A temporal coherent polarization optimization method, device, system, and storage medium

Abstract

The present invention discloses a time-coherent polarization optimization method and device, system, and storage medium, including: obtaining two polarization SLC images of VV and VH; selecting PSC candidate points by DA amplitude deviation index; for PSC candidate points with DA values ​​less than 0.25, connecting adjacent PSC candidate points to construct a spatial irregular triangular reference network, obtaining the best projection vector by signal-to-noise ratio calculation to maximize the temporal coherence coefficient of the arc, and selecting the optimal model parameters; for isolated points with ADI values ​​less than 0.4 outside the reference network, constructing a star network by searching nearby points, and estimating the arcs in the star network one by one to select the optimal model parameters; step S5, after orbit error correction and atmospheric phase estimation, the deformation estimation sequence of each PS point is obtained for analysis, and the corresponding sedimentation map is generated. The technical solution of the present invention can effectively improve the reliability and accuracy of PS point selection.

Description

Technical Field

[0001] The invention belongs to the technical field of synthetic aperture radar interferometry polarization optimization, and in particular relates to a time coherent polarization optimization method and device, a system, and a storage medium. Background Art

[0002] Persistent scattering InSAR (PSI) is a cutting-edge technology for dealing with decorrelation and atmospheric noise in radar interferometry. It was first proposed by Italian scholars in this field, Ferretti et al., in 1999. This method can identify scatterers that exhibit coherent scattering behavior over time, which are called persistent scatterers (PS). PSI measurement technology can utilize point targets characterized by a single PS, which have temporarily stable scattering behavior and high-quality interferometric phases, corresponding to artificial structures, exposed rocks, and artificial reflectors. Therefore, the selection of candidate PS (PSC) is a key step in PSI measurement technology, which directly affects the quality and coverage of the final product. Usually, PSI technology calculates ADI based on time series SAR amplitude information and identifies PSC under a certain threshold, which is generally set to 0.25-0.4.

[0003] Another PSI method for identifying a large number of PS pixels in all terrains (including non-urban areas lacking artificial buildings) has also been proposed. This method can not only identify PS pixels using the spatial correlation of the phase, but also characterize a parameter of phase stability, similar to the measure of temporal coherence, which is called temporal coherence. It is not exactly the same as the traditional ensemble phase coherence because it requires a predefined deformation model.

[0004] Before the advent of radar sensors with polarimetric configurations, SAR interferometry mostly used a single polarimetric channel. Radar polarimetric methods are a technique for extracting geophysical parameters from SAR images. Various methods for achieving this goal are based either on statistical analysis of polarimetric information or on scattering models, which provide a physical explanation of the scattering process. Therefore, the introduction of polarimetric optimization methods in radar interferometry can improve the performance of SAR interferometry. Cloude et al. proposed a general formula for coherent conventional interferometry using polarimetric methods in 1998. The method first establishes the spatial coherence optimization problem of different polarimetric channels, and then solves it to obtain the optimal linear combination of the channels to obtain the best phase estimate. As the spatial coherence is optimized, the decorrelation terms are reduced and the signal-to-noise ratio is improved. Since the density and quality of PS pixels are important factors in PSI technology, Navarro-Sanchez et al. adopted the concept of polarimetric optimization in PSI technology and proposed a polarimetric PSI method using spaceborne data sets, called the exhaustive search polarimetric method (ESPO). This method finds the best weight for each available polarimetric channel to obtain the best combination of these channels that maximizes the PS selection criterion. After ESPO was proposed, it was applied to polarization optimization of ADI and ensemble temporal coherence, with significant optimization results. The ESPO method must search for defined parameters one by one within a specified step size, and the computational disadvantages are quite obvious. Summary of the invention

[0005] The technical problem to be solved by the present invention is to provide a time-coherent polarization optimization method and device, system, and storage medium to optimize the amplitude spread index and time coherence, and apply them to the polarization optimization of ADI and time coherence, which can effectively improve the reliability and accuracy of PS point selection.

[0006] To achieve the above object, the present invention adopts the following technical solution:

[0007] A time-coherent polarization optimization method comprises the following steps:

[0008] Step S1, obtaining two polarization SLC images, VV and VH;

[0009] Step S2, selecting PSC candidate points according to the VV and VH polarized SLC images by using the DA amplitude deviation index;

[0010] Step S3, for PSC candidate points with DA values ​​less than 0.25, adjacent PSC candidate points are connected to construct a spatial irregular triangular reference network, the best projection vector is obtained by signal-to-noise ratio calculation to maximize the temporal coherence coefficient of the arc, and the optimal model parameters are estimated;

[0011] Step S4: For isolated points with ADI values ​​less than 0.4 outside the reference network, a star-shaped isolated network is constructed by searching for nearby pixels, and the isolated network is combined with the reference network to estimate the optimal model parameters of the isolated points one by one;

[0012] Step S5: After orbit error correction and atmospheric phase estimation, the deformation estimation sequence of each PS point is obtained for analysis to generate the corresponding sedimentation map.

[0013] Preferably, in step S1, differential interference processing and topographic phase removal are performed on the VV and VH polarization SLC images.

[0014] Preferably, in step S4, the maximum search distance is limited to within 1.5 kilometers to minimize the impact of atmospheric delay.

[0015] The present invention also provides a time coherent polarization optimization device, comprising:

[0016] An acquisition module, used for acquiring two polarization SLC images, VV and VH;

[0017] A selection module is used to select PSC candidate points according to the DA amplitude deviation index based on the two polarization SLC images of VV and VH;

[0018] The first processing module is used for connecting adjacent PSC candidate points with a DA value less than 0.25 to construct a spatial irregular triangular reference network, obtaining the best projection vector by signal-to-noise ratio calculation to maximize the temporal coherence coefficient of the arc, and selecting the optimal model parameters;

[0019] The second processing module, for isolated points with ADI values ​​less than 0.4 outside the reference network, constructs a star-shaped isolated network by searching for nearby pixels, and combines the isolated network with the reference network to estimate the optimal model parameters of the isolated points one by one;

[0020] The generation module is used to obtain the deformation estimation sequence of each PS point after orbit error correction and atmospheric phase estimation, and then analyze it to generate the corresponding sedimentation map.

[0021] Preferably, the acquisition module is also used to perform differential interference processing and topographic phase removal on the VV and VH polarization SLC images.

[0022] Preferably, the maximum search distance is limited to 1.5 km to minimize the effects of atmospheric delay.

[0023] The present invention also provides a time-coherent polarization optimization system, comprising: a memory and a processor, wherein the memory stores a computer program executed by the processor, and the computer program executes the time-coherent polarization optimization method when executed by the processor.

[0024] The present invention also provides a storage medium, on which a computer program is stored, and the computer program executes the time coherent polarization optimization method when running.

[0025] The present invention uses the conjugate gradient method to optimize the signal-to-noise ratio. Compared with the results obtained by using the PSI technology for existing single-polarization VV and VH images, it can greatly improve the reliability and number of PS pixels. Compared with the multi-polarization ESPO optimization method, its calculation time is greatly shortened, and the signal-to-noise ratio optimization result is similar to ESPO. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0027] Figure 1 This is a flow chart of a time-coherent polarization optimization method according to an embodiment of the present invention;

[0028] Figure 2 Optimize distribution map for ADI. DETAILED DESCRIPTION

[0029] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0030] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0031] Embodiment 1:

[0032] The embodiment of the present invention provides a method for optimizing time-coherent polarization based on signal-to-noise ratio, which improves the value of signal-to-noise ratio by iteratively searching parameters in one-dimensional space, thereby optimizing DA and time coherence coefficient. In dual-polarization interferometry, the pixel of Sentinel-1A image can be defined as the Pauli vector k:

[0033] k=[S VV ,2S VH ] T

[0034] Among them, S VV represents the vertical common polarity channel, S VH is the cross polarization channel, and T is the transpose operator. The polarization interferometry method projects each SAR image pixel k onto the corresponding projection vector ω to obtain a new pixel value μ. Therefore, when the two polarization channels are merged into one channel, the pixel μ is expressed as:

[0035] μ=ω * Tk

[0036] Wherein, * represents the conjugate operator.

[0037] Furthermore, the values ​​of the two polarization channels can be expressed as:

[0038]

[0039] in, and are the signal value and noise value of VV polarization data, respectively, while the definition of VH polarization data is the same as VV. Combining the above two formulas, we can get the following vector:

[0040]

[0041] Among them, k signal The signal part representing the Pauli vector k

[0042] This method is used to *T Parameterized, the expression is as follows:

[0043]

[0044] The value range of the real number x1 is from 0 to infinity, and a reasonable traversal range can be set as needed.

[0045] The ESPO method has a ω of *T is defined as:

[0046]

[0047] where α and θ1 represent the type of scattering mechanism and the direction of the target scatterer, respectively.

[0048] We can further get:

[0049]

[0050] Where A is μ signal is the amplitude, and β is the phase. signal Raise to the power of μ signal(p) =A 2Assume that the noise is independent of the signal and n0,n1~N(0,σ 2 ), then the power of the noise is μ noise(p) =(1+x 2 )σ 2 .

[0051] According to the definition of signal-to-noise ratio, we can get:

[0052]

[0053] Among them, if the value of x1 is fixed, μ noise(p) will be independent of θ1.

[0054] The SNR reaches its maximum value at α0=α1+θ1, so the optimal θ is defined as:

[0055] θ 1(opt) =α0-α1

[0056] The higher the signal-to-noise ratio, the higher the optimized value. The strategy for finding the best projection vector ω is to first set the initial value of x1, and then find the best θ by traversing θ1. 1(opt) Then use the obtained θ 1(opt) , optimize the value of x1 in the search space and get the best x 1(opt) . Repeating this several times will yield the best projection vector that maximizes the signal-to-noise ratio. Assuming that the number of calculations for θ1 and x1 are m and n respectively, the number of calculations for one iteration is (m+n). Assuming that the number of iterations is N, the total number of calculations is N*(m+n), and the number of calculations for the optimal projection vector ω obtained by searching for two parameters in two-dimensional space using the ESPO method is m*n. In general, when the number N is 2 or 3, the accuracy of the calculation is already very high, basically equal to the value obtained by the ESPO method. Nevertheless, when the values ​​of m and n are large, the cost and time of this method is much less than that of the ESPO method.

[0057] like Figure 1 As shown, an embodiment of the present invention provides a time-coherent polarization optimization method including:

[0058] Step 1: Select two polarization SLC images, VV and VH, from the Sentinel-1A image map, use the InSAR Scientific Computing Environment (ISCE) software to perform differential interference processing and terrain phase removal on the polarization data, and then generate an interferogram based on dual-polarization projection;

[0059] In order to obtain the interferometric image of VV / VH, an image is selected from the existing SLC images as the master image, and the other slave images are interfered with the master image to obtain:

[0060]

[0061] Among them, μ1 and μ2 as well as ω1 and ω2 are the pixel values ​​and projection vectors of the master image and the slave image respectively.

[0062] Step 2: Calculate the amplitude deviation index through the dual polarization channel, set the threshold to select the PSC candidate point, and obtain the optimized polarization phase information. The calculation formula of the amplitude deviation index in polarization form is as follows:

[0063]

[0064] Among them, σ a and a represent the standard deviation and the mean amplitude, respectively. N is the total number of SAR images in the image stack. The overline represents the mean. The embodiment of the present invention performs a grid search on each pixel in the entire polarization space to obtain an appropriate projection vector ω, thereby minimizing D A Value. D of pixel A The lower the value, the better the corresponding phase quality. During the period with large average amplitude, the phase quality is more stable and it is easier to become a PSC candidate point. In order to show the efficiency of the SNR method, Figure 2 The ESPO method and the SNR method are compared in Figure 2 As shown, the average D of the best channels obtained by the ESPO method and the SNR method A The values ​​are 0.42 and 0.43 respectively, while the average D A The values ​​are 0.52 and 0.51.

[0065] Step 3: Select PSC candidate points with DA values ​​less than 0.25, and use Delaunay triangulation to connect adjacent candidate points to build a triangular reference network. The phase difference between adjacent PSC candidate points on each coherent arc is expressed as the sedimentation rate Δv x,y and DEM error increment Δh x,y These increments are different on each arc.

[0066] The phase difference between two adjacent pixels is phase-unwrapped to obtain the observed phase of the arc in the reference network. And the model phase It is expressed as follows:

[0067]

[0068] Where Δv x,y and Δh x,y It is the model parameter of the arc, which represents the average velocity difference and residual elevation difference between x and y pixels.

[0069] Since the absolute parameters of PS are obtained from the conduction arc model parameters, it is crucial to ensure the accuracy of the model parameter estimation. Ferretti et al. introduced the temporal coherence γ x,y The concept of phase residual is equivalent to the root mean square (RMS) value of the phase residual. It provides a measure of how well the observed phase and displacement fit the model for each arc.

[0070]

[0071] In other words, γ x,y Can be used to evaluate the observed phase The model parameters (Δv x,y ,Δh x,y ) reliability.

[0072] In order to obtain the polarization form of the temporal coherence expression The interferogram phase and wrapped phase difference between pixel x and pixel y are given in polarization form:

[0073]

[0074] The higher the temporal coherence, the smaller the model parameter (Δv x,y ,Δh x,y After obtaining the model parameters of each arc in the reference network, spatial integration and integration testing are required to verify whether the data is correct.

[0075] Step 4: For isolated points with ADI values ​​less than 0.4 outside the reference network, a star-shaped isolated network is constructed by searching nearby pixels. The isolated network and the reference network are combined to estimate the optimal model parameters of the isolated points one by one. The present invention uses an adaptive estimation strategy to process isolated points, which includes two parts: adding a single pixel to the reference network and adding multiple pixels. First, the connection between a single isolated pixel and all pixels in the reference network is estimated within the initial spatial search window. The initial search window is the minimum search size plus the specified distance, where the minimum search window is set from the isolated pixel to the reference network. If the pixel passes the outlier detection process, its absolute parameters are estimated. In three cases, the new search window will repeat the same process. (1) The number of arcs estimated to fit the model is insufficient. (2) The search window pixels are insufficient. (3) The pixel fails the model test. The new search window is equal to the current search window plus the specified distance. Iteratively increase the current search window by the specified distance until the specified maximum search window size is reached or the pixel passes the outlier test.

[0076] Step 5. In order to improve the quality of the derived displacement model, it is necessary to estimate the atmospheric and orbital error components. To estimate the atmospheric and orbital error signals, it is necessary to unwrap the residual phase of each interferogram. Since the maximum phase component of each differential interferogram (i.e., linear displacement rate and DEM error) has been removed, the phase difference between adjacent pixels is unlikely to be greater than the interval [-π,π]. Therefore, the sparse MCF unwrapping algorithm can be directly used to unwrap the residual phase of each interferogram. The contribution of atmospheric perturbations and orbital errors to the phase can be estimated from the unwrapped residual phase. After orbital error correction and atmospheric phase estimation, the deformation estimation sequence of each PS point is obtained, and the corresponding settlement map is generated by drawing using ARCGIS software.

[0077] The embodiment of the present invention improves the time-series InSAR processing method, selects a projection vector different from the ESPO method, and iteratively calculates and estimates the parameters in the reference network by maximizing the temporal coherence. The degree of optimization is very close to the result calculated by the ESPO exhaustive method. The time required for parameter estimation can be greatly shortened, and the reliability and accuracy of the estimated parameters are significantly improved compared with the traditional VV single-polarization channel.

[0078] Embodiment 2:

[0079] The embodiment of the present invention further provides a time coherent polarization optimization device, comprising:

[0080] An acquisition module, used for acquiring two polarization SLC images, VV and VH;

[0081] A selection module is used to select PSC candidate points according to the DA amplitude deviation index based on the two polarization SLC images of VV and VH;

[0082] The first processing module is used for connecting adjacent PSC candidate points with a DA value less than 0.25 to construct a spatial irregular triangular reference network, obtaining the best projection vector by signal-to-noise ratio calculation to maximize the temporal coherence coefficient of the arc, and selecting the optimal model parameters;

[0083] The second processing module is used for constructing a star network for PSC candidate points with DA values ​​between 0.25 and 0.4 by searching nearby points, and estimating arcs in the star network one by one to select the optimal model parameters;

[0084] The generation module is used to obtain the deformation estimation sequence of each PS point after orbit error correction and atmospheric phase estimation, and then analyze it to generate the corresponding sedimentation map.

[0085] As an implementation of the embodiment of the present invention, the acquisition module is also used to perform differential interference processing and topographic phase removal on the VV and VH polarization SLC images.

[0086] As an implementation of the embodiment of the present invention, the maximum search distance is limited to within 1.5 kilometers to minimize the impact of atmospheric delay.

[0087] Embodiment 3:

[0088] An embodiment of the present invention further provides a time-coherent polarization optimization system, comprising: a memory and a processor, wherein the memory stores a computer program executed by the processor, and the computer program executes the time-coherent polarization optimization method when executed by the processor.

[0089] Embodiment 4:

[0090] An embodiment of the present invention further provides a storage medium, on which a computer program is stored, and the computer program executes the time-coherent polarization optimization method when running.

[0091] The embodiments described above are only descriptions of the preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should all fall within the protection scope determined by the claims of the present invention.

Claims

1. A time-coherent polarization optimization method, characterized in that: The following steps are involved: Step S1, obtaining two polarization SLC images, VV and VH; Step S2, selecting PSC candidate points according to the VV and VH polarized SLC images by using the DA amplitude deviation index; Step S3, for PSC candidate points with DA values ​​less than 0.25, adjacent PSC candidate points are connected to construct a spatial irregular triangular reference network, the best projection vector is obtained by signal-to-noise ratio calculation to maximize the temporal coherence coefficient of the arc, and the optimal model parameters are selected; Step S4: For isolated points with ADI values ​​less than 0.4 outside the reference network, a star-shaped isolated network is constructed by searching for nearby pixels, and the isolated network is combined with the reference network to estimate the optimal model parameters of the isolated points one by one; Step S5: After orbit error correction and atmospheric phase estimation, the deformation estimation sequence of each PS point is obtained for analysis to generate the corresponding sedimentation map.

2. The time-coherent polarization optimization method according to claim 1, characterized in that: In step S1, differential interference processing and topographic phase removal are performed on the VV and VH polarization SLC images.

3. The time-coherent polarization optimization method according to claim 2, characterized in that: In step S4, the maximum search distance is limited to within 1.5 kilometers.

4. A time-coherent polarization optimization device, characterized in that: include: An acquisition module, used for acquiring two polarization SLC images, VV and VH; A selection module is used to select PSC candidate points according to the DA amplitude deviation index based on the two polarization SLC images of VV and VH; The first processing module is used for connecting adjacent PSC candidate points with a DA value less than 0.25 to construct a spatial irregular triangular reference network, obtaining the best projection vector by signal-to-noise ratio calculation to maximize the temporal coherence coefficient of the arc, and selecting the optimal model parameters; The second processing module constructs a star-shaped isolated network for isolated points with ADI values ​​less than 0.4 outside the reference network by searching for nearby pixels. By combining the isolated network with the reference network, the optimal model parameters of the isolated points can be estimated one by one. The generation module is used to obtain the deformation estimation sequence of each PS point after orbit error correction and atmospheric phase estimation, and then analyze it to generate the corresponding sedimentation map.

5. The time-coherent polarization optimization device according to claim 4, characterized in that: The acquisition module is also used to perform differential interference processing and topographic phase removal on the VV and VH polarization SLC images.

6. The time-coherent polarization optimization device according to claim 5, characterized in that: The maximum search distance is limited to 1.5 km to minimize the effects of atmospheric delay.

7. A time-coherent polarization optimization system, characterized in that: include: A memory and a processor, wherein the memory stores a computer program executed by the processor, and when the computer program is executed by the processor, the time-coherent polarization optimization method according to any one of claims 1 to 3 is executed.

8. A storage medium, characterized in that: The storage medium stores a computer program, which executes the time-coherent polarization optimization method according to any one of claims 1 to 3 when running.

Images