjugate gradient algorithm to obtain a phase error estimation value, and generate a continuous phase correction function by using a segmented spline interpolation based on the phase error estimation value.

[0176] The third unit is configured to perform a wavelet domain convolution operation on the phase correction function and the compensated echo data to obtain corrected echo data, perform processing on the corrected echo data based on a variable scale kernel function azimuth compression algorithm to obtain a multi-band two-dimensional image, extract a local correlation coefficient, an edge intensity and a texture feature by using a deep neural network based on each frequency band in the multi-band two-dimensional image to construct a multi-dimensional feature vector, perform weight learning on the multi-dimensional feature vector, perform nonlinear weighted fusion on two-dimensional images of different frequency bands based on adaptive weight coefficients learned, and determine a synthetic aperture radar imaging result by using adaptive histogram equalization enhancement.

[0177] A third aspect of the embodiment of the application,

[0178] An electronic device is provided, comprising:

[0179] A processor;

[0180] A memory for storing processor-executable instructions;

[0181] The processor is configured to invoke the instructions stored in the memory to perform the method described above.

[0182] A fourth aspect of the embodiment of the application,

[0183] A computer readable storage medium is provided, which stores computer program instructions. The computer program instructions are executed by a processor to implement the method.

[0184] The present application can be a method, an apparatus, a system, and / or a computer program product. The computer program product can include a computer readable storage medium (or media) having computer readable program instructions thereon for performing various aspects of the present application.

[0185] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for optimizing the signal processing of a space-borne multi-frequency synthetic aperture radar, characterized in that, The method comprises the following steps: Obtaining original echo data of a spaceborne multi-frequency synthetic aperture radar, estimating Doppler center frequency of the original echo data by using an adaptive segmented polynomial fitting algorithm, constructing a frequency domain adaptive compensation factor matrix, performing two-dimensional matrix multiplication operation on the frequency domain adaptive compensation factor matrix and the original echo data to obtain compensated echo data, performing range compression on the compensated echo data based on a sub-band decomposition range matching filtering algorithm to generate a range image, and calculating range position information and scattering intensity information of a target scattering point based on the range image by using a maximum likelihood estimation method; Based on the range position information and the scattering intensity information of the target scattering point, the spatial distribution density of the target scattering point is determined, the compensated echo data is adaptively processed based on the spatial distribution density, the block size is dynamically adjusted, and in each block, the phase error of the compensated echo data is modeled by using a generalized polynomial function containing a quadratic phase error term, a cubic phase error term and a high-order cross phase error term, a composite optimization objective function is constructed by fusing image entropy and contrast, the composite optimization objective function is iterated quickly by using a conjugate gradient algorithm, and a phase error estimation value is obtained, and a continuous phase correction function is generated by using piecewise spline interpolation according to the phase error estimation value; The phase correction function and the compensated echo data are subjected to wavelet domain convolution operation to obtain corrected echo data, the corrected echo data is processed based on a variable scale kernel function azimuth compression algorithm to obtain a multi-frequency two-dimensional image, local correlation coefficients, edge intensities and texture features are extracted from each frequency band in the multi-frequency two-dimensional image by using a deep neural network to construct a multi-dimensional feature vector, weight learning is performed on the multi-dimensional feature vector, adaptive weight coefficients are learned based on the adaptive weight coefficients, nonlinear weighted fusion is performed on the two-dimensional images of different frequency bands, and synthetic aperture radar imaging results are determined by adaptive histogram equalization enhancement.

2. The method of claim 1, wherein, The adaptive segmented polynomial fitting algorithm is used to estimate the Doppler center frequency of the original echo data, which comprises the following steps: The original echo data is segmented, the original echo data is divided into a plurality of data segments according to a preset window length, the adjacent data segments have an overlap rate of 50%, and a to-be-processed data segment is obtained; Residual error of the to-be-processed data segment is evaluated, normalized residual error values of different order polynomial fitting are calculated, a ratio of the normalized residual error value to a normalized residual error value of a previous order polynomial fitting is compared with a preset threshold value, and when the ratio is less than the preset threshold value, an optimal polynomial order is determined; A Vandermonde matrix based on the optimal polynomial order is constructed, elements of the Vandermonde matrix are composed of different order power terms of time variables, a Hanning weighting matrix is constructed, diagonal elements of the Hanning weighting matrix are calculated by a cosine function, polynomial coefficients determined by the Vandermonde matrix and the Hanning weighting matrix are solved by using a least square method, and a fitting polynomial function is constructed based on the polynomial coefficients; The data signal-to-noise ratio of the to-be-processed sub-data segment is calculated, the preset window length is adaptively adjusted according to the data signal-to-noise ratio, when the data signal-to-noise ratio is greater than a first preset data signal-to-noise ratio threshold, the preset window length is reduced to one half of the original, when the data signal-to-noise ratio is less than a second preset data signal-to-noise ratio threshold, the preset window length is increased to twice the original, and the data fitting is performed again according to the adjusted preset window length and the fitting polynomial function to obtain an optimized fitting polynomial function; An error function between the to-be-processed sub-data segment and the optimized fitting polynomial function is calculated, a mean and a variance of the error function are obtained, a compensation term based on a Gaussian function is constructed, and the compensation term is added to an initial frequency estimation value calculated according to the optimized fitting polynomial function to obtain a final Doppler center frequency estimation result.

3. The method of claim 1, wherein, A distance direction matching filter algorithm based on sub-band decomposition is used to compress the compensation echo data in the distance direction to generate a range image, and a maximum likelihood estimation method is used to calculate the distance direction position information and the scattering intensity information of the target scattering point according to the range image, including: The spectrum of the compensation echo data is divided into a plurality of frequency sub-bands, adjacent frequency sub-bands have an overlapping region, the bandwidth of the overlapping region is 20% of the bandwidth of the sub-band, and a Hamming window function is used to weight the spectrum of each frequency sub-band; A reference signal is generated based on the transmission signal parameters, the spectrum of the reference signal contains the influence of the antenna pattern and the propagation attenuation, and the reference signal is multiplied with the signals of each frequency sub-band in the frequency domain to obtain a filtering result; The phase difference between the frequency sub-bands is calculated, a linear phase compensation method is used to correct the phase difference, the signal signal-to-noise ratio of the signal corresponding to each frequency sub-band is calculated, and the signal signal-to-noise ratio is used as a weight coefficient to weight and sum the filtering result to obtain a combined range image; The combined range image is energy-normalized, distance units are divided according to the system resolution, a mapping relationship between the amplitude and the distance is established, and the peak position of the combined range image is searched as a coarse positioning result of the scattering point; In the neighborhood of the coarse positioning result of the scattering point, the distance units are ten times subdivided by using a parabolic interpolation method, and the subdivided distance units are searched for peaks to obtain an accurate position of the scattering point. A likelihood function based on a Gaussian noise model is constructed, the likelihood function contains the distance direction position parameter and the scattering intensity parameter, and the likelihood function is optimized and solved by iteration to obtain the maximum value of the likelihood function as an initial estimation result. The Cramer-Rao bound of the initial estimation result is calculated, and it is judged whether the initial estimation result reaches a pre-determined theoretical precision limit, if not, a new search space is constructed in the neighborhood of the initial estimation result, and parameter iteration optimization is continued, and finally the position information and the scattering intensity information of the target scattering point are obtained.

4. The method of claim 1, wherein, In each block, a generalized polynomial function containing a quadratic phase error term, a cubic phase error term and a high-order cross-phase error term is used to model the phase error of the compensation echo data, and a composite optimization objective function fusing image entropy and contrast is constructed, including: In each sub-block, a quadratic term taking the range position information of the target scattering point as a variable is established as a quadratic phase error term to compensate for the phase error caused by the uniform acceleration of the target; a cubic term taking the range position information of the target scattering point as a variable is established as a cubic phase error term to compensate for the phase error caused by the variable acceleration of the target; a product term of the range position information and the azimuth position information of the target scattering point is established as a high-order cross-phase error term to determine the corresponding relationship of the phase error caused by the coupling of the range and the azimuth; the quadratic phase error term, the cubic phase error term and the high-order cross-phase error term are combined to construct a generalized polynomial function; The generalized polynomial function is applied to the compensation echo data for phase compensation to obtain a compensation image; the intensity distribution of the compensation image is calculated, the intensity distribution is normalized to obtain a probability distribution, and the image entropy is calculated based on the probability distribution; the maximum intensity and the minimum intensity of the compensation image are extracted, and the image contrast is calculated; the difference between the image contrast and 1 is calculated to obtain the image contrast loss; a first weight coefficient is determined based on the spatial distribution density of the target region, 1 is subtracted from the first weight coefficient to obtain a second weight coefficient; the image entropy is multiplied by the first weight coefficient to obtain a first weighted term, and the image contrast loss is multiplied by the second weight coefficient to obtain a second weighted term; the first weighted term and the second weighted term are added to construct a compound optimization objective function.

5. The method of claim 1, wherein, The conjugate gradient algorithm is used to quickly iterate the compound optimization objective function to obtain a phase error estimation value, and a continuous phase correction function is generated through piecewise spline interpolation based on the phase error estimation value, including: A numerical differentiation method is used to calculate the gradient of the compound optimization objective function with respect to each coefficient in the generalized polynomial function, the negative direction of the gradient is taken as the initial search direction, the optimal step size is determined through one-dimensional search in the initial search direction, the coefficients in the generalized polynomial function are updated based on the optimal step size to obtain updated coefficients, the updated gradient of the compound optimization objective function is calculated based on the updated coefficients, the inner product ratio of the updated gradient and the previous gradient is calculated to obtain a conjugate coefficient, and the negative direction of the updated gradient and the conjugate combination of the initial search direction are taken as a new search direction; the iteration is repeated until the difference between the adjacent two iterations of the compound optimization objective function is less than a preset convergence threshold to obtain a convergence coefficient, and the convergence coefficient is substituted into the generalized polynomial function to obtain a phase error estimation value; The center positions of each sub-block are determined as reference interpolation nodes, and supplementary interpolation nodes are added in the overlapping area of adjacent sub-blocks; a cubic spline basis function is selected as an interpolation basis function, and a spline coefficient equation set is constructed based on the reference interpolation nodes and the supplementary interpolation nodes; the boundary condition of the second derivative being zero at the end point is set for the spline coefficient equation set, and the spline coefficient equation set is solved to obtain a spline coefficient; the spline coefficient and the cubic spline basis function are combined to perform piecewise interpolation on the phase error estimation value to generate a continuous phase correction function.

6. The method of claim 1, wherein, The phase correction function is convolved with the compensated echo data in the wavelet domain to obtain corrected echo data, the corrected echo data is processed based on an azimuth compression algorithm of a variable scale kernel function to obtain multi-band two-dimensional images, based on each frequency band in the multi-band two-dimensional images, local correlation coefficients, edge intensities and texture features are extracted through a deep neural network to construct a multi-dimensional feature vector, weight learning is performed on the multi-dimensional feature vector, nonlinear weighted fusion is performed on the two-dimensional images of different frequency bands based on the adaptive weight coefficients learned, and adaptive histogram equalization enhancement is performed to determine the synthetic aperture radar imaging result, including: The phase correction function is subjected to discrete wavelet transform to obtain first wavelet coefficients, and the compensated echo data is subjected to discrete wavelet transform to obtain second wavelet coefficients, the first wavelet coefficients are multiplied with the second wavelet coefficients in the wavelet domain, and inverse wavelet transform is performed on the multiplication result to obtain corrected echo data; The kernel function parameters are determined according to the scattering characteristics of the corrected echo data, a variable scale kernel function is constructed based on the kernel function parameters, and the corrected echo data is compressed in the azimuth direction by using the variable scale kernel function to obtain multi-band two-dimensional images; A normalized cross-correlation operator is used to calculate the local correlation of the multi-band two-dimensional images to obtain a correlation coefficient matrix, a Canny edge detection operator is used to extract the edges of the multi-band two-dimensional images to obtain an edge intensity matrix, and a Gabor filter bank is used to filter the multi-band two-dimensional images to obtain a texture feature matrix; the correlation coefficient matrix, the edge intensity matrix and the texture feature matrix are concatenated in the feature dimension to obtain a multi-dimensional feature vector; A deep neural network including an input layer, a hidden layer and an output layer is constructed, the number of neurons of the input layer is the same as the dimension of the multi-dimensional feature vector, the number of neurons of the output layer is the same as the number of frequency bands, the multi-dimensional feature vector is input into the deep neural network for training to obtain adaptive weight coefficients, nonlinear weighted calculation is performed on the multi-band two-dimensional images according to the adaptive weight coefficients to obtain a fusion image; The gray statistical distribution of the fusion image is calculated, histogram equalization parameters are determined, and the fusion image is enhanced by using the histogram equalization parameters to obtain a final multi-band synthetic aperture radar image.

7. A space-borne multi-frequency synthetic aperture radar signal processing optimization system for implementing the method of any of the preceding claims 1-6, characterized in that, including: A first unit is configured to obtain original echo data of a space-borne multi-band synthetic aperture radar, estimate a Doppler center frequency of the original echo data by using an adaptive segmented polynomial fitting algorithm, construct a frequency domain adaptive compensation factor matrix, perform two-dimensional matrix multiplication between the frequency domain adaptive compensation factor matrix and the original echo data to obtain compensated echo data, compress the compensated echo data in the range direction by using a sub-band decomposition range direction matched filtering algorithm to generate a range image, and calculate range position information and scattering intensity information of a target scattering point by using a maximum likelihood estimation method based on the range image. The second unit is configured to determine the spatial distribution density of the target scattering points based on the distance-to-position information and the scattering intensity information of the target scattering points, perform adaptive block processing on the compensated echo data based on the spatial distribution density, dynamically adjust the block size, perform phase error modeling on the compensated echo data by using a generalized polynomial function including a quadratic phase error term, a cubic phase error term, and a high-order cross-phase error term in each block, construct a composite optimization objective function fusing image entropy and contrast, perform fast iteration on the composite optimization objective function by using a conjugate gradient algorithm, and obtain a phase error estimation value; and generate a continuous phase correction function by using piecewise spline interpolation according to the phase error estimation value. The third unit is configured to perform wavelet domain convolution operation on the phase correction function and the compensated echo data to obtain corrected echo data, perform processing on the corrected echo data based on an azimuth compression algorithm of a variable scale kernel function to obtain multi-frequency two-dimensional images, extract local correlation coefficients, edge intensities, and texture features by using a deep neural network based on each frequency band in the multi-frequency two-dimensional images, construct a multi-dimensional feature vector, perform weight learning on the multi-dimensional feature vector, perform nonlinear weighted fusion on the two-dimensional images of different frequency bands based on adaptive weight coefficients learned, and determine synthetic aperture radar imaging results by using adaptive histogram equalization enhancement.

8. An electronic device, comprising: The computer program instructions are executed by the processor to implement the method in any one of claims 1 to 6. The computer program instructions are executed by the processor to implement the method in any one of claims 1 to 6. ​ ​ 9. A computer-readable storage medium having stored thereon computer program instructions, wherein, ​

Images