A method and system for precise three-dimensional reconstruction using dual-frequency tomography SAR

Abstract

This invention provides a precise 3D reconstruction method and system for dual-frequency tomographic SAR. Addressing the challenges of accurate reconstruction and numerous false targets in single-frequency tomographic SAR under irregular, sparse, and undersampled conditions, this invention constructs two complex SAR image datasets in different frequency bands. Using adjacent pixels of equal height as priors, a joint sparse spectrum estimation model for dual-frequency tomographic SAR is built to reconstruct the scattering intensity profile of the reference pixels. Finally, the order of the dual-frequency tomographic SAR point cloud model is calculated, the scattering intensity values ​​of the dual-frequency tomographic SAR point cloud are estimated, and geocoding is performed to generate a dual-frequency 3D point cloud. This improves the density and reliability of 3D reconstruction of sparse array SAR. Compared to traditional Beamforming, Capon, and Music multi-look methods, and SVD-Wiener and SL1MMER single-look methods, this invention achieves the highest number of reconstructed point clouds. Compared to SVD-Wiener and SL1MMER super-resolution single-look methods, this invention better suppresses sidelobe noise, thereby improving the accuracy and completeness of tomographic SAR point cloud reconstruction.

Description

Technical Field

[0001] This invention belongs to the field of high-dimensional synthetic aperture radar imaging systems and data processing technology, specifically relating to a dual-frequency tomographic SAR precision three-dimensional reconstruction method and system. Background Technology

[0002] Synthetic Aperture Radar (SAR) is an all-weather, all-day active microwave imaging radar with a certain degree of penetration capability into the Earth's surface. Tomographic SAR, developed from SAR, is a high-efficiency, true-stereoscopic, and high-precision 3D reconstruction technology with broad application prospects in areas such as urban 3D scene reconstruction, field exploration and rescue, environmental reconnaissance in non-cooperative areas, and forest biomass and ice reserve estimation. Tomographic SAR achieves third-dimensional resolution and information recovery in overlaid areas by adding multiple antennas or multiple flight passes along the normal direction of the azimuth-slant range plane and performing a third-dimensional elevation-direction synthetic aperture.

[0003] However, when tomographic SAR performs third-dimensional synthetic aperture along the normal direction of the azimuth-slant range plane, multiple passes or multiple antenna samples form a non-uniform and sparse array. Traditional spectral analysis methods such as DFT, BF, and Capon require uniform elevation signal sampling and that the sampling frequency must satisfy the Nyquist sampling theorem for elevation inversion in the overlay region, preventing the tomographic resolution from exceeding the Rayleigh resolution limitation. While methods such as MUSIC and Singular Value Decomposition (SVD) can surpass Rayleigh resolution in elevation resolution, this comes at the cost of spatial resolution. Although tomographic SAR systems have elevation resolution, the limited number of antennas results in a smaller synthetic aperture length in the elevation direction, leading to a much lower elevation resolution than the azimuth and slant range resolutions. To meet the requirements of third-dimensional super-resolution imaging under sparse and irregular baseline distribution, the SL1MMER method based on compressed sensing (CS) has emerged. It utilizes the sparsity of the target signal structure to perform uncorrelated measurements on low-resolution, under-Nyquist sampled data in low-dimensional space to achieve accurate signal reconstruction. However, when the number of baselines is small and unevenly distributed, the SL1MMER method struggles to suppress third-dimensional sidelobe noise, thus introducing a large number of false targets.

[0004] As an extension of CS theory, in 2005, Baron et al. proposed the Distributed Compressed Sensing (DCS) theory to address the joint sparsity among multiple signals. Based on the principle in information theory that the joint entropy of two sources is less than the sum of their individual entropies, DCS, by defining the concept of joint sparsity, fully exploits the correlation between multiple signals to perform joint CS. Its performance is superior to that of individual signal CS, thereby reducing the system's sampling rate requirements. Simultaneously, by adding frequency bands to the SAR system, the third-dimensional sampling rate can be increased exponentially, effectively reducing the number of distributed satellites, which is of great significance for high-precision, high-time-efficiency 3D reconstruction.

[0005] Therefore, developing a dual-frequency tomographic SAR 3D reconstruction system and method based on DCS is of great significance for reducing the satellite number requirement of tomographic SAR and improving the timeliness of 3D reconstruction. Summary of the Invention

[0006] The technical problem to be solved by the present invention is to provide a method and system for precise three-dimensional reconstruction of dual-frequency tomographic SAR, which can improve the density and reliability of three-dimensional reconstruction of sparse array SAR.

[0007] The technical solution adopted by this invention to solve the above-mentioned technical problems is as follows: a dual-frequency tomographic SAR precise three-dimensional reconstruction method, comprising the following steps:

[0008] S1: Obtain two sparse array dual-frequency SAR complex image datasets of different frequency bands, and register them with the SAR complex image most coherent with other SAR complex images in the same dataset selected from each dataset to obtain two frequency band SAR complex image datasets; then obtain the preprocessed dual-frequency tomographic SAR complex image dataset through amplitude and phase preprocessing.

[0009] S2: Centered on the reference pixel, search and identify adjacent pixels of equal height in the pre-processed dual-frequency tomographic SAR complex image dataset. Perform SVD decomposition on the dual-frequency tomographic SAR data stack composed of adjacent pixels of equal height to obtain a dimension-reduced dual-frequency tomographic SAR data stack. Calculate the noise energy of the DCS spectrum estimation model through the dimension-reduced dual-frequency tomographic SAR data stack and reconstruct the scattering intensity profile of the reference pixel.

[0010] S3: Calculate the order of the dual-frequency tomographic SAR point cloud model and the location information of the scatterer based on the scattering intensity profile, estimate the scattering intensity value of the dual-frequency tomographic SAR point cloud, perform geocoding, and output the spatial location and scattering intensity data of the dual-frequency tomographic SAR point cloud.

[0011] According to the above scheme, the specific steps in step S1 are as follows:

[0012] S11: Collect a dual-frequency SAR complex image dataset, select a SAR complex image dataset of one frequency band as the main dataset, select the SAR complex image that is most coherent with other SAR complex images in the same dataset as the first main image; perform coherent complex image registration between the remaining dual-frequency SAR complex images in the same dataset and the first main image to obtain the registered first dual-frequency SAR complex image dataset.

[0013] S12: Select a SAR complex image dataset from the acquired dual-frequency SAR complex image dataset that is on the same platform as the main dataset but in a different frequency band. Select the SAR complex image that is most coherent with other SAR complex images in the same dataset as the second main image of the other frequency band SAR complex image dataset. Perform coherent complex image registration between the remaining dual-frequency SAR complex images in the same dataset and the second main image to obtain the registered second dual-frequency SAR complex image dataset.

[0014] S13: Input reference terrain data, and perform tomographic SAR deskewing and phase error correction on the registered dual-frequency SAR complex image datasets of the two frequency bands, based on the first main image and the second main image; perform amplitude error correction on the dual-frequency SAR complex image dataset based on the first main image; and obtain the preprocessed dual-frequency tomographic SAR complex image dataset.

[0015] According to the above scheme, the specific steps in step S2 are as follows:

[0016] S21: Centered on the reference pixel, within the window range defined by the preprocessed dual-frequency tomographic SAR complex image dataset, the method of average coherence threshold and average amplitude constraint criterion is used to search and identify adjacent pixels of equal height pixel one by one.

[0017] S22: Construct a dual-frequency tomographic SAR data stack using adjacent equal-height pixels and perform SVD decomposition to obtain the dimension-reduced dual-frequency tomographic SAR data stack.

[0018] S23: Substitute the dimension-reduced dual-frequency tomographic SAR data stack into the noise energy estimation model, calculate the noise energy of the DCS spectrum estimation model, and use it as the noise energy magnitude at the reference pixel.

[0019] S24: Input the dimensionality-reduced dual-frequency tomographic SAR data stack into the DCS spectrum estimation model to reconstruct the scattering intensity profile of the reference pixel.

[0020] Furthermore, in step S21, the specific steps are as follows:

[0021] S211: Based on the principle of weighted maximum likelihood coherence estimation, a similarity metric index between the reference pixel and its neighboring pixels is constructed.

[0022] S212: Within the coherence calculation window, calculate the similarity index between corresponding pixels of the reference pixel image block and the adjacent pixel image block one by one, calculate the average similarity index, and use it as the similarity index between the reference pixel image block and the adjacent pixel image block.

[0023] S213: Calculate the average amplitude of dual-frequency tomographic SAR data at the reference pixel and adjacent pixels;

[0024] S214: Identify the average amplitude of neighboring pixels whose average amplitude at the reference pixel is greater than 0.5 times, the average amplitude of neighboring pixels whose average amplitude at the reference pixel is less than 2 times, and the center pixels of neighboring pixel image blocks whose similarity metric index with the reference pixel image block is not less than the coherence threshold as adjacent equal-height pixels.

[0025] S215: Traverse all pixels as reference pixels, execute steps S211 to S214, and find the adjacent cells of equal height for all pixels.

[0026] Furthermore, in step S22, the specific steps are as follows:

[0027] S221: Decompose the dual-frequency tomographic SAR data stack composed of adjacent equal-height pixels into subspaces corresponding to signal and noise by using SVD, and retain only the signal subspace;

[0028] S222: Calculate the dimensionality-reduced dual-frequency tomographic SAR data stack through the signal subspace.

[0029] According to the above scheme, the specific steps in step S24 are as follows:

[0030] S241: Substitute the dimension-reduced dual-frequency tomographic SAR data stack to construct a DCS spectrum estimation model;

[0031] S242: The spectral matrix is ​​estimated using the spectral gradient projection method;

[0032] S243: Calculate the scattering intensity profile of the dual-frequency tomographic SAR at the reference pixel.

[0033] According to the above scheme, the specific steps in step S3 are as follows:

[0034] S31: Input the peak position information of the scattering intensity profile into the scatterer number detection model of the information constraint criterion to determine the order of the dual-frequency tomographic SAR point cloud model and the scatterer position information.

[0035] S32: Based on the model order and scatterer location information, the nonlocal least squares method is used to estimate the scattering intensity value of dual-frequency tomographic SAR point cloud;

[0036] S33: Geocode dual-frequency tomographic SAR point cloud data in the range-Doppler-elevation coordinate system, and output the spatial location and scattering intensity data of the dual-frequency tomographic SAR point cloud.

[0037] According to the above scheme, the specific steps in step S31 are as follows:

[0038] S311: Construct an information-theoretic constrained model for solving the model of the required order, i.e., a model for detecting the number of scatterers;

[0039] S312: The peak position information of the combined scattering intensity profile from large to small is obtained by exhaustively substituting the model order to solve the model and calculate the cost function. The effective number of scattering targets and the elevation position corresponding to the target distribution case with the minimum cost function are used as the effective number of targets and elevation position information at the reference pixel.

[0040] According to the above scheme, the specific steps in step S32 are as follows:

[0041] S321: Construct an optimization solution model for the target's three-dimensional complex scattering coefficients;

[0042] S322: The optimization solution model for solving the three-dimensional complex scattering coefficients of the target is obtained by using the least squares method, and the scattering intensity value of the dual-frequency tomographic SAR point cloud is estimated.

[0043] A dual-frequency tomographic SAR precision three-dimensional reconstruction system,

[0044] The dual-frequency tomographic SAR preprocessing submodule is used to acquire two sparse array dual-frequency SAR complex image datasets in two different frequency bands. It registers the SAR complex image most coherent with other SAR complex images in the same dataset selected from each dataset to obtain SAR complex image datasets in two frequency bands. Then, it obtains the preprocessed dual-frequency tomographic SAR complex image dataset through amplitude and phase preprocessing.

[0045] The dual-frequency tomographic SAR spectrum estimation submodule is used to search and identify adjacent equal-height pixels in the pre-processed dual-frequency tomographic SAR complex image dataset with the reference pixel as the center. It performs SVD decomposition on the dual-frequency tomographic SAR data stack composed of adjacent equal-height pixels to obtain a dimension-reduced dual-frequency tomographic SAR data stack. It calculates the noise energy of the DCS spectrum estimation model through the dimension-reduced dual-frequency tomographic SAR data stack and reconstructs the scattering intensity profile of the reference pixel.

[0046] The dual-frequency 3D point cloud generation submodule is used to calculate the order of the dual-frequency tomographic SAR point cloud model and the location information of the scatterer based on the scattering intensity profile, estimate the scattering intensity value of the dual-frequency tomographic SAR point cloud, perform geocoding, and output the spatial location and scattering intensity data of the dual-frequency tomographic SAR point cloud.

[0047] The beneficial effects of this invention are as follows:

[0048] 1. The present invention provides a dual-frequency tomographic SAR precise 3D reconstruction method and system, which addresses the problems of difficulty in accurate reconstruction and numerous false targets in single-frequency tomographic SAR under irregular, sparse, and insufficient sampling conditions. By constructing SAR complex image datasets in two different frequency bands, a dual-frequency tomographic SAR joint sparse spectrum estimation model is constructed using adjacent equal-height pixels as priors. The scattering intensity contour of the reference pixel is reconstructed. Finally, the order of the dual-frequency tomographic SAR point cloud model is calculated, the scattering intensity value of the dual-frequency tomographic SAR point cloud is estimated, and geocoding is performed to generate a dual-frequency 3D point cloud, thereby improving the density and reliability of sparse array SAR 3D reconstruction.

[0049] 2. Compared with traditional Beamforming, Capon, and Music multi-view methods, as well as SVD-Wiener and SL1MMER single-view methods, this invention has the highest number of point cloud reconstructions;

[0050] 3. Compared with the SVD-Wiener and SL1MMER super-resolution single-view methods, the present invention suppresses sidelobe noise better, thereby improving the accuracy and completeness of tomographic SAR point cloud reconstruction. Attached Figure Description

[0051] Figure 1 This is a flowchart of an embodiment of the present invention.

[0052] Figure 2 This is an optical image of a building according to an embodiment of the present invention.

[0053] Figure 3 This is an X-band SAR image of a building according to an embodiment of the present invention.

[0054] Figure 4 This is a Ku-band SAR image of a building according to an embodiment of the present invention.

[0055] Figure 5 This is a point cloud reconstruction result image from Ku-TomoSAR SVD-Wiener.

[0056] Figure 6 This is a result of X-TomoSAR SVD-Wiener point cloud reconstruction.

[0057] Figure 7 This is a point cloud reconstruction result image from Ku-TomoSAR SL1MMER.

[0058] Figure 8 This is a point cloud reconstruction result image from X-TomoSAR SL1MMER.

[0059] Figure 9This is a diagram showing the point cloud reconstruction results from Ku-TomoSAR Beamforming.

[0060] Figure 10 This is a diagram showing the point cloud reconstruction results from X-TomoSAR Beamforming.

[0061] Figure 11 This is a Ku-TomoSAR Capon point cloud reconstruction result image.

[0062] Figure 12 This is a result of X-TomoSAR Capon point cloud reconstruction.

[0063] Figure 13 This is a Ku-TomoSAR MUSIC point cloud reconstruction result image.

[0064] Figure 14 This is a result of X-TomoSAR MUSIC point cloud reconstruction.

[0065] Figure 15 This is a point cloud reconstruction result image of a dual-frequency TomoSAR DCS according to an embodiment of the present invention.

[0066] Figure 16 This is a diagram showing the actual results of an embodiment of the present invention.

[0067] Figure 17 This is an evaluation diagram of the three-dimensional reconstruction results of the single-frequency and dual-frequency tomographic SAR methods according to an embodiment of the present invention. Detailed Implementation

[0068] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0069] Example 1

[0070] See Figure 1 The specific steps of a dual-frequency tomographic SAR precise three-dimensional reconstruction method are as follows:

[0071] S1: Obtain two sparse array dual-frequency SAR complex image datasets of different frequency bands, and register them with the SAR complex image most coherent with other SAR complex images in the same dataset selected from each dataset to obtain two frequency band SAR complex image datasets; then obtain the preprocessed dual-frequency tomographic SAR complex image dataset through amplitude and phase preprocessing.

[0072] S2: Centered on the reference pixel, search and identify adjacent pixels of equal height in the pre-processed dual-frequency tomographic SAR complex image dataset. Perform SVD decomposition on the dual-frequency tomographic SAR data stack composed of adjacent pixels of equal height to obtain a dimension-reduced dual-frequency tomographic SAR data stack. Calculate the noise energy of the DCS spectrum estimation model through the dimension-reduced dual-frequency tomographic SAR data stack and reconstruct the scattering intensity profile of the reference pixel.

[0073] S3: Calculate the order of the dual-frequency tomographic SAR point cloud model and the location information of the scatterer based on the scattering intensity profile, estimate the scattering intensity value of the dual-frequency tomographic SAR point cloud, perform geocoding, and output the spatial location and scattering intensity data of the dual-frequency tomographic SAR point cloud.

[0074] Furthermore, in step S1, the specific steps are as follows:

[0075] S11: Collect a dual-frequency SAR complex image dataset, select a SAR complex image dataset of one frequency band as the main dataset, select the SAR complex image that is most coherent with other SAR complex images in the same dataset as the first main image; perform coherent complex image registration between the remaining dual-frequency SAR complex images in the same dataset and the first main image to obtain the registered first dual-frequency SAR complex image dataset.

[0076] S12: Select a SAR complex image dataset from the acquired dual-frequency SAR complex image dataset that is on the same platform as the main dataset but in a different frequency band. Select the SAR complex image that is most coherent with other SAR complex images in the same dataset as the second main image of the other frequency band SAR complex image dataset. Perform coherent complex image registration between the remaining dual-frequency SAR complex images in the same dataset and the second main image to obtain the registered second dual-frequency SAR complex image dataset.

[0077] S13: Input reference terrain data, and perform tomographic SAR deskewing and phase error correction on the registered dual-frequency SAR complex image datasets of the two frequency bands, based on the first main image and the second main image; perform amplitude error correction on the dual-frequency SAR complex image dataset based on the first main image; and obtain the preprocessed dual-frequency tomographic SAR complex image dataset.

[0078] In step S2, the specific steps are as follows:

[0079] S21: Centered on the reference pixel, within the window range defined by the preprocessed dual-frequency tomographic SAR complex image dataset, the method of average coherence threshold and average amplitude constraint criterion is used to search and identify adjacent pixels of equal height pixel one by one.

[0080] S22: Construct a dual-frequency tomographic SAR data stack using adjacent equal-height pixels and perform SVD decomposition to obtain the dimension-reduced dual-frequency tomographic SAR data stack.

[0081] S23: Substitute the dimension-reduced dual-frequency tomographic SAR data stack into the noise energy estimation model, calculate the noise energy of the DCS spectrum estimation model, and use it as the noise energy magnitude at the reference pixel.

[0082] S24: Input the dimensionality-reduced dual-frequency tomographic SAR data stack into the DCS spectrum estimation model to reconstruct the scattering intensity profile of the reference pixel.

[0083] Furthermore, in step S21, the specific steps are as follows:

[0084] S211: Based on the principle of weighted maximum likelihood coherence estimation, a similarity metric index between the reference pixel and its neighboring pixels is constructed.

[0085] S212: Within the coherence calculation window, calculate the similarity index between corresponding pixels of the reference pixel image block and the adjacent pixel image block one by one, calculate the average similarity index, and use it as the similarity index between the reference pixel image block and the adjacent pixel image block.

[0086] S213: Calculate the average amplitude of dual-frequency tomographic SAR data at the reference pixel and adjacent pixels;

[0087] S214: Identify the average amplitude of neighboring pixels whose average amplitude at the reference pixel is greater than 0.5 times, the average amplitude of neighboring pixels whose average amplitude at the reference pixel is less than 2 times, and the center pixels of neighboring pixel image blocks whose similarity metric index with the reference pixel image block is not less than the coherence threshold as adjacent equal-height pixels.

[0088] S215: Traverse all pixels as reference pixels, execute steps S211 to S214, and find the adjacent cells of equal height for all pixels.

[0089] In step S22, the specific steps are as follows:

[0090] S221: Decompose the dual-frequency tomographic SAR data stack composed of adjacent equal-height pixels into subspaces corresponding to signal and noise by using SVD, and retain only the signal subspace;

[0091] S222: Calculate the dimensionality-reduced dual-frequency tomographic SAR data stack through the signal subspace.

[0092] In step S24, the specific steps are as follows:

[0093] S241: Substitute the dimension-reduced dual-frequency tomographic SAR data stack to construct a DCS spectrum estimation model;

[0094] S242: The spectral matrix is ​​estimated using the spectral gradient projection method;

[0095] S243: Calculate the scattering intensity profile of the dual-frequency tomographic SAR at the reference pixel.

[0096] In step S3, the specific steps are as follows:

[0097] S31: Input the peak position information of the scattering intensity profile into the scatterer number detection model of the information constraint criterion to determine the order of the dual-frequency tomographic SAR point cloud model and the scatterer position information.

[0098] S32: Based on the model order and scatterer location information, the nonlocal least squares method is used to estimate the scattering intensity value of dual-frequency tomographic SAR point cloud;

[0099] S33: Geocode dual-frequency tomographic SAR point cloud data in the range-Doppler-elevation coordinate system, and output the spatial location and scattering intensity data of the dual-frequency tomographic SAR point cloud.

[0100] Furthermore, in step S31, the specific steps are as follows:

[0101] S311: Construct an information-theoretic constrained model for solving the model of the required order, i.e., a model for detecting the number of scatterers;

[0102] S312: The peak position information of the combined scattering intensity profile from large to small is obtained by exhaustively substituting the model order to solve the model and calculate the cost function. The effective number of scattering targets and the elevation position corresponding to the target distribution case with the minimum cost function are used as the effective number of targets and elevation position information at the reference pixel.

[0103] In step S32, the specific steps are as follows:

[0104] S321: Construct an optimization solution model for the target's three-dimensional complex scattering coefficients;

[0105] S322: The optimization solution model for solving the three-dimensional complex scattering coefficients of the target is obtained by using the least squares method, and the scattering intensity value of the dual-frequency tomographic SAR point cloud is estimated.

[0106] This embodiment addresses the problem of inaccurate reconstruction and numerous false targets in single-frequency tomographic SAR under irregular, sparse, and insufficient sampling conditions. By constructing SAR complex image datasets in two different frequency bands, the scattering intensity contour of the reference pixel is reconstructed using adjacent equal-height pixels and dual-frequency tomographic SAR spectrum estimation. Finally, the order of the dual-frequency tomographic SAR point cloud model is calculated to estimate the scattering intensity value of the dual-frequency tomographic SAR point cloud, and geocoding is performed to generate a dual-frequency 3D point cloud. This achieves the function of improving the density and reliability of sparse array SAR 3D reconstruction.

[0107] Example 2

[0108] The steps in this embodiment are the same as in Embodiment 1, except that each step is applied to a specific instance. Specifically, it includes the following steps:

[0109] S1: Obtain two sparse array dual-frequency SAR complex image datasets in two different frequency bands. Register the SAR complex image selected from each dataset as the most coherent SAR complex image with other SAR complex images in the same dataset to obtain two frequency band SAR complex image datasets. Then, obtain a preprocessed dual-frequency tomographic SAR complex image dataset through amplitude and phase preprocessing. The specific steps are as follows:

[0110] S11: Collect a dual-frequency SAR complex image dataset, select a SAR complex image dataset of one frequency band as the main dataset, select the SAR complex image most coherent with other SAR complex images as the first main image; register the remaining dual-frequency SAR complex images with it to obtain the registered dual-frequency SAR complex image dataset.

[0111] S12: Select a SAR complex image from the same platform as the main dataset but in a different frequency band, and use it as the second main image of another frequency band SAR complex image dataset;

[0112] S13: Input reference terrain data, and use the first and second main images as a reference to perform tomographic SAR deskewing and phase error correction on the registered two frequency band SAR complex image datasets respectively; then, use the first main image as a reference to perform amplitude error correction on the dual-frequency tomographic SAR complex image dataset; to obtain the amplitude and phase preprocessed dual-frequency tomographic SAR complex image dataset.

[0113] S2: Search and identify adjacent pixels of equal height within the preprocessed dual-frequency tomographic SAR complex image dataset. Perform SVD decomposition on the dual-frequency tomographic SAR data stack composed of adjacent pixels of equal height to obtain a dimension-reduced dual-frequency tomographic SAR data stack. Calculate the noise energy of the DCS spectrum estimation model and reconstruct the scattering intensity profile of the reference pixel using the dimension-reduced dual-frequency tomographic SAR data stack. The specific steps are as follows:

[0114] S21: Centered on the reference pixel, within the window defined by the preprocessed dual-frequency tomographic SAR complex image dataset, neighboring pixels of equal height are searched and identified pixel by pixel using the average coherence threshold and average amplitude constraint criteria; the specific steps are as follows:

[0115] S211: Based on the principle of weighted maximum likelihood coherence estimation, a similarity metric index is constructed between the reference pixel x and its neighboring pixel y.

[0116]

[0117] in

[0118]

[0119] in The distribution range is between (0,1], where M is the number of sparse array SAR images, and x i (m) represents the complex value of the m-th preprocessed dual-frequency tomographic SAR image at reference pixel i, y i (m) represents the complex value at the pixel adjacent to the reference pixel i in the m-th preprocessed dual-frequency tomographic SAR image. * Represents the conjugate operation for complex numbers.

[0120] S212: Within the coherence calculation window, calculate the similarity metric index between corresponding pixels of the reference pixel image block and its neighboring pixel image blocks one by one, calculate the average similarity metric index, and use it as the similarity metric index between the reference pixel image block and its neighboring pixel image blocks:

[0121]

[0122] Where N is the total number of pixels in the SAR image block.

[0123] S213: Calculate the average amplitude A of the dual-frequency tomographic SAR data at the reference pixel and its adjacent pixels. x 、A y .

[0124] S214: Average amplitude A at the reference pixel x The average amplitude A at adjacent pixels is greater than 0.5 times. y And the average amplitude A at the reference pixel x The average amplitude A at adjacent pixels is less than twice that of adjacent pixels. y Furthermore, the similarity metric index D(x,y) with the reference pixel image patch is not less than C. thresh The center pixel of an adjacent pixel image block is identified as an adjacent pixel of equal height, where a coherence threshold C is set. thresh =0.3.

[0125] S215: Traverse all pixels as reference pixels, and find the adjacent cells of equal height for all pixels according to steps 4.1 to 4.4.

[0126] S22: Perform SVD decomposition on the dual-frequency tomographic SAR data stack composed of adjacent equal-height pixels to obtain the dual-frequency tomographic SAR data stack with reduced data dimensionality; the specific steps are as follows:

[0127] S221: SVD is used to decompose the dual-frequency tomographic SAR data stack Y, composed of adjacent equal-height pixels, into subspaces corresponding to signal and noise, retaining only the signal subspace.

[0128] Y = USV H=[U M×K U M×(M-K) ]S M×L [V L×K V L×(L-K) ] H (4)

[0129] Where K is the rank of the matrix, L is the number of adjacent pixels of equal height plus 1, S is a diagonal matrix composed of singular value elements arranged in descending order, K < L, K < M, U M×K V L×K This represents the signal subspace formed by the eigenvectors corresponding to the first K singular values, while the noise subspace is formed by the eigenvectors corresponding to the remaining singular values.

[0130] S222: Order Among them I K×K For a K×K identity matrix, 0 (L-K)×K The zero matrix is ​​(LK)×L; the dimensionality-reduced dual-frequency tomographic SAR data stack is calculated as follows.

[0131] S23: Substitute the dimension-reduced dual-frequency tomographic SAR data stack into the noise energy estimation model to obtain the noise energy of the DCS spectrum estimation model;

[0132] The formula for calculating noise energy is as follows:

[0133]

[0134] in This represents the stack of dual-frequency tomographic SAR data after dimensionality reduction. The l-th column vector is used as the obtained parameters to determine the noise energy at the reference pixel, which is then used for noise energy constraints in DCS spectrum estimation and order determination of the dual-frequency tomographic SAR point cloud model.

[0135] S24: Input the dimension-reduced dual-frequency tomographic SAR data stack into the DCS spectrum estimation model to reconstruct the scattering intensity profile of the reference pixel.

[0136] The steps for dual-frequency tomographic SAR spectrum estimation based on DCS are as follows:

[0137] S241: Substitute the dimension-reduced dual-frequency tomographic SAR data stack to construct the DCS spectrum estimation model:

[0138]

[0139] Among them || || 2,1 、|| || F Let L1 and L2 represent the L1-L2 mixed norm and F norm of the matrix, respectively, and A1 and A2 are the observation matrices of the two frequency band tomographic SAR.

[0140] S242: Estimate the spectral matrix X using the spectral gradient projection method.

[0141] S243: Provides the scattering intensity results of the dual-frequency tomographic SAR at the final reference pixel. in Represents the estimated spectrum matrix The i-th row vector.

[0142] S3: Calculate the order of the dual-frequency tomographic SAR point cloud model and the location information of the scattering body based on the scattering intensity profile, estimate the scattering intensity value of the dual-frequency tomographic SAR point cloud, perform geocoding, and output the spatial location and scattering intensity data of the dual-frequency tomographic SAR point cloud; the specific steps are as follows:

[0143] S31: Input the peak location information of the scattering intensity profile into the scatterer number detection model based on the information constraint criterion to determine the order of the dual-frequency tomographic SAR point cloud model and the scatterer location information; the specific steps are as follows:

[0144] S311: Constructing an information-theoretic constraint model for solving problems of the required order:

[0145]

[0146] in To determine the number of valid targets at the reference pixel, z represents the number of targets at the reference pixel. The location of the maximum spectral peak Indicates the pre-duration dual-frequency tomographic SAR The three-dimensional scattering coefficients corresponding to the locations of the maximum spectral peaks, while the information constraint term uses the BIC and MDL models:

[0147]

[0148] Where c is the regularization parameter for model order discrimination.

[0149] S312: Based on the combinations of spectral peaks from largest to smallest, the cost function is calculated using an exhaustive method. The number of effective scattering targets and their elevation positions corresponding to the target distribution with the minimum cost function are used as the number of effective targets at the reference pixel. and elevation position information.

[0150] S32: Based on the model order and the location information of the scatterers, the scattering intensity values ​​of the dual-frequency tomographic SAR point cloud are estimated using the nonlocal least squares method; the specific steps are as follows:

[0151] S321: Constructing an optimization model for the target's three-dimensional complex scattering coefficients:

[0152]

[0153] S322: Solve the above equation using the least squares method, and estimate the scattering intensity of dual-frequency tomographic SAR point clouds using the following equation:

[0154]

[0155] S33: Geocode the dual-frequency tomographic SAR point cloud data in the range-Doppler-elevation coordinate system, and output the spatial location and scattering intensity data of the dual-frequency tomographic SAR point cloud.

[0156] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0157] The Bird's Nest simulation dataset includes four images each from the X-band and Ku-band SAR datasets. Figure 2 An optical image of a building. Figure 3 , Figure 4 These are X-band and Ku-band SAR images, respectively. The left side shows the overlapping pixels of buildings, and the center pixel circled in white is the simulation control point, used to objectively evaluate the accuracy and completeness of building elevation reconstruction and other indicators of different tomographic SAR 3D reconstruction algorithms.

[0158] Using X-band and Ku-band SAR datasets and comparing them with several classic single-frequency tomographic SAR 3D reconstruction methods, the superiority of the proposed algorithm in sparse array tomographic SAR 3D reconstruction is verified. Figure 5-16 The three-dimensional reconstruction results of several classic methods for X and Ku bands, as well as the dual-frequency tomographic SAR method of this embodiment, are compared. It can be seen that in the case of single-band X and Ku tomographic SAR, the single-look SVD-Wiener ( Figure 5 , 6 ), based on CS, the German aerospace SL1MMER ( Figure 7 , 8 The method has good reconstruction coverage, but it is accompanied by a large number of false targets; multi-view Beamforming ( Figure 9 , 10 ), Capon Figure 11 , 12 Music Figure 13 , 14 While this method significantly suppresses sidelobe noise and greatly reduces false targets, the reconstructed coverage is poor, especially in... Figure 16 The elliptical marking region; while the dual-frequency weighted DCS method of this embodiment ( Figure 15 It not only has the best tomographic SAR reconstruction accuracy and reconstruction completeness, but also effectively suppresses sidelobe noise and false targets. Figure 17 The table in the figure compares the evaluation results of three-dimensional reconstruction using single-frequency and dual-frequency tomographic SAR methods. It can be seen that all the index results of the dual-frequency DCS method proposed in this embodiment are better than those of single-frequency tomographic SAR.

[0159] Example 3

[0160] This embodiment is used to implement the principle construction system of the above method embodiment, including a dual-frequency tomographic SAR preprocessing submodule, a dual-frequency tomographic SAR spectrum estimation submodule, and a dual-frequency 3D point cloud generation submodule.

[0161] The dual-frequency tomographic SAR preprocessing submodule is used to acquire two sparse array dual-frequency SAR complex image datasets in two different frequency bands. It registers the SAR complex image most coherent with other SAR complex images in the same dataset selected from each dataset to obtain SAR complex image datasets in two frequency bands. Then, it obtains the preprocessed dual-frequency tomographic SAR complex image dataset through amplitude and phase preprocessing.

[0162] The dual-frequency tomographic SAR spectrum estimation submodule is used to search and identify adjacent equal-height pixels in the pre-processed dual-frequency tomographic SAR complex image dataset with the reference pixel as the center. It performs SVD decomposition on the dual-frequency tomographic SAR data stack composed of adjacent equal-height pixels to obtain a dimension-reduced dual-frequency tomographic SAR data stack. It calculates the noise energy of the DCS spectrum estimation model through the dimension-reduced dual-frequency tomographic SAR data stack and reconstructs the scattering intensity profile of the reference pixel.

[0163] The dual-frequency 3D point cloud generation submodule is used to calculate the order of the dual-frequency tomographic SAR point cloud model and the location information of the scatterer based on the scattering intensity profile, estimate the scattering intensity value of the dual-frequency tomographic SAR point cloud, perform geocoding, and output the spatial location and scattering intensity data of the dual-frequency tomographic SAR point cloud.

[0164] Each submodule is mainly used to implement the various steps of the method implementation, which will not be elaborated here.

[0165] It should be noted that, depending on the implementation needs, the various steps / components described in this application can be broken down into more steps / components, or two or more steps / components or parts of the operation of steps / components can be combined into new steps / components to achieve the purpose of this invention.

[0166] This embodiment also includes a processor, a communication interface, a memory, and a communication bus; wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; the memory stores a computer program, and when the program is executed by the processor, the processor performs the steps of a dual-frequency tomographic SAR precise three-dimensional reconstruction method.

[0167] This embodiment also provides a computer-readable storage medium storing executable instructions that, when executed by a processor, enable the processor to implement a dual-frequency tomographic SAR precise three-dimensional reconstruction method.

[0168] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects.

[0169] Furthermore, this application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0170] This application is described with reference to a block diagram of an apparatus (system) according to Embodiment 1 and a flowchart of a method and computer program product according to Embodiment 2. It should be understood that each step or block in the flowchart or block diagram, as well as combinations of steps or blocks in the flowchart or block diagram, can be implemented by computer program instructions.

[0171] These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing device to produce a machine, such that the instructions, which are executable by the processor of the computer or other programmable data processing device, produce instructions for implementing the process. Figure 1 One or more processes or boxes Figure 1 A system that specifies functions in one or more boxes.

[0172] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes or boxes Figure 1 The function specified in one or more boxes.

[0173] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes or boxes Figure 1 The steps of a dual-frequency tomographic SAR precise three-dimensional reconstruction method are specified in one or more boxes.

[0174] The above embodiments are only used to illustrate the design concept and features of the present invention, and their purpose is to enable those skilled in the art to understand the content of the present invention and implement it accordingly. The protection scope of the present invention is not limited to the above embodiments. Therefore, all equivalent changes or modifications made based on the principles and design ideas disclosed in the present invention are within the protection scope of the present invention.

Claims

1. A precise three-dimensional reconstruction method using dual-frequency tomographic SAR, characterized in that: Includes the following steps: S1: Obtain two sparse array dual-frequency SAR complex image datasets of different frequency bands, and register them with the SAR complex image most coherent with other SAR complex images in the same dataset selected from each dataset to obtain two frequency band SAR complex image datasets; then obtain the preprocessed dual-frequency tomographic SAR complex image dataset through amplitude and phase preprocessing. S2: Centered on the reference pixel, search and identify adjacent pixels of equal height in the pre-processed dual-frequency tomographic SAR complex image dataset. Perform SVD decomposition on the dual-frequency tomographic SAR data stack composed of adjacent pixels of equal height to obtain a dimension-reduced dual-frequency tomographic SAR data stack. Calculate the noise energy of the DCS spectrum estimation model through the dimension-reduced dual-frequency tomographic SAR data stack and reconstruct the scattering intensity profile of the reference pixel. S3: Calculate the order of the dual-frequency tomographic SAR point cloud model and the location information of the scatterer based on the scattering intensity profile, estimate the scattering intensity value of the dual-frequency tomographic SAR point cloud, perform geocoding, and output the spatial location and scattering intensity data of the dual-frequency tomographic SAR point cloud.

2. The dual-frequency tomographic SAR precise three-dimensional reconstruction method according to claim 1, characterized in that: The specific steps in step S1 are as follows: S11: Collect a dual-frequency SAR complex image dataset, select a SAR complex image dataset of one frequency band as the main dataset, select the SAR complex image that is most coherent with other SAR complex images in the same dataset as the first main image; perform coherent complex image registration between the remaining dual-frequency SAR complex images in the same dataset and the first main image to obtain the registered first dual-frequency SAR complex image dataset. S12: Select a SAR complex image dataset from the acquired dual-frequency SAR complex image dataset that is on the same platform as the main dataset but in a different frequency band. Select the SAR complex image that is most coherent with other SAR complex images in the same dataset as the second main image of the other frequency band SAR complex image dataset. Perform coherent complex image registration between the remaining dual-frequency SAR complex images in the same dataset and the second main image to obtain the registered second dual-frequency SAR complex image dataset. S13: Input reference terrain data, and perform tomographic SAR deskewing and phase error correction on the registered dual-frequency SAR complex image datasets of the two frequency bands, using the first main image and the second main image as references; perform amplitude error correction on the dual-frequency SAR complex image datasets using the first main image as references. The preprocessed dual-frequency tomographic SAR complex image dataset is obtained.

3. The dual-frequency tomographic SAR precise three-dimensional reconstruction method according to claim 1, characterized in that: The specific steps in step S2 are as follows: S21: Centered on the reference pixel, within the window range defined by the preprocessed dual-frequency tomographic SAR complex image dataset, the method of average coherence threshold and average amplitude constraint criterion is used to search and identify adjacent pixels of equal height pixel one by one. S22: Construct a dual-frequency tomographic SAR data stack using adjacent equal-height pixels and perform SVD decomposition to obtain the dimension-reduced dual-frequency tomographic SAR data stack. S23: Substitute the dimension-reduced dual-frequency tomographic SAR data stack into the noise energy estimation model, calculate the noise energy of the DCS spectrum estimation model, and use it as the noise energy magnitude at the reference pixel. S24: Input the dimensionality-reduced dual-frequency tomographic SAR data stack into the DCS spectrum estimation model to reconstruct the scattering intensity profile of the reference pixel.

4. The dual-frequency tomographic SAR precise three-dimensional reconstruction method according to claim 3, characterized in that: The specific steps in step S21 are as follows: S211: Based on the principle of weighted maximum likelihood coherence estimation, a similarity metric index between the reference pixel and its neighboring pixels is constructed. S212: Within the coherence calculation window, calculate the similarity index between corresponding pixels of the reference pixel image block and the adjacent pixel image block one by one, calculate the average similarity index, and use it as the similarity index between the reference pixel image block and the adjacent pixel image block. S213: Calculate the average amplitude of dual-frequency tomographic SAR data at the reference pixel and adjacent pixels; S214: Identify the average amplitude of neighboring pixels whose average amplitude at the reference pixel is greater than 0.5 times, the average amplitude of neighboring pixels whose average amplitude at the reference pixel is less than 2 times, and the center pixels of neighboring pixel image blocks whose similarity metric index with the reference pixel image block is not less than the coherence threshold as adjacent equal-height pixels. S215: Traverse all pixels as reference pixels, execute steps S211 to S214, and find the adjacent cells of equal height for all pixels.

5. The dual-frequency tomographic SAR precise three-dimensional reconstruction method according to claim 3, characterized in that: The specific steps in step S22 are as follows: S221: Decompose the dual-frequency tomographic SAR data stack composed of adjacent equal-height pixels into subspaces corresponding to signal and noise by using SVD, and retain only the signal subspace; S222: Calculate the dimensionality-reduced dual-frequency tomographic SAR data stack through the signal subspace.

6. The dual-frequency tomographic SAR precise three-dimensional reconstruction method according to claim 3, characterized in that: The specific steps in step S24 are as follows: S241: Substitute the dimension-reduced dual-frequency tomographic SAR data stack to construct a DCS spectrum estimation model; S242: The spectral matrix is ​​estimated using the spectral gradient projection method; S243: Calculate the scattering intensity profile of the dual-frequency tomographic SAR at the reference pixel.

7. The dual-frequency tomographic SAR precise three-dimensional reconstruction method according to claim 1, characterized in that: The specific steps in step S3 are as follows: S31: Input the peak position information of the scattering intensity profile into the scatterer number detection model of the information constraint criterion to determine the order of the dual-frequency tomographic SAR point cloud model and the scatterer position information. S32: Based on the model order and scatterer location information, the nonlocal least squares method is used to estimate the scattering intensity value of dual-frequency tomographic SAR point cloud; S33: Geocode dual-frequency tomographic SAR point cloud data in the range-Doppler-elevation coordinate system, and output the spatial location and scattering intensity data of the dual-frequency tomographic SAR point cloud.

8. The dual-frequency tomographic SAR precise three-dimensional reconstruction method according to claim 7, characterized in that: The specific steps in step S31 are as follows: S311: Construct an information-theoretic constrained model for solving the model of the required order, i.e., a model for detecting the number of scatterers; S312: The peak position information of the combined scattering intensity profile from large to small is obtained by exhaustively substituting the model order to solve the model and calculate the cost function. The effective number of scattering targets and the elevation position corresponding to the target distribution case with the minimum cost function are used as the effective number of targets and elevation position information at the reference pixel.

9. The dual-frequency tomographic SAR precise three-dimensional reconstruction method according to claim 7, characterized in that: The specific steps in step S32 are as follows: S321: Construct an optimization solution model for the target's three-dimensional complex scattering coefficients; S322: The optimization solution model for solving the three-dimensional complex scattering coefficients of the target is obtained by using the least squares method, and the scattering intensity value of the dual-frequency tomographic SAR point cloud is estimated.

10. A dual-frequency tomographic SAR precision three-dimensional reconstruction system, characterized in that: The dual-frequency tomographic SAR preprocessing submodule is used to acquire two sparse array dual-frequency SAR complex image datasets in two different frequency bands. It registers the SAR complex image most coherent with other SAR complex images in the same dataset selected from each dataset to obtain SAR complex image datasets in two frequency bands. Then, it obtains the preprocessed dual-frequency tomographic SAR complex image dataset through amplitude and phase preprocessing. The dual-frequency tomographic SAR spectrum estimation submodule is used to search and identify adjacent equal-height pixels in the pre-processed dual-frequency tomographic SAR complex image dataset with the reference pixel as the center. It performs SVD decomposition on the dual-frequency tomographic SAR data stack composed of adjacent equal-height pixels to obtain a dimension-reduced dual-frequency tomographic SAR data stack. It calculates the noise energy of the DCS spectrum estimation model through the dimension-reduced dual-frequency tomographic SAR data stack and reconstructs the scattering intensity profile of the reference pixel. The dual-frequency 3D point cloud generation submodule is used to calculate the order of the dual-frequency tomographic SAR point cloud model and the location information of the scatterer based on the scattering intensity profile, estimate the scattering intensity value of the dual-frequency tomographic SAR point cloud, perform geocoding, and output the spatial location and scattering intensity data of the dual-frequency tomographic SAR point cloud.

Images