Interference phase optimization method based on total power polarization and non-local phase connection

Abstract

This invention provides an interferometric phase optimization method based on total power polarization (TP) and nonlocal phase concatenation, belonging to the field of synthetic aperture radar (SAR) information processing. The method includes acquiring multi-polarization SAR images of the study area and constructing a total power interferogram (TP). Pixels are classified into DS (Distributed Dynamics) pixels and PS (Power-Side) pixels based on a classification threshold. The TP syntactic coherence matrix of the DS pixels is calculated based on homogeneous pixels. The similarity between the TP syntactic coherence matrix of the reference center pixel and homogeneous pixels is calculated. After removing pixels with abnormal similarity, nonlocal NL weighting is performed. Phase concatenation is then applied to the weighted TP syntactic coherence matrix to obtain the optimized temporal phase sequence for the DS pixels. For PS pixels, the TP syntactic phase is retained based on the TP syntactic coherence matrix, resulting in an optimized differential interferogram. Temporal deformation monitoring of the study area is achieved. Simultaneously considering both PS and DS pixels improves the quality of the interferometric phase.

Description

Technical Field

[0001] This invention belongs to the field of synthetic aperture radar information processing, specifically involving an interferometric phase optimization method based on total power polarization and nonlocal phase connection. Background Technology

[0002] Synthetic Aperture Radar Interferometry (InSAR) can measure subtle surface deformation over large areas, day and night, in all weather conditions. However, traditional differential InSAR techniques are easily affected by factors such as time decorrelation and atmospheric delay during the acquisition of deformation phases, and cannot solve the problem of monitoring slow, long-term surface deformation in the study area.

[0003] The development of temporal InSAR technology has opened up new avenues for subsidence detection, offering significant advantages in large-scale, high-precision, and long-term surface deformation monitoring. Among these, Permanent Scatterer InSAR (PS-InSAR) utilizes PS pixels with relatively stable scattering characteristics in SAR imagery for deformation monitoring. However, in suburban areas and other regions where stable scatterers are sparse, it is difficult to obtain a sufficient density of deformation monitoring points. Therefore, to better apply it to natural scenes in non-urban areas, Distributed Scatterer InSAR (DS-InSAR) technology has been developed. In DS-InSAR, based on the identification results of homogeneous pixel count (SHP), adaptive multi-look is used to improve the signal-to-noise ratio of DS pixels. Subsequently, the phase connection (PL) algorithm is used to estimate the time-series phase from the constructed interferometric coherence matrix. However, the phase connection algorithm is affected by the presence of heterogeneous samples within the selected homogeneous pixels, which inevitably leads to a decrease in the accuracy of phase optimization.

[0004] With the launch of satellites capable of acquiring multi-polarization SAR data, polarization SAR data has become increasingly abundant, promoting the development of Polarimetric Time-Series InSAR (PolMTI) technology. PolMTI can fully utilize the effective phase components of different polarization channels to optimize and improve phase quality. However, polarimetric coherence optimization algorithms, such as the Exhaustive Search Polarimetric Optimization (ESPO) algorithm, have a high computational burden when the number of polarization channels increases. Polarimetric Phase Connectivity (PPL) is a phase optimization algorithm proposed for DS pixels, neglecting the optimization of PS pixels. Furthermore, the selection threshold determination mechanism for DS pixels is unclear, affecting the quality of the interferometric phase and resulting in low reliability of the acquired coherent deformation monitoring points. Summary of the Invention

[0005] To address the problem that existing technologies do not simultaneously consider PS and DS pixels, resulting in low reliability of acquired coherent deformation monitoring points, this invention provides an interferometric phase optimization method based on total power polarization and nonlocal phase connection.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] The interferometric phase optimization method based on total power polarization and nonlocal phase connection includes the following steps:

[0008] Acquire multi-polarization time-series SAR images of the study area, calculate the total power interferogram based on the phase information in the multi-polarization time-series SAR images, represent the polarization data in the multi-polarization time-series SAR images as scattering vectors, and calculate the average polarization covariance matrix using the scattering vectors;

[0009] The similarity between pixels is calculated based on the average polarization covariance matrix to identify homogeneous pixels and obtain the number of homogeneous pixels. The total backscattering intensity and total backscattering polarization are calculated based on the total power interferogram and the average polarization covariance matrix. Based on the properties of distributed scatterer (DS) pixels and permanent scatterer (PS) pixels, classification thresholds are set for the number of homogeneous pixels, total backscattering intensity, and total backscattering polarization. According to the classification thresholds, the pixels in the multi-polarization temporal SAR image are divided into DS pixels and PS pixels.

[0010] Obtain the total power interference complex coherence matrix of the DS pixels. Using the center pixel as a reference, calculate the similarity of the total power interference complex coherence matrix between the center pixel and homogeneous pixels within the window. Remove pixels with abnormal similarity. Perform nonlocal NL weighting on the removed total power interference complex coherence matrix. Perform phase concatenation on the weighted interference complex coherence matrix to obtain the optimized temporal phase sequence of the DS pixels. For PS pixels, retain the TP total power phase based on the total power interferogram. Obtain the optimized differential interferogram based on the optimized temporal phase sequence of the DS pixels and the TP total power phase.

[0011] Preferably, the polarization data in the multi-polarization temporal SAR image is represented as a scattering vector. The multi-polarization temporal SAR image includes fully polarimetric SAR data and dual-polarimetric SAR data. The fully polarimetric SAR data is represented as a scattering vector, specifically as follows:

[0012] ;

[0013] in, , and Single-look complex SAR images representing HH horizontal polarization, VV vertical polarization, and VH cross-polarization channels, respectively; This represents the matrix transpose operation;

[0014] The dual-polarization SAR data is represented as a scattering vector, specifically:

[0015] ;

[0016] in, Represents single-look complex SAR images under HH or VV polarization.

[0017] Preferably, the acquisition of multi-polarization time-series SAR images of the study area specifically involves: using the Sentinel-1 satellite to acquire SLC image data with VV horizontal polarization and VH cross-polarization; and using a fully polarimetric satellite to acquire SAR image data with HH horizontal polarization, VV vertical polarization, and VH cross-polarization.

[0018] Preferably, for fully polarimetric SAR data, the total power interferogram is specifically as follows:

[0019] ;

[0020] in, The interferogram is obtained by the total power scattering mechanism method. , as well as Corresponding to , and Interference phase; , and These are the time-series average intensities corresponding to the channels; It is a natural constant;

[0021] For dual-polarization SAR data, the total power interferogram is specifically as follows:

[0022] ;

[0023] in, , The interference phase corresponding to the channel; , These represent the time-averaged intensity of the corresponding channels.

[0024] Preferably, the average polarization covariance matrix is ​​specifically:

[0025] ;

[0026] in, Let be the scattering vector of the i-th sample. This indicates the conjugate transpose. This represents the number of single-look complex SAR images, where n is the number of multilooks.

[0027] Preferably, the setting of classification thresholds for the number of homogeneous pixels, total backscattering intensity, and total backscattering polarizability is specifically calculated based on the Jaccard coefficient. The number of homogeneous pixels, total backscattering intensity, and total backscattering polarizability are specifically as follows:

[0028] ;

[0029] ;

[0030] ;

[0031] Where A is the set of pixels with TPC > 0.9, and B is the set of pixels under different threshold indicators. In this context, 'i' refers to the number of pixels in a homogeneous window; and In this context, 'i' refers to percentages. , They refer to , The Percentiles express The set of values ​​greater than the i-th percentile. express The set of values ​​less than the i-th percentile. It is the index function for finding the maximum value. , and These represent the number of homogeneous pixels and the total backscattering intensity, respectively. and polarization The classification threshold.

[0032] Preferably, the step of performing phase concatenation on the weighted interferometric complex coherence matrix to obtain the DS pixel-optimized temporal phase sequence specifically involves phase optimization using the PL algorithm based on EMI.

[0033] ;

[0034] in, This is the Hadamard product operator. For complex units, The weighted average TP coherence matrix for The absolute value, This indicates the search for the eigenvector that minimizes the entire expression. μ, To The eigenvector corresponding to the smallest eigenvalue obtained by eigenvalue decomposition. For phase, for The conjugate transpose of .

[0035] Preferably, the total backscattering intensity is specifically:

[0036] ;

[0037] Where q is The size of the matrix; The average polarization covariance matrix; For matrix traces; It is a natural constant;

[0038] The total backscattering polarizability is specifically:

[0039] ;

[0040] in, express The matrix.

[0041] This invention provides an interferometric phase optimization system based on total power polarization and nonlocal phase connection, specifically including:

[0042] The data processing module is used to acquire multi-polarization time-series SAR images of the study area, calculate the total power interferogram based on the phase information in the multi-polarization time-series SAR images, represent the polarization data in the multi-polarization time-series SAR images as scattering vectors, and calculate the average polarization covariance matrix using the scattering vectors.

[0043] The pixel classification module is used to calculate the similarity between pixels based on the average polarization covariance matrix, identify homogeneous pixels, and obtain the number of homogeneous pixels; calculate the total backscattering intensity and total backscattering polarization based on the total power interferogram and the average polarization covariance matrix; set classification thresholds for the number of homogeneous pixels, total backscattering intensity, and total backscattering polarization based on the properties of distributed scatterer (DS) pixels and permanent scatterer (PS) pixels, and classify the pixels in the multi-polarization temporal SAR image into DS pixels and PS pixels according to the classification thresholds.

[0044] The phase optimization module is used to obtain the total power interference complex coherence matrix of DS pixels. Taking the center pixel as a reference, it calculates the similarity of the total power interference complex coherence matrix of the center pixel and the homogeneous pixels within the window, removes pixels with abnormal similarity, performs nonlocal NL weighting on the removed total power interference complex coherence matrix, and performs phase concatenation on the weighted interference complex coherence matrix to obtain the optimized temporal phase sequence of DS pixels. For PS pixels, the total power phase of TP is retained based on the total power interferogram. Based on the optimized temporal phase sequence of DS pixels and the total power phase of TP, the optimized differential interferogram is obtained.

[0045] The interferometric phase optimization method based on total power polarization and nonlocal phase connection provided by this invention has the following beneficial effects:

[0046] This invention calculates scattering vectors and constructs a total power interferogram (TP) based on multi-polarization SAR images of the study area. It then calculates pixel similarity based on the average polarization covariance matrix (APC) to identify homogeneous pixels. Polarization Shannon entropy is calculated using the TP and APC, and classification is performed based on the properties of distributed scatterer (DS) pixels and permanent scatterer (PS) pixels. Targeted optimization is applied to different pixels to improve the noise resistance of the processing. Similarity calculations are performed on the center pixel of DS pixels and other homogeneous pixels to identify pixels with asymmetric similarity, reducing the error impact on homogeneous pixels. The weighted average of the remaining TP TP complex coherence matrix improves the accuracy of subsequent phase unwrapping. Phase concatenation is then performed on the weighted TP TP TP, yielding an optimized temporal phase sequence for DS pixels. For PS pixels, the TP TP total power phase is retained based on the TP TP. This classification optimization simultaneously improves the interferometric phase quality of PS and DS targets, resulting in a higher density of monitoring points. Based on the optimized total power phase of the DS and PS pixels, an optimized differential interferogram is obtained. High-quality coherent points are selected based on the time coherence (TPC) of the optimized differential interferogram, and three-dimensional phase unwrapping is performed to complete the time-series deformation monitoring. This results in less phase speckle noise and improved reliability of the monitoring results. Attached Figure Description

[0047] To more clearly illustrate the embodiments and design schemes of the present invention, the accompanying drawings required for this embodiment will be briefly described below. 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.

[0048] Figure 1 This is a flowchart of an interferometric phase optimization method based on total power polarization and nonlocal phase connection, according to an embodiment of the present invention.

[0049] Figure 2 This is a flowchart illustrating the classification process of PS and DS in an embodiment of the present invention.

[0050] Figure 3 This is a schematic diagram showing the real interference phase and simulated noise phase with the longest time baseline in an embodiment of the present invention. Wherein, Figure 3 (a) is the actual interferometric phase diagram; Figure 3 (b) is the simulated noise phase diagram.

[0051] Figure 4 This is a comparison chart of the interference phase optimization capabilities of different methods in the embodiments of the present invention.

[0052] Figure 5 In this embodiment of the invention, threshold values ​​for different classification indicators are determined based on changes in the Jaccard coefficient. Figure 5 (a) is a graph showing the change in the Jaccard coefficient between the classification sets with different PHP thresholds and the set with TPC > 0.9; Figure 5 (b) is a graph showing the change in the Jaccard coefficient between the classification sets with different SEI percentages and the set with TPC > 0.9; Figure 5 (c) is a graph showing the change in the Jaccard coefficient between the classification sets with different SEP percentages and the set with TPC > 0.9.

[0053] Figure 6 This is a comparison diagram of the phase coherence of different optimization methods in embodiments of the present invention. Figure 6 (a) is a schematic diagram of the original phase coherence; Figure 6 (b) is a schematic diagram of the phase coherence of a single-polarization phase connection; Figure 6 (c) is a schematic diagram of the total power phase connection coherence; Figure 6 (d) is a schematic diagram of the phase coherence of a single-polarization nonlocal phase connection; Figure 6 (e) is a schematic diagram of the phase coherence of the optimization method of the present invention.

[0054] Figure 7 The selected pixels with TPC>0.9 in this embodiment of the invention are obtained through different methods to achieve optimized interference phases. Figure 7 (a) is the optimized interferometric phase diagram obtained by single-polarization phase connection; Figure 7 (b) is the optimized interferometric phase diagram obtained by the total power phase connection; Figure 7 (c) is the optimized interferometric phase map obtained by single-polarization nonlocal phase connection; Figure 7 (d) is the optimized interference phase diagram obtained by the method of the present invention.

[0055] Figure 8 This is a schematic diagram illustrating deformation monitoring of a coal mine in a study area obtained using different methods in embodiments of the present invention. Figure 8 (a) is the original monitoring result diagram; Figure 8 (b) is a diagram showing the results of single-polarization phase connection; Figure 8 (c) is the total power phase connection result diagram; Figure 8 (d) is the result of single-polarization nonlocal phase connection; Figure 8 (e) is a result diagram of the optimization method of the present invention.

[0056] Figure 9 This section compares the deformation monitoring results from different methods with the projected GNSS measurement results in embodiments of the present invention. Figure 9(a) is a comparison diagram of the measurement results of single-polarization phase connection, total power phase connection, single-polarization nonlocal phase connection, and GNSS01. Figure 9 (b) is a comparison chart of measurement results of single-polarization phase connection, total power phase connection, single-polarization nonlocal phase connection, and GNSS02. Figure 9 (c) is a comparison diagram of the measurement results of the single-polarization phase connection, the total power phase connection, the single-polarization non-local phase connection, and the present invention and GNSS03. Detailed Implementation

[0057] To enable those skilled in the art to better understand and implement the technical solutions of the present invention, the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. The following embodiments are only used to more clearly illustrate the technical solutions of the present invention and should not be construed as limiting the scope of protection of the present invention.

[0058] Example

[0059] This invention provides an interferometric phase optimization method based on total power polarization and nonlocal phase connection, specifically an interferometric phase optimization method for synthetic aperture radar (SAR). An example verification is performed using 28 dual-polarized Sentinel-1 datasets covering a study area from July 22, 2023 to August 9, 2024. Figure 1 As shown, the specific steps include:

[0060] Step 1: Acquire multi-polarization time-series SAR images of the study area, register the SAR images of different polarizations, and perform conjugate multiplication on the registered SAR images; obtain the differential interferogram by subtracting the terrain phase from the original interferometric phase, mainly retaining deformation, atmospheric and noise information. Calculate and construct the total power interferogram based on the phase information of the differential interferogram.

[0061] For Sentinel-1 satellites, acquire SLC images with two polarizations: VV and VH. For all-polarization satellites such as ALOS PALSAR, acquire SAR data with three polarizations: HH horizontal polarization, VV vertical polarization, and VH cross-polarization.

[0062] For polarimetric data, the polarization information of a pixel can be described by a complex Pauli scattering vector. Under reciprocity conditions, the Pauli scattering vector of fully polarimetric data... It can be defined as:

[0063] ;

[0064] in, , and Single-look complex SAR images representing HH horizontal polarization, VV vertical polarization, and VH cross-polarization channels, respectively, according to the reciprocity theorem, Therefore, only Indicates cross-polarization, the same applies below. This represents the matrix transpose operation.

[0065] For dual-polarization SAR data containing one common polarization (HH / VV) and one cross-polarization (VH / HV), the Pauli scattering vector Represented as:

[0066] ;

[0067] in, Represents single-look complex SAR images under HH or VV polarization.

[0068] Total power (TP) interferograms for equal scattering mechanisms (ESM) are constructed based on scattering vectors, using time-averaged intensity as weights. For fully polarimetric SAR data, the total power interferogram is obtained using the following formula:

[0069] ;

[0070] in, The interferogram is obtained by the total power scattering mechanism method. , as well as Corresponding to , and The interference phase.

[0071] For dual-polarization SAR data, the total power interferogram is obtained by the following formula:

[0072] ;

[0073] in, , This represents the interference phase of the corresponding channel. , , and These represent the time-averaged intensities of the corresponding channels, calculated using the following formula:

[0074] ;

[0075] ;

[0076] ;

[0077] ;

[0078] Where N is the number of PolSAR images, Indicates the first SLC data of a PolSAR image.

[0079] Step 2: Construct the average polarization covariance matrix to classify DS and PS pixels.

[0080] S21. Construct the average polarization covariance matrix and identify homogeneous pixels. To suppress speckle noise, the average polarization covariance matrix is ​​obtained by averaging the polarization covariance matrices of N time series using the following formula. :

[0081] ;

[0082] in, Let be the scattering vector of the i-th sample. This represents the conjugate transpose, where n is the multi-view number.

[0083] according to Calculate the similarity between pixel i and the center pixel j in the sliding window, and select homogeneous pixels. X and Y are the PolSAR values ​​of pixel i and pixel j, respectively. The time-series average covariance matrix for fully polarized data For bipolar data At the significance level In this case, the similarity threshold is calculated as follows:

[0084] ;

[0085] in, This is the chi-square distribution function, whose value is obtained by consulting the chi-square distribution table. The significance level is set to [value missing]. .

[0086] S22. Calculate the polarized Shannon entropy. Defined as relative to the total backscattering intensity ( ) and polarization ( The sum of the two related components:

[0087] ;

[0088] in, Given the average polarization coherence matrix, and considering that its covariance matrix has a similar distribution, we replace it with the average polarization covariance matrix. .but and The expressions for the two items are as follows:

[0089] ;

[0090] ;

[0091] Where q is Given the matrix dimension, set the bipolar data q=2. It reflects the amount of information given by the total scattering power (or intensity) of the target. This reflects the diversity of contributions from various scattering mechanisms. High values ​​indicate a complex and varied polarization scattering mechanism of the target, while low values ​​suggest a relatively simple or uniform scattering mechanism. The temporal phase coherence (TPC) of the total power interferogram is calculated as follows:

[0092] ;

[0093] in, It is the number of interferograms. It is the first The amplitude interferogram corresponds to the noise phase. The set of pixels where TP-TPC > 0.9 is taken as A.

[0094] S23. The Jaccard index is used to measure the overlap of pixel sets with a time coherence ratio (TPC) greater than 0.9 for different indicators. The Jaccard index is a statistic used to compare the similarity between finite sample sets. The Jaccard index is the proportion of the number of elements in the intersection of two sets A and B in the union of A and B, as shown in the following formula:

[0095] ;

[0096] Where A is the set of pixels with TPC > 0.9, and B is the set of pixels under different thresholds for a certain index. The number of homogeneous pixels (PHPnum) and total backscattering intensity (PHPnum) are calculated based on the Jaccard coefficient. ) and polarization ( Classification threshold , and As shown in the following formula:

[0097] ;

[0098] ;

[0099] ;

[0100] in, In this context, 'i' refers to the number of pixels in the homogeneous window. The default window size is 15×15, and here 'i' has a maximum value of 225. and In this context, 'i' refers to percentages. , They refer to , The Percentiles To find The set of values ​​greater than the i-th percentile. That is, find The set of values ​​less than the i-th percentile. It is the maximum value index function. DS pixel scattering is dispersed, with many homogeneous pixels and high total backscattering polarization; DS pixel scattering is stable, with few homogeneous pixels and low total backscattering polarization.

[0101] like Figure 2 As shown in the flowchart, the classified PS and DS pixels are obtained. In this example, three thresholds are obtained based on the changes in the Jaccard coefficient, as follows: Figure 5 As shown, we obtain 2, as well as .

[0102] Step 3: Perform phase connection on DS pixels based on total power nonlocality.

[0103] S31. Obtain the total power interference complex coherence matrix of the DS pixels. Based on the above classification results, further phase optimization is performed on the DS pixels. First, the TP interference phase of the DS pixels is considered as the observation value of the phase connection. Therefore, the TP complex coherence matrix can be expressed as:

[0104] ;

[0105] in, To fuse the TP interferometric phase of dual-polarization information, This represents the coherence coefficient between the m-th and n-th TP interferograms.

[0106] S32. Using the center pixel as a reference, measure the similarity of the total power interference complex coherence matrix between the center pixel and other homogeneous pixels. The center pixel is the pixel at the very center of the homogeneous pixel window. Kullback-Leibler divergence (KLD) is used to quantify the similarity of the TP coherence matrix between the reference pixel i and other similar pixels j within the search window:

[0107] ;

[0108] in, Let KLD be the symmetric KLD between pixels i and j. KLD is non-negative. When the coherence matrix distributions of the two pixels are completely similar, KLD is infinitely close to 0. trace(·) represents the trace operation of the matrix. N is the number of time-series images.

[0109] S33. After removing anomalous similarity pixels, perform nonlocal NL weighted filtering on the total power interference complex coherence matrix. Since homogeneous pixel identification results are susceptible to Type II errors, potentially leading to abnormally large KLD values, pixels with KLD values ​​greater than a certain threshold are first removed before calculating the weights. Homogeneous pixels. Here, It represents the absolute deviation from the median. Specifically, it is The median of a set of values, where It is all The median value. Therefore, after removing outliers, similarity is converted to a spatial average weight:

[0110] ;

[0111] in, For the symmetric KLD between pixel i and pixel j, This represents the average KLD similarity between the reference pixel and the pixel after removing outliers. Then, a weighted average is calculated for the TP coherence matrix of pixels of the same type within the search window.

[0112] ;

[0113] in, The TP coherence matrix is ​​the weighted average of pixel i. This is to remove homogeneous pixel sets with abnormal similarity.

[0114] S34. Perform phase concatenation on the weighted averaged interferometric complex coherence matrix to obtain the optimized time-series phase sequence. Utilize the EMI-based PL algorithm to optimize the phase of the TP-NL coherence matrix:

[0115] ;

[0116] in, To The eigenvector corresponding to the smallest eigenvalue obtained by eigenvalue decomposition. , This is the Hadamard product operator. It is a complex unit. The weighted average TP coherence matrix The absolute value, express The eigenvector μ corresponding to the smallest eigenvalue after eigenvalue decomposition. for The conjugate transpose of . After the main image is determined, from The optimized interference phase is obtained in .

[0117] Step 4: For PS pixels, retain the TP total power phase based on the total power interferogram; for DS pixels, obtain the TP-NL-PL optimized interferometric phase to obtain the optimized differential interferogram. Calculate temporal coherence based on the optimized differential interferogram, select high-quality coherent points based on temporal coherence, and perform three-dimensional phase unwrapping to complete the temporal deformation monitoring of the study area.

[0118] To evaluate the polarization optimization effect of differential interferograms, the time coherence (TPC), an evaluation index of interferogram phase quality, was calculated using the following formula. Interferograms obtained by different optimization methods in the study area were then compared and analyzed.

[0119] ;

[0120] in, It is the number of interferograms. It is the first Amplitude-optimized interferogram corresponding to noise phase, The window size is selected here (considering SAR image resolution and land cover in the study area). The window is obtained based on the estimation of the interferometric phase value of the neighborhood pixels. For example... Figure 6 As shown, the phase coherence of the original phase, single-polarization phase connection, total power phase connection, single-polarization nonlocal phase connection, and the phase coherence of the optimization method of this invention are obtained respectively. The larger the TPC, the better the phase quality. Figure 7 As shown, the interference phase of the longest time baseline interferogram 20231002-20240809 with TPC > 0.9 pixels calculated using the optimization method of this invention is also demonstrated. It can be clearly seen that the optimization method of this invention has better phase quality, less phase speckle noise, and a significant improvement in phase quality compared to other methods.

[0121] In this embodiment, 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. Finally, the deformation monitoring results obtained by the above five methods were obtained. The results were then magnified to a coal mine area covered by the study area, as shown in the figure. Figure 8 As shown in the figure, compared with the original monitoring results, the deformation monitoring points obtained by the single-polarization phase connection, total power phase connection, and single-polarization nonlocal phase connection methods increased to 81,951, 84,853, and 302,423 respectively, with the method of this invention obtaining a maximum of 312,408 monitoring points. Furthermore, it can be seen from the figure that the overall settlement trend monitored by the three methods is consistent. To further illustrate the reliability of the method of this invention, deformation measurements of three GNSS points in the coal mine area were obtained and compared with the deformation monitoring results obtained by the above four optimization methods. Before the comparison, the GNSS measurements in the three directions were projected onto the LOS direction based on the satellite's heading angle and the radar's incident angle. Figure 9As shown, the four monitoring results are basically consistent with the measurement results of different GNSS, but the time series fluctuation of the method of the present invention is smaller and more stable, and is more consistent with the measurement results of GNSS.

[0122] To evaluate the optimized performance of this invention on the DS target, Monte Carlo simulation was also used to simulate N=30 fully polarized interferometric data. The polarization-time-series interferometric coherence matrix is ​​shown. It can be derived from the polarization coherence matrix and temporal interference coherence matrix express:

[0123] ;

[0124] in, For Kronecker product. This can be described by the extended Bragg model:

[0125] ;

[0126] in, , . Modeled as a single master image interferogram and coherence amplitude matrix constitute: . Generally, related models are used for generation:

[0127] ;

[0128] in, , and These are the initial coherence and long-term coherence values, which are set in this example. , . The time baseline length, This is the time constant controlling the intensity of coherence attenuation. The intensity and phase of the simulated interferometric data are generated using the fracsurf function developed by Professor Ramon Hanssen, with a size of 256×256 pixels. Phase components related to orbit, topography, and atmosphere are set to zero. The deformation phase follows a linear variation according to a specified pattern. This is based on the polarization-time-series interferometry matrix. Interferograms of the three polarization channels of the first SLC main image are extracted. The temporal phase sequence of the first main image is selected. Figure 3 (a) and Figure 3 (b) shows the real interferograms and simulated noisy interferograms with time baselines of 1–30. To compare the phase recovery capabilities of the interferograms, Figure 4 of (a) Figure 4(b) Figure 4 (c) and Figure 4 (d) shows, from top to bottom, the optimized interference phase, residual phase, and posterior coherence of the longest baseline using the single-polarization phase connection, the total power phase connection, the single-polarization nonlocal phase connection, and the optimization method of the present invention. It can be seen that the optimized method of the present invention produces clearer and smoother interference fringes, less residual anomalous noise, and greater posterior coherence.

[0129] This invention also provides an interferometric phase optimization system based on total power polarization and nonlocal phase connection, specifically including:

[0130] The data processing module is used to acquire multi-polarization time-series SAR images of the study area, calculate the total power interferogram based on the phase information in the multi-polarization time-series SAR images, represent the polarization data in the multi-polarization time-series SAR images as scattering vectors, and calculate the average polarization covariance matrix using the scattering vectors.

[0131] The pixel classification module is used to calculate the similarity between pixels based on the average polarization covariance matrix, identify homogeneous pixels, and obtain the number of homogeneous pixels; calculate the total backscattering intensity and total backscattering polarization based on the total power interferogram and the average polarization covariance matrix; set classification thresholds for the number of homogeneous pixels, total backscattering intensity, and total backscattering polarization based on the properties of distributed scatterer (DS) pixels and permanent scatterer (PS) pixels, and classify pixels in multi-polarization temporal SAR images into DS pixels and PS pixels according to the classification thresholds.

[0132] The phase optimization module is used to obtain the total power interference complex coherence matrix of DS pixels. Taking the center pixel as a reference, it calculates the similarity of the total power interference complex coherence matrix of the center pixel and homogeneous pixels within the window, removes anomalous similarity pixels, performs nonlocal NL weighting on the removed total power interference complex coherence matrix, and performs phase concatenation on the weighted interference complex coherence matrix to obtain the optimized temporal phase sequence of DS pixels. For PS pixels, the total power phase of TP is retained based on the total power interferogram. Based on the optimized temporal phase sequence of DS pixels and the total power phase of TP, the optimized differential interferogram is obtained.

[0133] It should be noted that the specific embodiments described above enable those skilled in the art to more fully understand the present invention, but do not limit the present invention in any way. Therefore, although the present invention has been described in detail in this specification and embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the present invention; and all technical solutions and improvements that do not depart from the spirit and scope of the present invention are covered within the protection scope of the present invention patent. No reference numerals in the claims should be construed as limiting the scope of the claims. Any simple variations or equivalent substitutions of technical solutions that can be readily obtained by those skilled in the art within the scope of the technology disclosed in the present invention are within the protection scope of the present invention.

Claims

1. An interferometric phase optimization method based on total power polarization and nonlocal phase connection, characterized in that, Includes the following steps: Acquire multi-polarization time-series SAR images of the study area, calculate the total power interferogram based on the phase information in the multi-polarization time-series SAR images, represent the polarization data in the multi-polarization time-series SAR images as scattering vectors, and calculate the average polarization covariance matrix using the scattering vectors; The similarity between pixels is calculated based on the average polarization covariance matrix to identify homogeneous pixels and obtain the number of homogeneous pixels; the total backscattering intensity and total backscattering polarization are calculated based on the total power interferogram and the average polarization covariance matrix. Based on the properties of distributed scatterer (DS) pixels and permanent scatterer (PS) pixels, classification thresholds are set for the number of homogeneous pixels, total backscattering intensity, and total backscattering polarization. According to the classification thresholds, the pixels in the multi-polarization temporal SAR image are divided into DS pixels and PS pixels. Obtain the total power interference complex coherence matrix of the DS pixels. Using the center pixel as a reference, calculate the similarity of the total power interference complex coherence matrix between the center pixel and the homogeneous pixels in the window. Remove the pixels with abnormal similarity. Perform nonlocal NL weighting on the removed total power interference complex coherence matrix. Perform phase connection on the weighted interference complex coherence matrix to obtain the optimized temporal phase sequence of the DS pixels. For PS pixels, the TP total power phase is preserved based on the total power interferogram; Based on the optimized temporal phase sequence of the DS pixels and the total power phase of TP, the optimized differential interferogram is obtained.

2. The interferometric phase optimization method based on total power polarization and nonlocal phase connection according to claim 1, characterized in that, The polarization data in the multi-polarization temporal SAR image is represented as a scattering vector. The multi-polarization temporal SAR image includes fully polarimetric SAR data and dual-polarimetric SAR data. The fully polarimetric SAR data is represented as a scattering vector, specifically as follows: ； in, , and Single-look complex SAR images representing HH horizontal polarization, VV vertical polarization, and VH cross-polarization channels, respectively; This represents the matrix transpose operation; The dual-polarization SAR data is represented as a scattering vector, specifically: ； in, Represents single-look complex SAR images under HH or VV polarization.

3. The interferometric phase optimization method based on total power polarization and nonlocal phase connection according to claim 1, characterized in that, The acquisition of multi-polarization time-series SAR images of the study area specifically involves: using the Sentinel-1 satellite to acquire SLC image data with VV horizontal polarization and VH cross-polarization; and using a fully polarimetric satellite to acquire SAR image data with HH horizontal polarization, VV vertical polarization, and VH cross-polarization.

4. The interferometric phase optimization method based on total power polarization and nonlocal phase connection according to claim 3, characterized in that, For fully polarimetric SAR data, the total power interferogram is specifically as follows: ； in, The interferogram is obtained by the total power scattering mechanism method. , as well as Corresponding to , and Interference phase; , and These are the time-series average intensities corresponding to the channels; It is a natural constant; For dual-polarization SAR data, the total power interferogram is specifically as follows: ； in, , The interference phase corresponding to the channel; , These represent the time-averaged intensity of the corresponding channels.

5. The interferometric phase optimization method based on total power polarization and nonlocal phase connection according to claim 1, characterized in that, The average polarization covariance matrix is ​​specifically: ； in, Let be the scattering vector of the i-th sample. This indicates the conjugate transpose. This represents the number of single-look complex SAR images, where n is the number of multilooks.

6. The interferometric phase optimization method based on total power polarization and nonlocal phase connection according to claim 1, characterized in that, The classification thresholds for the number of homogeneous pixels, total backscattering intensity, and total backscattering polarization are specifically calculated based on the Jaccard coefficient. The number of homogeneous pixels, total backscattering intensity, and total backscattering polarization are specifically as follows: ； ； ； Where A is the set of pixels with TPC > 0.9, and B is the set of pixels under different threshold indicators. In this context, 'i' refers to the number of pixels in a homogeneous window; and In this context, 'i' refers to percentages. , They refer to , The Percentiles express The set of values ​​greater than the i-th percentile. express The set of values ​​less than the i-th percentile. It is the index function for finding the maximum value. , and These represent the number of homogeneous pixels and the total backscattering intensity, respectively. and polarization The classification threshold.

7. The interferometric phase optimization method based on total power polarization and nonlocal phase connection according to claim 1, characterized in that, The phase concatenation of the weighted interferometric complex coherence matrix yields the DS-pixel optimized temporal phase sequence, specifically using the PL algorithm based on EMI for phase optimization. ； in, This is the Hadamard product operator. For complex units, The weighted average TP coherence matrix for The absolute value, This indicates the search for the eigenvector that minimizes the entire expression. μ, To The eigenvector corresponding to the smallest eigenvalue obtained by eigenvalue decomposition. For phase, for The conjugate transpose of .

8. The interferometric phase optimization method based on total power polarization and nonlocal phase connection according to claim 1, characterized in that, The total backscattering intensity is specifically: ； Where q is The size of the matrix; The average polarization covariance matrix; For matrix traces; It is a natural constant; The total backscattering polarizability is specifically: ； in, express The matrix.

9. An interferometric phase optimization system based on total power polarization and nonlocal phase connection, characterized in that, include: The data processing module is used to acquire multi-polarization time-series SAR images of the study area, calculate the total power interferogram based on the phase information in the multi-polarization time-series SAR images, represent the polarization data in the multi-polarization time-series SAR images as scattering vectors, and calculate the average polarization covariance matrix using the scattering vectors. The pixel classification module is used to calculate the similarity between pixels based on the average polarization covariance matrix, identify homogeneous pixels, and obtain the number of homogeneous pixels; and to calculate the total backscattering intensity and total backscattering polarization based on the total power interferogram and the average polarization covariance matrix. Based on the properties of distributed scatterer (DS) pixels and permanent scatterer (PS) pixels, classification thresholds are set for the number of homogeneous pixels, total backscattering intensity, and total backscattering polarization. According to the classification thresholds, the pixels in the multi-polarization temporal SAR image are divided into DS pixels and PS pixels. The phase optimization module is used to obtain the total power interference complex coherence matrix of DS pixels. Taking the center pixel as a reference, it calculates the similarity of the total power interference complex coherence matrix of the center pixel and the homogeneous pixels in the window, removes the pixels with abnormal similarity, performs nonlocal NL weighting on the removed total power interference complex coherence matrix, and performs phase concatenation on the weighted interference complex coherence matrix to obtain the optimized temporal phase sequence of DS pixels. For PS pixels, the TP total power phase is preserved based on the total power interferogram; Based on the optimized temporal phase sequence of the DS pixels and the total power phase of TP, the optimized differential interferogram is obtained.

Images