A high-resolution tomographic SAR four-dimensional imaging method and device with space-time baseline decoupling processing and a storage medium

Abstract

The application discloses a kind of high-resolution tomography SAR four-dimensional imaging methods of space-time baseline decoupling processing, which comprises the following steps: step 1, reconstruct original satellite data into uniform baseline observation data, and utilize the low rank characteristic of Hankel matrix, solve its noiseless estimation;Step 2, elevation information is inverted based on noiseless estimation, and elevation information is used as priori, deformation spectrum estimation dictionary is constructed by space-time decoupling, and deformation information estimation is carried out;Step 3, the elevation and deformation variable estimated are matched using maximum likelihood method to realize four-dimensional imaging.The method realizes high-precision imaging while exponentially reducing the algorithm complexity of four-dimensional imaging, and the differential tomography SAR four-dimensional imaging result verification is carried out based on domestic land exploration No.1 satellite data.

Description

Technical Field

[0001] This invention belongs to the field of radar signal processing technology, and particularly relates to a high-resolution tomographic SAR four-dimensional imaging method, device and storage medium with spatiotemporal baseline decoupling processing. Background Technology

[0002] Microwave remote sensing technology, represented by Interferometric Synthetic Aperture Radar (InSAR), has become a key method for achieving high-precision Earth parameter observation. It plays an indispensable role in various fields such as surveillance, topographic mapping, environmental monitoring, geological exploration, and disaster assessment. However, traditional InSAR is limited by SAR imaging limitations. When facing densely built-up urban areas, scattering points at different heights overlap due to equal distances, causing three-dimensional objects to overlap on the range-azimuth plane. This results in the scattering and deformation characteristics of these objects at different heights being mixed and overlapping, making them difficult to distinguish and extract. In contrast, differential SAR four-dimensional imaging technology can effectively overcome the distortion problems caused by traditional imaging mechanisms, such as image compression, overlay, and top-bottom inversion. It can extract elevation information while effectively distinguishing deformation information in overlay areas, achieving range-azimuth-elevation-time four-dimensional imaging. This technology combines the advantages of traditional three-dimensional SAR imaging and interferometric SAR, and is of great value for constructing accurate four-dimensional urban environmental models, achieving high-precision target interpretation, urban area mapping, and assessing disaster impacts.

[0003] Currently, traditional D-TomoSAR four-dimensional imaging methods include array signal processing methods and compressed sensing methods. The former models multiple satellite observations as a two-dimensional spatial-temporal observation array, applying classical Capon and Music methods to solve for elevation information in the spatial dimension and deformation information in the temporal dimension, respectively. The latter, represented by compressed sensing, uses spectral estimation methods to decompose the observation signal model into elevation and deformation frequency components, applying spectral estimation algorithms to analyze the spectral peak positions to obtain elevation and deformation information. Because the accuracy of results obtained by array methods is relatively low, spectral estimation algorithms remain the mainstream approach. However, since four-dimensional imaging requires the estimation of two spectral components, the steering vector dimension on which traditional spectral estimation algorithms rely grows exponentially, leading to an exponential increase in computational complexity. This poses a significant challenge to large-scale spaceborne monitoring. Summary of the Invention

[0004] The purpose of this invention is to address the problems of severe noise pollution and excessive computational complexity in four-dimensional imaging observation data. This invention proposes a high-resolution tomographic SAR four-dimensional imaging method with spatiotemporal baseline decoupling processing. The method obtains noiseless signal estimation by using the low-rank constraint of the Hankel matrix and performs deformation spectrum estimation in a hierarchical manner using elevation priors, thereby achieving high-quality four-dimensional SAR imaging with low computational complexity. This method can significantly reduce computational complexity while maintaining accuracy.

[0005] Technical Solution: To achieve the above-mentioned objectives, this invention provides a high-resolution tomographic SAR four-dimensional imaging method with spatiotemporal baseline decoupling processing, the method comprising the following steps:

[0006] Step 1: Reconstruct the original satellite data into uniform baseline observation data, and use the low-rank property of the Hankel matrix to solve for its noiseless estimate;

[0007] Step 2: Based on noiseless estimation, the elevation information is inverted and used as a priori. A deformation spectrum estimation dictionary is constructed through spatiotemporal decoupling to estimate the deformation information.

[0008] Step 3: Use the maximum likelihood method to match the estimated elevation and deformation to achieve four-dimensional imaging.

[0009] Furthermore, the specific method for step 1 is as follows:

[0010] (1.1) Suppose the satellite performs N transit observations, then the observation vector for the a-th pixel in the region is: ,in, This represents the observation value of pixel a at the nth row, where n ranges from 1 to N. This represents a vector composed of all observations, which is a non-uniform linear array signal. It is represented by filling in virtual observations of 0, thus transforming the 1×N dimensional array... Reconstructing a 1×M dimensional uniform linear array signal If M > N, record the filling positions in the virtual observation dictionary. M and N are positive integers;

[0011] (1.2) Let for For noise-free estimation, first construct the computation Let the Hankel matrix W be... The initial value is a vector of all zeros. Matrix W is calculated through multiple iterations, with the following iterative expression:

[0012]

[0013] in, This represents the d-th iteration of W. express The d-th iteration, Indicates based on virtual observation dictionary The position index in the parentheses is used to set the position corresponding to the target vector within the parentheses to zero, p=N / M. This indicates the process of finding the Hankel matrix. The constructor is shown below:

[0014]

[0015] Among them, M s =floor(M / 2), where floor(.) represents rounding down. This represents the vector obtained after the (d+1)th iteration. The i-th element, where i ranges from 0 to M. s ;

[0016] (1.3) Find the mapping W in the low-dimensional subspace. s The expression to be solved is as follows:

[0017]

[0018] Where U is the right matrix obtained after W's SVD decomposition, U * Let be the conjugate matrix of U;

[0019] (1.4) Regarding W s W is obtained by applying threshold constraints. L The calculation expression is as follows:

[0020]

[0021] in, Indicates W s The q-th singular value obtained by performing SVD decomposition. Indicates W s The q-th eigenvector obtained after SVD decomposition express The conjugate;

[0022] (1.5) Regarding W L Inverse Hankel algorithm is used to obtain Noiseless estimation The calculation expression is as follows:

[0023]

[0024] in, This represents the inverse Hankel operator. W L The m-th inverse diagonal element, This means summing up all combinations of i and j that satisfy i + j = m. W L The element in the i-th row and j-th column, This means that each value obtained during the process of m taking values ​​from 0 to M-1 forms a column vector in sequence.

[0025] Furthermore, the method for step 2 is as follows:

[0026] (2.1) Noiseless estimation based on compressed sensing method Super-resolution elevation estimation is performed to obtain the elevation information of C overlapping targets within a single pixel of a SAR image. Construct a spatial spectrum based on known elevation information:

[0027]

[0028] in, For spatial frequency, It is the vertical baseline of the m-th observation. is the wavelength of the satellite radar system, and r is the distance from the satellite to the pixel, both in meters;

[0029] (2.2) Preset deformation rate grid , -1×10 -5 Up to 1×10 -5 Divide the interval into Q-1 equal parts, i.e. =-1×10 -5 , =1×10 -5 ,and The difference between each adjacent element is equal, and the value of h ranges from 1 to Q, where Q is the number of deformable meshes, which is a positive integer.

[0030] (2.3) Constructing a time spectrum for estimating the deformation rate:

[0031]

[0032] in, For time frequency, Represents the time baseline of the m-th observation, in days, and constructs a spectral estimation dictionary:

[0033]

[0034] in, Represents the Kronecker product.

[0035] Furthermore, the specific method for step 3 is as follows:

[0036] Traverse the deformation rate mesh in step 2.2, and Substitution Calculate the following formulas respectively:

[0037]

[0038] in, express The conjugate matrix, , This represents the noise-free estimate of the a-th pixel. The k-th element in express Conjugate;

[0039] The calculated Sort from highest to lowest, select the first C. For each selected Numerical and elevation information Compare the numerical values ​​of the elements and select the element with the closest value. At the same time, select the value of h corresponding to this value. This forms an elevation-deformation pair. For all C Performing the same operation yields C pairs of elevation deformations. Thus, the elevation and deformation information of the C stacked shelters within the a-th pixel unit are obtained. Using the abscissa of the a-th pixel unit in the original SAR image as the first dimension, the ordinate as the second dimension, the elevation information as the third dimension, and the deformation information as the fourth dimension, the point difference layering analysis four-dimensional imaging result is obtained. Traversing all pixels in the image, the elevation and deformation information of all targets in the entire image are solved, thereby obtaining the full scene difference layering analysis four-dimensional SAR image.

[0040] Furthermore, this invention proposes a high-resolution tomographic SAR four-dimensional imaging device with spatiotemporal baseline decoupling processing. The device includes a processor, a memory, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, it implements the steps of any of the high-resolution tomographic SAR four-dimensional imaging methods with spatiotemporal baseline decoupling processing described in the present invention.

[0041] Furthermore, the present invention proposes a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of a high-resolution tomographic SAR four-dimensional imaging method for spatiotemporal baseline decoupling processing as described in any one of the claims.

[0042] Beneficial effects: Compared with the prior art, the technical solution of the present invention has the following beneficial technical effects:

[0043] This invention proposes a high-resolution tomographic SAR four-dimensional imaging method with spatiotemporal baseline decoupling. Noiseless signal estimation is obtained through low-rank constraints of the Hankel matrix. Then, a traditional spectral estimation algorithm is used to first estimate the elevation frequency. Finally, the estimated elevation is used as prior information to estimate the deformation frequency and match it with the elevation using the maximum likelihood estimation principle. This method can exponentially reduce computational complexity, significantly improve observation efficiency in large-scale scenarios, enhance the real-time performance of deformation monitoring, and provide reliable technical support for applications such as urban disaster early warning. Attached Figure Description

[0044] Figure 1 This is a flowchart of the method of the present invention;

[0045] Figure 2 This is the third-dimensional elevation map of the four-dimensional imaging results from the LuTan-1 satellite;

[0046] Figure 3 This is the fourth-dimensional deformation map of the four-dimensional imaging results from the LuTan-1 satellite. Detailed Implementation

[0047] The embodiments of the present invention will now be described with reference to the accompanying drawings. The embodiments shown in the drawings are merely exemplary and intended to explain the principles of the present invention, and are not intended to limit the scope of the present invention.

[0048] like Figure 1 As shown, this invention provides a high-resolution tomographic SAR four-dimensional imaging method with spatiotemporal baseline decoupling processing, which includes the following steps:

[0049] Step 1: Reconstruct the original satellite data into uniform baseline observation data, and use the low-rank property of the Hankel matrix to solve for its noiseless estimate;

[0050] Step 2: Based on noiseless estimation, the elevation information is inverted and used as a priori. A deformation spectrum estimation dictionary is constructed through spatiotemporal decoupling to estimate the deformation information.

[0051] Step 3: Use the maximum likelihood method to match the estimated elevation and deformation to achieve four-dimensional imaging.

[0052] Furthermore, the specific method for step 1 is as follows:

[0053] (1.1) Suppose the satellite performs N transit observations, then the observation vector for the a-th pixel in the region is: ,in, This represents the observation value of pixel a at the nth row, where n ranges from 1 to N. This represents a vector composed of all observations, which is a non-uniform linear array signal. It is represented by filling in virtual observations of 0, thus transforming the 1×N dimensional array... Reconstructing a 1×M dimensional uniform linear array signal If M > N, record the filling positions in the virtual observation dictionary. ;

[0054] (1.2) Let for For noise-free estimation, first construct the computation Let the Hankel matrix W be... The initial value is a vector of all zeros. Matrix W is calculated through multiple iterations, with the following iterative expression:

[0055]

[0056] in, This represents the d-th iteration of W. express The d-th iteration, Indicates based on virtual observation dictionary The position index in the parentheses is used to set the position corresponding to the target vector within the parentheses to zero, p=N / M. This indicates the process of finding the Hankel matrix. The constructor is shown below:

[0057]

[0058] Among them, M s =floor(M / 2), where floor(.) represents rounding down. This represents the vector obtained after the (d+1)th iteration. The i-th element, where i ranges from 0 to M. s ;

[0059] (1.3) Find the mapping W in the low-dimensional subspace. s The expression to be solved is as follows:

[0060]

[0061] Where U is the right matrix obtained after W's SVD decomposition, U * Let be the conjugate matrix of U;

[0062] (1.4) Regarding W s W is obtained by applying threshold constraints. L The calculation expression is as follows:

[0063]

[0064] in, Indicates W s The q-th singular value obtained by performing SVD decomposition. Indicates W s The q-th eigenvector obtained after SVD decomposition express The conjugate;

[0065] (1.5) Regarding W L Inverse Hankel algorithm is used to obtain Noiseless estimation The calculation expression is as follows:

[0066]

[0067] in, This represents the inverse Hankel operator. W L The m-th inverse diagonal element, This means summing up all combinations of i and j that satisfy i + j = m. W L The element in the i-th row and j-th column, This means that each value obtained during the process of m taking values ​​from 0 to M-1 forms a column vector in sequence.

[0068] Furthermore, the method for step 2 is as follows:

[0069] (2.1) Noiseless estimation based on compressed sensing method Super-resolution elevation estimation is performed to obtain the elevation information of C overlapping targets within a single pixel of a SAR image. Construct a spatial spectrum based on known elevation information:

[0070]

[0071] in, For spatial frequency, It is the vertical baseline of the m-th observation. is the wavelength of the satellite radar system, and r is the distance from the satellite to the pixel, both in meters;

[0072] (2.2) Preset deformation rate grid , -1×10 -5 Up to 1×10 -5 Divide the interval into Q-1 equal parts, i.e. =-1×10 -5 , =1×10 -5 ,and The difference between each adjacent element is equal, and the value of h ranges from 1 to Q, where Q is the number of deformable meshes, which is a positive integer.

[0073] (2.3) Constructing a time spectrum for estimating the deformation rate:

[0074]

[0075] in, For time frequency, Represents the time baseline of the m-th observation, in days, and constructs a spectral estimation dictionary:

[0076]

[0077] in, Represents the Kronecker product.

[0078] Furthermore, the specific method for step 3 is as follows:

[0079] Traverse the deformation rate mesh in step 2.2, and Substitution Calculate the following formulas respectively:

[0080]

[0081] in, express The conjugate matrix, , This represents the noise-free estimate of the a-th pixel. The k-th element in express Conjugate;

[0082] The calculated Sort from highest to lowest, select the first C. For each selected Numerical and elevation information Compare the numerical values ​​of the elements and select the element with the closest value. At the same time, select the value of h corresponding to this value. This forms an elevation-deformation pair. For all C Performing the same operation yields C pairs of elevation deformations. Thus, the elevation and deformation information of the C stacked shelters within the a-th pixel unit are obtained. Using the abscissa of the a-th pixel unit in the original SAR image as the first dimension, the ordinate as the second dimension, the elevation information as the third dimension, and the deformation information as the fourth dimension, the point difference layering analysis four-dimensional imaging result is obtained. Traversing all pixels in the image, the elevation and deformation information of all targets in the entire image are solved, thereby obtaining the full scene difference layering analysis four-dimensional SAR image.

[0083] Furthermore, this invention proposes a high-resolution tomographic SAR four-dimensional imaging device with spatiotemporal baseline decoupling processing. The device includes a processor, a memory, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, it implements the steps of any of the high-resolution tomographic SAR four-dimensional imaging methods with spatiotemporal baseline decoupling processing described in the present invention.

[0084] Furthermore, the present invention proposes a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of a high-resolution tomographic SAR four-dimensional imaging method for spatiotemporal baseline decoupling processing as described in any one of the claims.

[0085] like Figure 2 and Figure 3 As shown, Figure 2 This is the third-dimensional elevation map of the four-dimensional imaging results from the LuTan-1 satellite; Figure 3 This is the fourth-dimensional deformation map of the four-dimensional imaging results from the LuTan-1 satellite.

[0086] The above description is merely a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention are included within the scope of protection of the present invention.

Claims

1. A high-resolution tomographic SAR four-dimensional imaging method with spatiotemporal baseline decoupling processing, characterized in that, The method includes the following steps: Step 1: Reconstruct the original satellite data into uniform baseline observation data, and use the low-rank property of the Hankel matrix to solve for its noiseless estimate; Step 2: Based on noiseless estimation, the elevation information is inverted and used as a priori. A deformation spectrum estimation dictionary is constructed through spatiotemporal decoupling to estimate the deformation information. Step 3: Use the maximum likelihood method to match the estimated elevation and deformation to achieve four-dimensional imaging; The specific method for step 2 is as follows: (2.1) Noiseless estimation based on compressed sensing method Super-resolution elevation estimation is performed to obtain the elevation information of C overlapping targets within a single pixel of a SAR image. Construct a spatial spectrum based on known elevation information: ； in, For spatial frequency, It is the vertical baseline of the m-th observation. is the wavelength of the satellite radar system, and r is the distance from the satellite to the pixel, both in meters; (2.2) Preset deformation rate grid , -1×10 -5 Up to 1×10 -5 Divide the interval into Q-1 equal parts, i.e. =-1×10 -5 , =1×10 -5 ,and The difference between each adjacent element is equal, and the value of h ranges from 1 to Q, where Q is the number of deformable meshes, which is a positive integer. (2.3) Constructing a time spectrum for estimating the deformation rate: ； in, For time frequency, Represents the time baseline of the m-th observation, in days, and constructs a spectral estimation dictionary: ； in, Represents the Kronecker product; The specific method for step 3 is as follows: Traverse the deformation rate mesh in step 2.2, and Substitution Calculate the following formulas respectively: ； in, express The conjugate matrix, , This represents the noise-free estimate of the a-th pixel. The k-th element in express Conjugate; The calculated Sort from highest to lowest, select the first C. For each selected Numerical and elevation information Compare the numerical values ​​of the elements and select the element with the closest value. At the same time, select the value of h corresponding to this value. This forms an elevation-deformation pair. For all C Performing the same operation yields C pairs of elevation deformations. Thus, the elevation and deformation information of the C stacked shelters within the a-th pixel unit are obtained. Using the abscissa of the a-th pixel unit in the original SAR image as the first dimension, the ordinate as the second dimension, the elevation information as the third dimension, and the deformation information as the fourth dimension, the difference-layer analysis four-dimensional imaging result of this point is obtained. Traversing all pixels in the image, the elevation and deformation information of all targets in the entire image are solved to obtain the full-scene difference-layer analysis four-dimensional SAR image.

2. The high-resolution tomographic SAR four-dimensional imaging method with spatiotemporal baseline decoupling processing according to claim 1, characterized in that, The specific method for step 1 is as follows: (1.1) Suppose the satellite performs N transit observations, then the observation vector for the a-th pixel in the region is: ,in, This represents the observation value of pixel a at the nth row, where n ranges from 1 to N. This represents a vector composed of all observations, which is a non-uniform linear array signal. It is represented by filling in virtual observations of 0, thus transforming the 1×N dimensional array... Reconstructing a 1×M dimensional uniform linear array signal If M > N, record the filling positions in the virtual observation dictionary. M and N are positive integers; (1.2) Let for The noiseless estimate is constructed by computation. Let the Hankel matrix W be... The initial value is a vector of all zeros. Matrix W is calculated through multiple iterations, with the following iterative expression: ； in, This represents the d-th iteration of W. express The d-th iteration, Indicates based on virtual observation dictionary The position index in the parentheses is used to set the position corresponding to the target vector within the parentheses to zero, p=N / M. This indicates the process of finding the Hankel matrix. The constructor is shown below: ； Among them, M s =floor(M / 2), where floor(.) represents rounding down. This represents the vector obtained after the (d+1)th iteration. The i-th element, where i ranges from 0 to M. s ; (1.3) Find the mapping W in the low-dimensional subspace. s The expression to be solved is as follows: ； Where U is the right matrix obtained after W's SVD decomposition, U * Let be the conjugate matrix of U; (1.4) Regarding W s W is obtained by applying threshold constraints. L The calculation expression is as follows: ； in, Indicates W s The q-th singular value obtained by performing SVD decomposition. Indicates W s The q-th eigenvector obtained after SVD decomposition express The conjugate; (1.5) Regarding W L Inverse Hankel algorithm is used to obtain Noiseless estimation The calculation expression is as follows: ； in, This represents the inverse Hankel operator. W L The m-th inverse diagonal element, This means summing up all combinations of i and j that satisfy i + j = m. W L The element in the i-th row and j-th column, This means that each value obtained during the process of m taking values ​​from 0 to M-1 forms a column vector in sequence.

3. A high-resolution tomographic SAR four-dimensional imaging device with spatiotemporal baseline decoupling processing, characterized in that, The device includes a processor, a memory, and a computer program stored in the memory and executable on the processor. When executed by the processor, the computer program implements the steps of the high-resolution tomographic SAR four-dimensional imaging method for spatiotemporal baseline decoupling processing as described in claim 1 or 2.

4. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of a high-resolution tomographic SAR four-dimensional imaging method for spatiotemporal baseline decoupling processing as described in claim 1 or 2.

Images