A method for suppressing sidelobes of terahertz SAR images based on image post-processing

Abstract

The present invention discloses a terahertz SAR image sidelobe suppression method based on image post-processing. Since the pulse repetition frequency of the radar signal is higher than the clutter bandwidth, there are clear areas within a certain range on both sides of the SAR image. The areas contain only noise and no echo energy of the imaging scene. If the high sidelobes of a strong point run through the entire azimuth direction of the SAR image, a clear area with a certain azimuth width is selected, a noise floor value and a numerical suppression function of each column of pixels are calculated, a numerically suppressed template image is generated, and the azimuth high sidelobes are suppressed. If the high sidelobes of the strong point extend to a limited distance range, a numerical suppression function is calculated for each strong point in the image, and a numerically suppressed template image is generated. The high sidelobes of the strong point are suppressed, the angular effect of the SAR image in the terahertz frequency band is reduced, the strong point sidelobes are suppressed, and the overall quality of the SAR image is improved. The suppression multiple is automatically calculated, and the azimuth high sidelobes and the high sidelobes of the strong point are suppressed simultaneously. The robustness is good, and good image quality is provided for subsequent target detection and recognition.

Description

Technical Field

[0001] The present invention belongs to the technical field of signal processing, and in particular relates to a sidelobe suppression technology. Background Art

[0002] The higher the radar frequency, the more pronounced the angular effect. This is reflected in SAR images, where the high sidelobes of a strong point obscure surrounding areas of lower value. Due to signal spurious and phase noise, strong points in SAR images also produce high sidelobes. The sidelobes of some strong points can extend across a certain area in the image azimuth, or even cover the entire azimuth.

[0003] In the terahertz frequency band, SAR images exhibit particularly pronounced angular effects, with the sidelobe intensity of certain strong points in the image significantly higher than that of surrounding targets. Due to the immaturity of terahertz radar devices, the sidelobes of these strong points in the image are further elevated. These strong sidelobes are difficult to suppress through autofocusing or windowing. Consequently, at the same resolution, terahertz SAR image quality is inferior to that of lower-frequency SAR images.

[0004] Image enhancement methods for SAR include:

[0005] In the paper "Research on Regularization Methods for Synthetic Aperture Radar Image Enhancement," a SAR image enhancement technique is disclosed. This technique utilizes a regularization algorithm to achieve image reconstruction characteristics based on prior knowledge. Based on a radar observation model, the regularization algorithm is used to enhance SAR images. The enhanced image shows a reduction in the 3dB main lobe width, demonstrating the effectiveness of the algorithm.

[0006] In "A Fast Algorithm for SAR Image Enhancement Based on Noise Visibility Function," a fast algorithm for SAR image enhancement is disclosed. This method combines the visual characteristics of the human eye and introduces a noise visibility function to achieve gain control of detail layer images. This method fully utilizes the parallel computing capabilities of GPUs to effectively improve the real-time performance of SAR image enhancement.

[0007] In "An Improved SAR Image Enhancement Method Based on Anisotropic Differentiation," a SAR image enhancement method is disclosed. This method constructs two diffusion coefficient distribution functions in the image gradient direction and perpendicular to the gradient direction. No smoothing is performed in the gradient direction, but smoothing is performed perpendicular to the gradient direction. This method denoises and enhances the image while maintaining image contour clarity.

[0008] Existing SAR image enhancement methods mainly focus on denoising and enhancing the entire image. In view of the specific characteristics of terahertz video SAR images, new targeted image enhancement algorithms are needed. Summary of the Invention

[0009] To address the problems of the prior art, the present invention proposes a terahertz SAR image sidelobe suppression method based on image post-processing. The method suppresses strong point sidelobes in SAR images, improves the overall quality and display effect of SAR images, and provides a good image foundation for subsequent detection and recognition of SAR images. To achieve the above objectives, the present invention adopts the following technical solutions.

[0010] Assume that the SAR image is I and the range dimension is N r , that is, the number of rows of the image, the azimuth dimension is N a , that is, the number of columns of the image; since the pulse repetition frequency of the radar signal is higher than the clutter bandwidth, there is a certain range of clear areas on both sides of the SAR image. This area contains only noise and no echo energy of the imaging scene. If the high sidelobe of the strong point runs through the entire azimuth direction of the SAR image, a clear area with a certain azimuth width is selected, and the noise floor value and the numerical suppression function of each column of pixels are calculated to generate a numerically suppressed template image to suppress the high sidelobes in azimuth; if the high sidelobes of the strong point extend to a limited distance range, the numerical suppression function of each strong point in the image is calculated to generate a numerically suppressed template image to suppress the high sidelobes of the strong point.

[0011] Furthermore, suppressing azimuth high side lobes includes: selecting an area in a clear area of the SAR image, averaging in azimuth to obtain a one-dimensional vector in range, and using the median of the one-dimensional vector as a noise floor value; using 1.2 times the noise floor value as a threshold, using the range gate position greater than the threshold in the clear area as the position requiring high side lobe suppression, selecting an area of a certain width before and after each column of pixels, averaging in azimuth as the projection vector of the column of pixels, deleting the pixels corresponding to the position requiring high side lobe suppression, re-interpolating to obtain a vector with the same length as the projection vector, dividing the corresponding elements of the two vectors to obtain a suppression multiple vector; generating a suppression template image with the same size as the SAR image, using a bell-shaped function definition according to the suppression multiple vector to obtain a bell-shaped function corresponding to each position requiring high side lobe suppression, taking the minimum value of the corresponding elements of the two, and assigning it to the suppression template image; multiplying the suppression template image with the corresponding elements of the SAR image to obtain the SAR image after azimuth high side lobe suppression.

[0012] Calculating the noise floor value includes: extracting N rows on one side of image I r , the number of columns is N a_part Region I part , find the average in the azimuth direction and get the length N r The one-dimensional vector V part , as the projection vector of the clear area, since the vector has a high sidelobe area, the mean of the vector cannot be used as the noise floor, and the median function mid(·) is used to calculate the median T noise =mid(V part), as the noise floor value of the SAR image.

[0013] Calculating the suppression multiple vector includes: suppressing the high side lobes of SAR images cannot use a uniform suppression value, but should calculate the suppression multiple based on the values of each area in the image, and set the suppression threshold of the high side lobe area to T lobe =1.2T noise , V part The median value is greater than T lobe The corresponding range gate position is defined as the high sidelobe area, and the range position information vector S of the high sidelobe is obtained lobe , suppress the range gate in SAR image S lobe The suppression multiples at different directions are different.

[0014] Specifically, let the j-th column pixel of image I be I(:,j), and the value range of j is [N p ,N a -N p ], expand N on both sides of position j p pixels, and the distance dimension is N r , the azimuth dimension is j+2N p The image block I(:,jN p :j+N p ), find the average value in its azimuth direction, and get the length N r The distance vector V r , as the projection vector of column pixel I(:,j), delete V r The high sidelobe position S lobe , use linear interpolation to fill the original position, keep the number of elements the same as V r The same, the corrected one-dimensional vector V is obtained r ′, divide by the corresponding elements. / V r and V r ' divided by N depress =V r . / V r ′, calculate the suppression multiple vector corresponding to each column of pixels in the clutter area of image I.

[0015] Generating a template image with suppressed azimuth high side lobes includes: generating a template image I with all zeros of the same size as I m , if there is a high side lobe in the i-th row of image I, then calculate the bell-shaped function where N e is the half width of the bell-shaped function, independent variable x=[iN e ,i+N e ] T The length is 2N e +1 column vector, assigning the sidelobe suppression function y(x) to Im Each column of I m The element in row i and column j is I m (i,j), as the area that needs to be assigned, the value range of j is

[0016] [1,N a ], use the minimum value MIN(·) of the two vectors to separate y(x) and I m The corresponding element of (x,j) takes the minimum value I m (x,j)=MIN(I m (x,j),y(x)) is assigned to I m (x,j), calculate I m All columns of I are assigned the high sidelobe suppression function at each range gate to I m , get the suppression template image I of the azimuth high side lobe m .

[0017] Furthermore, suppressing the high side lobes of strong points includes: converting the image to the logarithmic domain, setting the strong point threshold according to the noise floor value, and extracting the position of the strong point in the image; generating a suppression template image with the same size as the SAR image, calculating the suppression function of each strong point in the azimuth and range directions, taking the minimum value of the corresponding elements of the two, and assigning the value to the suppression template image; subtracting the logarithmic domain SAR image from the suppression template image, performing exponential calculation on the SAR image, and obtaining the SAR image after the strong point high side lobes are suppressed.

[0018] Extracting the strong point position includes: assuming that the value T above the noise floor in the SAR image I′ after suppressing the azimuth high side lobe is noise Strong points above 30dB need to suppress high side lobes in azimuth and range directions, using db(·) as 10log 10 (·) function, converting the image I′ into a dB image in db(T noise )+30dB is the threshold, record all I′ dB = db(I′) position information.

[0019] Generating a template image with high side lobes suppressed at a strong point includes: setting the number of bits N to be suppressed around the strong point e , as the half width of the bell-shaped function, expand multiple pixels in the azimuth and distance directions of the strong point to generate dB All-zero template image I′ of the same size m , assign the suppression function of each strong point to I′ m , and obtain the suppression template image of strong points and high side lobes.

[0020] Calculate the strong point I′ in row i and column j dB (i,j) row-wise suppression function Let the independent variable x1=[jNe ,j+N e ] is of length 2N e +1 row vector, use the two vectors to find the minimum value MIN(·) of the corresponding elements to convert y row (x1) and I′ m The corresponding element of (i,x1) takes the minimum value I′ m (i,x1)=MIN(I′ m (i,x1),y row (x1)) is assigned to I′ m (i,x1), calculate the column-wise suppression function Let the independent variable x2 = [iN e ,i+N e ] T The length is 2N e +1 column vector, use the two vectors to find the minimum value MIN(·) of the corresponding elements to convert y col (x2) and I′ m The corresponding element of (x2,j) takes the minimum value I′ m (x2,j)=MIN(I′ m (x2,j),y col (x2)) is assigned to I′ m (x2, j), get the suppression template image I′ of strong point and high side lobe m , where N depress is the suppression multiple vector.

[0021] Suppressing the high side lobes of strong points includes: dB Perform matrix difference operation with the numerical suppression template I″ dB =I′ dB -I′ m , calculated using the exponential The SAR image I″ is obtained after the strong point and high sidelobe are suppressed.

[0022] The beneficial effects of the present invention include utilizing an image post-processing method to reduce the angular effect of terahertz frequency band SAR images, suppressing strong point sidelobes, and improving the overall quality of SAR images. The suppression multiple is automatically calculated based on the strong point values in the SAR images, and high azimuth sidelobes and strong point high sidelobes are suppressed simultaneously, thereby providing good robustness and providing good image quality for subsequent target detection and recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 is the original SAR image, Figure 2 It is a one-dimensional vector curve diagram after taking the average of the clear area orientation. Figure 3 is the azimuth high sidelobe suppression template image, Figure 4 SAR image after azimuth high sidelobe suppression, Figure 5It is a strong point high sidelobe suppression template image, Figure 6 It is a SAR image after the strong point high sidelobe is suppressed. DETAILED DESCRIPTION

[0024] The technical solution of the present invention is described in detail below with reference to the accompanying drawings.

[0025] Assume that the SAR image is I, and the range and azimuth dimensions are N respectively. r =800 and N a =2000, the SAR image before sidelobe suppression is as follows Figure 1 shown.

[0026] For the case where high side lobes run through the entire azimuth image, a clear area of a certain width is extracted to calculate the noise floor value, and the distance dimension N is extracted from the left side of image I. r =800, azimuth dimension N a_part =100 image area I part , find the average in the azimuth direction and get the length N r The one-dimensional vector V part ,like Figure 2 As shown, the noise floor value is calculated as T noise =893110.

[0027] Set the high sidelobe area suppression threshold to T lobe =1.2T noise , V part The median value is greater than T lobe The corresponding range gate position is defined as the high sidelobe area, and the distance position information vector recording the high sidelobe is S lobe .

[0028] For the pixel I(:,i) in the i-th column of image I, expand N on both sides of position i p = 20 pixels, and the distance dimension is N r , the azimuth dimension is i+2N p The image block is averaged in the azimuth direction to obtain a length of N r The distance vector is V r , as the projection vector of column pixel I(:,i).

[0029] Delete vector V r The high sidelobe position S lobe , use linear interpolation to supplement the original position so that the number of elements is the same as V r Same, get a new one-dimensional vector V r ′, calculate the high sidelobe suppression multiple vector N depress , the value range of i is [N p ,N a -N p], calculate the high sidelobe suppression factor vector corresponding to each column of pixels in the SAR image.

[0030] Generate an all-0 template image I with the same size as I m , calculate the suppression function of each high sidelobe position and obtain the template image I that suppresses the high sidelobe in the azimuth m ,like Figure 3 shown.

[0031] Combine the SAR image I with the template image I m The corresponding elements are multiplied to obtain the SAR image I′ after azimuth high sidelobe suppression, as shown in Figure 4 shown.

[0032] In image I′, db(T noise )+30dB is the threshold value, and the value is greater than db(T noise )+30dB pixels are taken as the strong point positions that need to be suppressed, and the position information of all strong points is recorded, and the image I′ is converted into a dB image I′ dB , i.e. I′ dB =10log 10 (I′).

[0033] The strong point is expanded by multiple pixels in the direction and distance, and the number of pixel expansion bits N is set. e =15 is the number of bits to be suppressed around the strong point, and numerical suppression is performed in the azimuth and distance directions respectively to generate the same value as I′ dB All-zero template image I′ of the same size m , assign the suppression function of each strong point to I′ m , and obtain the template image that suppresses strong points, such as Figure 5 shown.

[0034] Take the dB SAR image I′ dB Subtract the matrix from the numerical suppression template matrix to suppress the high side lobes of the strong points and obtain the SAR image after strong point suppression, such as Figure 6 shown.

[0035] The above are embodiments of the present invention and do not limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention are included in the protection scope of the present invention.

Claims

1. A terahertz SAR image sidelobe suppression method based on image post-processing, characterized in that: include: Assume that the SAR image is I and the range dimension is N r , that is, the number of rows of the image, the azimuth dimension is N a , that is, the number of columns of the image; If the high side lobes of the strong point run through the entire azimuth direction of the SAR image, a clear area with a certain azimuth width is selected, the noise floor value and the numerical suppression function of each column of pixels are calculated, and a numerical suppression template image is generated to suppress the high side lobes in azimuth. The method for suppressing high azimuth side lobes includes: selecting an area in a clear area of the SAR image, averaging the values in azimuth to obtain a one-dimensional vector in range, and using the median of the one-dimensional vector as a noise floor value; using 1.2 times the noise floor value as a threshold, using range gate positions in the clear area that are greater than the threshold as positions requiring high side lobe suppression, selecting an area of a certain width before and after each column of pixels, averaging the values in azimuth as a projection vector for the column of pixels, deleting pixels corresponding to the positions requiring high side lobe suppression, re-interpolating the values to obtain a vector with the same length as the projection vector, and dividing corresponding elements of the two vectors to obtain a suppression multiple vector; generating a suppression template image with the same size as the SAR image, defining a bell-shaped function according to the suppression multiple vector, obtaining a bell-shaped function corresponding to each position requiring high side lobe suppression, taking the minimum value of corresponding elements of the two, and assigning the result to the suppression template image; and multiplying the suppression template image by corresponding elements of the SAR image to obtain a SAR image with high azimuth side lobe suppression. If the high side lobes of the strong points extend to a limited distance range, the numerical suppression function of each strong point in the image is calculated to generate a numerically suppressed template image to suppress the high side lobes of the strong points; The method for suppressing strong point high side lobes includes: converting an image into a logarithmic domain, setting a strong point threshold according to a noise floor value, and extracting strong point positions in the image; generating a suppression template image with the same size as the SAR image, calculating the suppression function of each strong point in azimuth and range, taking the minimum value of corresponding elements of the two, and assigning the value to the suppression template image; performing a matrix difference operation and an exponential calculation on the logarithmic domain SAR image and the suppression template image to obtain a SAR image after suppressing strong point high side lobes.

2. The terahertz SAR image sidelobe suppression method based on image post-processing according to claim 1, characterized in that: The calculation of the noise floor value includes: extracting N rows on one side of the image I r , the number of columns is N a_part Region I part , find the average in the azimuth direction and get the length N r The one-dimensional vector V part , as the projection vector of the clear area, use the median function mid(·) to calculate the median T noise =mid(V part ), as the noise floor value of the SAR image.

3. The terahertz SAR image sidelobe suppression method based on image post-processing according to claim 2, characterized in that: Calculate the suppression multiple vector, including: setting the suppression threshold of the high sidelobe area to T lobe =1.2T noise , V part The median value is greater than T lobe The corresponding range gate position is defined as the high sidelobe area, and the range position information vector S of the high sidelobe is obtained lobe , different azimuths are used to suppress the range gate in SAR images at different times lobe The numerical value at .

4. The terahertz SAR image sidelobe suppression method based on image post-processing according to claim 3, characterized in that: The calculation of the suppression multiple vector also includes: assuming that the j-th column pixel of the image I is I(:, j), and the value range of j is [N p ,N a -N p ], expand N on both sides of position j p pixels, and the distance dimension is N r , the azimuth dimension is j+2N p The image block I(:,jN p :j+N p ), find the average value in its azimuth direction, and get the length N r The distance vector V r , as the projection vector of column pixel I(:,j), delete V r The high sidelobe position S lobe , use linear interpolation to fill the original position, keep the number of elements the same as V r The same, the corrected one-dimensional vector V′ is obtained r , divide by the corresponding elements. / V r and V′ r Divide, that is, N depress =V r . / V′ r , get the suppression multiple vector N at I(:,j) depress , calculate the suppression multiple vector corresponding to each column of pixels in the clutter area of image I.

5. The terahertz SAR image sidelobe suppression method based on image post-processing according to claim 4, characterized in that: Generate a template image with suppressed azimuth high side lobes, including: generating a template image I with all zeros of the same size as I m , if there is a high side lobe in the i-th row of image I, then calculate the bell-shaped function where N e is the half width of the bell-shaped function, independent variable x=[iN e ,i+N e ] T The length is 2N e +1 column vector, assigning the sidelobe suppression function y(x) to I m Each column of I m The element in row i and column j is I m (i,j), as the area that needs to be assigned, the value range of j is [1,N a ], use the minimum value MIN(·) of the two vectors to separate y(x) and I m The corresponding element of (x,j) takes the minimum value I m (x,j)=MIN(I m (x,j),y(x)) is assigned to I m (x,j), calculate I m All columns of I are assigned the high sidelobe suppression function at each range gate to I m , get the suppression template image I of the azimuth high side lobe m .

6. The terahertz SAR image sidelobe suppression method based on image post-processing according to claim 1, characterized in that: Extracting the strong point position includes: assuming that the value T above the noise floor in the SAR image I′ after suppressing the azimuth high side lobe is noise Strong points above 30dB need to suppress high side lobes in azimuth and range directions, using db(·) as 10log 10 (·) function, converting the image I′ into a dB image in db(T noise )+30dB is the threshold, record all I′ dB = db(I′) position information.

7. The method for suppressing side lobes of terahertz SAR images based on image post-processing according to claim 6, characterized in that: Generate a template image with high side lobes that suppresses strong points, including: setting the number of bits N that need to be suppressed around the strong point e , as the half width of the bell-shaped function, expand multiple pixels in the azimuth and distance directions of the strong point to generate dB All-zero template image I′ of the same size m , assign the suppression function of each strong point to I′ m , and obtain the suppression template image of strong points and high side lobes.

8. The method for suppressing side lobes of terahertz SAR images based on image post-processing according to claim 7, characterized in that: Generating a template image with suppressed strong points and high side lobes also includes: calculating the strong point I′ in the i-th row and j-th column dB (i,j) row-wise suppression function Let the independent variable x1=[jN e ,j+N e ] is of length 2N e +1 row vector, use the two vectors to find the minimum value MIN(·) of the corresponding elements to convert y row (x1) and I′ m The corresponding element of (i,x1) takes the minimum value I′ m (i,x1)=MIN(I′ m (i,x1),y row (x1)) is assigned to I′ m (i,x1), calculate the column-wise suppression function Let the independent variable x2 = [iN e ,i+N e ] T The length is 2N e +1 column vector, use the two vectors to find the minimum value MIN(·) of the corresponding elements to convert y col (x2) and I′ m The corresponding element of (x2,j) takes the minimum value I′ m (x2,j)=MIN(I′ m (x2,j),y col (x2)) is assigned to I′ m (x2, j), get the suppression template image I′ of strong point and high side lobe m ; where N depress is the suppression multiple vector.

Images