Sidelobe suppression method for SAR (Synthetic Aperture Radar) image

Abstract

The invention discloses a side lobe suppression method for an SAR image, and the method comprises the steps: obtaining an imaging result containing a high side lobe through a CS algorithm; secondly, re-arranging a scene only containing one point target with the intensity of 1, performing echo simulation on the newly arranged scene and imaging based on the CS algorithm, and obtaining an imaging result of the point target under the CS algorithm as a convolution kernel; and finally, a real scene image is obtained through Lucy-Richardson-based time domain filtering deconvolution. According to the invention, the sidelobe interference in the orientation in the imaging process can be inhibited, the robustness and universality of the imaging method are improved, and the imaging quality is further improved.

Description

Technical Field

[0001] This invention belongs to the field of remote sensing imaging technology, specifically relating to a method for suppressing sidelobes in SAR images. Background Technology

[0002] With the increasingly widespread application of Synthetic Aperture Radar (SAR) images, improving SAR image quality has become a key research focus. After collecting signals in the azimuth and range directions, SAR performs coherent superposition imaging based on matched filtering, ultimately obtaining impulse response functions (sinc functions) in both directions. Due to the nature of the sinc function, the final SAR image exhibits sidelobe interference. Strong sidelobe interference can obscure weak scattering targets, affecting target detection and significantly reducing SAR image quality. For normal observation of weak targets, radar requires strict suppression of peak sidelobes; sidelobe suppression is a crucial step in ensuring SAR image quality, and its effectiveness is an important standard in the SAR image quality technical evaluation system. Currently, widely used methods for SAR image sidelobe suppression include frequency domain windowing and Spatial Variant Apoization (SVA). Frequency domain windowing uses a simple and effective window function to process the signal, smoothing and reducing sidelobes. However, windowing widens the main lobe, affecting resolution. The spatial apodization algorithm is an improvement on the dual apodization and triple spatial apodization algorithms. It uses an adaptive cosine weighting function to suppress sidelobes, and the algorithm generally performs well and has high computational efficiency. However, this algorithm can only handle signals sampled at integer multiples of Nyquist, and some sidelobes cannot be suppressed when there is excessive mixing between the main and sidelobes. Summary of the Invention

[0003] In view of this, the present invention provides a sidelobe suppression method for SAR images, which can suppress sidelobe interference in the azimuth direction during imaging, improve the robustness and universality of the imaging method, and further improve the imaging quality.

[0004] The technical solution for implementing the present invention is as follows:

[0005] A sidelobe suppression method for SAR images is proposed. First, an imaging result with high sidelobes is obtained through the Chirp Scaling (CS) algorithm. Second, a scene containing only a point target with an intensity of 1 is re-deployed. Echo simulation and imaging based on the CS algorithm are performed on the newly deployed scene to obtain the imaging result of the point target under the CS algorithm, which is used as the convolution kernel. Finally, the real scene image is obtained by deconvolution based on Lucy-Richardson temporal filtering.

[0006] Furthermore, the method of the present invention includes the following steps:

[0007] Step 1: Obtain radar operating parameters;

[0008] Step 2: Acquire SAR image of observation scene I

[0009] Step 3: Generate echoes and perform CS imaging based on radar operating parameters and the deployed scene to obtain the imaging function.

[0010] Step 4: Utilize and Perform Lucy-Richardson filtering to obtain the real scene image I1.

[0011] Furthermore, the radar operating parameters include scene center reference slant range, satellite flight speed, equivalent slant angle, azimuth antenna size, satellite orbital altitude, operating wavelength, frequency modulation of the transmitted signal, frequency modulation of the range frequency modulation signal in the azimuth frequency domain, speed of light, range and azimuth time vectors, and range and azimuth frequency vectors.

[0012] Furthermore, step 3 specifically includes:

[0013] Step 3.1: Set up scene I0;

[0014] Step 3.2: Generate scene echoes based on the parameters from Step 1 and the deployed scene I0.

[0015] Step 3.3: Generate the echo from the scene Perform CS imaging to obtain the imaging function.

[0016] Furthermore, step 4 specifically involves:

[0017] For images containing sidelobe interference Perform Lucy-Richardson filtering. Each column and As input, the final result is the real scene image I1; the Lucy-Richardson filter solution process is an iterative process, represented as:

[0018]

[0019] Where i is the iteration number. This represents the convolution operation. This represents the real-world image obtained during the i-th iteration of the Lucy-Richardson filter. This represents the real-world image obtained during the (i+1)th iteration of the Lucy-Richardson filter.

[0020] Beneficial effects:

[0021] 1. This invention addresses the problem of high sidelobe interference in SAR imaging results based on the matched filtering principle, overcoming the issue that sidelobe interference from strong targets can mask the detection of weak scattering targets.

[0022] 2. The method of the present invention constructs a temporal convolution mathematical model in the image domain, and then performs echo simulation and imaging to obtain the convolution kernel by setting up a scene containing only a point target with an intensity of 1. Furthermore, it deconvolves the high sidelobe SAR and the convolution kernel to obtain the real scene and improve the SAR image quality.

[0023] 3. This invention can suppress sidelobe interference in the azimuth direction during imaging, improve the robustness and universality of the imaging method, and further improve the imaging quality. Attached Figure Description

[0024] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0025] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0026] like Figure 1 As shown, the present invention provides a sidelobe suppression method for SAR images, comprising:

[0027] Step 1: Obtain radar operating parameters.

[0028] The radar operating parameters include scene center reference slant range, satellite flight speed, equivalent slant angle, azimuth antenna size, satellite orbital altitude, operating wavelength, frequency modulation of the transmitted signal, frequency modulation of the range frequency modulation signal in the azimuth frequency domain, speed of light, range and azimuth time vectors, and range and azimuth frequency vectors.

[0029] Radar operating parameters are R ref V is the reference slant distance at the scene center. e For the satellite's flight speed, For reference to the equivalent oblique angle, L a Where λ is the azimuth antenna size, H is the orbital altitude of the satellite, λ is the operating wavelength, Kr is the frequency modulation of the transmitted signal, and c is the speed of light.

[0030] Let τ and t be the range and azimuth time vectors, respectively. Define the range time series as: τ = T s_RWC +(0:N r -1) / F sThe colon operator 0:N r -1 indicates that the sequence {0,1,2,...,N} is generated. r -1}。 T s_RWC N represents the equivalent window opening time after the person moves around. r F represents the range number of echo signal points. s Let be the range sampling frequency. Define the azimuth time series: t = [(0:N] a -1) / PRF] T , [·] T This represents the transpose operation. N a The echo signal azimuth point number is given, PRF is the pulse repetition frequency, and the colon operator 0:N is used. a -1 indicates that the sequence {0,1,2,...,N} is generated. a -1}.

[0031] f τ Let f be the range and azimuth frequency vectors, respectively. Define the range frequency vector as: f τ =-F s / 2+(0:N r -1)·F s / N r Define the azimuth frequency domain vector: f a =[-PRF / 2+(0:N a -1)·PRF / N a ] T , [·] T This indicates the transpose operation.

[0032] Step 2: Acquire SAR image of observation scene I

[0033] First, the original echo signal matrix of the observed scene is subjected to azimuth-to-Fourier transform. Second, the transformed echo signal is subjected to range migration consistency correction, followed by range-to-Fourier transform. The signal obtained from the range-to-Fourier transform is subjected to range migration correction and range compression. The echo signal after range migration correction and range compression is subjected to inverse range-to-Fourier transform. The obtained echo signal is subjected to azimuth processing and residual phase compensation. Finally, the obtained signal is subjected to inverse azimuth-to-Fourier transform.

[0034] Step 2.1: Obtain the original echo signal matrix of observation scene I.

[0035] The echo pixels in scene I are a two-dimensional matrix, and the echo matrix is ​​represented as:

[0036]

[0037] Among them, S echo,m,kThe azimuth sampling time is t. m The distance sampling time is τ k The echo. That is, the echo matrix S. echo The element in the m-th row and k-th column.

[0038] Step 2.2: Azimuth to Fourier Transform

[0039] Perform an azimuth-to-Fourier transform on the echo matrix to obtain the azimuth frequency domain echo signal S. echo,a_fft

[0040] The Fourier transform in this embodiment of the invention differs from the Fast Fourier Transform (FFT). The specific operation method is as follows: Extract S... echo The first column, take the first N of this column matrix. a / 2 elements and the following N a Swap the positions of two elements, then perform an FFT on the resulting column matrix. After the FFT, swap the first half of the elements with the second half of the resulting column matrix again, thus completing the distance-to-Fourier transform of the first column of the echo signal. Performing the above operation on each column of the echo matrix sequentially completes the azimuth-to-Fourier transform of the original echo signal matrix. Let the signal after the azimuth-to-Fourier transform be S. echo,a_fft .

[0041] Step 2.3 Distance migration consistency correction

[0042] In the range-Doppler domain, the signal S echo,a_fft Multiplying by the CS factor Φ1 completes the distance migration consistency correction, and the signal after the distance migration consistency correction transformation is denoted as S1. All matrix multiplication and division operations in this invention are point-by-point operations. [The last sentence appears to be incomplete and requires further context.] echo,a_fft Perform a dot product operation with Φ1, and denote the resulting two-dimensional matrix as S1.

[0043] Where, matrix Φ1=exp{-jπZ1Z2[T τ -T ref ] 2}

[0044] Matrix T τ =[τ,…τ,] T Na×Nr , [·] T This represents the transpose operation. τ is N. r A row vector of n elements, T τ It is N a N consisting of τ vectors a ×N r matrix.

[0045] As can be seen from the following, since f is N aA column vector of n elements, so C s b r τ ref Also N a A column vector of n elements.

[0046] T ref =[τ ref ,…τ ref , Na×Nr τ ref It is N a A column vector of n elements, T ref It is N r τ ref N composed of vectors a ×N r matrix.

[0047] Z1 = [C s , ...C s , Na×Nr C s It is N a Z1 is a column vector of N elements. r C s N composed of vectors a ×N r matrix.

[0048] Z2=[b r , ...b r , Na×Nr b r It is N a Z2 is a column vector of N elements. r b r N composed of vectors a ×N r matrix.

[0049] in

[0050]

[0051] Because T τ T ref Z1 and Z2 are N a ×N r Matrix, so Φ1 is N a ×N r Matrix, therefore S1 is also N a ×N r matrix.

[0052] Step 2.4: Distance to Fourier Transform

[0053] Perform a distance-to-Fourier transform on signal S1 to transform the signal into a two-dimensional frequency domain. Ignore complex constants to obtain signal S2.

[0054] The specific operation method is as follows: Extract the first row of signal S1, and then extract the first N rows of this matrix. r / 2 elements and the following N r Swap the positions of the first two elements, then perform an FFT on the resulting row matrix. After the FFT, swap the first half of the elements with the second half of the resulting row matrix again, thus completing the distance-to-Fourier transform of the first row of the echo signal data. Performing the above operation on each row of the echo matrix sequentially completes the distance-to-Fourier transform of the original echo signal matrix, transforming the signal to the two-dimensional frequency domain. Ignoring complex constants, we obtain signal S2.

[0055] Step 2.5: Distance migration correction and distance compression

[0056] Signal S2 undergoes range migration correction and range focusing in the two-dimensional frequency domain by multiplying it by a range compensation factor Φ2. The signal after range migration correction and range compression is denoted as S3, and the compensation factor is...

[0057]

[0058] Define matrix Z3 = [f τ …f τ ] T Na×Nr , [·] T f represents the transpose operation. τ It is N r A row vector of N elements, Z3 is formed by N elements. a f τ N composed of vectors a ×N r Matrices. Matrices Z1 and Z2 are defined in step 2.3. R ref c is defined in step 1. Because Z1, Z2, Z3 are N a ×N r Matrix, so Φ2 is N a ×N r Matrix, therefore S3 is also N a ×N r matrix.

[0059] The first term in the formula performs secondary range compression and range-oriented focusing, while the second term performs range migration correction.

[0060] Step 2.6: Inverse Fourier Transform of Distance

[0061] Perform an inverse Fourier transform on the distance-compressed frequency domain echo signal S3 to obtain the time-domain distance-compressed echo signal S4.

[0062] The method for inverse distance-to-Fourier transform is the same as the steps for distance-to-Fourier transform described above, except that the FFT function used in the process is replaced with the IFFT function.

[0063] Step 2.7: Multiply by the azimuth compensation factor in the range-Doppler domain.

[0064] The echo signal S4 is multiplied by the azimuth compensation factor Φ3 in the range-Doppler domain to complete azimuth processing and residual phase compensation, resulting in the echo signal S5. The compensation factor is...

[0065]

[0066] Z4 = [f, ..., f, ] Na×Nr f is N a Z4 is a column vector of N elements. r N consisting of f vectors a ×N r matrix.

[0067] The definitions of matrices Z1 and Z2 are given in step 2.3.

[0068]

[0069] Matrix T τ See step 2.3 for the definition.

[0070] Θ1,Θ2,R,Z4 are N a ×N r Matrix, therefore Φ3 is N a ×N r Matrix, S5 is N a ×N r matrix.

[0071] Step 2.8: Inverse Fourier Transform of Azimuth

[0072] Perform an inverse Fourier transform on the echo signal S5 to obtain the echo signal, which is the SAR image of observation scene I.

[0073] The method for inverse azimuth to Fourier transform is the same as the steps for azimuth to Fourier transform described above, except that the FFT function used in the process is replaced with the IFFT function.

[0074] Step 3: Generate echoes and perform CS imaging based on the parameters from Step 1 and the scene layout.

[0075] Step 3.1: Set up scene I0.

[0076] Set up a scene I0 with N×N point coordinates, where N is typically 20-30. The center coordinates of scene I0 are (0, y0, 0), and the coordinates of each scene point are...

[0077]

[0078] Among them, China PRT stands for Pulse Repetition Period.

[0079] The radar backscattering coefficient of the center coordinate point (0, y0, 0) of the scene is set to σ = 1, and the coefficients of the other coordinate points are set to 0.

[0080] Step 3.2: Generate the echo of scene I0 based on the parameters from step 1 and the deployed scene.

[0081] The coordinates of the satellite's sequential sampling positions are as follows The echo signal of the target at the center point of the scene (0, y0, 0) is represented as follows:

[0082] Where σ is the backscattering characteristic of the point target, which is 1, and W a It is the azimuth antenna directivity function.

[0083] W a (R m )=sinc(((acos(R ref / R m -0.886·λ / L a ))), ε(·) represents the step function, which is the envelope of the emitted pulse, R m It is the slant range variation matrix function from the point target to the radar.

[0084]

[0085] Among them, T t = [t, ..., t,] Na×Nr t is N a A column vector of n elements, T t It is N r N consisting of t vectors a ×N r matrix.

[0086] Step 3.3: Generate the echo from the scene Perform CS imaging in step 2 to obtain the imaging function.

[0087] Step 4: Obtain the real scene image I1.

[0088] For images containing sidelobe interference Perform Lucy-Richardson filtering. Each column and As input, the final result is the real-world image I1. The Lucy-Richardson filter solution process is an iterative process, represented as:

[0089]

[0090] Where i is the iteration number. This represents the convolution operation. This represents the real-world scene result obtained during the i-th iteration of the Lucy-Richardson filter. This represents the real-world scene result obtained during the (i+1)th iteration of the Lucy-Richardson filter.

[0091] In summary, the above are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for suppressing sidelobes in SAR images, characterized in that, First, an imaging result with high sidelobes is obtained through the CS algorithm. Second, a scene containing only a point target with an intensity of 1 is re-deployed, and echo simulation and imaging based on the CS algorithm are performed on the newly deployed scene to obtain the imaging result of the point target under the CS algorithm, which is used as the convolution kernel. Finally, the real scene image is obtained by deconvolution based on Lucy-Richardson temporal filtering.

2. The sidelobe suppression method for SAR images as described in claim 1, characterized in that, Specifically, the following steps are included: Step 1: Obtain radar operating parameters; Step 2: Acquire SAR image of observation scene I Step 3: Generate echoes and perform CS imaging based on radar operating parameters and the deployed scene to obtain the imaging function. Step 4: Utilize and Perform Lucy-Richardson filtering to obtain the real scene image I1.

3. The sidelobe suppression method for SAR images as described in claim 2, characterized in that, The radar operating parameters include scene center reference slant range, satellite flight speed, equivalent slant angle, azimuth antenna size, satellite orbital altitude, operating wavelength, frequency modulation of the transmitted signal, frequency modulation of the range frequency modulation signal in the azimuth frequency domain, speed of light, range and azimuth time vectors, and range and azimuth frequency vectors.

4. A sidelobe suppression method for SAR images as described in claim 2 or 3, characterized in that, Step 3 specifically includes: Step 3.1: Set up scene I0; Step 3.2: Generate scene echoes based on the parameters from Step 1 and the deployed scene I0. Step 3.3: Generate the echo from the scene Perform CS imaging to obtain the imaging function.

5. The sidelobe suppression method for SAR images as described in claim 4, characterized in that, Step 4 is as follows: For images containing sidelobe interference Perform Lucy-Richardson filtering. Each column and As input, the final result is the real scene image I1; the Lucy-Richardson filter solution process is an iterative process, represented as: Where i is the iteration number. This represents the convolution operation. This represents the real-world image obtained during the i-th iteration of the Lucy-Richardson filter. This represents the real-world image obtained during the (i+1)th iteration of the Lucy-Richardson filter.

Images