A double-layer iterative image domain target three-dimensional scattering center position extraction method

Abstract

The application discloses a kind of double-layer iteration's image domain target three-dimensional scattering center position extraction method, this method contains: S1, target scattering is modeled and is carried out first layer iteration respectively to obtain the scattering center position of target frequency domain scattering distance direction, azimuth and pitch direction one-dimensional imaging;S2, the data of S1 is combined as target three-dimensional scattering center candidate position, based on target three-dimensional scattering center candidate position carries out second layer iteration and obtains target three-dimensional scattering center position.Its advantages are: this method is based on the scattering center position of target frequency domain scattering distance direction, azimuth and pitch direction one-dimensional imaging target three-dimensional scattering center candidate position, and then based on target three-dimensional scattering center candidate position carries out second layer iteration and obtains target three-dimensional scattering center position, both realize the extraction precision of target scattering center position, and significantly reduce the storage of target overall three-dimensional imaging, improve scattering center position estimation efficiency.

Description

Technical Field

[0001] This invention relates to the field of radar signal feature information processing, specifically to a method for extracting the three-dimensional scattering center position of a target in the image domain through a two-layer iterative process, and further to extracting the scattering center position in the image domain using a two-layer iterative CLEAN algorithm based on a greedy strategy. Background Technology

[0002] Radar imaging features depend on the electromagnetic scattering characteristics of the target. Electromagnetic scattering of electrically large targets is mainly contributed by some local scattering points, resulting in discrete or distributed scattering centers in high-resolution radar imaging. The electromagnetic scattering of complex targets can be modeled as a combination of multiple scattering centers. The scattering center model can provide a sparse representation of the target's scattering features. By simply storing the scattering center model parameters, features such as the target's radar cross section, one-dimensional range profile, and two-dimensional / three-dimensional radar imaging can be quickly reconstructed. However, in existing technologies, the accuracy of extracting the three-dimensional scattering center position from the electromagnetic scattering of complex targets is relatively low. Modifying the result to improve accuracy would significantly increase the storage requirements of the imaging results, severely reducing the efficiency of estimating the scattering center position of such complex targets. Summary of the Invention

[0003] The purpose of this invention is to provide a two-layer iterative method for extracting the three-dimensional scattering center position of a target in the image domain. This method models the target scattering and performs a first-layer iteration to obtain the scattering center positions of the target's scattering in the range, azimuth, and elevation directions in the frequency domain. Based on this data, candidate positions for the target's three-dimensional scattering center are obtained. Then, a second-layer iteration is performed based on these candidate positions to obtain the final three-dimensional scattering center position. This method achieves high accuracy in extracting the target's scattering center position while significantly reducing the storage requirements of the overall three-dimensional image of the target and improving the efficiency of scattering center position estimation.

[0004] To achieve the above objectives, the present invention is implemented through the following technical solution:

[0005] A two-layer iterative method for extracting the location of the three-dimensional scattering center of a target in the image domain, comprising:

[0006] S1. Model the target scattering and perform the first layer iteration to obtain the scattering center positions of the one-dimensional imaging in the range, azimuth and elevation directions of the target frequency domain scattering.

[0007] S2. Combine the data from S1 to form candidate positions for the target's three-dimensional scattering center. Based on these candidate positions, perform a second iteration to obtain the target's three-dimensional scattering center position.

[0008] Optionally, S1 includes:

[0009] S11. The target scattering is modeled using the point scattering center, resulting in the three-dimensional scattering center imaging model s(x) of the target. l ,y m ,z n ), where x l For distance-oriented variables, y m z is the azimuth variable. n For pitch direction;

[0010] S12, Based on the target three-dimensional scattering center imaging model s(x) l ,ym,z n The scattering center positions of the target's frequency domain scattering in the range, azimuth, and elevation directions are extracted respectively.

[0011] Optionally, in S11:

[0012] When the radar performs small-angle observations in the azimuth and elevation directions along the positive x-axis, a three-dimensional frequency domain scattering center model of the target is adopted, modeled using a point scattering center. for:

[0013]

[0014] Where k is the space wavenumber and θ is the pitch angle. Let a be the azimuth angle. t Let x be the complex amplitude at the scattering center, j be an imaginary number, and x be an imaginary number. t ,yt,z t Let be the coordinates of the scattering center along the x, y, and z axes, respectively, where t = 1, 2, ..., T, and T is the number of scattering centers, k. x ,k y ,k z Let k be the space wavenumber in the x, y, and z axes, respectively. Based on the small-angle approximations of azimuth and elevation angles, we can obtain k. x ≈k, k z ≈kθ;

[0015] Based on formula (1), the one-dimensional range-oriented frequency domain scattering sampling data S(k x,i This can be represented as:

[0016]

[0017] Among them, a x,t T is the complex amplitude of the distance-to-scattering center. x N represents the number of distance-oriented scattering centers. x Indicates the number of frequency domain sampling points in the distance direction;

[0018] For the one-dimensional range-direction frequency domain scattering sampling data S(k) of formula (2) x,iPerforming a discrete inverse Fourier transform yields the range-direction one-dimensional scattering center imaging model s(x). l ):

[0019]

[0020] Distance spread function h(x) to one-dimensional imaging point l -x t )for:

[0021]

[0022] Where, k x,0 x is the initial wave value for frequency domain sampling. l Represents the range-oriented imaging sampling points, l = 1, ..., L, W x Indicates the imaging width in the range direction;

[0023] Similarly, the azimuth-oriented one-dimensional scattering center imaging model s(y m )for:

[0024]

[0025] Azimuth spread function h(y) to one-dimensional imaging point m -y t )for:

[0026]

[0027] Among them, T y a is the number of azimuth scattering centers. y,t k is the complex amplitude of the azimuth scattering center. y,0 For the initial wavenumber of the azimuth frequency domain sampling, y m This represents the azimuth imaging sampling points, m = 1, ..., M, N y W represents the number of sampling points in the azimuth frequency domain. y Indicates the azimuth imaging width;

[0028] Pitch-to-one scattering center imaging model s(z) n )for:

[0029]

[0030] pitch-to-one-dimensional imaging point spread function h(z) n -z t )for:

[0031]

[0032] Among them, T z a is the number of pitch scattering centers. z,t k is the complex amplitude of the pitch-to-scattering center.z,0 To sample the initial wavenumber in the pitch frequency domain, z n This represents the elevation imaging sampling points, n = 1, ..., N, N z W represents the number of frequency domain sampling points in the pitch direction. z Indicates the width of the image in the pitch direction;

[0033] Based on the range-directed one-dimensional scattering center imaging model s(x) l ), azimuth-oriented one-dimensional scattering center imaging model s(y m ) and pitch-to-one scattering center imaging model s(z n ), thus obtaining the target three-dimensional scattering center imaging model s(x) l ,y m ,z n )for:

[0034]

[0035] Where, x l ,y m ,z n For 3D imaging position sampling, k x,0 ,k y,0 ,k z,0 These represent the initial wavenumbers for frequency domain sampling in the range, azimuth, and elevation directions, respectively; a t x is the complex amplitude of the three-dimensional scattering center. t ,yt,z t These are the coordinates of the scattering center along the x, y, and z axes, respectively.

[0036] Based on the distance-to-one-dimensional imaging point spread function h(x) l -x t ), azimuth-directed one-dimensional imaging point spread function h(y) m -y t ) and pitch-to-one imaging point spread function h(z) n -z t ), thus obtaining the three-dimensional imaging point spread function h(x) l ,y m ,z n )for:

[0037]

[0038] Among them, W x W y W z N represents the imaging width in the range, azimuth, and elevation directions, respectively. x N y N z These represent the number of frequency domain scattering samples in the range, azimuth, and elevation directions, respectively.

[0039] Optionally, in S12, a one-dimensional CLEAN algorithm based on a greedy strategy is used for extraction.

[0040] The extraction of the scattering center location in range-oriented one-dimensional imaging includes:

[0041] The one-dimensional imaging of the initial range-to-frequency scattering data is I0(x) l Let the iteration number t = 1, and select one-dimensional imaging I in the image domain. t-1 (x l The point with the maximum squared amplitude is taken as the location of the t-th scattering center.

[0042]

[0043] Calculate the position of the t-th scattering center One-dimensional imaging point spread function

[0044]

[0045] According to one-dimensional imaging in the image domain t-1 (x l Estimate the location of the t-th scattering center. Scattering center amplitude at the location

[0046]

[0047]

[0048] I t-1 (x)=[I t-1 (x1),...,I t-1 (x L )] T (15)

[0049] Where H represents the conjugate transpose and T represents the transpose;

[0050] Define the cutoff threshold for one-dimensional scattering center amplitude extraction as ε1, when At that time, from one-dimensional imaging I t-1 (x l Subtract the estimated scattering center distance from the image.

[0051]

[0052] And let t = t + 1, continue to extract the one-dimensional image I using formula (11) - formula (15). t (x l The scattering center of )

[0053] when At this point, the CLEAN algorithm terminates, completing the extraction of the scattering center for one-dimensional imaging, and obtaining the location of the scattering center for the range-oriented one-dimensional imaging. Among them, T x The number of one-dimensional scattering centers in the range direction;

[0054] Similarly, the location of the scattering center in the azimuth-oriented one-dimensional imaging is extracted as follows: The location of the scattering center in the pitch-to-one imaging is Among them, T y With T z These represent the number of one-dimensional scattering centers in the azimuth and elevation directions, respectively.

[0055] Optionally, after extracting a scattering center position at the same location, if the amplitude of the remaining residual scattering center at this location or in the nearby area still exceeds the one-dimensional scattering center amplitude extraction cutoff threshold ε1, the scattering center position at this location will continue to be extracted in the subsequent CLEAN algorithm. Then, the extracted scattering center positions will be merged and clustered to obtain a complex amplitude of a scattering center position, and then merged into a composite scattering center.

[0056] Optionally, S2 includes:

[0057] S21. Combine the scattering center positions of each one-dimensional image extracted in S1 into candidate positions of the target's three-dimensional scattering center.

[0058] S22. Perform three-dimensional imaging on each sub-region centered on the candidate position of the target's three-dimensional scattering center;

[0059] S23. Extract the amplitude and position of the target's three-dimensional scattering center from all sub-regions centered on the candidate positions of the target's three-dimensional scattering center after three-dimensional imaging.

[0060] Optionally, in S21,

[0061] By arranging and combining the scattering center positions from range-oriented one-dimensional imaging, azimuth-oriented one-dimensional imaging, and elevation-oriented one-dimensional imaging, candidate positions for the target's three-dimensional scattering center are obtained. Where T = T x ·T y ·T z .

[0062] Optionally, in S22,

[0063] For three-dimensional frequency domain scattering data with varying incident wave frequency, azimuth angle, and elevation angle, the candidate positions of the target's three-dimensional scattering center are determined. Three-dimensional imaging is performed on the sub-region centered on the scattering center; the imaging widths of the sub-regions centered on the scattering center in the range, azimuth, and elevation directions are D, respectively. x D y and D z The three-dimensional image of the t-th scattering center sub-region is obtained as I. t (x l ,y m ,z n ), where x l ,y m ,z n The range of variation are respectively and

[0064] Three-dimensional imaging is performed on the sub-region centered on all candidate locations of the target's three-dimensional scattering centers to obtain T three-dimensional images I1, I2, ..., I T .

[0065] Optionally, step S23 employs a three-dimensional CLEAN algorithm to extract the amplitude and location of the target's three-dimensional scattering center, specifically including:

[0066] The initial three-dimensional imaging set of the sub-region centered on the candidate location of the target's three-dimensional scattering center is Y0 = {I1, I2, ..., I...} T Let the number of iterations q = 1, and choose set Y. q-1 (x l ,x m ,z n The point with the maximum squared amplitude in the middle is used as the location estimate for the q-th scattering center.

[0067]

[0068] calculate The 3D point spread function for imaging the t-th sub-region at the location:

[0069]

[0070] Let I be the residual image of the t-th sub-region after q-1 iterations. t,q-1 ,estimate The amplitude of the scattering center is:

[0071]

[0072] Among them, I t,q-1 For residual image I t,q-1 A column vector composed of pixels Let be the column vector composed of the pixels imaged by the spread function of the t-th sub-region;

[0073] Set the cutoff threshold for the three-dimensional scattering center amplitude extraction to ε2, when At that time, the three-dimensional sub-region imaging set Y q-1 (x l ,x m ,z n For each sub-region in the image, subtract the extracted 3D scattering center image:

[0074]

[0075] Update the set of 3D imaging sub-regions as follows:

[0076] Y q (x l ,x m ,z n )={I 1,q (x l ,x m ,z n ),...,I T,q (x l ,x m ,z n )} (twenty one)

[0077] Let q = q + 1, and continue to extract the three-dimensional scattering center position of the target using formulas (17)-(19);

[0078] when At this point, the CLEAN algorithm for extracting the three-dimensional scattering center terminates, yielding the target's three-dimensional scattering center position parameters. Q represents the number of extracted three-dimensional scattering centers.

[0079] Optionally, in the CLEAN algorithm, let... For the scattering measurement of the t-th scattering center and the phase variance of the incoherent scattering, a t G is the complex amplitude of the scattering center in the frequency domain. t Let the amplitude of the scattering center in the image domain be denoted as , then we have

[0080]

[0081] Let the signal-to-noise ratio (SNR) of the frequency domain scattering measurement data be SNR. For scattering measurements with varying frequency, the variance of the phase measurement error is expressed as: in, To measure the wavenumber variance of the system's frequency stability, x t Estimate the location of the scattering center;

[0082] For scattering measurement data with angular variations, the variance of the phase measurement error with azimuth angular variations is expressed as: k c y is the wavenumber of the incident wave at its center frequency.t To estimate the location of the azimuth scattering center, The variance of the angle measurement;

[0083] The variance σ of the phase measurement error for scattering measurement data 2 This will lead to a decrease in main lobe gain. Consequently, the average sidelobe level of target imaging increases. Where K is the number of scattering measurement data samples;

[0084] The cutoff threshold ε, extracted from the scattering center in the image domain, is defined by the average sidelobe of the target imaging and the signal-to-noise ratio of the measurement data. t for

[0085]

[0086] Among them, B p The peak sidelobe gain is denoted by ε, the SNR is the power ratio of the scattered measurement signal to the noise, and the cutoff threshold ε is the power ratio of the peak sidelobe to the average sidelobe. t Extract a cutoff threshold ε1 for the amplitude of one-dimensional scattering center or a cutoff threshold ε2 for the amplitude of three-dimensional scattering center;

[0087] After extracting the locations of all image domain scattering centers, the amplitude of the image domain scattering centers is multiplied by an amplitude correction term to obtain the amplitude of the frequency domain scattering centers.

[0088]

[0089] Compared with the prior art, the present invention has the following advantages:

[0090] In this invention, a two-layer iterative method for extracting the three-dimensional scattering center position of a target in the image domain is presented. Point scattering center modeling is used for target scattering. The first iteration employs a one-dimensional CLEAN algorithm based on a greedy strategy to extract the scattering center positions of the target's scattering in the frequency domain, specifically in the range, azimuth, and elevation directions. The second iteration spatially arranges the one-dimensional scattering centers extracted in the first iteration into candidate three-dimensional scattering center positions, and performs three-dimensional imaging on the sub-regions containing these candidate scattering centers. Then, a three-dimensional CLEAN algorithm is used to extract the three-dimensional scattering center from the three-dimensional imaging of these sub-regions. This method achieves high accuracy in extracting the target scattering center position while significantly reducing the storage requirements of the overall three-dimensional target imaging, thus improving the efficiency of scattering center position estimation. Attached Figure Description

[0091] Figure 1 This is a schematic diagram of a two-layer iterative method for extracting the location of a target's three-dimensional scattering center in the image domain according to the present invention;

[0092] Figure 2 A schematic diagram of a small-angle observation of the target;

[0093] Figure 3 A three-dimensional geometric model of the vehicle target;

[0094] Figure 4 Extract the scattering center from the one-dimensional image of the vehicle target distance;

[0095] Figure 5 Extract the scattering center from the one-dimensional image of the vehicle target's orientation;

[0096] Figure 6 Extract the scattering center from the one-dimensional image of the vehicle target in pitch direction;

[0097] Figure 7 Extract convergence curves for the three-dimensional scattering centers of vehicle targets;

[0098] Figure 8 The three-dimensional scattering center distribution of the vehicle target;

[0099] Figures 9(a)-9(c) Reconstruct the radar cross section for the vehicle's three-dimensional scattering center. Detailed Implementation

[0100] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0101] Extracting the location of the target scattering center from frequency domain scattering data requires estimating the amplitude, location, and number of scattering centers. The location of the scattering center can be estimated through three-dimensional inverse Fourier transform imaging of the frequency domain data, selecting the location of the strongest scattering point in the image as the scattering center. To improve the accuracy of scattering center location estimation, zero-padding is required during inverse Fourier transform imaging of the frequency domain scattering data. However, this significantly increases the storage size of the imaging results. For example, performing five times zero-padding on three-dimensional frequency domain scattering data before inverse Fourier transform imaging results in an image size increase of 125 times, severely reducing the efficiency of scattering center estimation.

[0102] Therefore, under small-angle radar observation conditions, assuming that the amplitude, position, and number of target scattering centers remain constant, one-dimensional imaging of the scattering data in the range, azimuth, and elevation directions is calculated separately. The positions of the scattering centers in the range, azimuth, and elevation dimensions are extracted, and these three dimensions are combined to form candidate positions for the target's three-dimensional scattering centers. For example, if 10 one-dimensional scattering centers are extracted from each of the range, azimuth, and elevation dimensions, then 1000 candidate positions for the three-dimensional scattering centers are formed. Performing three-dimensional imaging only on the sub-regions containing these candidate positions significantly reduces the storage requirements for the overall three-dimensional imaging of the target.

[0103] Based on the above, such as Figure 1 As shown, this invention provides a method for extracting the location of the three-dimensional scattering center of a target in the image domain through a two-layer iterative process. This method includes:

[0104] S1. Model the target scattering and perform the first layer iteration to obtain the scattering center positions of the one-dimensional imaging in the range, azimuth and elevation directions of the target frequency domain scattering.

[0105] S2. Combine the data from S1 to form candidate positions for the target's three-dimensional scattering center. Based on these candidate positions, perform a second iteration to obtain the target's three-dimensional scattering center position.

[0106] In this invention, S1 includes:

[0107] S11. The target scattering is modeled using the point scattering center, resulting in the three-dimensional scattering center imaging model s(x) of the target. l ,y m ,z n ), where x l For the range-oriented imaging position variable, y m z is the azimuth imaging position variable. n For pitch imaging position variables;

[0108] S12, Based on the target three-dimensional scattering center imaging model s(x) l ,y m ,z n The scattering center positions of the target's frequency domain scattering in the range, azimuth, and elevation directions are extracted respectively.

[0109] Specifically, in S11, as Figure 2 As shown, when the radar performs small-angle observations in the azimuth and elevation directions along the positive x-axis, a three-dimensional frequency domain scattering center model of the target is adopted, modeled using a point scattering center. for:

[0110]

[0111] Where k is the space wavenumber and θ is the pitch angle. Let a be the azimuth angle. t Let x be the complex amplitude at the scattering center, j be an imaginary number, and x be an imaginary number. t ,y t ,z t Let k be the coordinates of the t-th scattering center along the x, y, and z axes, respectively, where t = 1, 2, ..., T, and T is the number of scattering centers. x ,k y ,k z Let k be the space wavenumber in the x, y, and z axes, respectively. Based on the small-angle approximations of azimuth and elevation angles, we can obtain k. x ≈k, k z ≈kθ. Here, the smaller angle is a variable, not a fixed value, and can be set according to the required imaging resolution and incident wave frequency. Optionally, the angle range of the smaller angle is less than 10°. Of course, it can also be other numerical angle ranges, which are not limited in this invention.

[0112] In this invention, when the radar observation azimuth and elevation angles are 0° and the radar incident wave frequency sweep changes, based on formula (1), the one-dimensional range-direction frequency domain scattering sampling data S(k) is obtained. x,i This can be represented as:

[0113]

[0114] Among them, a x,t T is the complex amplitude of the distance-to-scattering center. x N represents the number of distance-oriented scattering centers. x This represents the number of sampling points in the frequency domain along the distance.

[0115] For the one-dimensional range-direction frequency domain scattering sampling data S(k) of formula (2) x,i Performing a discrete inverse Fourier transform yields the range-direction one-dimensional scattering center imaging model s(x). l ):

[0116]

[0117] Distance spread function h(x) to one-dimensional imaging point l -x t )for:

[0118]

[0119] Where, k x,0 x is the initial wave value for frequency domain sampling. l Represents the range imaging sampling points, l = 1, ..., L, where L is the total number of range imaging sampling points, W x Indicates the imaging width in the range direction;

[0120] Similarly, when the radar center frequency is fixed, the incident elevation angle is 0°, and the radar observation azimuth changes, the target azimuth is reflected in the one-dimensional scattering center imaging model s(y). m )for:

[0121]

[0122] Azimuth spread function h(y) to one-dimensional imaging point m -y t )for:

[0123]

[0124] Among them, T y a is the number of azimuth scattering centers. y,t k is the complex amplitude of the azimuth scattering center. y,0 For the initial wavenumber of the azimuth frequency domain sampling, y m This represents the azimuth imaging sampling points, m = 1, ..., M, where M is the total number of azimuth imaging sampling points, and N... y W represents the number of sampling points in the azimuth frequency domain. y Indicates the azimuth imaging width.

[0125] Similarly, when the radar center frequency is fixed, the incident azimuth angle is 0°, and the radar observation elevation changes, the target elevation is reflected in the one-dimensional scattering center imaging model s(z). n )for:

[0126]

[0127] pitch-to-one-dimensional imaging point spread function h(z) n -z t )for:

[0128]

[0129] Among them, T z a is the number of pitch scattering centers. z,t k is the complex amplitude of the pitch-to-scattering center. z,0 To sample the initial wavenumber in the pitch frequency domain, z n This represents the elevation imaging sampling points, n = 1, ..., N, where N is the total number of elevation imaging sampling points. z W represents the number of frequency domain sampling points in the pitch direction. z This indicates the width of the image in the pitch direction.

[0130] Based on the above one-dimensional scattering center imaging model s(x) l ), azimuth-oriented one-dimensional scattering center imaging model s(y m ) and pitch-to-one scattering center imaging model s(z nThe target three-dimensional scattering center imaging model s(xl,y) is obtained. m ,z n )for:

[0131]

[0132] Where, x l ,y m ,z n For 3D imaging position sampling, k x,0 ,k y,0 ,k z,0 These represent the initial wavenumbers for frequency domain sampling in the range, azimuth, and elevation directions, respectively; a t x is the complex amplitude of the three-dimensional scattering center. t ,y t ,z t These are the coordinates of the t-th scattering center along the x, y, and z axes, respectively.

[0133] Based on the distance-to-one-dimensional imaging point spread function h(x) l -x t ), azimuth-directed one-dimensional imaging point spread function h(y) m -y t ) and pitch-to-one imaging point spread function h(z) n -z t ), thus obtaining the three-dimensional imaging point spread function h(x) l ,y m ,z n )for:

[0134]

[0135] Among them, W x W y W z N represents the imaging width in the range, azimuth, and elevation directions, respectively. x N y N z These represent the number of frequency domain scattering samples in the range, azimuth, and elevation directions, respectively.

[0136] Furthermore, in S12, a one-dimensional CLEAN algorithm based on a greedy strategy is used for extraction. The CLEAN algorithm is an iterative algorithm based on a greedy strategy. In each iteration, it finds the location and amplitude of the strongest scattering point in the radar image and subtracts the scattering field of the strong scattering point from the frequency domain data, or subtracts the imaging spread function of the strong scattering point from the image domain, until all strong scattering points have been extracted and the residual data is reduced below the noise level.

[0137] In this embodiment, the one-dimensional CLEAN algorithm is used to extract the scattering center positions of the one-dimensional images in the range, azimuth, and elevation directions, respectively. Since the one-dimensional scattering center imaging models in the range, elevation, and azimuth directions are all the same (as shown in formulas (3), (5), and (7)), the extraction of the scattering center position of the range one-dimensional image using the one-dimensional CLEAN algorithm is described below. This method is also applicable to the extraction of the scattering center positions of the one-dimensional images in the azimuth and elevation directions.

[0138] Specifically, the extraction of the scattering center location in range-oriented one-dimensional imaging includes:

[0139] The one-dimensional imaging of the initial range-to-frequency scattering data is I0(x) l Let the iteration number t = 1, and select one-dimensional imaging I in the image domain. t-1 (x l The point with the maximum squared amplitude is taken as the location of the t-th scattering center. estimate:

[0140]

[0141] Calculate the position of the t-th scattering center One-dimensional imaging point spread function

[0142]

[0143] According to one-dimensional imaging in the image domain t-1 (x l Estimate the location of the t-th scattering center. Scattering center amplitude at the location

[0144]

[0145]

[0146] I t-1 (x)=[I t-1 (x1),...,I t-1 (x L )] T (15)

[0147] Where H represents the conjugate transpose and T represents the transpose.

[0148] Define the cutoff threshold for one-dimensional scattering center amplitude extraction as ε1, when At that time, from one-dimensional imaging I t-1 (x l Subtract the estimated scattering center distance from the image.

[0149]

[0150] And let t = t + 1, continue to extract the one-dimensional image I using formula (11) - formula (15). t (x l The scattering center of )

[0151] when At this point, the CLEAN algorithm terminates, completing the extraction of the scattering center for one-dimensional imaging, and obtaining the location of the scattering center for the range-oriented one-dimensional imaging. Among them, T x The number of one-dimensional scattering centers in the range direction.

[0152] Similarly, the scattering center position of the azimuth-oriented one-dimensional image extracted using the above one-dimensional CLEAN algorithm is: The location of the scattering center in the pitch-to-one imaging is Among them, T y With T z These represent the number of one-dimensional scattering centers in the azimuth and elevation directions, respectively.

[0153] In practical applications, during the extraction of scattering centers using the CLEAN algorithm, due to measurement errors in the scattering data and the influence of unstable coherent scattering, if the amplitude of the residual scattering center at or near the extracted location still exceeds the one-dimensional scattering center amplitude extraction cutoff threshold ε1, subsequent CLEAN algorithms will continue to extract scattering center locations at that location or in the vicinity. This is equivalent to extracting more than two scattering centers at the same location or within a single resolution cell. Therefore, it is necessary to merge and cluster the extracted scattering center locations to obtain a complex amplitude of a single scattering center location, which is then merged into a composite scattering center. In subsequent applications, this composite scattering center is used as the scattering center location for that iteration.

[0154] For example, suppose M scattering center locations are extracted within a resolution cell, with amplitudes of respectively... The locations of the scattering centers are respectively Clustering them into a complex amplitude of a scattering center has the following value:

[0155]

[0156] Merging into a single composite scattering center

[0157]

[0158] In this embodiment, S2 includes:

[0159] S21. Combine the scattering center positions of each one-dimensional image extracted in S1 into candidate positions of the target's three-dimensional scattering center.

[0160] Specifically, in step S21, the scattering center positions of the range-oriented one-dimensional imaging, the scattering center positions of the azimuth-oriented one-dimensional imaging, and the scattering center positions of the pitch-oriented one-dimensional imaging are arranged and combined to obtain the candidate positions of the target's three-dimensional scattering center. Where T = T x ·T y ·T z .

[0161] S22. Perform three-dimensional imaging on each sub-region centered on the candidate position of the target's three-dimensional scattering center.

[0162] Specifically, in S22, for the three-dimensional frequency domain scattering data with varying radar incident wave frequency, azimuth angle, and elevation angle, the candidate positions of the target's three-dimensional scattering center are determined. Three-dimensional imaging is performed on the sub-region centered on the scattering center. The imaging widths of the sub-regions selected for the range, azimuth, and elevation directions are D, respectively. x D y and D z The three-dimensional image of the t-th scattering center sub-region is obtained as I. t (x l ,y m ,z n ), where x l ,y m ,z n The range of variation are respectively and Three-dimensional imaging is performed on the sub-region centered on all candidate locations of the target's three-dimensional scattering centers to obtain T three-dimensional images I1, I2, ..., I T .

[0163] S23. Extract the amplitude and position of the target's three-dimensional scattering center from all sub-regions centered on the candidate positions of the target's three-dimensional scattering center after three-dimensional imaging.

[0164] In this embodiment, step S23 employs a three-dimensional CLEAN algorithm to extract the amplitude and position of the target's three-dimensional scattering center. Specifically, it includes: initializing a three-dimensional imaging set of the sub-region centered on the candidate position of the target's three-dimensional scattering center as Y0 = {I1, I2, ..., I...}. T Let the number of iterations q = 1, and choose set Y. q-1 (x l ,x m ,z n The point with the maximum squared amplitude in the middle is used as the location estimate for the q-th scattering center.

[0165]

[0166] calculate The 3D point spread function for imaging the t-th sub-region at the location:

[0167]

[0168] Let I be the residual image of the t-th sub-region after q-1 iterations. t,q-1 ,estimate The amplitude of the scattering center is:

[0169]

[0170] Among them, I t,q-1 For residual image I t,q-1 A column vector composed of pixels Let be the column vector composed of the pixels of the spread function imaging of the t-th sub-region.

[0171] Set the cutoff threshold for the three-dimensional scattering center amplitude extraction to ε2, when At that time, the three-dimensional sub-region imaging set Y q-1 (x l ,x m ,z n For each sub-region in the image, subtract the extracted 3D scattering center image:

[0172]

[0173] Update the set of 3D imaging sub-regions as follows:

[0174] Y q (x l ,x m ,z n )={I 1,q (x l ,x m ,z n ),...,I T,q (x l ,x m ,z n )} (twenty one)

[0175] Let q = q + 1, and continue to extract the three-dimensional scattering center position of the target using formulas (17)-(19);

[0176] when At this point, the CLEAN algorithm for extracting the three-dimensional scattering center terminates, yielding the target's three-dimensional scattering center position parameters. Q represents the number of extracted three-dimensional scattering centers.

[0177] As mentioned above, the cutoff threshold for scattering center amplitude extraction is crucial in the CLEAN algorithm implementation. An effective cutoff threshold is needed, ensuring that all scattering centers are extracted while preventing the extraction of false targets such as signal sidelobes and noise as true scattering centers. Factors influencing the selection of the cutoff threshold mainly include measurement errors in the target scattering data and contributions from unstable coherent scattering. Measurement errors include measurement system frequency stability, observation angle deviation, and measurement noise, while unstable coherent scattering includes multipath scattering, medium dispersion, and cavity scattering. Scattering measurement errors and incoherent scattering introduce phase errors into the scattering data, leading to a decrease in the main lobe and an increase in the sidelobe level of the scattering center distance-image system response function. This results in the image domain scattering center amplitude extracted by the CLEAN algorithm being lower than the true scattering center amplitude.

[0178] Due to factors such as scattering measurement errors and incoherent scattering, the image amplitude decreases after frequency domain scattering measurement data is transformed to the image domain using inverse Fourier transform. In the CLEAN algorithm, let... For the scattering measurement of the t-th scattering center and the phase variance of the incoherent scattering, a t G is the complex amplitude of the scattering center in the frequency domain. t Let the amplitude of the scattering center in the image domain be denoted as , then we have

[0179]

[0180] Let the signal-to-noise ratio (SNR) of the frequency domain scattering measurement data be SNR. For scattering measurements with varying frequency, the variance of the phase measurement error is expressed as: in, To measure the wavenumber variance of the system's frequency stability, x t Estimate the location of the scattering center.

[0181] For scattering measurement data with angular variations, such as phase measurement error with azimuth angular variations, the variance is expressed as: k c y is the wavenumber of the incident wave at its center frequency. t To estimate the location of the azimuth scattering center, This represents the variance of the angle measurements. The calculation method for the correlation data of pitch angle changes is the same as that for the correlation data of azimuth angle changes.

[0182] The variance σ of the phase measurement error for scattering measurement data 2 This will lead to a decrease in main lobe gain. Consequently, the average sidelobe level of target imaging increases. Where K is the number of scattering measurement data samples.

[0183] To prevent false targets from being extracted as true scattering centers, a cutoff threshold ε for image domain scattering center extraction is defined based on the target imaging average sidelobes and the signal-to-noise ratio of the measurement data. t for

[0184]

[0185] Among them, B p ε is the gain of the peak sidelobe relative to the average sidelobe, SNR is the power ratio of the scattered measurement signal to the noise, and ε is the gain of the peak sidelobe relative to the average sidelobe. t Extract a cutoff threshold ε1 for the amplitude of one-dimensional scattering center or a cutoff threshold ε2 for the amplitude of three-dimensional scattering center.

[0186] After extracting the locations of all image domain scattering centers, the amplitude of the image domain scattering centers is multiplied by an amplitude correction term to obtain the amplitude of the frequency domain scattering centers.

[0187]

[0188] by Figure 3 Using the geometry of civilian vehicle targets as an application example, this paper introduces the implementation process of extracting the three-dimensional scattering center of a vehicle target using the image domain target three-dimensional scattering center extraction method of the present invention with two-layer iterative method. Figure 3 The three-dimensional geometric model of the vehicle target has a length, width, and height of 5.08m, 2.12m, and 1.56m, respectively, and a total of 35,000 triangular facets, each of which is set to a metallic material. The vehicle geometry is mainly contributed by single scattering. The physical optics method is used to simulate the vehicle's frequency domain scattering data, and the simulation parameter settings are shown in Table 1.

[0189]

[0190] Table 1 Electromagnetic Simulation Parameters for Vehicle Targets

[0191] The one-dimensional CLEAN algorithm was used to extract the scattering center positions of one-dimensional imaging in the range, azimuth, and elevation directions. 101 frequency domain sampling data points with azimuth and elevation angles of 90° and a frequency variation of 23GHz–25GHz were selected, zero-padding was performed at 10x, and then inverse Fourier transform imaging was performed to obtain the one-dimensional image I0(x) in the image domain. l The scattering center position of the range-direction one-dimensional imaging is extracted using the one-dimensional CLEAN algorithm based on formulas (11)-(15). The number of range-scattering centers T is obtained. x =7, the result is as follows Figure 4 As shown.

[0192] 101 frequency domain sampling data points with an incident wave frequency of 24 GHz, an elevation angle of 90°, and an azimuth angle variation of 87°–93° were selected. These data points were then zero-paddinged tenfold, and inverse Fourier transform imaging was performed to obtain the one-dimensional image domain image I0(y). m The scattering center position of the azimuth direction one-dimensional imaging is extracted using the one-dimensional CLEAN algorithm of formulas (11)-(15). The number of azimuth scattering centers T is obtained y =44, the result is as follows Figure 5 As shown.

[0193] Eighty-one frequency domain sampling data points with an incident wave frequency of 24 GHz, an azimuth angle of 90°, and an elevation angle variation of 87°–93° were selected. These data points were then zero-paddinged tenfold before inverse Fourier transform imaging was performed to obtain a one-dimensional image in the image domain, I0(z). n The scattering center position of the pitch-direction one-dimensional imaging is extracted using the one-dimensional CLEAN algorithm of formulas (11)-(15). The number of pitch scattering centers T is obtained. z =17, the result is as follows Figure 6 As shown.

[0194] The three-dimensional CLEAN algorithm is used to extract the three-dimensional scattering center positions of the vehicle target. Candidate three-dimensional scattering center positions are calculated by combining the scattering center positions from one-dimensional imaging in the range, azimuth, and pitch directions. Total T = 5236 (T = T x T y T z (Number) Select the imaging width D for the range, azimuth, and elevation sub-regions. x =2δ x D y =2δ y D z =2δ z , where δ x ,δ y ,δ z These represent the imaging resolutions in the range, azimuth, and elevation directions, respectively. For the three-dimensional frequency domain scattering data with varying radar incident wave frequency, azimuth angle, and elevation angle, a 10-fold zero-padding operation is performed on each. Three-dimensional inverse Fourier transform imaging is performed on the sub-region centered on the image to obtain the three-dimensional image I of the sub-region. t (x l ,y m ,z n ),in,

[0195] Initialize the target 3D sub-region imaging set as Y0 = {I1, I2, ..., I...} T The three-dimensional center of the target is extracted using the three-dimensional CLEAN algorithm based on formulas (17)-(19). The SNR of the frequency domain scattering data is set to 20 dB, and the gain B of the peak sidelobe relative to the average sidelobe is... p =2, and the cutoff threshold during the CLEAN algorithm iteration process is calculated using formula (23), thus obtaining the convergence curve of the three-dimensional scattering center amplitude intensity versus the cutoff threshold. Figure 7 As shown. When At that point, the CLEAN algorithm for extracting the three-dimensional scattering centers terminated, yielding a total of Q = 271 three-dimensional scattering centers. The target's three-dimensional scattering center parameters... Amplitude and location distribution as follows Figure 8 As shown.

[0196] The three-dimensional scattering center location extracted from the vehicle is used, as shown in Figures 9(a)-(c), to reconstruct the radar cross section (RCS) of the vehicle. Figure 9(a) reconstructs the RCS at 90° azimuth, 90° elevation, and a frequency of 23 GHz–25 GHz; Figure 9(b) reconstructs the RCS at 87°–93° azimuth, 90° elevation, and a frequency of 24 GHz; and Figure 9(c) reconstructs the RCS at 90° azimuth, 87°–93° elevation, and a frequency of 24 GHz, and compares these results with the RCS directly simulated for the vehicle.

[0197] The accuracy of the simulated RCS and the reconstructed RCS of the scattering center in Figures 9(a)-(c) was evaluated. Taking the median error of the reconstructed RCS of the scattering center, the RCS modeling accuracy in the frequency direction, azimuth direction, and elevation direction was obtained as 0.58dB, 2.32dB, and 1.56dB, respectively. The average of these three values ​​yielded an overall RCS modeling accuracy of 1.49dB. This invention uses vehicle target frequency domain simulation data as an application example. A two-layer iterative CLEAN algorithm was used to extract the three-dimensional scattering center of the vehicle in the lateral positive incident direction. Based on the three-dimensional scattering center, the radar scattering cross sections in the range, azimuth, and elevation directions were reconstructed. The effectiveness of the three-dimensional scattering center extraction algorithm of this invention was verified through comparison. As can be seen from the above, the two-layer iterative image domain target three-dimensional scattering center extraction method of this invention not only achieves high accuracy in extracting the target scattering center position but also significantly reduces the storage requirements of the overall three-dimensional target imaging and improves the efficiency of scattering center position estimation.

[0198] In summary, the present invention provides a two-layer iterative method for extracting the three-dimensional scattering center position of a target in the image domain. This method models the target scattering and performs a first-layer iteration to obtain the scattering center positions of the target's scattering in the range, azimuth, and elevation directions in the frequency domain. Based on this data, candidate positions for the target's three-dimensional scattering center are obtained, and then a second-layer iteration is performed based on these candidate positions to obtain the final three-dimensional scattering center position. This method achieves high accuracy in extracting the target scattering center position while significantly reducing the storage requirements of the overall three-dimensional target imaging, thus improving the efficiency of scattering center position estimation.

[0199] Furthermore, the method of this invention adopts point scattering center modeling for target scattering. The first iteration uses a one-dimensional CLEAN algorithm based on a greedy strategy to extract the scattering center positions of the target scattering in the range, azimuth, and elevation directions in the frequency domain. The second iteration arranges the one-dimensional scattering centers extracted in the first iteration into candidate three-dimensional scattering centers in space, and performs three-dimensional imaging on the candidate scattering center sub-regions. Then, the three-dimensional CLEAN algorithm is used to extract the three-dimensional scattering centers from the three-dimensional imaging of the sub-regions. This method uses the CLEAN algorithm to extract the scattering center positions, which significantly reduces the amount of data storage.

[0200] Although the present invention has been described in detail through the preferred embodiments above, it should be understood that the above description should not be considered as a limitation of the present invention. Various modifications and substitutions to the present invention will be apparent to those skilled in the art after reading the above description. Therefore, the scope of protection of the present invention should be defined by the appended claims.

Claims

1. A method for extracting the location of the three-dimensional scattering center of a target in the image domain through a two-layer iterative process, characterized in that, Include: S1. Model the target scattering and perform the first layer iteration to obtain the scattering center positions of the one-dimensional imaging in the range, azimuth and elevation directions of the target frequency domain scattering. S2. Combine the data from S1 to form candidate locations for the target's three-dimensional scattering center, and perform a second iteration based on these candidate locations to obtain the target's three-dimensional scattering center location. S1 includes: S11. The target scattering is modeled using the point scattering center, resulting in the three-dimensional scattering center imaging model s(x) of the target. l ,y m ,z n ), where x l For distance-oriented variables, y m z is the azimuth variable. n For pitch direction; S12, Based on the target three-dimensional scattering center imaging model s(x) l ,y m ,z n The positions of the scattering centers in the range, azimuth, and elevation directions of the target's frequency domain scattering are extracted respectively. In S11: When the radar performs small-angle observations in the azimuth and elevation directions along the positive x-axis, a three-dimensional frequency domain scattering center model of the target is adopted, modeled using a point scattering center. for: Where k is the space wavenumber and θ is the pitch angle. Let a be the azimuth angle. t Let x be the complex amplitude at the scattering center, j be an imaginary number, and x be an imaginary number. t ,y t ,z t Let be the coordinates of the scattering center along the x, y, and z axes, respectively, where t = 1, 2, ..., T, and T is the number of scattering centers, k. x ,k y ,k z Let k be the space wavenumber in the x, y, and z axes, respectively. Based on the small-angle approximations of azimuth and elevation angles, we can obtain k. x ≈k, k z ≈kθ; Based on formula (1), the one-dimensional range-oriented frequency domain scattering sampling data S(k x,i This can be represented as: Among them, a x,t T is the complex amplitude of the distance-to-scattering center. x N represents the number of scattering centers at a distance. x Indicates the number of frequency domain sampling points in the distance direction; For the one-dimensional range-direction frequency domain scattering sampling data S(k) of formula (2) x,i Performing a discrete inverse Fourier transform yields the range-direction one-dimensional scattering center imaging model s(x). l ): Distance spread function h(x) to one-dimensional imaging point l -x t )for: Where, k x,0 x is the initial wave value for frequency domain sampling. l Represents the range-oriented imaging sampling points, l = 1, ..., L, W x Indicates the imaging width in the range direction; Similarly, the azimuth-oriented one-dimensional scattering center imaging model s(y m )for: Azimuth spread function h(y) to one-dimensional imaging point m -y t )for: Among them, T y a is the number of azimuth scattering centers. y,t k is the complex amplitude of the azimuth scattering center. y,0 For the initial wavenumber of the azimuth frequency domain sampling, y m This represents the azimuth imaging sampling points, m = 1, ..., M, N y W represents the number of sampling points in the azimuth frequency domain. y Indicates the azimuth imaging width; Pitch-to-one scattering center imaging model s(z) n )for: pitch-to-one-dimensional imaging point spread function h(z) n -z t )for: Among them, T z a is the number of pitch scattering centers. z,t k is the complex amplitude of the pitch-to-scattering center. z,0 To sample the initial wavenumber in the pitch frequency domain, z n This represents the elevation imaging sampling points, n = 1, ..., N, N z W represents the number of frequency domain sampling points in the pitch direction. z Indicates the width of the image in the pitch direction; Based on the range-directed one-dimensional scattering center imaging model s(x) l ), azimuth-oriented one-dimensional scattering center imaging model s(y m ) and pitch-to-one scattering center imaging model s(z n ), thus obtaining the target's three-dimensional scattering center imaging model s(x) l ,y m ,z n )for: Where, x l ,y m ,z n For 3D imaging position sampling, k x,0 ,k y,0 ,k z,0 These represent the initial wavenumbers for frequency domain sampling in the range, azimuth, and elevation directions, respectively; a t x is the complex amplitude of the three-dimensional scattering center. t ,y t ,z t These are the coordinates of the scattering center along the x, y, and z axes, respectively. Based on the distance-to-one-dimensional imaging point spread function h(x) l -x t ), azimuth-directed one-dimensional imaging point spread function h(y) m -y t ) and pitch-to-one imaging point spread function h(z) n -z t ), thus obtaining the three-dimensional imaging point spread function h(x) l ,y m ,z n )for: Among them, W x W y W z N represents the imaging width in the range, azimuth, and elevation directions, respectively. x N y N z These represent the number of frequency domain scattering samples in the range, azimuth, and elevation directions, respectively.

2. The method for extracting the three-dimensional scattering center position of a target in the image domain through a two-layer iterative process as described in claim 1, characterized in that, In S12, a one-dimensional CLEAN algorithm based on a greedy strategy is used for extraction. The extraction of the scattering center location in range-oriented one-dimensional imaging includes: The one-dimensional imaging of the initial range-to-frequency scattering data is I0(x) l Let the iteration number t = 1, and select one-dimensional imaging I in the image domain. t-1 (x l The point with the maximum squared amplitude is taken as the location of the t-th scattering center. Calculate the position of the t-th scattering center One-dimensional imaging point spread function According to one-dimensional imaging in the image domain t-1 (x l Estimate the location of the t-th scattering center. Amplitude of the scattering center at the location I t-1 (x)=[I t-1 (x1),...,I t-1 (x) L )] T (15) Where H represents the conjugate transpose and T represents the transpose; Define the cutoff threshold for one-dimensional scattering center amplitude extraction as ε1, when At that time, from one-dimensional imaging I t-1 (x l Subtract the estimated scattering center distance from the image. And let t = t + 1, continue to extract the one-dimensional image I using formula (11) - formula (15). t (x l The scattering center of ) when At this point, the CLEAN algorithm terminates, completing the extraction of the scattering center for one-dimensional imaging, and obtaining the location of the scattering center for the range-oriented one-dimensional imaging. Among them, T x The number of one-dimensional scattering centers in the range direction; Similarly, the location of the scattering center in the azimuth-oriented one-dimensional imaging is extracted as follows: The location of the scattering center in the pitch-to-one imaging is Among them, T y With T z These represent the number of one-dimensional scattering centers in the azimuth and elevation directions, respectively.

3. The method for extracting the three-dimensional scattering center position of a target in the image domain through a two-layer iterative process as described in claim 2, characterized in that... After extracting a scattering center position for the same location, if the amplitude of the residual scattering center at this location or in the nearby area still exceeds the cutoff threshold ε for extracting the amplitude of the one-dimensional scattering center, the scattering center position is extracted again at this location in the subsequent CLEAN algorithm. Then, the extracted scattering center positions are merged and clustered to obtain a complex amplitude of a scattering center position, and then merged into a composite scattering center position.

4. The method for extracting the location of the three-dimensional scattering center of a target in the image domain through a two-layer iterative process as described in claim 1 or 2, characterized in that, S2 includes: S21. Combine the scattering center positions of each one-dimensional image extracted in S1 into candidate positions of the target's three-dimensional scattering center. S22. Perform three-dimensional imaging on each sub-region centered on the candidate position of the target's three-dimensional scattering center; S23. Extract the amplitude and position of the target's three-dimensional scattering center from all sub-regions centered on the candidate positions of the target's three-dimensional scattering center after three-dimensional imaging.

5. The method for extracting the three-dimensional scattering center position of a target in the image domain through a two-layer iterative process as described in claim 4, characterized in that... In S21, By arranging and combining the scattering center positions from range-oriented one-dimensional imaging, azimuth-oriented one-dimensional imaging, and elevation-oriented one-dimensional imaging, candidate positions for the target's three-dimensional scattering center are obtained. Where T = T x ·T y ·T z .

6. The method for extracting the three-dimensional scattering center position of a target in the image domain through a two-layer iterative process as described in claim 5, characterized in that... In S22, For three-dimensional frequency domain scattering data with varying incident wave frequency, azimuth angle, and elevation angle, the candidate positions of the target's three-dimensional scattering center are determined. Three-dimensional imaging is performed on the sub-region centered on the scattering center; the imaging widths of the sub-regions centered on the scattering center in the range, azimuth, and elevation directions are D, respectively. x D y and D z The three-dimensional image of the t-th scattering center sub-region is obtained as I. t (x l ,y m ,z n ), where x l ,y m ,z n The range of variation are respectively and Three-dimensional imaging is performed on the sub-region centered on all candidate locations of the target's three-dimensional scattering centers to obtain T three-dimensional images I1, I2, ..., I T .

7. The method for extracting the three-dimensional scattering center position of a target in the image domain through a two-layer iterative process as described in claim 6, characterized in that... S23 employs a three-dimensional CLEAN algorithm to extract the amplitude and location of the target's three-dimensional scattering center, specifically including: The initial three-dimensional imaging set of the sub-region centered on the candidate location of the target's three-dimensional scattering center is Y0={I1,I2,...,I... T Let the number of iterations q = 1, and choose set Y. q-1 (x l ,x m ,z n The point with the maximum squared amplitude in the middle is used as the location estimate for the q-th scattering center. calculate The 3D point spread function for imaging the t-th sub-region at the location: Let I be the residual image of the t-th sub-region after q-1 iterations. t,q-1 ,estimate The amplitude of the scattering center is: Among them, I t,q-1 For residual image I t,q-1 A column vector composed of pixels Let be the column vector composed of the pixels imaged by the spread function of the t-th sub-region; Set the cutoff threshold for the three-dimensional scattering center amplitude extraction to ε2, when At that time, the three-dimensional sub-region imaging set Y q-1 (x l ,x m ,z n For each sub-region in the image, subtract the extracted 3D scattering center image: Update the set of 3D imaging sub-regions as follows: Y q (x l ,x m ,z n )={I 1,q (x l ,x m ,z n ),...,I T,q (x l ,x m ,z n )} (21) Let q = q + 1, and continue to extract the three-dimensional scattering center position of the target using formulas (17)-(19); when At this point, the CLEAN algorithm for extracting the three-dimensional scattering center terminates, yielding the target's three-dimensional scattering center position parameters. Q represents the number of extracted three-dimensional scattering centers.

8. The method for extracting the three-dimensional scattering center position of a target in the image domain through a two-layer iterative process as described in claim 2 or 7, characterized in that, In the CLEAN algorithm, let For the scattering measurement of the t-th scattering center and the phase variance of the incoherent scattering, a t G is the complex amplitude of the scattering center in the frequency domain. t Let the amplitude of the scattering center in the image domain be denoted as , then we have Let the signal-to-noise ratio (SNR) of the frequency domain scattering measurement data be SNR. For scattering measurements with varying frequency, the variance of the phase measurement error is expressed as: in, To measure the wavenumber variance of the system's frequency stability, x t Estimate the location of the scattering center; The variance of the phase measurement error due to azimuth angle variation is expressed as: k c y is the wavenumber of the incident wave at its center frequency. t To estimate the location of the azimuth scattering center, The variance of the angle measurement; The variance σ of the phase measurement error for scattering measurement data 2 This will lead to a decrease in main lobe gain. Consequently, the average sidelobe level of target imaging increases. Where K is the number of scattering measurement data samples; The cutoff threshold ε, extracted from the scattering center in the image domain, is defined by the average sidelobe of the target imaging and the signal-to-noise ratio of the measurement data. t for Among them, B p The peak sidelobe gain is denoted by ε, the SNR is the power ratio of the scattered measurement signal to the noise, and the cutoff threshold ε is the power ratio of the peak sidelobe to the average sidelobe. t Extract a cutoff threshold ε1 for the amplitude of one-dimensional scattering center or a cutoff threshold ε2 for the amplitude of three-dimensional scattering center; After extracting the locations of all image domain scattering centers, the amplitude of the image domain scattering centers is multiplied by an amplitude correction term to obtain the amplitude of the frequency domain scattering centers.

Images