CN104991241B - Target signal extraction and super-resolution enhancement processing method in strong clutter condition - Google Patents

Abstract

The invention discloses a target signal extraction and super-resolution enhancement processing method in a strong clutter condition. The main idea is that after an imaging result of point targets in a ground scene is obtained, target rectangular region slicing data including the point targets and a ground clutter is selected; a strong scattering point corresponding to a scattering point two-dimension position corresponding to the Fourier primary function correlation maximum value in the target rectangular region slicing data is extracted; weak scattering points nearby the strong scattering point are extracted using a gradient descent method, so as to obtain a weak scattering point area nearby the strong scattering point corresponding to the scattering point two-dimension position which meets conditions of the gradient descent method; subtraction of the weak scattering point area nearby the strong scattering point corresponding to the scattering point two-dimension position which meets conditions of the gradient descent method from the target rectangular region slicing data is performed, so as to obtain a complete SAR (Synthetic Aperture Radar) image; and the complete SAR image is subjected to the super-resolution enhancement processing using a regularization method, so as to obtain a final super-resolution SAR image.

Description

Target signal extraction and super-resolution enhancement processing method under strong clutter background

Technical Field

The invention belongs to the technical field of radar signal processing, and particularly relates to a target signal extraction and super-resolution enhancement processing method under a strong clutter background, which is suitable for target signal extraction under the strong clutter background or regularization super-resolution SAR imaging processing of sparse signals.

Background

Synthetic Aperture Radar (SAR) is a high-resolution microwave imaging radar, compared with the traditional optical imaging, the microwave imaging radar is not limited by weather conditions, can observe a target signal or a scene where the target signal is located all day long and all weather, and the research of automatically identifying the target signal in a battlefield by utilizing the SAR image is highly regarded at present. Generally, the higher the resolution of the SAR image, the higher the accuracy of automatically recognizing the target signal.

However, when the Synthetic Aperture Radar (SAR) is in an operating state, strong ground clutter in the radiation range of the Synthetic Aperture Radar (SAR) antenna beam enters a Synthetic Aperture Radar (SAR) receiver together with a target signal through the Synthetic Aperture Radar (SAR) antenna, and a SAR image with a low signal-to-noise ratio is formed. In the post-processing process, if the target signal is not extracted accurately, but the SAR image with a low signal-to-noise ratio is subjected to super-resolution enhancement directly, a large amount of false target signals are introduced, so that the accuracy of automatic identification of the target signals is reduced.

Therefore, when the signal-to-noise ratio of the SAR image is low, accurate extraction of the target signal is very necessary. For target signal extraction, a two-dimensional constant false alarm method is used for processing in the traditional method, and can obviously inhibit weak ground clutter and accurately extract a target signal; however, for a complex environment with strong ground clutter or large ground clutter fluctuation, a large amount of false alarms can be generated by adopting the two-dimensional constant false alarm method, so that a target signal cannot be accurately extracted; although the constant false alarm threshold in the two-dimensional constant false alarm method is processed by a sliding window, the false alarm can be reduced, the operation can cause loss to weak scattering points of the target signal, and the integrity of the target signal cannot be ensured.

Disclosure of Invention

In view of the above deficiencies of the prior art, the present invention provides a method for extracting a target signal and enhancing the super-resolution in a strong clutter background, so as to finally obtain a super-resolution SAR image.

The realization idea of the invention is as follows: according to the correlation between the target signal and the Fourier basis height and the advantage that the improved regularization method can enhance the resolution of the target signal, the target signal is extracted by constructing a self-adaptive clutter threshold, and finally the resolution enhancement processing is carried out on the extracted target signal, so that the super-resolution SAR image is finally obtained.

In order to achieve the technical purpose, the invention is realized by adopting the following technical scheme.

A method for extracting a target signal and enhancing the super-resolution under a strong clutter background is characterized by comprising the following steps:

step 1, sequentially carrying out motion compensation and conventional linear frequency modulation label imaging processing on SAR (synthetic aperture radar) echo signals of a ground scene to obtain an imaging result of a p point targetObtaining an imaging result Img containing P point targets, wherein P ∈ {1,2, …, P } represents the total number of the point targets in the ground scene;representing the distance fast time, t_mIndicating an azimuth slow time;

step 2, extracting target rectangular region slice data D containing P point targets and ground clutter from an imaging result Img containing P point targets_aRespectively obtaining slice data D of the target rectangular region_aAnd the target rectangular region slice data D_aIs shown, wherein N ∈ [1, N [ ]]，m∈[1,M]，k∈[1,N]，h∈[1,M]And N denotes target rectangular region slice data D_aM represents target rectangular region slice data D_aNumber of azimuth sampling units of (D)_aRepresenting target rectangular region slice data comprising P point targets and ground clutter, wherein P represents the total number of the point targets in the ground scene;

step 3, slicing the target rectangular region into data D_aFour corner clutter rectangular regions of uniform size and Fourier basis function W_kh(N, m) are respectively subjected to correlation processing, and the self-adaptive clutter threshold is obtained through calculation, wherein N ∈ [1, N]，m∈[1,M]And N denotes target rectangular region slice data D_aM represents target rectangular region slice data D_aNumber of azimuth sampling units of (D)_aRepresenting target rectangular region slice data comprising P point targets and ground clutter, wherein P represents the total number of the point targets in the ground scene;

step 4, slicing the target rectangular area into data D_aAnd Fourier basis function W_kh(n, m) carrying out correlation processing, and extracting to obtain target rectangular region slice data D_aMedium and fourier basis function W_khThe two-dimensional position (k, h) of the scattering point corresponding to the maximum value of the (n, m) correlation^qCorresponding strong scattering point G (k, h)^q(ii) a Wherein,k∈[1,N]，h∈[1,M]and N denotes target rectangular region slice data D_aM represents target rectangular region slice data D_aNumber of azimuth sampling units of (D)_aRepresenting target rectangular region slice data comprising P point targets and ground clutter, wherein P represents the total number of the point targets in the ground scene, and q represents the iteration times;

step 5, adopting a gradient descent method to slice data D of the target rectangular region_aAnd Fourier basis function W_khThe two-dimensional position (k, h) of the scattering point corresponding to the maximum value of the (n, m) correlation^qCorresponding strong scattering point G (k, h)^qExtracting the adjacent weak scattering point region to obtain the strong scattering point meeting the condition of the gradient descent methodAdjacent weak scattering spot areaWherein N ∈ [1, N]，m∈[1,M]，k∈[1,N]，h∈[1,M]，N denotes target region slice data D_aM represents target region slice data D_aNumber of azimuth sampling units of (D)_aRepresenting target region slice data comprising P point targets and ground clutter, wherein P represents the total number of the point targets in the ground scene, and q represents the iteration times;

step 6, slicing the target rectangular region into data D_aSubtracting the strong scattering point satisfying the gradient descent method conditionAdjacent weak scattering spot areaObtaining slice data of the residual target rectangular regionAnd slicing the data for the remaining target rectangular regionUsing step 4 and step 5 to carry out iteration operation until the target rectangular region slice data is remainedAll scattering points in (A) and Fourier basis function W_khThe maximum value of the (n, m) correlation degree is lower than the self-adaptive clutter threshold, iteration is stopped, P strong scattering points and weak scattering areas adjacent to the P strong scattering points are obtained, and a complete SAR image G containing P point targets is further combined and formed

Wherein,representing two-dimensional positions of scattering points satisfying a gradient descent method conditionAll the corresponding scattering points are located at the same time,n denotes target rectangular region slice data D_aM represents target rectangular region slice data D_aNumber of azimuth sampling units of (D)_aRepresenting target rectangular region slice data including P point targets and ground clutter, P representing total number of point targets in ground scene, and P also representing target rectangular region slice data D_aThe total number of the strong scattering points in the (1) and q represents the iteration times;

step 7, performing super-resolution enhancement processing on a complete SAR image G containing P point targets by utilizing a regularization method to obtain a final super-resolution SAR imageWherein P represents the total number of point targets in the ground scene.

The invention has the beneficial effects that: extracting strong scattering points of a target signal by using the strong correlation between the target signal and a Fourier basis function, further extracting weak scattering points adjacent to the strong scattering points by using a gradient descent method in order to protect the weak scattering points, and extracting a complete target signal from a strong clutter by using iterative operation; in addition, the invention utilizes the improved regularization method to carry out resolution enhancement processing on the target, can enhance the target resolution, inhibit side lobes and noise, improve the SAR image contrast and realize the super-resolution SAR imaging of the target signal.

Drawings

The present invention will be described in further detail with reference to the accompanying drawings and specific embodiments.

FIG. 1 is a schematic flow chart of a method for extracting a target signal and enhancing super-resolution in a strong clutter background according to the present invention;

FIG. 2 is slice data D of a target rectangular region including four corner rectangular regions according to the present invention_aA schematic diagram of (a);

FIG. 3 is a schematic diagram of the method for extracting weak scattering points adjacent to a strong scattering point G (k, h) corresponding to a two-dimensional position of the scattering points by using a gradient descent method; wherein, a first weak scattering point of a weak scattering point region adjacent to a strong scattering point G (k, h) corresponding to the two-dimensional position of the scattering point is G (k, h-1), a second weak scattering point of a weak scattering point region adjacent to a strong scattering point G (k, h) corresponding to the two-dimensional position of the scattering point is G (k-1, h-1), a third weak scattering point of a weak scattering point region adjacent to a strong scattering point G (k, h) corresponding to the two-dimensional position of the scattering point is G (k-1, h), a fourth weak scattering point of a weak scattering point region adjacent to a strong scattering point G (k, h) corresponding to the two-dimensional position of the scattering point is G (k-1, h +1), a fifth weak scattering point of a weak scattering point region adjacent to a strong scattering point G (k, h) corresponding to the two-dimensional position of the scattering point is G (k, h +1), a strong scattering point G (k) corresponding to the two-dimensional position of the scattering point, h) a sixth weak scattering point of the adjacent weak scattering point region is G (k +1, h +1), a seventh weak scattering point of the weak scattering point region adjacent to the strong scattering point G (k, h) corresponding to the two-dimensional position of the scattering point is G (k +1, h), and an eighth weak scattering point of the weak scattering point region adjacent to the strong scattering point G (k, h) corresponding to the two-dimensional position of the scattering point is G (k +1, h-1);

FIG. 4 is a schematic view of a measured scene of a corner reflector target of the present invention;

fig. 5(a) is a schematic diagram of a corner reflector image in a strong clutter background, in which the horizontal axis represents azimuth direction and the unit is a sampling unit, the vertical axis represents distance direction and the unit is a sampling unit,

fig. 5(b) is a schematic diagram of a result obtained by performing corner reflector extraction on the corner reflector image in the strong clutter background shown in fig. 5(a) in combination with a target strong scattering point extraction criterion and a target region extraction criterion, wherein a horizontal axis represents an azimuth direction, and a unit is a sampling unit, and a vertical axis represents a distance direction, and a unit is a sampling unit;

fig. 6(a) is a schematic diagram showing the result of 8-fold interpolation after extracting the corner reflectors, in which the horizontal axis represents the azimuth direction in units of sampling units and the vertical axis represents the distance direction in units of sampling units,

FIG. 6(b) is a schematic diagram of the result of performing resolution enhancement processing on FIG. 6(a) by using the improved regularization method proposed by the present invention, wherein the horizontal axis represents the azimuth direction and the unit is a sampling unit, and the vertical axis represents the distance direction and the unit is a sampling unit;

FIG. 7(a) is a schematic cross-sectional view of the No. 1 corner reflector target and the No. 2 corner reflector target of FIG. 6(b) taken along the azimuth direction before and after being respectively super-resolved, wherein the horizontal axis represents the azimuth direction in units of azimuth sampling units, the vertical axis represents normalized amplitude in units of dB,

FIG. 7(b) is a schematic cross-sectional view of the corner reflector No. 3 target and the corner reflector No. 2 target of FIG. 6(b) taken along the distance direction before and after being respectively super-resolved, wherein the horizontal axis represents the distance direction, the unit is a distance sampling unit, and the vertical axis represents the normalized amplitude, the unit is dB;

fig. 8(a) is a schematic diagram of an image of a motor vehicle target in a strong clutter background, in which the horizontal axis represents azimuth, the unit is a sampling unit, the vertical axis represents distance, the unit is a sampling unit,

FIG. 8(b) is a diagram of the results of the motor vehicle target extraction performed on FIG. 8(a) in conjunction with the strong scattering point extraction criterion, wherein the horizontal axis represents the azimuth direction in units of sampling units, and the vertical axis represents the distance direction in units of sampling units;

fig. 9(a) is a schematic diagram showing the result of 12-fold interpolation after extracting the vehicle target, in which the horizontal axis represents the azimuth direction in units of sampling units, the vertical axis represents the distance direction in units of sampling units,

fig. 9(b) is a schematic diagram of the super-resolution enhancement processing result of fig. 9(a) by using the improved regularization method proposed by the present invention, wherein the horizontal axis represents the azimuth direction and the unit is a sampling unit, and the vertical axis represents the distance direction and the unit is a sampling unit.

Detailed Description

Referring to fig. 1, a schematic flow chart of a method for extracting a target signal and enhancing super-resolution in a strong clutter background according to the present invention is shown, where the method for extracting a target signal and enhancing super-resolution in a strong clutter background includes the following steps:

step 1, sequentially performing motion compensation and conventional line frequency modulation (CS) imaging processing on SAR (synthetic aperture radar) echo signals in a ground scene to obtain an imaging result of a p point targetFurther obtaining an imaging result Img containing the target with the P points; wherein,representing the distance fast time, t_mIndicating azimuth slow time, P ∈ {1,2, …, P }, where P indicates the total number of point objects in the ground scene.

Specifically, the SAR radar echo signals in the ground scene are sequentially subjected to motion compensation and conventional line frequency modulation (CS) imaging processing to obtain the imaging result of the p point targetImaging result of the p-th point targetThe expression is as follows:

wherein, △ f_rRepresents the frequency bandwidth of the signal after the distance pulse pressure,representing the distance fast time, t_mIndicating slow time of orientation, △ f_aRepresenting the Doppler bandwidth, σ_pRepresents the scattering center amplitude of the target at the p point, c represents the speed of light, lambda represents the radar center wavelength, R (t)_m) Representing the slope distance process of the SAR to the scene center of the p point target,

and isR_BRepresents the shortest slope distance, t, of the scene center of the p point target_mRepresenting azimuth slow time, sinc [. C]Representing the impulse response function, X_nAnd the x-axis coordinate of the P-th point target in the direction of the air route in the rectangular coordinate system is shown, V is the speed of the radar loader, P ∈ {1,2, …, P } is shown, and P is the total number of point targets in the ground scene.

Step 2, extracting target rectangular region slice data D containing P point targets and ground clutter from an imaging result Img containing P point targets_aRespectively obtaining slice data D of the target rectangular region_aAnd the target rectangular region slice data D_aIs shown, wherein N ∈ [1, N [ ]]，m∈[1,M]，k∈[1,N]，h∈[1,M]And N denotes target rectangular region slice data D_aM represents target rectangular region slice data D_aNumber of azimuth sampling units of (D)_aTarget rectangular region slice data comprising P point targets and ground clutter is represented, P representing the total number of point targets in the ground scene.

Specifically, in an imaging result Img including P point targets, target rectangular region slice data D including the P point targets and the ground clutter is extracted_aRespectively obtaining slice data D of the target rectangular region_aAnd the target rectangular region slice data D_aThe target rectangular region slice data D, the time domain expression G (k, h) of_aAnd the target rectangular region slice data D_aThe time domain expressions G (k, h) of (a) may be expressed as:

S(n,m)＝E(n,m)+C(n,m)

G(k,h)＝IFFT2{E(n,m)+C(n,m)}

where E (n, m) represents target rectangular region slice data D_aAll scattering points of the medium P point targetIn the form of a discrete form of (a),slice data D representing a rectangular region of interest_aThe middle P points are all the scattering points of the target,

and isD_aRepresenting target rectangular region slice data including P point targets and ground clutter, P representing the total number of point targets in the ground scene, P ∈ {1,2, …, P },represents the imaging result of the p-th point target, and C (n, m) represents target rectangular region slice data D_aA clutter signal in, andB_nmslice data D representing a rectangular region of interest_aThe amplitude of the clutter in (1) is,slice data D representing a rectangular region of interest_aN represents target rectangular region slice data D_aM represents target rectangular region slice data D_aThe IFFT2 {. denotes a two-dimensional inverse Fourier transform,representing the distance fast time, t_mIndicating a slow time to azimuth.

Step 3, slicing the target rectangular region into data D_aFour corner ground clutter rectangular regions of consistent size and Fourier basis function W_kh(N, m) are respectively subjected to correlation processing, and the self-adaptive clutter threshold is obtained through calculation, wherein N ∈ [1, N]，m∈[1,M]And N denotes target rectangular region slice data D_aM represents target rectangular region slice data D_aNumber of azimuth sampling units of (D)_aTarget rectangular region slice data comprising P point targets and ground clutter is represented, P representing the total number of point targets in the ground scene.

Specifically, referring to FIG. 2, the rectangular region is a rectangular region including four corners with the same sizeTarget region slice data D_aSchematic representation of (a). Typically, the four uniformly sized corner clutter rectangular regions are all 5% D_a～15％D_aThe corner ground clutter rectangular region is calculated to be too small when being too small, point targets or scattering points (including strong scattering points and weak scattering points) can be introduced when the corner ground clutter rectangular region is too large, and four corner ground clutter rectangular regions with the same size are selected by the method according to the calculation of a large amount of data, wherein the size of the four corner ground clutter rectangular regions is 10% D_a。

Slicing target region into data D_aFour corner clutter rectangular regions and Fourier basis function W_khAnd (n, m) respectively carrying out correlation processing, and calculating to obtain the self-adaptive clutter threshold.

Adaptive clutter threshold and Fourier basis function W_khThe expressions of (n, m) are respectively:

＝max{_i|_i＝max[Coh(D_i)]}

wherein, Coh (D)_i) Slice data D representing a rectangular region of interest_aThe ith corner clutter rectangular region D in the four corner clutter rectangular regions with the same size_iI ∈ {1,2,3,4}, andrepresenting the ith corner clutter rectangular region D_iOf the energy sum of (max [ ·)]Denotes the maximum of the evaluation function, N ∈ [1, N]，m∈[1,M]，k∈[1,N]，h∈[1,M]And N denotes target rectangular region slice data D_aM represents target rectangular region slice data D_aNumber of azimuth sampling units of (D)_aRepresenting target rectangular region slice data including P point targets and ground clutter, P representing the total number of point targets in the ground scene, W_kh(n, m) denotes the Fourier basis function, W_kh ^*(n, m) represents a Fourier basis function W_khConjugation of (n, m).

Step 4, slicing the target rectangular area into data D_aAnd Fourier basis function W_kh(n, m) carrying out correlation processing, and calculating to obtain target rectangular region slice data D_aMedium and fourier basis function W_khThe two-dimensional position (k, h) of the scattering point corresponding to the maximum value of the (n, m) correlation^qCorresponding strong scattering point G (k, h)^qAnd further extracting target rectangular region slice data D_aMedium and fourier basis function W_khStrong scattering point corresponding to maximum value of (n, m) correlationWherein, k ∈ [1, N]，h∈[1,M]And N denotes target rectangular region slice data D_aM represents target rectangular region slice data D_aNumber of azimuth sampling units of (D)_aAnd representing target rectangular region slice data comprising P point targets and ground clutter, wherein P represents the total number of the point targets in the ground scene, and q represents the iteration number.

The specific substeps of step 4 are:

4.1 slicing data D of a rectangular region of interest_aAnd Fourier basis function W_kh(N, m) to obtain a correlation coefficient Coh (k, h), wherein N ∈ [1, N ]]，m∈[1,M]，k∈[1,N]，h∈[1,M]And N denotes target rectangular region slice data D_aM represents target rectangular region slice data D_aNumber of azimuth sampling units of (D)_aTarget rectangular region slice data comprising P point targets and ground clutter is represented, P representing the total number of point targets in the ground scene.

Specifically, the target rectangular region slice data D is sliced by utilizing the characteristic that the point target signal has high correlation with Fourier base, and the ground clutter shows weaker correlation_aAnd Fourier basis function W_kh(n, m) after the correlation processing, a correlation coefficient Coh is obtained (k, h) expressed as:

wherein E is_SSlice data D representing a rectangular region of interest_aEnergy of, ands (n, m) represents target rectangular region slice data D_aN ∈ [1, N]，m∈[1,M]，k∈[1,N]，h∈[1,M]And N denotes target rectangular region slice data D_aM represents target rectangular region slice data D_aNumber of azimuth sampling units, W_kh(n, m) represents a Fourier basis function, andk∈[1,N]，h∈[1,M]，D_arepresenting target rectangular region slice data comprising P point targets and ground clutter, wherein P represents the total number of the point targets in the ground scene;

4.2 obtaining target rectangular region slice data D by calculation_aMedium and fourier basis function W_khThe two-dimensional position (k, h) of the scattering point corresponding to the maximum value of the (n, m) correlation^qWherein k ∈ [1, N]，h∈[1,M]And N denotes target rectangular region slice data D_aM represents target rectangular region slice data D_aNumber of azimuth sampling units of (D)_aRepresenting target rectangular region slice data comprising P point targets and ground clutter, wherein P represents the total number of the point targets in the ground scene, and q represents the iteration times;

4.3 calculating to obtain target rectangular region slice data D_aMedium and fourier basis function W_khThe two-dimensional position (k, h) of the scattering point corresponding to the maximum value of the (n, m) correlation^qCorresponding strong scattering point G (k, h)^qAnd further extracting target rectangular region slice data D_aMedium and fourier basis functionsW_khStrong scattering point corresponding to maximum value of (n, m) correlationWherein, k ∈ [1, N]，h∈[1,M]And N denotes target rectangular region slice data D_aM represents target rectangular region slice data D_aNumber of azimuth sampling units of (D)_aAnd representing target rectangular region slice data comprising P point targets and ground clutter, wherein P represents the total number of the point targets in the ground scene, and q represents the iteration number.

Specifically, the extracted target rectangular region slice data D_aMedium and fourier basis function W_khStrong scattering point corresponding to maximum value of (n, m) correlationThe extraction criteria are:

s.t.max(Coh(k,h)^q)≥

wherein, G (k, h)^qSlice data D representing target region_aMedium and fourier basis function W_khThe two-dimensional position (k, h) of the scattering point corresponding to the maximum value of the (n, m) correlation^qCorresponding strong scatter points, representing adaptive clutter threshold, Coh (k, h)^qDenotes the correlation coefficient, D_aAnd representing target rectangular region slice data comprising P point targets and ground clutter, wherein P represents the total number of the point targets in the ground scene, and q represents the iteration number.

Step 5, adopting a gradient descent method to slice data D of the target rectangular region_aAnd Fourier basis function W_khStrong scattering point corresponding to maximum value of (n, m) correlationExtracting the adjacent weak scattering point region to obtainTo a strong scattering point satisfying the conditions of the gradient descent methodAdjacent weak scattering spot areaWherein N ∈ [1, N]，m∈[1,M]，k∈[1,N]，h∈[1,M]，N denotes target rectangular region slice data D_aM represents target rectangular region slice data D_aNumber of azimuth sampling units of (D)_aAnd representing target rectangular region slice data comprising P point targets and ground clutter, wherein P represents the total number of the point targets in the ground scene, and q represents the iteration number.

Specifically, referring to FIG. 3, a gradient descent method is used for extracting target rectangular region slice data D for the present invention_aMedium and fourier basis function W_khStrong scattering point corresponding to maximum value of (n, m) correlationA schematic diagram of adjacent weak scattering points, and a gradient descent method is adopted to slice data D of a target rectangular region_aMedium and fourier basis function W_khStrong scattering point corresponding to maximum value of (n, m) correlationExtracting the adjacent weak scattering point region to obtain the strong scattering point meeting the condition of the gradient descent methodAdjacent weak scattering spot areaThe expression is as follows:

wherein,representing the q iteration to extract strong scattering points satisfying the gradient descent method condition, grad (-) representing gradient operation,strong scattering points satisfying gradient descent method condition and extracted by representing q-th iterationOne of the surrounding eight adjacent weak scattering points, i is more than or equal to-1 and less than or equal to 1, j is more than or equal to-1 and less than or equal to 1, k ∈ [1, N ]]，h∈[1,M]，N denotes target rectangular region slice data D_aM represents target rectangular region slice data D_aNumber of azimuth sampling units of (D)_aAnd representing target rectangular region slice data comprising P point targets and ground clutter, wherein P represents the total number of the point targets in the ground scene, and q represents the iteration number.

Step 6, slicing the target rectangular region into data D_aSubtracting the strong scattering point satisfying the gradient descent method conditionAdjacent weak scattering spot areaObtaining slice data of the residual target rectangular regionAnd slicing the data for the remaining target rectangular regionUsing step 4 and step 5 to carry out iteration operation until the target rectangular region slice data is remainedAll scattering points in (A) and Fourier basis function W_khThe maximum value of the (n, m) correlation degree is lower than the self-adaptive clutter threshold, iteration is stopped, P strong scattering points and weak scattering areas adjacent to the P strong scattering points are obtained, and a complete SAR image G containing P point targets is further combined and formed

Wherein,representing two-dimensional positions of scattering points satisfying a gradient descent method conditionAll the corresponding scattering points are located at the same time,n denotes target rectangular region slice data D_aM represents target rectangular region slice data D_aNumber of azimuth sampling units of (D)_aRepresenting target rectangular region slice data including P point targets and ground clutter, P representing total number of point targets in ground scene, and P also representing target rectangular region slice data D_aAnd q represents the iteration number.

Step 7, performing super-resolution enhancement processing on a complete SAR image G containing P point targets by utilizing a regularization method to obtain a final super-resolution SAR imageWherein P represents the total number of point targets in the ground scene.

The specific substeps of step 7 are:

7.1 processing a complete SAR image G containing P point targets by using a regularization method to obtain a high-resolution SAR image to be reconstructedUtilizing the high-resolution SAR image to be reconstructedAnd the high resolution SAR image to be reconstructedThe gradient distribution sparsity of the image to be reconstructed is vectorized to the high-resolution SAR image to be reconstructedObtaining an optimization functionThe expression is as follows:

where Φ denotes the imaging operator, λ₁，λ₂Are all representative of the regularization parameters,representing a high resolution SAR image to be reconstructed, G representing a complete SAR image containing P point targets,the expression is given in the 2-norm,representing the k-norm.

Specifically, if the size of a complete SAR image G including P point targets is M × N, a high-resolution SAR image to be reconstructed is obtainedIs (M × N) × 1, the optimization functionThe first term in the expression is a data fidelity term that characterizes the minimization of the actual observation and the high resolution SAR image to be reconstructedThe square error between; the second term represents the sparse prior of the target, the sparse prior of the target is properly selected to be beneficial to inhibiting a pseudo target, the side lobe of the final super-resolution imaging result is reduced, and the resolution of a scattering point of the target can be protected and enhanced; the third term represents the sparse prior of the target edge, which is a smoothness penalty term, and the term is properly selected to reserve the strong scattering gradient (such as the image edge) of the final super-resolution imaging result, so as to maintain the target shape.

7.2 ignoring optimization functionThe third item in the expression is calculated to obtain an optimization function by utilizing a regularization methodFurther obtaining a final super-resolution SAR image

Specifically, the gradient descent method used by the invention is used for extracting and processing the point target under the strong clutter background, and the edge weak scattering point information of the point target can be retained, so that the shape of the point target can be maintained.

Using an improved regularization method, i.e. ignoring the optimisation functionThe third term in the expression is used for obtaining an optimized function imageObtaining a final super-resolution SAR image

In particular, the optimization function imageThe expression is as follows:

wherein phi represents an imaging operator, G represents a complete SAR image containing P point targets, and lambda represents₁A regularization parameter is represented as a function of,the expression is given in the 2-norm,representing the k-norm.

The effectiveness of the present invention was further verified by simulation experiments on the following measured data.

(I) point target measured data

Three triangular reflectors with the side length of 15 cm are placed in an actually measured scene, as shown in an actually measured scene schematic diagram of a corner reflector target shown in fig. 4, the distance between the corner reflectors is 0.94 m in the direction, the distance between the corner reflectors is 0.72 m in the direction, and basic parameters of an airborne SAR are shown in a table I.

Watch 1

Actually measuring 1, accurately extracting a corner reflector target in a strong clutter background by using the method of the invention, wherein the result is shown in fig. 5(a) and 5 (b); fig. 5(a) is a schematic diagram of a corner reflector image in a strong clutter background, where the horizontal axis represents an azimuth direction and the unit is a sampling unit, and the vertical axis represents a distance direction and the unit is a sampling unit; fig. 5(b) is a schematic diagram of a result obtained by performing corner reflector extraction on the corner reflector image in the strong clutter background shown in fig. 5(a) by combining a target strong scattering point extraction criterion and a target region extraction criterion, wherein the horizontal axis represents an azimuth direction, and the unit is a sampling unit, and the vertical axis represents a distance direction, and the unit is a sampling unit.

As can be seen from the two diagrams of fig. 5(a) and 5(b), the method of the present invention can successfully extract the corner reflector targets from the ground clutter.

Actually measuring 2, carrying out resolution enhancement processing on the extracted corner reflector by using the method of the invention, and the result is shown in fig. 6(a) and 6 (b); fig. 6(a) is a schematic diagram of a result obtained by performing 8-fold interpolation after extracting a corner reflector, wherein the horizontal axis represents an azimuth direction, the unit is a sampling unit, the vertical axis represents a distance direction, and the unit is a sampling unit; fig. 6(b) is a schematic diagram of the result of performing resolution enhancement processing on fig. 6(a) by using the improved regularization method proposed by the present invention, wherein the horizontal axis represents the azimuth direction and the unit is the sampling unit, and the vertical axis represents the distance direction and the unit is the sampling unit.

As is apparent from fig. 6(a) and 6(b), the method of the present invention has an effect of enhancing the resolution of the corner reflector, thereby achieving the purpose of super-resolution enhancement.

Actually measuring 3, respectively carrying out resolution comparison analysis on the corner reflectors before and after resolution enhancement treatment by using the method of the invention, wherein the results are shown in fig. 7(a) and fig. 7 (b); fig. 7(a) is a schematic cross-sectional view in the azimuth direction before and after performing super-resolution processing on the corner reflector No. 1 target and the corner reflector No. 2 target in fig. 6(b), respectively, in which a straight line indicates the schematic cross-sectional view in the azimuth direction resolution before the super-resolution processing on the corner reflector No. 1 target and the corner reflector No. 2 target, a dotted line indicates the schematic cross-sectional view in the azimuth direction resolution after the super-resolution processing on the corner reflector No. 1 target and the corner reflector No. 2 target, a horizontal axis indicates the azimuth direction, a unit is an azimuth sampling unit, and a vertical axis indicates a normalized amplitude, a unit is dB; fig. 7(b) is a schematic sectional view of the 3 rd corner reflector target and the 2 nd corner reflector target of fig. 6(b) in the distance direction before and after being super-resolved, wherein a straight line represents a sectional view of the 3 rd corner reflector target and the 2 nd corner reflector target in the distance direction before being super-resolved, a dotted line represents a sectional view of the 3 rd corner reflector target and the 2 nd corner reflector target in the distance direction after being super-resolved, a horizontal axis represents the distance direction, a unit is a distance sampling unit, and a vertical axis represents a normalized amplitude, and a unit is dB.

As is apparent from fig. 7(a) and 7(b), after the super-resolution enhancement processing is performed on the corner reflector target, the resolution of the corner reflector target is greatly reduced, and the measured data of the corner reflector target verifies the effectiveness of the present invention.

(II) measured data of target

A motor vehicle target is placed in an actual measurement scene, the distance length is 5.5 meters, the azimuth length is 3 meters, and basic radar parameters are shown in a table II.

Watch two

Actually measuring 1, accurately extracting the motor vehicle target under the strong clutter background by using the method, wherein the result is shown in fig. 8(a) and 8 (b); FIG. 8(a) is a schematic diagram of an image of a motor vehicle target in a strong clutter background, wherein the horizontal axis represents azimuth direction in units of sampling units, and the vertical axis represents distance direction in units of sampling units; fig. 8(b) is a schematic diagram of the result of extracting the motor vehicle target from fig. 8(a) by combining the strong scattering point extraction criterion, wherein the horizontal axis represents the azimuth direction, the unit is a sampling unit, and the vertical axis represents the distance direction, the unit is a sampling unit.

As can be seen from the two diagrams of fig. 8(a) and 8(b), the method of the present invention can successfully extract the motor vehicle target from the strongly clutter background.

Actually measuring 2, performing super-resolution enhancement processing on the extracted motor vehicle target by using the method disclosed by the invention, and obtaining results as shown in fig. 9(a) and 9 (b); fig. 9(a) is a schematic diagram of the result obtained by performing 12-fold interpolation after extracting the motor vehicle target, wherein the horizontal axis represents the azimuth direction and the unit is a sampling unit, and the vertical axis represents the distance direction and the unit is a sampling unit; fig. 9(b) is a schematic diagram of the super-resolution enhancement processing result of fig. 9(a) by using the improved regularization method proposed by the present invention, wherein the horizontal axis represents the azimuth direction and the unit is a sampling unit, and the vertical axis represents the distance direction and the unit is a sampling unit.

It is obvious from circles 1 and 2 in fig. 9(a) and 9(b), respectively, that the using method of the present invention has an enhancement effect on the resolution of the motor vehicle target, wherein taking circle 2 as an example, since the resolution of three point targets of the motor vehicle target is low before the super-resolution enhancement processing is performed on the motor vehicle target, the three point target specific display points are focused together and appear as a bright point; after the super-resolution enhancement processing, the three point targets appear as three bright points. Therefore, the super-resolution enhancement processing greatly improves the resolution of the point target in the image, more scattering information can be seen, and the contrast of the image and the positioning accuracy of the scattering center are effectively enhanced.

In conclusion, the simulation experiment verifies the correctness, the effectiveness and the reliability of the method.

It will be apparent to those skilled in the art that various changes and modifications may be made in the present invention without departing from the spirit and scope of the invention; thus, if such modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include such modifications and variations.

Claims (9)

1. A method for extracting a target signal and enhancing the super-resolution under a strong clutter background is characterized by comprising the following steps:

step 1, sequentially carrying out motion compensation and conventional linear frequency modulation label imaging processing on SAR (synthetic aperture radar) echo signals of a ground scene to obtain an imaging result of a p point targetAnd obtaining an imaging result Img containing P point targets, wherein P ∈ {1,2, …, P } represents PThe total number of point targets in the ground scene;representing the distance fast time, t_mIndicating an azimuth slow time;

step 2, extracting target rectangular region slice data D containing P point targets and ground clutter from an imaging result Img containing P point targets_aRespectively obtaining slice data D of the target rectangular region_aAnd the target rectangular region slice data D_aIs shown, wherein N ∈ [1, N [ ]]，m∈[1,M]，k∈[1,N]，h∈[1,M]And N denotes target rectangular region slice data D_aM represents target rectangular region slice data D_aNumber of azimuth sampling units of (D)_aRepresenting target rectangular region slice data comprising P point targets and ground clutter, wherein P represents the total number of the point targets in the ground scene;

step 3, slicing the target rectangular region into data D_aFour corner clutter rectangular regions of uniform size and Fourier basis function W_kh(N, m) are respectively subjected to correlation processing, and the self-adaptive clutter threshold is obtained through calculation, wherein N ∈ [1, N]，m∈[1,M]And N denotes target rectangular region slice data D_aM represents target rectangular region slice data D_aNumber of azimuth sampling units of (D)_aRepresenting target rectangular region slice data comprising P point targets and ground clutter, wherein P represents the total number of the point targets in the ground scene;

step 4, slicing the target rectangular area into data D_aAnd Fourier basis function W_kh(n, m) carrying out correlation processing, and extracting to obtain target rectangular region slice data D_aMedium and fourier basis function W_khThe two-dimensional position (k, h) of the scattering point corresponding to the maximum value of the (n, m) correlation^qCorresponding strong scattering point G (k, h)^qWherein k ∈ [1, N]，h∈[1,M]And N denotes target rectangular region slice data D_aM represents target rectangular region slice data D_aAzimuth sampling unit ofNumber, D_aRepresenting target rectangular region slice data comprising P point targets and ground clutter, wherein P represents the total number of the point targets in the ground scene, and q represents the iteration times;

step 5, adopting a gradient descent method to slice data D of the target rectangular region_aAnd Fourier basis function W_khThe two-dimensional position (k, h) of the scattering point corresponding to the maximum value of the (n, m) correlation^qCorresponding strong scattering point G (k, h)^qExtracting the adjacent weak scattering point region to obtain the strong scattering point meeting the condition of the gradient descent methodAdjacent weak scattering spot areaWherein N ∈ [1, N]，m∈[1,M]，k∈[1,N]，h∈[1,M]，N denotes target region slice data D_aM represents target region slice data D_aNumber of azimuth sampling units of (D)_aRepresenting target region slice data comprising P point targets and ground clutter, wherein P represents the total number of the point targets in the ground scene, and q represents the iteration times;

step 6, slicing the target rectangular region into data D_aSubtracting the strong scattering point satisfying the gradient descent method conditionAdjacent weak scattering spot areaObtaining slice data of the residual target rectangular regionAnd slicing the data for the remaining target rectangular regionUsing step 4 and step 5 to carry out iteration operation until the target rectangular region slice data is remainedAll scattering points in (A) and Fourier basis function W_khThe maximum value of the (n, m) correlation degree is lower than the self-adaptive clutter threshold, iteration is stopped, P strong scattering points and weak scattering areas adjacent to the P strong scattering points are obtained, and a complete SAR image G containing P point targets is further combined and formed

Wherein,representing two-dimensional positions of scattering points satisfying a gradient descent method conditionAll the corresponding scattering points are located at the same time,n denotes target rectangular region slice data D_aM represents target rectangular region slice data D_aNumber of azimuth sampling units of (D)_aRepresenting target rectangular region slice data including P point targets and ground clutter, P representing total number of point targets in ground scene, and P also representing target rectangular region slice data D_aThe total number of the strong scattering points in the (1) and q represents the iteration times;

step 7, performing super-resolution enhancement processing on a complete SAR image G containing P point targets by utilizing a regularization method to obtain a final super-resolution SAR imageWhere P represents a point in a ground sceneThe total number of targets.

2. The method for extracting target signal and enhancing super-resolution as claimed in claim 1, wherein in step 1, the imaging result of the p-th point target is obtainedImaging result of the p-th point targetThe expression is as follows:

$s_p(\hat{t},t_m)={\mathrm{\σ}}_p\sin c\[{\mathrm{\Δf}}_r(\hat{t}-2R(t_m)/c)\]\sin c\[{\mathrm{\Δf}}_a(t_m-X_n/V)\]$

wherein, △ f_rRepresents the frequency bandwidth of the signal after the distance pulse pressure,representing the distance fast time, t_mIndicating slow time of orientation, △ f_aRepresenting the Doppler bandwidth, σ_pRepresents the scattering center amplitude of the target at the p-th point, c represents the speed of light, R (t)_m) Representing the slope distance process of the SAR to the scene center of the p point target,

and isR_BRepresents the shortest slope distance, t, of the scene center of the p point target_mRepresenting azimuth slow time, sinc [. C]Representing the impulse response function, X_nAnd the x-axis coordinate of the P-th point target in the direction of the air route in the rectangular coordinate system is shown, V is the speed of the radar loader, P ∈ {1,2, …, P } is shown, and P is the total number of point targets in the ground scene.

3. The method as claimed in claim 1, wherein in step 2, the target rectangular region slice data D is obtained_aAnd the target rectangular region slice data D_aThe time domain expressions G (k, h) of (a) are respectively:

S(n,m)＝E(n,m)+C(n,m)

G(k,h)＝IFFT2{E(n,m)+C(n,m)}

where E (n, m) represents target rectangular region slice data D_aAll scattering points of the midpoint targetIn the form of a discrete form of (a),slice data D representing a rectangular region of interest_aTarget all scattering points, anP represents the total number of point objects in the ground scene, P ∈ {1,2, …, P },represents the imaging result of the p-th point target, and C (n, m) represents the rectangular region slice data D of the target_aA clutter signal in, andB_nmslice data D representing the target rectangular region_aThe amplitude of the clutter in (1) is,slice data D representing the target rectangular region_aN represents target rectangular region slice data D_aM represents target rectangular region slice data D_aNumber of azimuth sampling units of (D)_aRepresenting target rectangular region slice data comprising P point targets and ground clutter, P representing the total number of point targets in the ground scene, IFFT2 {. cndot.) representing a two-dimensional inverse Fourier transform,representing the distance fast time, t_mIndicating a slow time to azimuth.

4. The method as claimed in claim 1, wherein in step 3, the fourier basis function W is used to extract the target signal under a strong clutter background and enhance super resolution_kh(n, m) represented by:

$W_{kh}(n,m)=\exp \{-j2\mathrm{\π}(\frac{nk}{N}+\frac{mh}{M})\}$

wherein N ∈ [1, N]，m∈[1,M]，k∈[1,N]，h∈[1,M]And N denotes target rectangular region slice data D_aM represents target rectangular region slice data D_aNumber of azimuth sampling units of (D)_aTarget rectangular region slice data comprising P point targets and ground clutter is represented, P representing the total number of point targets in the ground scene.

5. The method as claimed in claim 1, wherein in step 3, the four corner clutter rectangular areas with the same size are all 10% D_a(ii) a Wherein D is_aRepresenting target rectangular region slice data including P point targets and ground clutter.

6. The method for extracting and super-resolution enhancement of target signal under strong clutter background according to claim 1, wherein in step 3, said adaptive clutter threshold is expressed as:

＝max{_i|_i＝max[Coh(D_i)]}

wherein, Coh (D)_i) Slice data D representing a rectangular region of interest_aThe ith corner clutter rectangular region D in the four corner clutter rectangular regions with the same size_iI ∈ {1,2,3,4}, and

representing the ith corner clutter rectangular region D_iOf the energy sum of (max [ ·)]Denotes the maximum of the evaluation function, N ∈ [1, N]，m∈[1,M]，k∈[1,N]，h∈[1,M]And N denotes target rectangular region slice data D_aM represents target rectangular region slice data D_aNumber of azimuth sampling units of (D)_aRepresenting target rectangular region slice data including P point targets and ground clutter, P representing the total number of point targets in the ground scene, W_kh(n, m) denotes the Fourier basis function, W_kh ^*(n, m) represents a Fourier basis function W_khConjugation of (n, m).

7. The method for extracting and super-resolution enhancement of target signal under strong clutter background of claim 1, wherein in step 4, the target rectangular region slice data D_aMedium and fourier basis function W_khThe two-dimensional position (k, h) of the scattering point corresponding to the maximum value of the (n, m) correlation^qCorresponding strong scattering point G (k, h)^qObtaining target region slice data D_aMedium and fourier basis function W_khThe two-dimensional position (k, h) of the scattering point corresponding to the maximum value of the (n, m) correlation^qCorresponding strong scattering point G (k, h)^qThe substeps of (A) are:

7.1 slicing data D of a rectangular region of interest_aAnd Fourier basis function W_kh(N, m) to obtain a correlation coefficient Coh (k, h), wherein N ∈ [1, N ]]，m∈[1,M]，k∈[1,N]，h∈[1,M]And N denotes target rectangular region slice data D_aM represents target rectangular region slice data D_aNumber of azimuth sampling units of (D)_aRepresenting target rectangular slice data comprising P point targets and ground clutter, wherein P represents the total number of the point targets in the ground scene;

7.2 calculation to obtain target rectangular slice data D_aMedium and fourier basis function W_kh(n, m) phaseTwo-dimensional position (k, h) of scattering point corresponding to correlation maximum^qWherein k ∈ [1, N]，h∈[1,M]And N denotes target rectangular slice data D_aM represents target rectangular slice data D_aNumber of azimuth sampling units of (D)_aRepresenting target rectangular slice data comprising P point targets and ground clutter, wherein P represents the total number of the point targets in the ground scene, and q represents the iteration times;

7.3 obtaining target rectangular slice data D_aMedium and fourier basis function W_khThe two-dimensional position (k, h) of the scattering point corresponding to the maximum value of the (n, m) correlation^qCorresponding strong scattering point G (k, h)^qExtracting the strong scattering point G (k, h)^qStrong scattering point ofWherein, k ∈ [1, N]，h∈[1,M]And N denotes target rectangular slice data D_aM represents target rectangular slice data D_aNumber of azimuth sampling units of (D)_aTarget rectangular slice data comprising P point targets and ground clutter is represented, P represents the total number of point targets in the ground scene, and q represents the number of iterations.

8. The method for extracting and super-resolution enhancement of target signal under strong clutter background of claim 1, wherein in step 5, the strong scattering point satisfying the gradient descent method conditionAdjacent weak scattering spot areaThe expression is as follows:

$G_{ext}^q(\tilde{k},\tilde{h})=G_{ext}^q(k,h)+\ΣG_{ext}^q(k+i,h+j)$

$\begin{array}{cc}st.& grad\left(G_{ext}^q\right(k,h\left)\right)>grad\left(G_{ext}^q\right(k+i,h+j\left)\right);(-1\≤i\≤1,-1\≤j\≤1)\end{array}$

wherein,representing the q iteration to extract strong scattering points satisfying the gradient descent method condition, grad (-) representing gradient operation,strong scattering points satisfying gradient descent method condition and extracted by representing q-th iterationOne of the surrounding eight adjacent weak scattering points, i is more than or equal to-1 and less than or equal to 1, j is more than or equal to-1 and less than or equal to 1, k ∈ [1, N ]]，h∈[1,M]，N denotes target rectangular region slice data D_aM represents target rectangular region slice data D_aNumber of azimuth sampling units of (D)_aIndicating bagAnd P point targets and target rectangular region slice data of the ground clutter, wherein P represents the total number of the point targets in the ground scene, and q represents the iteration times.

9. The method of claim 1, wherein in step 7, the final super-resolution image is processedObtaining the final super-resolution imageThe substeps of (A) are:

9.1 processing a complete SAR image G containing P point targets by using a regularization method to obtain a high-resolution SAR image to be reconstructedUtilizing the high-resolution SAR image to be reconstructedAnd the high resolution SAR image to be reconstructedThe gradient distribution sparsity of the image to be reconstructed is vectorized to the high-resolution SAR image to be reconstructedObtaining an optimization functionThe expression is as follows:

$\min J\left(\stackrel{\‾}{G}\right)=\left|\right|G-\mathrm{\Φ}\stackrel{\‾}{G}|{|}_2^2+{\mathrm{\λ}}_1|\left|\stackrel{\‾}{G}\right|{|}_k^k+{\mathrm{\λ}}_2\left|\right|D\left|\stackrel{\‾}{G}\right||{|}_k^k$

where Φ denotes the imaging operator, λ₁，λ₂Are all representative of the regularization parameters,representing a high resolution SAR image to be reconstructed, G representing a complete SAR image containing P point targets,the expression is given in the 2-norm,represents a k-norm;

9.2 ignoring optimization functionThe third item in the expression is calculated to obtain an optimization function by utilizing a regularization methodFurther obtaining a final super-resolution SAR image