Self-adaptive sequential phase enhancement method and system based on phase stripe trend suppression

Abstract

The invention discloses an adaptive sequential phase enhancement method and system based on phase fringe trend suppression, and the method comprises the steps: S1, obtaining a time sequence SAR image, constructing a nearest interference network, and extracting a time sequence interference phase fringe trend of a nearest interference phase diagram; s2, estimating an SLC phase fringe trend from the time sequence interference phase fringe trend by using a least square method to obtain SLC residual phase information; s3, estimating a sample covariance matrix based on the time sequence amplitude information of the time sequence SAR image and the SLC residual phase information, dividing sub-sample covariance matrix blocks, and performing local phase enhancement on the sub-sample covariance matrix blocks to obtain an SLC residual enhanced phase sub-sequence; and S4, performing adjustment correction processing on the same-name image of the SLC residual enhanced phase sub-sequence, and compensating the SLC phase fringe trend to the processed SLC residual enhanced phase sequence to obtain an SLC enhanced phase sequence.

Description

Technical Field

[0001] The present invention belongs to the technical field of time series interferometric synthetic aperture radar data processing, and particularly relates to an adaptive sequential phase enhancement method and system based on phase fringe trend suppression. Background Art

[0002] Time Series Interferometric Synthetic Aperture Radar (TSInSAR) technology gives full play to the advantages of SAR satellites in all-weather, all-day, wide coverage and short revisit time. By using time series analysis technology, it can realize the inversion of millimeter-level surface time series deformation information, and is widely used in the fields of geological disaster monitoring such as landslides, earthquakes, debris flows, urban health monitoring and resource exploration, etc., and shows excellent application prospects. In a conventional scene, the number of pixels of Distributed Scatterer (DS) in the corresponding SAR image is usually much higher than that of Persistent Scatterer (PS), especially in non-artificial surface areas such as farmland, cultivated land, and bare land. Therefore, compared with the traditional PS point-based TSInSAR technology, the DSInSAR technology can provide a rich density of deformation monitoring points and interpret more complete and reliable deformation field detail information. However, different from the PS points where there is a dominant scatterer in the pixel resolution unit to keep the phase stable in a long time series, the DS phase is usually jointly affected by multiple similar scatterers, which makes the DS phase more vulnerable to various decoherence factors, reducing the quality of the DS phase and seriously restricting the accuracy of subsequent deformation phase interpretation. In view of this, the phase enhancement processing to improve the signal-to-noise ratio of the DS phase has become one of the key steps in DSInSAR. A large number of domestic and foreign research scholars have successively proposed phase enhancement methods such as PTA, PL, least squares, eigenvalue decomposition, EMI and Sequential, etc., and extensive research has been carried out in aspects such as accurate estimation of the covariance matrix, non-positive definite regularization processing of the covariance matrix, and refinement of the phase enhancement model, and remarkable research results have been achieved.

[0003] However, for the current phase enhancement methods, there are still some key problems to be solved: (1) Traditional phase enhancement methods are difficult to effectively balance the recovery of dense fringe information in large-gradient deformation regions and the improvement of phase signal-to-noise ratio in high-noise deformation regions. Then, research scholars proposed phase enhancement models based on adaptive spatio-temporal filtering fusion and prior information-driven. Although the performance of phase enhancement has been effectively improved, they still face problems such as time-consuming full-interference pair traversal processing, strong dependence on prior data, and poor accuracy of dynamic prior models; (2) The Sequential phase enhancement model can effectively balance phase enhancement performance and processing efficiency and is one of the most excellent phase enhancement methods at present. However, it still faces problems such as fixed sub-matrix block dimensions, complicated matrix block compression estimation, and poor phase continuity after enhancement. To sum up, the above existing problems at present still significantly restrict the performance of phase enhancement in complex scenarios corresponding to large-gradient deformations, and severely limit the high-precision and rich interpretation of deformation information. Therefore, how to improve the key problems existing in the current phase enhancement model, improve the performance of phase enhancement methods in terms of phase detail information recovery, phase signal-to-noise ratio improvement, and efficient processing, and achieve high-precision deformation information interpretation in complex surface regions is an important research work to be carried out at present. Summary of the Invention

[0004] The present invention aims to solve the deficiencies of the prior art and provides the following solutions:

[0005] An adaptive sequential phase enhancement method based on phase fringe trend suppression, comprising the following steps:

[0006] S1. Obtain time-series SAR images, construct a nearest-neighbor interference network, and adaptively extract the time-series interference phase fringe trend of the nearest-neighbor interference phase map according to the frequency-domain phase spectrum information extraction method;

[0007] S2. Estimate the SLC phase fringe trend from the time-series interference phase fringe trend by using the least squares method to obtain the SLC residual phase information;

[0008] S3. Estimate the sample covariance matrix based on the time-series amplitude information of the time-series SAR images and the SLC residual phase information. According to the coherence quality and the matrix block overlap principle, divide sub-sample covariance matrix blocks from the sample covariance matrix, and perform local phase enhancement on the sub-sample covariance matrix blocks to obtain an SLC residual enhanced phase subsequence;

[0009] S4. Perform adjustment and correction processing on the homonymous images of the SLC residual enhanced phase subsequence, and compensate the SLC phase fringe trend to the processed SLC residual enhanced phase sequence to obtain an SLC enhanced phase sequence.

[0010] Preferably, the S1 includes:

[0011] Obtain the time-series SAR image;

[0012] Perform image registration preprocessing on the time-series SAR image, construct the nearest neighbor interference network based on the registered time-series SAR image information, and conjugate multiply to obtain the nearest neighbor interference phase map;

[0013] According to the frequency-domain phase spectrum information extraction method, adaptively extract the main phase fringe trend in the nearest neighbor interference phase map, including main phase information such as flat-earth phase, terrain phase, and deformation phase;

[0014] Traverse all the nearest neighbor interference phase maps and extract the time-series interference phase fringe trends of all the nearest neighbor interference maps.

[0015] Preferably, the S2 includes:

[0016] Combined with the least squares theory, taking any SLC image as a reference, estimate and extract N SLC phase fringe trends from the time-series interference phase fringe trends corresponding to N-1 nearest neighbor interference maps;

[0017] Conjugate multiply the SLC phase fringe trend with the original SLC image to suppress the phase fringe trend and obtain the SLC residual phase information.

[0018] Preferably, the S3 includes:

[0019] According to the time-series amplitude information of the time-series SAR image, use the non-hypothesis testing algorithm to identify homogeneous pixel points. Based on the homogeneous pixel set and the SLC residual phase information, use the maximum likelihood estimation method to traverse each pixel in the homogeneous pixel set one by one to estimate the sample covariance matrix;

[0020] Take the modulus of the sample covariance matrix, extract the sample coherence matrix, and divide the sub-sample covariance matrix blocks according to the coherence quality and sub-matrix block overlap principle;

[0021] According to the phase enhancement model estimated by coherence and phase coupling, perform local phase enhancement processing on each sub-sample covariance matrix block one by one, and estimate the local SLC residual enhanced phase subsequences corresponding to all the sub-sample covariance matrix blocks.

[0022] Preferably, the S4 includes:

[0023] For each SLC residual enhanced phase subsequence, perform homonymous image SLC residual enhanced phase matching adjustment processing pairwise, and take the previous matrix block as a reference, substitute the adjusted value into the subsequent matrix block, and estimate the SLC corrected residual enhanced phase subsequence after the adjusted value correction;

[0024] Traverse all the SLC residual enhanced phase subsequences corresponding to the sub-sample covariance matrix blocks, complete the adjustment and correction processing of each subsequence, and merge to obtain a complete N×1-dimensional SLC residual enhanced phase sequence;

[0025] Compensate the SLC phase fringe trend into the SLC residual enhanced phase sequence to obtain the final SLC enhanced phase sequence.

[0026] The present invention also provides an adaptive sequential phase enhancement system based on phase fringe trend suppression. The system applies the method described in any one of the above, and includes: a phase fringe trend extraction module, a residual phase information acquisition module, a residual enhanced phase acquisition module, and an enhanced phase acquisition module;

[0027] The phase fringe trend extraction module is used to obtain a time-series SAR image, construct a nearest neighbor interferometric network, and adaptively extract the time-series interferometric phase fringe trend of the nearest neighbor interferometric phase map according to the frequency-domain phase spectrum information extraction method;

[0028] The residual phase information acquisition module estimates the SLC phase fringe trend from the time-series interferometric phase fringe trend by using the least squares method to obtain the SLC residual phase information;

[0029] The residual enhanced phase acquisition module estimates the sample covariance matrix based on the time-series amplitude information of the time-series SAR image and the SLC residual phase information, divides the sub-sample covariance matrix blocks from the sample covariance matrix according to the coherence quality and matrix block overlap principle, and performs local phase enhancement on the sub-sample covariance matrix blocks to obtain SLC residual enhanced phase subsequences;

[0030] The enhanced phase acquisition module is used to perform adjustment and correction processing on the homologous images of the SLC residual enhanced phase subsequences, and compensate the SLC phase fringe trend into the processed SLC residual enhanced phase sequence to obtain the SLC enhanced phase sequence.

[0031] Preferably, the phase fringe trend extraction module includes: an image acquisition unit, an interferometric phase map acquisition unit, a main phase fringe trend extraction unit, and a phase fringe trend extraction unit;

[0032] The image acquisition unit is used to obtain the time-series SAR image;

[0033] The interferometric phase map acquisition unit performs image registration preprocessing on the time-series SAR image, constructs the nearest neighbor interferometric network according to the registered time-series SAR image information, and conjugately multiplies to obtain the nearest neighbor interferometric phase map;

[0034] The main phase fringe trend extraction unit adaptively extracts the main phase fringe trend in the nearest interferometric phase map according to the frequency-domain phase spectrum information extraction method, including main phase information such as flat phase, terrain phase, and deformation phase;

[0035] The phase fringe trend extraction unit traverses all the nearest interferometric phase maps and extracts the temporal interferometric phase fringe trends of all the nearest interferograms.

[0036] Preferably, the residual phase information acquisition module includes: an SLC phase fringe trend extraction unit and an SLC residual phase information extraction unit;

[0037] The SLC phase fringe trend extraction unit combines the least squares theory, takes any SLC image as a reference, and estimates and extracts N SLC phase fringe trends from the temporal interferometric phase fringe trends corresponding to N-1 nearest interferograms;

[0038] The SLC residual phase information extraction unit conjugately multiplies the SLC phase fringe trend with the original SLC image to suppress the phase fringe trend and obtain the SLC residual phase information.

[0039] Preferably, the residual enhanced phase acquisition module includes: a sample covariance matrix acquisition unit, a sample covariance matrix block division unit, and an SLC residual enhanced phase subsequence acquisition unit;

[0040] The sample covariance matrix acquisition unit uses a non-hypothesis testing algorithm to identify homogeneous pixel points based on the temporal amplitude information of the temporal SAR images, and uses the maximum likelihood estimation method to traverse each pixel in the homogeneous pixel set one by one to estimate the sample covariance matrix on the basis of the homogeneous pixel set and the SLC residual phase information;

[0041] The sample covariance matrix block division unit takes the modulus of the sample covariance matrix, extracts the sample coherence matrix, and divides the sub-sample covariance matrix blocks according to the coherence quality and the sub-matrix block overlap principle;

[0042] The SLC residual enhanced phase subsequence acquisition unit performs local phase enhancement processing on each of the sub-sample covariance matrix blocks according to the phase enhancement model estimated by the coherence and phase coupling, and estimates the local SLC residual enhanced phase subsequences corresponding to all the sub-sample covariance matrix blocks.

[0043] Preferably, the enhanced phase acquisition module includes: a flat value correction unit, a sequence merging unit, and a compensation unit;

[0044] The flat value correction unit performs pairwise SLC residual enhanced phase matching adjustment processing on each of the SLC residual enhanced phase subsequences, and takes the previous matrix block as a reference, substitutes the flat value into the subsequent matrix block, and estimates the SLC corrected residual enhanced phase subsequence after flat value correction;

[0045] The sequence merging unit traverses the SLC residual enhanced phase subsequences corresponding to all the sub-sample covariance matrix blocks, completes the adjustment and correction processing of each subsequence, and merges to obtain a complete N×1-dimensional SLC residual enhanced phase sequence;

[0046] The compensation unit compensates the SLC phase fringe trend into the SLC residual enhanced phase sequence to obtain the final SLC enhanced phase sequence.

[0047] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0048] The present invention adopts a frequency-domain phase spectrum information extraction method to estimate and suppress the SLC phase fringe trend and obtain SLC residual phase information; adaptively divides sub-covariance matrix blocks according to the coherence quality and matrix block overlap principle; combines the coherence and phase coupling enhancement model and the overlapping SLC homonymous image enhanced phase matching adjustment model to estimate the SLC enhanced phase sequence, effectively improving the DSInSAR phase enhancement performance and processing efficiency, and improving the deformation interpretation accuracy. Description of the Drawings

[0049] In order to more clearly illustrate the technical solutions of the present invention, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0050] Figure 1 It is a schematic flowchart of the method of the embodiment of the present invention;

[0051] Figure 2 It is an interferometric phase diagram before and after phase enhancement at an interval of 36 days in the Bulianta mining area of Inner Mongolia in the embodiment of the present invention. Among them, a is the original interferometric phase, b is the interferometric phase after EMI phase enhancement, c is the interferometric phase after Sequential phase enhancement, and d is the interferometric phase after the phase enhancement of the present invention;

[0052] Figure 3 It is an interferometric phase diagram before and after phase enhancement at an interval of 84 days in the Bulianta mining area of Inner Mongolia in the embodiment of the present invention. a is the original interferometric phase, b is the interferometric phase after EMI phase enhancement, c is the interferometric phase after Sequential phase enhancement, and d is the interferometric phase after the phase enhancement of the present invention;

[0053] Figure 4 This is the interferometric phase diagram before and after phase enhancement with an 84-day time interval in the Yungang mining area group in Datong, Shanxi Province in the embodiment of the present invention. a is the original interferometric phase, b is the interferometric phase after EMI phase enhancement, c is the interferometric phase after Sequential phase enhancement, and d is the interferometric phase after the phase enhancement method of the present invention;

[0054] Figure 5 This is the temporal coherence corresponding to different phase enhancement methods in the Yungang mining area group in Datong, Shanxi Province in the embodiment of the present invention. a is the temporal coherence corresponding to EMI phase enhancement, b is the temporal coherence corresponding to Sequential phase enhancement, and c is the temporal coherence corresponding to the phase enhancement of the present invention. Detailed implementation manners

[0055] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0056] To make the above objects, features, and advantages of the present invention more obvious and understandable, the present invention will be further described in detail below in conjunction with the accompanying drawings and specific implementation manners.

[0057] Embodiment 1

[0058] In this embodiment, as Figure 1 shown, an adaptive sequential phase enhancement method based on phase fringe trend suppression includes the following steps:

[0059] S1. Obtain temporal SAR images, construct a nearest neighbor interferometric network, and adaptively extract the temporal interferometric phase fringe trend of the nearest neighbor interferometric phase diagram according to the frequency domain phase spectrum information extraction method.

[0060] S1 includes: obtaining temporal SAR images; performing image registration preprocessing on the temporal SAR images, constructing a nearest neighbor interferometric network according to the information of the registered temporal SAR images, and obtaining the nearest neighbor interferometric phase diagram by conjugate multiplication; adaptively extracting the main phase fringe trend in the nearest neighbor interferometric phase diagram according to the frequency domain phase spectrum information extraction method, including main phase information such as flat earth phase, terrain phase, and deformation phase; traversing all nearest neighbor interferometric phase diagrams and extracting the temporal interferometric phase fringe trend of all nearest neighbor interferometric diagrams.

[0061] In this embodiment, the frequency domain phase spectrum information extraction method is defined as:

[0062]

[0063] where f (u，v) is the pixel block centered at the spatial domain pixel (i, j) after being converted by the fast Fourier transform into the pixel block centered at the frequency domain pixel (u, v). is the phase spectrum information, || is the modulus operation, and f thr is the spectral threshold for dividing the phase signal and the noise signal, max( ) is the matrix maximum value extraction operation, and λ is the spectral power division value. Based on the theory of extracting phase spectrum information in the frequency domain, the size of the spectral calculation window and the parameter of the spectral power division value are reasonably set to strictly control the phenomena of pseudo fringes and phase information loss. The larger the spectral estimation window, the more effectively the phase spectrum signal and the noise spectrum signal can be separated, but an overly large estimation window is likely to reduce the phase spectrum peak and smooth the phase spectrum signal, resulting in phase information loss; the smaller the spectral power division value, the more the interference of the noise spectrum information can be avoided. Furthermore, the trend of the temporal interference phase fringes in the interference phase map is adaptively extracted, including the main phase information such as the flat phase, the terrain phase, and the deformation phase.

[0064] S2. Estimate the SLC phase fringe trend from the trend of the temporal interference phase fringes by using the least squares method to obtain the SLC residual phase information.

[0065] S2 includes: Combining the least squares theory, taking any SLC image as a reference, estimating and extracting N SLC phase fringe trends from the trends of the temporal interference phase fringes corresponding to N - 1 nearest interferograms; multiplying the SLC phase fringe trend by the conjugate of the original SLC image to suppress the phase fringe trend and obtain the SLC residual phase information.

[0066] S3. Estimate the sample covariance matrix based on the temporal amplitude information and the SLC residual phase information of the temporal SAR images. According to the coherence quality and the matrix block overlap principle, divide the sub - sample covariance matrix blocks from the sample covariance matrix, and perform local phase enhancement on the sub - sample covariance matrix blocks to obtain the SLC residual enhanced phase subsequences.

[0067] S3 includes: According to the temporal amplitude information of the temporal SAR images, use the non - parametric test algorithm to identify the homogeneous pixel points. Based on the homogeneous pixel set and the SLC residual phase information, use the maximum likelihood estimation method to traverse each pixel in the homogeneous pixel set one by one to estimate the sample covariance matrix; take the modulus of the sample covariance matrix to extract the sample coherence matrix, and divide the sub - sample covariance matrix blocks according to the coherence quality and the sub - matrix block overlap principle; according to the phase enhancement model estimated by the coherence and phase coupling, perform local phase enhancement processing on each sub - sample covariance matrix block one by one to estimate the local SLC residual enhanced phase subsequences corresponding to all sub - sample covariance matrix blocks.

[0068] In this embodiment, the method for partitioning the sub-sample covariance matrix block includes: taking the modulus of the sample covariance matrix, extracting the sample coherence matrix, and based on the coherence index of the phase quality evaluation, using a 180-day time baseline as the threshold, the Nth max scene image that forms the time baseline closest to 180 days with the first scene image is set as the maximum dimension of the sub-matrix block, and the minimum dimension of the sub-matrix block is set to 5; extracting the first N max ×N-dimensional initial sub-matrix from the N×N-dimensional sample covariance matrix formed by N SLC images, calculating the average coherence value under the same time baseline, that is, the mean of the diagonal vectors of the sub-matrix, to obtain a 1×N-dimensional average coherence vector; extracting the Nth patch1 pixel positions that are stably greater than 0.4 in the 1×N-dimensional average coherence vector. If 5 ≤ N patch1 ≤ N max , then the dimension of the first matrix block is set to N patch1 ×N patch1 ; if N patch1 > N max , then the dimension of the first matrix block is set to N max ×N max ; if N patch1 < 5, then the dimension of the first matrix block is set to 5×5; assuming the dimension of the first matrix block is N patch1 ×N patch1 , according to the sub-matrix block overlap principle, the overlap rate is set to 20%, and the overlap dimension is calculated as N O = round(N patch1 ×20%), where round() is the rounding operation; combined with the first sub-matrix block, according to the sequential extraction principle, the starting pixel position of the initial sub-matrix for calculating the dimension estimation of the second matrix block is (N patch1 - N O + 1, N patch1 - N O + 1), and then an N max ×(N - N patch1 + N O )-dimensional initial sub-matrix is extracted. According to the above sub-matrix average coherence vector calculation and matrix block dimension judgment theory, the dimension N patch2 ×N patch2 of the second matrix block is estimated; repeat the above processing flow of block overlap degree calculation, initial sub-matrix starting pixel position estimation, initial sub-matrix extraction corresponding to the dimension estimation of the mth matrix block, and dimension estimation of the mth matrix block to divide the N×N-dimensional sample covariance matrix into multiple sub-sample covariance matrix blocks with inconsistent dimensions.

[0069] The method for estimating the local SLC residual enhanced phase subsequence includes: According to the complex circular Gaussian distribution hypothesis, the logarithmic function of the probability density function corresponding to the temporal phase is:

[0070]

[0071] where N SHP is the number of homogeneous pixels, N is the number of SAR images, ln() is the logarithmic operation, det() is the determinant operation, G is the true coherence matrix, z is the SLC residual phase vector, Θ is the complex phase matrix containing the true phase sequence θ, that is, Θ = diag(e iθ ), H is the transpose operation, and k is a natural number; furthermore, the coherence model is derived as:

[0072] Γ = ΘGΘ H

[0073] The phase enhancement model is:

[0074]

[0075] where Re() is the real part operation, is the true coherence matrix. Based on the above phase enhancement model of coherence and phase coupling estimation, local phase enhancement processing is carried out on each sub-sample covariance matrix block one by one to estimate the local SLC residual enhanced phase corresponding to all sub-sample covariance matrix blocks.

[0076] S4. Perform adjustment and correction processing on the homologous images of the SLC residual enhanced phase subsequence, and compensate the SLC phase fringe trend to the processed SLC residual enhanced phase sequence to obtain the SLC enhanced phase sequence.

[0077] S4 includes: For each SLC residual enhanced phase subsequence, perform SLC residual enhanced phase matching and adjustment processing on the homologous images in pairs, and taking the previous matrix block as a reference, substitute the adjusted value into the subsequent matrix block to estimate the SLC corrected residual enhanced phase subsequence after adjusting the value; traverse the SLC residual enhanced phase subsequences corresponding to all sub-sample covariance matrix blocks, complete the adjustment and correction processing of each subsequence, and combine to obtain the complete N×1-dimensional SLC residual enhanced phase sequence; compensate the SLC phase fringe trend to the SLC residual enhanced phase sequence to obtain the final SLC enhanced phase sequence:

[0078] θ op = θ fringe + θ res_op

[0079] where θ op is the final N×1-dimensional SLC enhanced phase sequence, θ fringe is the SLC phase fringe trend, θres_op It is the SLC residual enhanced phase sequence.

[0080] Example 2

[0081] To verify the technical effects of the present invention, this example selects the Bulianta mining area in Inner Mongolia as the test area, uses 24 scenes of Sentinel-1A SAR satellite data, and conducts phase enhancement processing and analysis based on EMI, Sequential, and the method of the present invention under the same conditions. For the SLC phase sequence after phase enhancement, SLC enhanced phases with time intervals of 36 days and 84 days are respectively selected for interference processing. The enhanced interference phase diagram at a 36-day time interval is as Figure 2 shown, Figure 2 (a) is the original interference phase, Figure 2 (b) is the interference phase after EMI phase enhancement, Figure 2 (c) is the interference phase after Sequential phase enhancement, Figure 2 (d) is the interference phase after the phase enhancement of the present invention. Through Figure 2 the interference phase results processed by different phase enhancement methods shown, it can be found that the interference phase corresponding to the 36-day time interval is more severely affected by noise. Through local magnification, it can be found that the phase fringes can be roughly identified at the dense fringes in the mining area; after EMI and Sequential phase enhancement processing, although the phase noise is significantly suppressed, significant phase fringe blurring appears at the dense fringes in the mining area; while the method of the present invention not only effectively suppresses the phase noise, improves the phase signal-to-noise ratio, but also effectively restores the phase fringe detail information at the dense fringes in the mining area, and the continuity of the phase fringe edges is stronger. The enhanced interference phase diagram at an 84-day time interval is as Figure 3 shown, Figure 3 (a) is the original interference phase, Figure 3 (b) is the interference phase after EMI phase enhancement, Figure 3 (c) is the interference phase after Sequential phase enhancement, Figure 3 (d) is the interference phase after the phase enhancement of the present invention; compared with the original interference phase at a 36-day time interval, the interference phase at an 84-day time interval is severely affected by phase noise, and the phase fringe information in the mining area can hardly be identified. However, after phase enhancement processing, the phase is smoother and the phase noise is significantly suppressed. However, the phase information in the fringe-dense area corresponding to the interference phase after EMI and Sequential phase enhancement still cannot be effectively restored, and the continuity of the phase fringe edges is poor. While the interference phase after the phase enhancement of the method of the present invention effectively balances the recovery of phase information and the improvement of the signal-to-noise ratio, has better noise robustness and self-adaptability, and effectively verifies the better phase enhancement performance of the method of the present invention.

[0082] Example 3

[0083] In order to further verify the universal technical effect of the present invention in various scenarios, this embodiment selects the mining area group in Yungang District, Datong City, Shanxi Province as the test area, and also uses 24 scenes of Sentinel-1ASAR satellite data to carry out phase enhancement processing analysis under the same conditions based on EMI, Sequential and the method of the present invention. The experiment selects the results before and after interferometric phase enhancement at an interval of 84 days for comparative analysis, such as Figure 4 As shown, Figure 4 (a) is the original interference phase, Figure 4 (b) is the interference phase after EMI phase enhancement, Figure 4 (c) is the interference phase after Sequential phase enhancement, Figure 4 (d) is the interference phase after phase enhancement of the present invention. Analysis shows that the interference phase after EMI and Sequential phase enhancement only effectively suppresses phase noise, but does not restore the phase fringe edges in the densely packed areas of the mining area. The method of the present invention not only effectively improves the phase signal-to-noise ratio, but also effectively restores the edge information of the phase fringe. In addition, the experiment also selected the time domain coherence used to screen DS candidate points as the evaluation index of phase enhancement performance. The calculation results are shown in Figure 5 As shown, Figure 5 (a) is the time domain coherence corresponding to EMI phase enhancement, Figure 5 (b) is the time domain coherence corresponding to Sequential phase enhancement. Figure 5 (c) is the time domain coherence corresponding to the phase enhancement of the present invention. Statistical analysis shows that, with the time domain coherence of 0.8 as the threshold, the corresponding pixel proportion of EMI is about 52.88%, the corresponding pixel proportion of Sequential is about 47.89%, and the corresponding pixel proportion of the present invention is about 59.08%; for the local mining area A, the corresponding pixel proportions of EMI, Sequential and the present invention method greater than 0.8 are 40.75%, 35.24% and 71.74% respectively; for the local mining area B, the corresponding pixel proportions of EMI, Sequential and the present invention method greater than 0.8 are 48.25%, 43.56% and 62.10% respectively; the experimental results show that the present invention method can effectively improve the density of deformation monitoring points, especially in the large gradient deformation of the mining area, the pixel proportion is at least increased by about 13.85% compared with the existing method.

[0084] In summary, it can be seen that the method of the present invention can effectively restore the phase fringe information at the large gradient deformation in the mining area, and at the same time can effectively take into account the improvement of the phase signal-to-noise ratio and improve the accuracy of phase information interpretation; in addition, the method of the present invention can effectively increase the density of monitoring points at the large gradient deformation in the mining area and improve the richness of the interpretation of the deformation field information in the mining area.

[0085] Embodiment 4

[0086] In this embodiment, an adaptive sequential phase enhancement system based on phase fringe trend suppression includes: a phase fringe trend extraction module, a residual phase information acquisition module, a residual enhanced phase acquisition module, and an enhanced phase acquisition module;

[0087] The phase fringe trend extraction module is used to obtain temporal SAR images, construct a nearest neighbor interference network, and adaptively extract the temporal interference phase fringe trend of the nearest neighbor interference phase map according to the frequency domain phase spectrum information extraction method.

[0088] The phase fringe trend extraction module includes: an image acquisition unit, an interference phase map acquisition unit, a main phase fringe trend extraction unit, and a phase fringe trend extraction unit. The image acquisition unit is used to obtain temporal SAR images; the interference phase map acquisition unit performs image registration preprocessing on the temporal SAR images, constructs a nearest neighbor interference network according to the information of the registered temporal SAR images, and obtains the nearest neighbor interference phase map by conjugate multiplication; the main phase fringe trend extraction unit adaptively extracts the main phase fringe trend in the nearest neighbor interference phase map according to the frequency domain phase spectrum information extraction method, including main phase information such as flat earth phase, terrain phase, and deformation phase; the phase fringe trend extraction unit traverses all the nearest neighbor interference phase maps and extracts the temporal interference phase fringe trend of all the nearest neighbor interference maps.

[0089] The residual phase information acquisition module estimates the SLC phase fringe trend from the temporal interference phase fringe trend using the least squares method to obtain the SLC residual phase information.

[0090] The residual phase information acquisition module includes: an SLC phase fringe trend extraction unit and an SLC residual phase information extraction unit. The SLC phase fringe trend extraction unit combines the least squares theory, takes any SLC image as a reference, and estimates and extracts N SLC phase fringe trends from the temporal interference phase fringe trends corresponding to N - 1 nearest neighbor interference maps; the SLC residual phase information extraction unit conjugately multiplies the SLC phase fringe trend with the original SLC image to suppress the phase fringe trend and obtain the SLC residual phase information.

[0091] The residual enhanced phase acquisition module estimates the sample covariance matrix based on the temporal amplitude information and the SLC residual phase information of the temporal SAR images, divides the sub-sample covariance matrix blocks from the sample covariance matrix according to the coherence quality and the matrix block overlap principle, and performs local phase enhancement on the sub-sample covariance matrix blocks to obtain the SLC residual enhanced phase subsequence.

[0092] The residual enhanced phase acquisition module includes: a sample covariance matrix acquisition unit, a sample covariance matrix block division unit, and an SLC residual enhanced phase subsequence acquisition unit. The sample covariance matrix acquisition unit identifies homogeneous pixel points by using a non-hypothesis testing algorithm based on the temporal amplitude information of the temporal SAR image. Based on the homogeneous pixel set and the SLC residual phase information, the maximum likelihood estimation method is used to traverse each pixel in the homogeneous pixel set one by one to estimate the sample covariance matrix. The sample covariance matrix block division unit takes the modulus of the sample covariance matrix, extracts the sample coherence matrix, and divides the sub-sample covariance matrix blocks according to the coherence quality and the sub-matrix block overlap principle. The SLC residual enhanced phase subsequence acquisition unit performs local phase enhancement processing on each sub-sample covariance matrix block one by one according to the phase enhancement model estimated by coherence and phase coupling, and estimates the local SLC residual enhanced phase subsequences corresponding to all sub-sample covariance matrix blocks.

[0093] The enhanced phase acquisition module is used to perform adjustment and correction processing on the homologous images of the SLC residual enhanced phase subsequences, and compensate the SLC phase fringe trend to the processed SLC residual enhanced phase sequence to obtain the SLC enhanced phase sequence.

[0094] The enhanced phase acquisition module includes: an adjustment value correction unit, a sequence merging unit, and a compensation unit. The adjustment value correction unit performs SLC residual enhanced phase matching adjustment and correction processing on the homologous images of each SLC residual enhanced phase subsequence in pairs, and takes the previous matrix block as a reference, substitutes the adjustment value into the subsequent matrix block, and estimates the SLC corrected residual enhanced phase subsequence after adjustment value correction. The sequence merging unit traverses the SLC residual enhanced phase subsequences corresponding to all sub-sample covariance matrix blocks, completes the adjustment and correction processing of each subsequence, and merges to obtain a complete N×1-dimensional SLC residual enhanced phase sequence. The compensation unit compensates the SLC phase fringe trend to the SLC residual enhanced phase sequence to obtain the final SLC enhanced phase sequence.

[0095] The embodiments described above are only descriptions of the preferred embodiments of the present invention, and do not limit the scope of the present invention. Without departing from the design spirit of the present invention, various deformations and improvements made by those of ordinary skill in the art to the technical solutions of the present invention shall fall within the protection scope determined by the claims of the present invention.

Claims

1. An adaptive sequential phase enhancement method based on phase fringe trend suppression, characterized in that: The following steps are involved: S1. Acquire time-series SAR images, construct a nearest neighbor interferometric network, and adaptively extract the time-series interferometric phase fringe trend of the nearest neighbor interferometric phase map based on the frequency domain phase spectrum information extraction method; S2. Using the least squares method to estimate the SLC phase fringe trend from the timing interference phase fringe trend, the SLC residual phase information is obtained; S3. Estimate a sample covariance matrix based on the time-series amplitude information of the time-series SAR image and the SLC residual phase information, divide the sample covariance matrix into sub-sample covariance matrix blocks according to the coherence quality and matrix block overlap principle, and perform local phase enhancement on the sub-sample covariance matrix blocks to obtain an SLC residual enhanced phase sub-sequence; S4. performing adjustment correction processing on the image with the same name of the SLC residual enhanced phase subsequence, and compensating the SLC phase fringe trend to the processed SLC residual enhanced phase sequence to obtain the SLC enhanced phase sequence.

2. According to claim 1, the method for adaptive sequential phase enhancement based on phase fringe trend suppression is characterized in that: The S1 includes: Acquiring the time-series SAR image; Performing image registration preprocessing on the time-series SAR images, constructing the nearest neighbor interference network based on the registered time-series SAR image information, and obtaining the nearest neighbor interference phase map by conjugate multiplication; According to the frequency domain phase spectrum information extraction method, the main phase fringe trend in the nearest neighbor interference phase image is adaptively extracted, including the main phase information such as flat ground phase, terrain phase and deformation phase; All the nearest interference phase images are traversed to extract the temporal interference phase fringe trends of all the nearest interference images.

3. According to claim 2, the method for adaptive sequential phase enhancement based on phase fringe trend suppression is characterized in that: The S2 includes: In combination with the least squares theory, taking any SLC image as a reference, estimating and extracting N SLC phase fringe trends from the temporal interference phase fringe trends corresponding to N-1 nearest interference patterns; The SLC phase fringe trend is conjugate-multiplied with the original SLC image to suppress the phase fringe trend and obtain the SLC residual phase information.

4. According to claim 3, the method for adaptive sequential phase enhancement based on phase fringe trend suppression is characterized in that: The S3 includes: According to the time series amplitude information of the time series SAR image, a non-hypothesis test algorithm is used to identify homogeneous pixel points, and based on the homogeneous pixel set and the SLC residual phase information, a maximum likelihood estimation method is used to traverse the pixels in the homogeneous pixel set one by one to estimate the sample covariance matrix; Taking a module of the sample covariance matrix, extracting a sample coherence matrix, and dividing the sub-sample covariance matrix blocks according to the coherence quality and the principle of sub-matrix block overlap; According to the phase enhancement model of coherence and phase coupling estimation, local phase enhancement processing is performed on the sub-sample covariance matrix blocks one by one, and the local SLC residual enhancement phase subsequences corresponding to all the sub-sample covariance matrix blocks are estimated.

5. The adaptive sequential phase enhancement method based on phase fringe trend suppression according to claim 4, characterized in that: The S4 includes: For each of the SLC residual enhanced phase subsequences, the SLC residual enhanced phase matching adjustment processing is carried out for the images with the same name in pairs, and the adjustment value is substituted into the rear matrix block based on the previous matrix block to estimate the SLC corrected residual enhanced phase subsequence after the adjustment value correction; Traversing the SLC residual enhanced phase subsequences corresponding to all the subsample covariance matrix blocks, completing the adjustment correction processing of each subsequence, and merging to obtain a complete N×1 dimensional SLC residual enhanced phase sequence; The SLC phase fringe trend is compensated to the SLC residual enhanced phase sequence to obtain the final SLC enhanced phase sequence.

6. An adaptive sequential phase enhancement system based on phase fringe trend suppression, the system applying the method according to any one of claims 1 to 5, characterized in that: include: Phase fringe trend extraction module, residual phase information acquisition module, residual enhanced phase acquisition module and enhanced phase acquisition module; The phase fringe trend extraction module is used to acquire time-series SAR images, construct a nearest neighbor interferometric network, and adaptively extract the time-series interferometric phase fringe trend of the nearest neighbor interferometric phase image based on the frequency domain phase spectrum information extraction method; The residual phase information acquisition module estimates the SLC phase fringe trend from the timing interference phase fringe trend using the least square method to obtain the SLC residual phase information; The residual enhancement phase acquisition module estimates the sample covariance matrix based on the time series amplitude information of the time series SAR image and the SLC residual phase information, divides the sub-sample covariance matrix blocks from the sample covariance matrix according to the coherence quality and matrix block overlap principle, and performs local phase enhancement on the sub-sample covariance matrix blocks to obtain the SLC residual enhancement phase sub-sequence; The enhanced phase acquisition module is used to perform adjustment correction processing on the image with the same name of the SLC residual enhanced phase subsequence, and compensate the SLC phase fringe trend to the processed SLC residual enhanced phase sequence to obtain the SLC enhanced phase sequence.

7. The adaptive sequential phase enhancement system based on phase fringe trend suppression according to claim 6, characterized in that: The phase fringe trend extraction module includes: an image acquisition unit, an interference phase image acquisition unit, a main phase fringe trend extraction unit and a phase fringe trend extraction unit; The image acquisition unit is used to acquire the time-series SAR image; The interferometric phase image acquisition unit performs image registration preprocessing on the time-series SAR image, constructs the nearest neighbor interferometric network according to the registered time-series SAR image information, and obtains the nearest neighbor interferometric phase image by conjugate multiplication; The main phase fringe trend extraction unit adaptively extracts the main phase fringe trend in the nearest neighbor interference phase image according to the frequency domain phase spectrum information extraction method, including main phase information such as flat ground phase, terrain phase and deformation phase; The phase fringe trend extraction unit traverses all the nearest interference phase images and extracts the temporal interference phase fringe trends of all the nearest interference images.

8. The adaptive sequential phase enhancement system based on phase fringe trend suppression according to claim 7, characterized in that: The residual phase information acquisition module includes: an SLC phase fringe trend extraction unit and an SLC residual phase information extraction unit; The SLC phase fringe trend extraction unit combines the least squares theory and takes any SLC image as a reference to estimate and extract N SLC phase fringe trends from the temporal interference phase fringe trends corresponding to N-1 nearest interference images; The SLC residual phase information extraction unit conjugate-multiplies the SLC phase fringe trend with the original SLC image to suppress the phase fringe trend and obtain the SLC residual phase information.

9. The adaptive sequential phase enhancement system based on phase fringe trend suppression according to claim 8, characterized in that: The residual enhancement phase acquisition module includes: a sample covariance matrix acquisition unit, a sample covariance matrix block division unit and an SLC residual enhancement phase subsequence acquisition unit; The sample covariance matrix acquisition unit uses a non-hypothesis test algorithm to identify homogeneous pixel points based on the time series amplitude information of the time series SAR image, and uses a maximum likelihood estimation method to traverse the pixels in the homogeneous pixel set one by one on the basis of the homogeneous pixel set and the SLC residual phase information to estimate the sample covariance matrix; The sample covariance matrix block division unit modulos the sample covariance matrix, extracts the sample coherence matrix, and divides the sub-sample covariance matrix blocks according to the coherence quality and the sub-matrix block overlap principle; The SLC residual enhanced phase subsequence acquisition unit performs local phase enhancement processing on the subsample covariance matrix blocks one by one according to the phase enhancement model estimated by coherence and phase coupling, and estimates the local SLC residual enhanced phase subsequences corresponding to all the subsample covariance matrix blocks.

10. The adaptive sequential phase enhancement system based on phase fringe trend suppression according to claim 9, characterized in that: The enhanced phase acquisition module includes: an adjustment value correction unit, a sequence merging unit and a compensation unit; The adjustment value correction unit performs SLC residual enhancement phase matching adjustment processing on the same-name images in pairs for each of the SLC residual enhancement phase subsequences, and substitutes the adjustment value into the rear matrix block based on the previous matrix block to estimate the SLC correction residual enhancement phase subsequence after the adjustment value correction; The sequence merging unit traverses the SLC residual enhanced phase subsequences corresponding to all the subsample covariance matrix blocks, completes the adjustment correction processing of each subsequence, and merges to obtain a complete N×1 dimensional SLC residual enhanced phase sequence; The compensation unit compensates the SLC phase fringe trend into the SLC residual enhancement phase sequence to obtain the final SLC enhancement phase sequence.

Images