A polarimetric time-series InSAR method based on adaptive decomposition of coherence matrix

Abstract

The application discloses a polarimetric time-series InSAR method based on coherent matrix adaptive decomposition, which comprises the following steps: acquiring time-series multi-polarization time-series SAR images of a research area, pre-processing the SAR images and identifying homogeneous pixels; performing polarimetric optimization under different criteria according to PS and DS targets; calculating time-series coherence by using the polarimetric optimized interference phase of eigenvalue decomposition, selecting high-quality pixels, using the polarimetric optimized differential interferogram, performing polarimetric time-series interferometric measurement, obtaining deformation monitoring points, and completing deformation monitoring of the research area; the application firstly distinguishes PS and DS targets through homogeneous pixel identification, searches for an optimal scattering mechanism by using eigenvalue decomposition and the BEST algorithm, optimizes the PS and DS targets by using different strategies, so that the quality of the interferogram phase is obviously improved, the noise of the interferogram phase is reduced, and the interference fringes are clearer.

Description

Technical Field

[0001] This invention relates to the field of surface deformation monitoring technology, specifically to a polarization time-series InSAR method based on adaptive decomposition of the coherence matrix. Background Technology

[0002] Surface deformation includes ground subsidence and land collapse. Traditional surface deformation monitoring technologies and methods mainly include leveling, GPS, electronic distance measuring instruments, and underground bedrock and stratification markers. However, these "contact" monitoring methods require significant manpower, material resources, and financial investment. They typically require surveyors to enter the monitoring area, increasing the difficulty of the monitoring work and even posing certain safety hazards. Furthermore, these traditional point-based monitoring technologies and methods suffer from drawbacks such as low spatial resolution of monitoring results, long monitoring cycles, small monitoring ranges, and high costs.

[0003] The rapid development of Synthetic Aperture Radar (SAR) systems has provided abundant data sources for Earth observation and has been widely applied in deformation monitoring. Traditional PS-InSAR technology, by selecting permanent scattering targets (PS), can achieve good results in urban areas dominated by man-made structures such as buildings, bridges, and dams. However, for distributed scattering targets (DS) with complex and variable scattering characteristics, low echo signal strength, and weak backscattering ability, such as vegetation, bare soil, deserts, and roads, the results of PS-InSAR technology are often less than ideal. In reality, the surface features are complex and diverse, and PS and DS targets often coexist on the ground. Among existing technologies, Polarimetric Interferometry SAR (PolInSAR) has been proposed. Compared with single-polarization data for monitoring surface deformation, multi-polarization data can make full use of the different responses of different scatterers to polarization signals, improve interferometric coherence, optimize interferometric phase quality, increase monitoring point density, and improve monitoring point quality. Summary of the Invention

[0004] To address the aforementioned technical shortcomings, the present invention aims to provide a polarization-time-series InSAR method based on adaptive coherence matrix decomposition. This method first distinguishes between PS and DS targets through homogeneous pixel identification, then utilizes eigenvalue decomposition and the BEST algorithm to search for the optimal scattering mechanism, applying different strategies to optimize the PS and DS targets. This results in a significant improvement in the phase quality of the interferogram, a reduction in phase noise, and clearer interference fringes.

[0005] To solve the above-mentioned technical problems, the present invention adopts the following technical solution:

[0006] This invention provides a polarization-time series InSAR method based on adaptive coherence matrix decomposition, comprising the following methods:

[0007] S1. Acquire and study multipolar time-series SAR images of the monitoring area, preprocess the SAR images and perform homogeneous pixel identification to obtain PS and DS targets;

[0008] S2. Polarization optimization under different criteria based on PS and DS objectives, including the following steps:

[0009] S21. Construct polarization scattering vectors for multipolarization data under the Pauli basis;

[0010] S22. Construct the PS target polarization coherence matrix T PS And the polarization vector interference phase of the PS target;

[0011] S23. The polarization coherence matrix T constructed for the DS target DS Perform polarization adaptive MMSE filtering;

[0012] S24, Eigenvalue decomposition of polarization coherence matrix;

[0013] S25. For the PS and DS targets, respectively, D is used. A Polarization optimization is performed using the criterion and the γ criterion;

[0014] S3: Utilize the polarization-optimized interferometric phase obtained by eigenvalue decomposition to calculate temporal coherence, select high-quality pixels, and use the polarization-optimized differential interferogram to perform polarization-temporal interferometry to obtain deformation monitoring points and complete deformation monitoring of the study area.

[0015] Preferably, the specific method of step S1 is as follows: After acquiring and studying multi-polarization time-series SAR images of the monitoring area, the SAR images are preprocessed by registration, cropping, and geocoding. Using the intensity information of the multi-polarization time-series SAR images, the multi-polarization SAR images are divided into PS and DS targets under their respective polarizations through a homogeneous pixel identification method based on statistical tests. Then, the amplitude deviation D is used... A Criteria for D in DS objectives A Pixels with values ​​less than 0.33 are converted into PS targets. Since different ground features have different intensity information in different polarizations, after identifying homogeneous pixels in multipolar SAR images, the union of the identification results of homogeneous points in different polarizations is taken to obtain the final PS and DS targets.

[0016] Preferably, the specific method of step S21 is as follows:

[0017] Polarimetric radar transmits and receives electromagnetic pulse signals in two different ways: vertical and horizontal. Depending on the signal transmission and reception method during antenna imaging, it records images as different polarizations, including four types: VV, VH, HH, and HV. A scattering matrix S is constructed based on the polarization data.

[0018]

[0019] In the formula, S represents the scattering matrix, S hh S represents a single-look complex SAR image under HH polarization. vv S represents a single-view complex SAR image under VV polarization. vh S represents a single-look complex SAR image under VH polarization. hv Represents a single-look complex SAR image under HV polarization;

[0020] For fully polarized data, construct the scattering vector k under the Pauli basis. q as follows:

[0021]

[0022] In the formula, q represents total polarization, according to the reciprocity theorem S vh =S hv Therefore, only S is used. vh This indicates cross-polarization, and the same applies below. T represents the matrix transpose operation.

[0023] For dual-polarized data, there are several different polarization combinations:

[0024] For SAR image data combining HH and VV, construct the scattering vector k under the Pauli basis. d as follows:

[0025]

[0026] For dual-polarization SAR image data with a combination of co-polarization (HH, VV) and cross-polarization (VH, HV), construct the scattering vector k under the Pauli basis. d as follows:

[0027] k d =[S xx 2S vh ] T

[0028] In the formula, d represents dual polarization; S xx Represents single-look complex SAR images under HH or VV polarization;

[0029] Polarization vector interferometry differs from single-polarization scalar interferometry. Single-polarization scalar interferometry involves the conjugate multiplication between the first and second registered images, while polarization vector interferometry requires constructing the scattering vectors of each of the two polarized images and introducing a scattering mechanism to perform an outer product to obtain the interference phase. The scattering mechanism is represented by the unit complex projection vector ω.

[0030] For fully polarized data, ω q We obtain it from the following formula:

[0031]

[0032] In the formula, a, β and ψ represent the polarization parameters of the scattering mechanism vector, respectively, and j and e represent the imaginary unit and the natural constant, respectively;

[0033] For bipolar data, ω d We obtain it from the following formula:

[0034]

[0035] To perform polarization vector interferometry, the unit complex vector ω is used based on the specific scattering mechanism of the two images. i The scattering vector k of the first and second images i μ is obtained by projecting the following formulas respectively. i :

[0036] μ i =ω i H ·k i i = 1, 2

[0037] In the formula, i represents the i-th image, H represents the conjugate transpose, and μ i Each component represents a specific scattering mechanism, similar to a single-view complex image in single polarization.

[0038] Preferably, the specific method of step S22 is as follows:

[0039] Coherence matrices can effectively reduce the impact of speckle noise, while spatially averaged coherence matrices lose spatial resolution. Therefore, to preserve spatial resolution, this study uses temporal coherence matrices and directly constructs polarization coherence matrices for PS targets:

[0040]

[0041] In the formula k j This represents the j-th polarization scattering vector;

[0042] The polarization vector interference phase is generated based on the two complex scalars:

[0043] I = μ1·μ2 *

[0044] In the formula, * represents the complex conjugate operator.

[0045] Preferably, the specific method of step S23 is as follows:

[0046] DS targets are mostly homogeneous objects such as vegetation, bare soil, and roads. Their echo signal strength is low, backscattering ability is weak, and phase stability is poor. Therefore, after identifying homogeneous pixels, homogeneous filtering is required for DS targets to improve their phase quality. There are various homogeneous pixel identification algorithms based on statistical tests, such as BWS hypothesis testing, two-sample t hypothesis testing, generalized likelihood ratio hypothesis testing, FaSHPS test algorithm, and HTCI algorithm.

[0047] Construct the polarization coherence matrix T for the DS target DS :

[0048]

[0049] In the formula, k1 and k2 represent the scattering vectors under the Pauli basis, and T 11 and T 22 Ω represents the coherence matrix. 12 Representing the polarization coherence matrix, in MMSE polarization filtering of homogeneous pixels, the span image of the polarization coherence matrix T needs to be calculated first:

[0050] span = Trace(T) 11 )+Trace(T 22 )

[0051] In the formula, Trace represents the trace of the matrix, and the filter weight coefficient b is calculated using the value of span according to the following formula:

[0052]

[0053] In the formula, SHPS represents homogeneous pixels, and the coef value is 0.51. Finally, the MMSE filtering of homogeneous pixels is completed by the following formula, and the polarization coherence matrix T of the DS target is obtained. DS Complete MMSE filtering:

[0054] T DS =T mean +b(T ori -T mean )

[0055] In the formula T mean T represents the average coherence matrix of homogeneous pixels within the window. ori This represents the original coherence matrix.

[0056] Preferably, the specific method of step S24 is as follows:

[0057] In this step, T can be PS and T DS The polarization coherence matrices of the two targets can be combined into one for eigenvalue decomposition, or they can be decomposed separately. In this invention, the two are combined to form the polarization coherence matrix T of the entire image for eigenvalue decomposition; the time-averaged polarization coherence matrix can be decomposed as follows:

[0058]

[0059] In the formula λ represents the average time-polarization coherence matrix. i Represents the eigenvalue, u i This represents the eigenvector, and num represents the number of polarization channels;

[0060] When using fully polarized data, num = 3, we obtain three eigenvalues ​​λ1≥λ2≥λ3≥0 and the corresponding eigenvectors u1, u2, u3;

[0061] When using bipolarized data, num = 2, we obtain two eigenvalues ​​λ1 ≥ λ2 ≥ 0 and corresponding eigenvectors u1 and u2.

[0062] Preferably, the specific method of step S24 is as follows:

[0063] Step 1: PS target D A Criterion optimization:

[0064] The phase quality evaluation index D is obtained by comparing the standard deviation and mean of the amplitude of temporal SAR images. A Used to optimize PS targets, D A The smaller the value, the better the phase quality, as shown by the following formula:

[0065]

[0066] In the formula The standard deviation of the amplitude under Pol polarization. This represents the mean amplitude under Pol polarization, where Pol is the VV, HH, or VH polarization channel. For polarization time-series images, D under different scattering mechanisms ω represents... A We obtain it from the following formula:

[0067]

[0068] in

[0069]

[0070] In the formula, SM represents the scattering mechanism, N represents the number of SAR images, the upper horizontal line indicates that the average value is taken, and the complex vector ω of the scattering mechanism is the eigenvector u obtained above. i ;

[0071] Applying the BEST algorithm at the cell level to fully polarized PS targets:

[0072]

[0073] In the formula This is the pixel with the best phase quality at that point;

[0074] The BEST algorithm is used for bipolar PS targets:

[0075]

[0076] Step 2: Optimization of the γ criterion for the DS objective:

[0077] Coherence γ has a corresponding functional relationship with the standard deviation of the interferometric phase and is the evaluation standard for the quality of the interferometric phase. The γ standard is used to optimize the DS target. The larger the γ value, the better the phase quality.

[0078] γ for the Pol-polarized interferogram is calculated using the following formula:

[0079]

[0080] In the formula, S1 and S2 represent the two SAR images that constitute the interferogram, E represents the expectation operator, and the mathematical expression of the multipolarization γ is:

[0081]

[0082] Polarization optimization is performed on the DS objective using the γ criterion, and the average coherence is calculated.

[0083]

[0084] In the formula, M represents the number of interferograms, and j represents the j-th interferogram;

[0085] Applying the BEST algorithm at the cell level to fully polarimetric DS targets:

[0086]

[0087] In the formula This is the pixel with the best phase quality at that point;

[0088] Applying the BEST algorithm at the cell level to dual-polarized DS targets:

[0089]

[0090] The phase values ​​of the PS and DS pixels with the best phase quality under each pixel are combined to form the final polarization optimized interferometric phase.

[0091] The beneficial effects of this invention are as follows: First, it distinguishes between PS and DS targets through homogeneous pixel identification. Then, using eigenvalue decomposition and the BEST algorithm, it searches for the optimal scattering mechanism and applies different strategies to optimize the PS and DS targets. By fully utilizing the different responses of different scatterers to polarization signals and leveraging multi-polarization information, the phase quality of the interferogram is significantly improved, phase noise is reduced, and interference fringes are clearer. The number of deformation monitoring points is significantly increased, and through the optimized algorithm, monitoring points can be obtained in areas not detected in the original image. Attached Figure Description

[0092] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0093] Figure 1 This is a flowchart of a polarization-time series InSAR method based on adaptive coherence matrix decomposition provided in an embodiment of the present invention.

[0094] Figure 2 These are comparison images of differential interference details before and after polarization optimization provided in this embodiment of the invention;

[0095] Figure 3 This is a comparison chart of TPC values ​​before and after polarization optimization provided in an embodiment of the present invention;

[0096] Figure 4 This is a statistical curve of TPC before and after optimization provided in an embodiment of the present invention;

[0097] Figure 5 This is a comparison chart of time-series settlement monitoring results before and after polarization optimization provided in an embodiment of the present invention;

[0098] Figure 6 This is a detailed comparison diagram of the time-series sedimentation results before and after polarization optimization provided in the embodiments of the present invention. Detailed Implementation

[0099] 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.

[0100] like Figure 1 A polarization time-series InSAR method based on coherence matrix adaptive decomposition is proposed. This embodiment uses 34 dual-polarization Sentinel-1 images covering Shanghai Pudong Airport from May 15, 2016 to December 24, 2017 (33 interferograms were obtained by using a single master image method), with a total of 700×2400 image pixels.

[0101] The specific implementation plan is as follows:

[0102] Step 1: Acquire dual-polarized Sentinel-1 data, preprocess the Sentinel-1 SAR image and divide it into PS and DS targets.

[0103] The original VV and VH polarimetric SAR images were preprocessed by registration, cropping, geocoding, etc., and statistical verification algorithms were used to identify homogeneous pixels and classify PS and DS targets.

[0104] Step Two:

[0105] Based on the PS and DS targets obtained in step one, construct the polarization scattering matrix S:

[0106]

[0107] In the formula, S represents the scattering matrix, S vv S represents a single-view complex SAR image under VV polarization. vh This represents a single-look complex SAR image under VH polarization.

[0108] The polarization scattering vector k is formed under the Pauli basis. i :

[0109] k i =[S vv，i 2S vh，i ] T

[0110] In the formula S vv，i S represents the i-th single-look complex SAR image under VV polarization. vh，i Let represent the i-th single-look complex SAR image under VH polarization, and T denote the matrix transpose operation.

[0111] Based on the polarization scattering vector, the polarization coherence matrix of the PS target and the polarization vector interference phase of the PS target are obtained:

[0112] In the formula k j This represents the j-th polarization scattering vector.

[0113]

[0114] In the formula, H represents the conjugate transpose, and k j This represents the j-th polarization scattering vector.

[0115] I = μ1·μ2 *

[0116] In the formula, * is the complex conjugate operator, and I is the polarization vector interference phase.

[0117] Obtain the DS polarization coherence matrix and perform MMSE filtering:

[0118]

[0119] In the formula, k1 and k2 represent the scattering vectors under the Pauli basis, and T 11 and T 22 Ω represents the coherence matrix. 12 This represents the polarization coherence matrix. In MMSE polarization filtering of homogeneous pixels, the span image of the polarization coherence matrix T needs to be calculated first:

[0120] span = Trace(T) 11 )+Trace(T 22 )

[0121] In the formula, Trace represents the trace of the matrix, and the filter weight coefficient b is calculated using the value of span according to the following formula:

[0122]

[0123] In the formula, SHPS represents homogeneous pixels, and the coef value is 0.51. Finally, the MMSE filtering of homogeneous pixels is completed by the following formula, and the polarization coherence matrix T of the DS target is obtained. DS Complete MMSE filtering:

[0124] T DS =T mean +b(T ori -T mean )

[0125] After merging the PS and DS polarization coherence matrices, eigenvalue decomposition is performed.

[0126]

[0127] In the formula λ represents the average time-polarization coherence matrix. i Represents the eigenvalue, u i Let num represent the eigenvectors, and num represent the number of polarization channels. Sentinel-1 data is bipolar data, so num = 2, which will yield two eigenvalues ​​λ1 ≥ λ2 ≥ 0 and corresponding eigenvectors u1 and u2.

[0128] D for PS target A Criterion optimization:

[0129] Calculate the original VV and VH polarimetric SAR images respectively, and use the obtained eigenvectors u1 and u2 as the D of the PS target in the complex vector ω of the scattering mechanism. A value.

[0130]

[0131] In the formula The standard deviation of the amplitude under Pol polarization. This represents the mean amplitude under Pol polarization, where Pol can be either the VV or VH polarization channel. For polarization time-series images, D under different scattering mechanisms ω... A It can be obtained through the following formula:

[0132]

[0133] in

[0134]

[0135] In the formula, SM represents the scattering mechanism, N represents the number of SAR images, the upper horizontal line indicates that the average value is taken, and the complex vector ω of the scattering mechanism is the eigenvector u obtained above. i .

[0136] Applying the BEST algorithm at the cell level to fully polarized PS targets:

[0137]

[0138] In the formula This is the pixel with the best phase quality at that point.

[0139] Optimize the DS objective using the γ criterion:

[0140] For interferograms polarized by VV or VH, γ can be calculated using the following formula:

[0141]

[0142] In the formula, S1 and S2 represent the two SAR images constituting the interferogram, and E represents the expectation operator, which is achieved by averaging homogeneous pixels in this invention. The mathematical expression for multipolarization γ is:

[0143]

[0144] Polarization optimization is performed on the DS objective using the γ criterion, and the average coherence is calculated.

[0145]

[0146] In the formula, M represents the number of interferograms, and j represents the j-th interferogram.

[0147] Applying the BEST algorithm at the cell level to fully polarimetric DS targets:

[0148]

[0149] In the formula This is the pixel with the best phase quality at that point.

[0150] The phase values ​​of the PS and DS pixels with the best phase quality under each pixel are combined to form the final polarization optimized interferometric phase.

[0151] Step 3: Monitor surface deformation based on polarization-optimized differential interferometry;

[0152] By utilizing the polarization-optimized interferometric phase obtained from eigenvalue decomposition, the temporal average coherence is calculated, high-quality pixel selection is performed, and polarization-optimized differential interferograms are used to conduct polarization-time interferometry measurements to obtain deformation monitoring points, thus completing deformation monitoring of the study area. To demonstrate the monitoring effect of this invention, a comparison is made with the monitoring effect of the commonly used PS-InSAR method (based on Sentinel-1 VV polarization data). In this comparative embodiment, the same parameters are set for the time-series processing of the differential interferogram.

[0153] First, the interferogram results are compared. The original VV polarization interferogram is shown below. Figure 2 The interference phase noise level is relatively high, and the quality is poor, with problems such as blurred interference phase and incoherence in the outline and edge areas of ground objects. Interference fringes exist in the edge area of ​​the airport terminal building, but they are not smooth and continuous. There are almost no interference fringes in the distributed scatterer area, and there is a relatively serious incoherence phenomenon. After polarization optimization by the method proposed in this invention, the interference pattern fringes are clearer and more continuous, preserving the edge outline details of buildings and other structures, and the interference phase noise is significantly reduced. The spatial distribution is more continuous, smooth, and stable.

[0154] Then, temporal phase coherence (TPC) is used to quantitatively evaluate the interferogram quality. The phase quality of the differential interferogram is calculated by using the noise level of each pixel across all interferograms, which can be obtained by the following formula:

[0155]

[0156] In the formula, M is the number of interferograms, and ψ noisei ψ is the noise phase corresponding to the i-th interferogram. noisei The window size (considering SAR image resolution and land cover in the study area, a 15×15 window was used in this study) was obtained based on the interferometric phase value estimation of the neighborhood pixels.

[0157] like Figure 3 and Figure 4 As shown, a larger TPC value corresponds to better phase quality of the interferogram. It can be clearly observed that the method proposed in this invention yields a greater number of high-quality pixels, further demonstrating that the proposed method can effectively improve the phase quality of differential interferogram sets and increase the number of high-quality pixels when selecting subsequent monitoring points.

[0158] In this embodiment, a TPC > 0.9 was selected as the threshold for high-quality pixel selection. The commonly used open-source software StaMPS was used for time-series analysis and deformation calculation, resulting in the deformation monitoring results shown below. Figure 5 As shown in the figure, the deformation results obtained by the two methods show a basically consistent trend. The monitoring points for the two methods are 20845 and 58697, respectively. The method of this invention increases the number of monitoring points obtained by the original VV polarization method by 182%. The method proposed in this invention obtains more details of surface deformation. Some detailed surface deformation monitoring results are shown in the figure. Figure 6 As shown, take Figure 5 Comparing the two local images, designated as Local 1 and Local 2, a significant increase in the density of monitoring points can be observed in both regions C and D.

[0159] The method of this invention has the characteristics of significant optimization effect, high monitoring point density, and fast calculation efficiency. The results are consistent with those of commonly used single-polarization monitoring methods, which shows the effectiveness and reliability of the method of this invention.

[0160] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A polarization-time series InSAR method based on adaptive decomposition of the coherence matrix, characterized in that, Including the following methods: S1. Acquire and study multipolar time-series SAR images of the monitoring area, preprocess the SAR images and perform homogeneous pixel identification to obtain PS and DS targets; S2. Polarization optimization under different criteria based on PS and DS objectives, including the following steps: S21. Construct polarization scattering vectors for multipolarization data under the Pauli basis; S22. Construct the PS target polarization coherence matrix T PS And the polarization vector interference phase of the PS target; S23. The polarization coherence matrix T constructed for the DS target DS Perform polarization adaptive MMSE filtering; S24, Eigenvalue decomposition of polarization coherence matrix; S25. For the PS and DS targets, respectively, D is used. A Polarization optimization is performed using the criterion and the γ criterion; S3. Using the polarization-optimized interference phase obtained from eigenvalue decomposition, calculate the temporal coherence, select high-quality pixels, use the polarization-optimized differential interferogram to perform polarization temporal interferometry, obtain deformation monitoring points, and complete the deformation monitoring of the study area.

2. The polarization-time series InSAR method based on adaptive decomposition of the coherence matrix as described in claim 1, characterized in that, The specific method for step S1 is as follows: After acquiring and studying multi-polarization temporal SAR images of the monitoring area, the SAR images are preprocessed by registration, cropping, and geocoding. Utilizing the intensity information of the multi-polarization temporal SAR images, a homogeneous pixel identification method based on statistical tests is used to classify the multi-polarization SAR images into PS and DS targets under their respective polarizations. Then, the amplitude deviation D is used... A Criteria for D in DS objectives A Pixels with values ​​less than 0.33 are converted into PS targets. Since different ground features have different intensity information in different polarizations, after identifying homogeneous pixels in multipolar SAR images, the union of the identification results of homogeneous points in different polarizations is taken to obtain the final PS and DS targets.

3. The polarization-time series InSAR method based on adaptive decomposition of the coherence matrix as described in claim 1, characterized in that, The specific method for step S21 is as follows: Polarimetric radar transmits and receives electromagnetic pulse signals in two different ways: vertical and horizontal. Depending on the signal transmission and reception method during antenna imaging, it records different polarization images, resulting in four polarization modes: VV, VH, HH, and HV. A scattering matrix S is constructed based on the polarization data. In the formula, s represents the scattering matrix, s hh S represents a single-look complex SAR image under HH polarization. vv S represents a single-view complex SAR image under VV polarization. vh S represents a single-look complex SAR image under VH polarization. hv Represents a single-look complex SAR image under HV polarization; For fully polarized data, construct the scattering vector k under the Pauli basis. q as follows: In the formula, q represents total polarization, according to the reciprocity theorem S vh =S hv Therefore, only S is used. vh This indicates cross-polarization, and the same applies below. T represents the matrix transpose operation. For dual-polarized data, there are several different polarization combinations: For SAR image data combining HH and VV, construct the scattering vector k under the Pauli basis. d as follows: For dual-polarization SAR image data with a combination of co-polarization (HH, VV) and cross-polarization (VH, HV), construct the scattering vector k under the Pauli basis. d as follows: k d =[S xx ,2S vh ] T In the formula, d represents dual polarization; S xx Represents single-look complex SAR images under HH or VV polarization; Polarization vector interferometry differs from single-polarization scalar interferometry. Single-polarization scalar interferometry involves the conjugate multiplication between the first and second registered images, while polarization vector interferometry requires constructing the scattering vectors of each of the two polarized images and introducing a scattering mechanism to perform an outer product to obtain the interference phase. The scattering mechanism is represented by the unit complex projection vector ω. For fully polarized data, ω q We obtain it from the following formula: In the formula, α, β and ψ represent the polarization parameters of the scattering mechanism vector, respectively, and j and e represent the imaginary unit and the natural constant, respectively; For bipolar data, ω d We obtain it from the following formula: To perform polarization vector interferometry, the unit complex vector ω is used based on the specific scattering mechanism of the two images. i The scattering vector k of the first and second images i μ is obtained by projecting the following formulas respectively. i : m i =ω i H ·k i ,i=1,2 In the formula, i represents the i-th image, H represents the conjugate transpose, and μ i Each component represents a specific scattering mechanism, similar to a single-view complex image in single polarization.

4. The polarization-time series InSAR method based on adaptive decomposition of the coherence matrix as described in claim 3, characterized in that, The specific method for step S22 is as follows: Using the time coherence matrix, the polarization coherence matrix is ​​directly constructed for the PS target: In the formula k j This represents the j-th polarization scattering vector; The polarization vector interference phase is generated based on the two complex scalars: I=μ1·μ2 * In the formula, * represents the complex conjugate operator.

5. The polarization-time series InSAR method based on adaptive decomposition of the coherence matrix as described in claim 4, characterized in that, The specific method for step S23 is as follows: Construct the polarization coherence matrix T for the DS target DS In the formula, k1 and k2 represent the scattering vectors under the Pauli basis, and T 11 and T 22 Ω represents the coherence matrix. 12 Representing the polarization coherence matrix, in MMSE polarization filtering of homogeneous pixels, the span image of the polarization coherence matrix T needs to be calculated first: span=Trace(T 11 )+Trace(T 22 ) In the formula, Trace represents the trace of the matrix, and the filter weight coefficient b is calculated using the value of span according to the following formula: In the formula, SHPS represents homogeneous pixels, and the coef value is 0.

51. Finally, the MMSE filtering of homogeneous pixels is completed by the following formula, and the polarization coherence matrix T of the DS target is obtained. DS Complete MMSE filtering: T DS =T mean +b(T ori -T mean ) In the formula T mean T represents the average coherence matrix of homogeneous pixels within the window. ori This represents the original coherence matrix.

6. The polarization-time series InSAR method based on adaptive decomposition of the coherence matrix as described in claim 5, characterized in that, The specific method for step S24 is as follows: T PS and T DS The polarization coherence matrices of the two targets are merged to form the polarization coherence matrix T of the entire image. Eigenvalue decomposition is then performed on the time-series averaged polarization coherence matrix, which is then decomposed as follows: In the formula λ represents the average time-polarization coherence matrix. i Represents the eigenvalue, u i This represents the eigenvector, and num represents the number of polarization channels; When using fully polarized data, num = 3, we obtain three eigenvalues ​​λ1≥λ2≥λ3≥0 and the corresponding eigenvectors u1, u2, u3; When using bipolarized data, num = 2, we obtain two eigenvalues ​​λ1 ≥ λ2 ≥ 0 and corresponding eigenvectors u1 and u2.

7. The polarization-time series InSAR method based on adaptive decomposition of the coherence matrix as described in claim 6, characterized in that, The specific method for step S24 is as follows: Step 1: PS target D A Criterion optimization: The phase quality evaluation index D is obtained by comparing the standard deviation and mean of the amplitude of temporal SAR images. A Used to optimize PS targets, D A The smaller the value, the better the phase quality, as shown by the following formula: In the formula The standard deviation of the amplitude under Pol polarization. This represents the mean amplitude under Pol polarization, where Pol is the VV, HH, or VH polarization channel. For polarization time-series images, D under different scattering mechanisms ω represents... A We obtain it from the following formula: in In the formula, SM represents the scattering mechanism, N represents the number of SAR images, the upper horizontal line indicates that the average value is taken, and the complex vector ω of the scattering mechanism is the eigenvector u obtained above. i ; Applying the BEST algorithm at the cell level to fully polarized PS targets: In the formula This is the pixel with the best phase quality at that point; The BEST algorithm is used for bipolar PS targets: Step 2: Optimization of the γ criterion for the DS objective: Coherence γ has a corresponding functional relationship with the standard deviation of the interferometric phase and is the evaluation standard for the quality of the interferometric phase. The γ standard is used to optimize the DS target. The larger the γ value, the better the phase quality. γ for the Pol-polarized interferogram is calculated using the following formula: In the formula, S1 and S2 represent the two SAR images that constitute the interferogram, E represents the expectation operator, and the mathematical expression of the multipolarization γ is: Polarization optimization is performed on the DS objective using the γ criterion, and the average coherence is calculated. In the formula, M represents the number of interferograms, and j represents the j-th interferogram; Applying the BEST algorithm at the cell level to fully polarimetric DS targets: In the formula This is the pixel with the best phase quality at that point; Applying the BEST algorithm at the cell level to dual-polarized DS targets: The phase values ​​of the PS and DS pixels with the best phase quality under each pixel are combined to form the final polarization optimized interferometric phase.

Images