Large-area psinsar processing method, device and equipment based on block strategy

Abstract

The application discloses a large-area PSInSAR processing method, device and equipment based on a block strategy, which comprises the following steps: automatically dividing data blocks by using a regular grid division method with overlapping degrees according to the row and column numbers of original data and sub-block parameters to obtain a plurality of sub-blocks; performing block PS point selection, block network construction and block parameter solving processing on each sub-block; and adjusting parameters between the sub-blocks by using overlapping degree information and connection relationship between the sub-blocks to realize result smoothing and accurate splicing. The application firstly divides data blocks and then performs block processing, thereby improving memory occupation in the steps of point selection, network construction, parameter solving and the like during PSInSAR processing and improving parallel processing efficiency; the number of PS points and the number of network redundant arc segments can be increased without reducing the calculation efficiency under the current hardware conditions, which is beneficial to improving the precision, density and coverage area of the result; and the information of the homonymic points considering the sub-block overlapping degree can smoothly and accurately splice the sub-blocks and inhibit the propagation of global errors.

Description

Technical Field

[0001] This invention relates to the field of temporal InSAR deformation monitoring technology, and in particular to a large-area PSInSAR processing method, apparatus and equipment based on a block strategy. Background Technology

[0002] Temporal InSAR technology, developed based on InSAR, is a technique that utilizes multiple SAR images to achieve high-precision, high-density monitoring of land surface deformation. It suppresses the effects of temporal decorrelation and atmospheric delay by extracting temporally coherent targets. Permanent scatterer InSAR (PSInSAR) is one of the most classic methods. Within a scattering cell, there is a dominant ground scatterer, such as man-made structures or exposed rock, which maintains phase stability over a long period; this is called a permanent scatterer. Therefore, PSInSAR is suitable for monitoring land subsidence and infrastructure deformation in areas with many man-made features, especially in urban areas. With the launch of new high-resolution SAR satellites (such as TerraSAR-X, Fucheng-1, and LuTan-1), we can acquire SAR data with higher spatial resolution, shorter revisit periods, and wider coverage areas. How to utilize PSInSAR technology to process these massive amounts of SAR data for rapid, accurate, and detailed monitoring of land surface deformation is a problem that urgently needs to be solved.

[0003] For large-area PSInSAR deformation monitoring, the impact of hardware configuration, computing speed, result density, and result accuracy needs to be considered. Layered or block-based methods are commonly used for large-area PSInSAR processing. The layered method constructs a two-layer network, employing a "control first, then fragmentation" approach. The first layer uses stricter point selection thresholds, redundant network construction methods, and parameter calculation thresholds for processing, while the second layer densifies the observation points. The drawback is the need to comprehensively consider the global connectivity of the first layer network and the accuracy of the parameter calculation results. If the first layer network lacks global connectivity or has low calculation accuracy, it will affect the final coverage, density, and calculation accuracy of the deformation monitoring points. The block-based method involves PSInSAR processing and stitching of blocks first. Considering the complex distribution characteristics of actual ground features, complex block segmentation based on the distribution characteristics of PS points is required, which significantly impacts automated processing and accurate and effective result stitching. Furthermore, these methods perform block segmentation after point selection, without considering the memory consumption and processing efficiency during PS point selection.

[0004] Currently, layered or block-based methods are commonly used for large-area PSInSAR processing. However, these methods lack comprehensive consideration of memory usage, PSInSAR processing efficiency, and the accuracy and density of results, which will affect the overall results of PSInSAR processing. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the prior art and provide a large-area PSInSAR processing method, apparatus and equipment based on a block strategy, which provides an effective solution to the problems of excessive memory usage, low processing efficiency, low density and coverage of results, and suppression of result errors in large-area PSInSAR processing.

[0006] To achieve the above objectives, the technical solution of the present invention is as follows:

[0007] In a first aspect, the present invention provides a large-area PSInSAR processing method based on a block-based strategy, comprising:

[0008] Based on the number of rows and columns and the sub-block parameters of the original data, the data is automatically divided into several sub-blocks using a regular grid partitioning method with overlap.

[0009] Each sub-block undergoes PS point selection, sub-network construction, and sub-parameter calculation.

[0010] By utilizing the overlap information and connectivity between sub-blocks, the parameters between sub-blocks are adjusted to achieve smooth and accurate splicing of the results.

[0011] Optionally, the method for dividing the regular grid with overlap is as follows:

[0012] Set the size and overlap of the sub-blocks according to the configuration of the computer used, calculate the step size for moving each sub-block, move within the matrix of the original data size by step size, determine the position and boundary range of each block, and adjust the size of the sub-blocks in the last row and last column to fit the boundary of the matrix, ensuring that all sub-blocks are divided according to the specified size and overlap requirements, and that the last sub-block can fit the actual size of the matrix.

[0013] Optionally, asynchronous parallel computing, multi-threaded processing, or distributed systems can be used to perform block PS point selection, block network construction, and block parameter calculation for each sub-block.

[0014] Optionally, the selection of the block PS points includes:

[0015] In each sub-block, the intensity value sequence of each pixel is read sequentially, relative radiometric correction is performed, and the amplitude deviation value is calculated. The coherence value sequence of each pixel is read sequentially to calculate the average coherence. The PS point of the sub-block is selected using the amplitude deviation threshold and the average coherence threshold.

[0016] Take the union of the PS points of the overlapping regions and retain all PS points of the overlapping regions. Update the PS points of each sub-block.

[0017] Optionally, the segmented network includes:

[0018] Determine the number of PS points in the sub-block. If it is less than the specified number, skip the sub-block. Otherwise, read the position information, phase sequence, time and space baseline, incident angle and slant range information of the PS point set in the sub-block, construct a local Delaunay triangulation, and delete arc segments with excessive distance using a distance threshold. Perform arc segment parameter calculation and retain arc segments with a distance greater than the overall coherence threshold. Remove isolated arc segments and points to generate the maximum connected network.

[0019] Optionally, the formula for calculating the overall coherence γ is as follows:

[0020]

[0021] Where, ΔΦ i Phase difference between the two endpoints of the arc segment It is the vertical baseline, T i Time baseline, wavelength λ, slant range R, incident angle θ, total number of images N, and elevation residual increment and deformation rate increment at the two endpoints of the arc segment, respectively.

[0022] Optionally, the block parameter calculation includes:

[0023] Based on the arc segment parameter solution results and the maximum connected network in the sub-block, the point with the smallest amplitude deviation value in the sub-block is selected as the reference point. The elevation residual and deformation rate parameter of the PS point are calculated and nonlinear deformation decomposition is performed using the weighted least squares adjustment method.

[0024] Optionally, the adjustment of parameters between sub-blocks using inter-block overlap information and connectivity includes:

[0025] Based on the block-based strategy, find the corresponding points in the overlapping areas of each pair of sub-blocks in turn. If the number of corresponding points is greater than a specified threshold, calculate the difference between the deformation rate and the elevation residual of each corresponding point.

[0026] Outliers are removed using three times the mean square error. The average difference between the deformation rate and elevation residual of the remaining corresponding points after outlier removal is calculated and used as the adjustment value between sub-blocks.

[0027] Based on the calculated adjustment values ​​between sub-blocks and the corresponding sub-block relationships, establish coefficient matrix B. m×n and observation matrix L m×1 m represents the number of adjustment values, n represents the total number of sub-blocks, and the coefficient matrix B m×n There are only three elements: 1, -1, and 0. 1 and -1 represent the positions of the two sub-blocks of the overlapping block with adjustment values; L m×1 This indicates the adjustment value for the elevation residual or deformation rate of the corresponding overlapping block;

[0028] Find and delete columns in the coefficient matrix where all elements are 0, to obtain the updated coefficient matrix B. m×n

[0029] For the updated coefficient matrix B m×n The connectivity relationships of overlapping sub-blocks are found using connected component analysis.

[0030] Based on the connectivity relationship, from the coefficient matrix B m×n and observation matrix L m×1 Extract the coefficient matrix B of each connected component subset mi×ni and observation matrix L mi×1 , where mi and ni represent the total number of overlapping sub-blocks and the total number of sub-blocks in the i-th connected component;

[0031] Least square adjustment is performed in each connected component subset, where the block with the most overlapping sub-blocks with other sub-blocks is selected as the reference block with zero adjustment value, and the other sub-blocks are adjusted relatively.

[0032] Secondly, the present invention provides a large-area PSInSAR processing device based on a block-based strategy, comprising:

[0033] The data block partitioning module is used to automatically partition the data into several sub-blocks based on the number of rows and columns of the original data and the sub-block parameters using a regular grid partitioning method with overlap.

[0034] The block processing module is used to perform block PS point selection, block mesh construction, and block parameter calculation for each sub-block.

[0035] The block splicing module is used to adjust the parameters between sub-blocks by utilizing the overlap information and connectivity between sub-blocks, so as to achieve smooth and accurate splicing of the results.

[0036] Thirdly, the present invention provides an electronic device, including a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus;

[0037] Memory, used to store computer programs;

[0038] When the processor executes the program stored in the memory, it implements the steps of the large-area PSInSAR processing method based on the block strategy as described in any one of claims 1-8.

[0039] Compared with the prior art, the advantages of this invention are as follows:

[0040] The large-area PSInSAR processing method based on a block-based strategy provided in this embodiment first divides the data into blocks and then processes them in blocks. This reduces memory usage and improves parallel processing efficiency in the point selection, network construction, and parameter calculation stages of PSInSAR processing. It can increase the number of PS points and redundant arc segments in the network construction under current hardware conditions without reducing computational efficiency, which is beneficial for improving the accuracy, density, and coverage area of ​​the results. Information from corresponding points considering sub-block overlap can smoothly and accurately stitch sub-blocks together, suppressing the propagation of global errors. The entire method has a clear workflow structure and achieves full automation. Attached Figure Description

[0041] Figure 1 This is an implementation flowchart of the large-area PSInSAR processing method based on a block-based strategy provided in this application embodiment.

[0042] Figure 2 This is a diagram showing the result of data block partitioning for the instance data.

[0043] Figure 3 This is a diagram showing the result of selecting points in the block data of the instance.

[0044] Figure 4 This is a graph showing the deformation rate and elevation residuals of the instance data after block processing.

[0045] Figure 5 This is a diagram showing the results of the sub-block connectivity relationships and the corrections for the deformation rate and elevation residuals of each sub-block.

[0046] Figure 6 This is a comparison chart of the deformation rate results of instance data splicing and traditional methods.

[0047] Figure 7 This is a comparison chart of the elevation residual results obtained by stitching together example data and using traditional methods.

[0048] Figure 8 It is a scatter plot of the deformation rate of instance data splicing and traditional methods.

[0049] Figure 9 It is a scatter plot of the elevation residuals from the example data splicing and the traditional method.

[0050] Figure 10 This is a schematic diagram of the composition of a large-area PSInSAR processing device based on a block-based strategy provided in an embodiment of this application.

[0051] Figure 11 This is a schematic diagram illustrating the composition of an electronic device provided in an embodiment of this application. Detailed Implementation

[0052] Example:

[0053] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0054] See Figure 1 As shown, this embodiment provides a large-area PSInSAR processing method based on a block-based strategy, which mainly includes the following steps:

[0055] 110. Based on the number of rows and columns and the sub-block parameters of the original data, the data is automatically divided into several sub-blocks using a regular grid partitioning method with overlap. This step is the data block partitioning step.

[0056] 120. Perform block PS point selection, block mesh construction, and block parameter calculation for each sub-block; this step is the block processing step.

[0057] In this step, to further improve efficiency, asynchronous parallel computing, multi-threaded processing, or distributed systems can be used to perform block PS point selection, block network construction, and block parameter calculation for each sub-block.

[0058] 130. Adjust the parameters between sub-blocks using the overlap information and connectivity relationship between sub-blocks to achieve smooth and accurate splicing of the results; this step is the block splicing step.

[0059] Therefore, the large-area PSInSAR processing method based on a block-based strategy provided in this embodiment first divides the data into blocks and then processes them in blocks. This reduces memory usage and improves parallel processing efficiency in the point selection, network construction, and parameter calculation stages of PSInSAR processing. It can increase the number of PS points and redundant arc segments in the network construction under current hardware conditions without reducing computational efficiency, which is beneficial for improving the accuracy, density, and coverage area of ​​the results. Information from corresponding points considering sub-block overlap can smoothly and accurately stitch sub-blocks together, suppressing the propagation of global errors. The entire method has a clear workflow structure and achieves full automation.

[0060] In one specific embodiment, step 110 includes:

[0061] First, prepare the necessary files (i.e., raw data), obtain SAR data and external DEM of the study area, use SAR processing software to generate a single-view complex image set, select one image as the main image, and perform registration and interferometric processing on the other images to generate the files required for PSInSAR: spatiotemporal baseline file, differential interferometric atlas with reference ellipsoid and DEM removed, coherence atlas, intensity atlas, slant range and incident angle files.

[0062] Then, based on the computer configuration and the row and column data of the original data, the overlap is set, and regular grid blocks are formed. The size and overlap of the sub-blocks are set according to the actual configuration of the computer. The step size for each block movement is calculated as step = block_size - overlap, representing the number of rows or columns moved in each block movement. By moving within the matrix of the original data size using the step size, the position and boundary range (rows and columns) of each block are determined. Furthermore, the block size is adjusted at the last row and last column to fit the matrix boundaries, ensuring that all blocks are divided according to the specified size and overlap requirements, and that the last block fits the actual size of the matrix, preventing omissions or exceeding the matrix's range due to block size limitations.

[0063] In one specific embodiment, the selection of block PS points includes:

[0064] The intensity value sequence of each pixel in the sub-block is read sequentially, relative radiometric correction is performed and amplitude deviation is calculated. The coherence value sequence of each pixel is read sequentially to calculate the average coherence. The PS point of the sub-block is selected using the amplitude deviation threshold and the average coherence threshold.

[0065] The PS points of each sub-block are updated by taking the union of the PS points in the overlapping areas of the sub-blocks. Since the relative radiation correction is performed on each sub-block separately, the selection results in different overlapping areas of the sub-blocks are not completely the same when using a uniform amplitude deviation selection threshold. After the block selection is completed, the PS points of the overlapping areas are taken into union to retain all the PS points in the overlapping areas and the PS points of each sub-block are updated.

[0066] In one specific embodiment, the segmented network construction includes:

[0067] The process involves determining the number of PS points in a sub-block. If the number is less than a specified number, the sub-block is skipped. Otherwise, the position information, phase sequence, temporal and spatial baselines, incident angle, and slant range information of the PS point set are read from the sub-block. A local Delaunay triangulation is constructed, and arc segments with excessively large distances are removed using a distance threshold. Arc segment parameters are calculated, retaining arc segments with distances greater than the overall coherence threshold. Isolated arc segments and points are eliminated to generate the maximum connected network. The formula for calculating the overall coherence γ is as follows:

[0068]

[0069] Where, ΔΦ i Phase difference between the two endpoints of the arc segment It is the vertical baseline, T i Time baseline, wavelength λ, slant range R, incident angle θ, total number of images N, and elevation residual increments and deformation rate increments at the two endpoints of the arc segment, respectively.

[0070] In one specific embodiment, the block parameter calculation includes:

[0071] Based on the arc segment parameter calculation results and the maximum connected network in the sub-block, the point with the smallest amplitude deviation threshold in the sub-block is selected as the reference point. The elevation residual, deformation rate parameter and nonlinear deformation decomposition of the PS point are calculated using the weighted least squares adjustment method.

[0072] In one specific embodiment, step 130 includes:

[0073] Based on the initial block division strategy, corresponding points in the overlapping areas of each pair of sub-blocks are sequentially searched. If the number of corresponding points exceeds a specified threshold, the difference between the deformation rate and the elevation residual for each corresponding point is calculated.

[0074] Δvv k =v i -v j

[0075] Δhh k =h i -h j

[0076] Where i and j represent the index numbers of the two sub-blocks, Δvv k and Δhh k This represents the difference between the deformation rate and the elevation residual of the kth corresponding point in two sub-blocks.

[0077] Using a three-fold standard error to eliminate outliers, the following calculation formula takes the calculation of the deformation rate inter-block adjustment value as an example; the calculation method for the elevation residual inter-block adjustment value is the same:

[0078] Calculate the average value

[0079]

[0080] Calculate the difference between the observed values ​​and the mean:

[0081]

[0082] Obtain the mean square error:

[0083]

[0084] Determine if the limit is exceeded:

[0085]

[0086] Calculate the average of the differences in deformation rates of the remaining corresponding points after removing outliers, and use this as the adjustment value between sub-blocks.

[0087] Based on the calculated adjustment values ​​between sub-blocks, establish coefficient matrix B. m×n and observation matrix L m×1m represents the number of adjustment values, n represents the total number of sub-blocks, and the coefficient matrix B m×n There are only three elements: 1, -1, and 0. 1 and -1 represent the positions of the two child blocks of the overlapping block with adjustment values. L m×1 This indicates the adjustment value for the elevation residual or deformation rate of the corresponding overlapping block.

[0088] Find and delete columns in the coefficient matrix where all elements are 0. These columns represent sub-blocks without PS points and independent sub-blocks without overlapping areas, and do not need to be spliced.

[0089] For the updated B m×n Connected Component Analysis (CCA) is used to find the connectivity relationships between overlapping sub-blocks. CCA is a graph theory method used to identify interconnected subsets of a graph. It transforms the non-zero values ​​of the coefficient matrix into edges by constructing an adjacency matrix, thus treating columns as nodes and non-zero values ​​in rows as connections between nodes.

[0090] Based on the connectivity, from matrix B m×n and observation matrix L m×1 Extract the coefficient matrix B of each connected component subset mi×ni and observation matrix L mi×1 Where mi and ni represent the total number of overlapping sub-blocks and the total number of sub-blocks in the i-th connected component; least squares adjustment is performed in each connected component subset, and the block with the most overlapping sub-blocks with other sub-blocks is selected as the reference block (the sub-block with zero deformation and elevation residual adjustment values), which is the coefficient matrix B. mi×ni The column with the most non-zero elements is used to make relative adjustments to other sub-blocks.

[0091] X = (B T PB) -1 B T PL

[0092] Where B is the coefficient matrix, L is the observation matrix, and P is the weight matrix. The identity matrix can be used as the weight, or the ratio of the number of corresponding points in overlapping sub-blocks to the total number of corresponding points in all overlapping sub-blocks can be used as the weight.

[0093] The following example application scenario will further illustrate the method of the present invention:

[0094] (1) Data block partitioning

[0095] In this example, the SAR data consists of 23 TerraSAR-X strip pattern (SM) images with a resolution of 3m. The external DEM used is CopDEM GLO-30, and the data, after cropping, registration, and interferometry, has a row and column count of 13500*11500. Figure 2 As shown, using a regular grid partitioning method with overlap, the regular grid is divided into 3000*3000 sub-blocks with an overlap of 500, resulting in 20 sub-blocks. The rightmost sub-block has 4000 columns and the bottommost sub-block has 3500 rows, adapting to the boundaries of the original data.

[0096] (2) Block processing

[0097] First, based on the data block division method described above, the intensity and coherence values ​​of each pixel in each sub-block are read in. Relative radiometric correction is performed on the intensity values, and the amplitude deviation and average coherence values ​​are calculated. Pixels with an amplitude deviation threshold less than 0.35 and an average coherence greater than 0.6 are selected as PS points. Finally, the union of the selected points in each sub-block is taken, as shown below. Figure 3 As shown, the total number of PS points in the sub-block is 4.5 million, of which 1.1 million are overlapping PS points, and the number of PS points after merging is 3.4 million.

[0098] Within each block, the location information, deformation sequence, incident angle, and slant distance of each candidate point are read. A local Delaunay triangulation network is constructed, and arc segments with distances greater than a threshold are removed, generally assuming that the atmospheric influence of two points within a 1-kilometer radius is similar. The solution space search method is used to calculate parameters such as the elevation residual increment and deformation rate increment at the two endpoints of the arc segment when the overall coherence is maximized. The formula for calculating the overall coherence γ is:

[0099]

[0100] Where, ΔΦ i Phase difference between the two endpoints of the arc segment It is the vertical baseline, T i Time baseline, wavelength λ, slant range R, incident angle θ, total number of images N, and elevation residual increment and deformation rate increment at the two endpoints of the arc segment, respectively.

[0101] Arc segments with an overall coherence less than 0.75 are deleted. The connectivity of the remaining arc segments is calculated, and isolated arc segments and subnetworks are removed, retaining only the globally maximally connected network. The point with the smallest amplitude deviation in the network is selected as the reference point for zero deformation and zero elevation residual. The deformation rate and elevation residual value of each PS point are calculated using the weighted least squares adjustment method. Subsequent iterative processing of the results can be performed to separate nonlinear deformation and perform two-layer network construction, etc. Figure 4As shown, a) is the result of deformation rate after block solution, and b) is the result of elevation residual after block solution. There are obvious boundaries between different sub-blocks.

[0102] (3) Segmented splicing

[0103] To unify the results across different sub-blocks, a stitching process is required. The deformation rate and elevation residual information of corresponding points in the overlapping areas between sub-blocks are used for overall adjustment. First, corresponding points are found between different sub-blocks. If the number of corresponding points is less than a specified number, the two sub-blocks are considered disconnected; otherwise, the difference in deformation rate and elevation residual for corresponding points in different sub-blocks is calculated. Corresponding points with a value greater than three times the mean square error are removed. The average of the remaining differences in deformation rate and elevation residual for corresponding points is calculated as the overall adjustment value for deformation rate and elevation residual between sub-blocks. Because there are multiple overlapping sub-blocks, the final adjustment amount for each sub-block is calculated using the least squares adjustment method and redundant observation information from the overlapping sub-blocks.

[0104] Because the distribution of PS points within sub-blocks can be highly uneven—a sub-block may not contain any PS points, or a sub-block may contain PS points but overlapping regions may not have corresponding points—it is necessary to determine the connectivity of sub-blocks containing PS points. Connected Component Analysis (CCA) is used to find the connectivity relationships between overlapping sub-blocks. CCA is a graph theory method used to identify interconnected subsets of a graph. It transforms the non-zero values ​​of the coefficient matrix into edges by constructing an adjacency matrix, thus treating columns as nodes and non-zero values ​​in rows as connections between nodes. Least squares adjustment is then performed in each connected component to obtain the adjustment number for each sub-block within that component. For example... Figure 5 As shown, a) represents the relationship between connected components. There are two connected components. ① contains 19 sub-blocks, and ② contains one sub-block. There are no sub-blocks without a PS point. b) represents the overall adjustment of the elevation residuals for each sub-block, and c) represents the overall adjustment of the deformation rate for each sub-block.

[0105] Finally, the results obtained from the block processing are adjusted using the adjustment amount of each sub-block. For example... Figure 6 As shown, a) is the final deformation rate obtained using this method. b) is the result of processing the entire image using the traditional method. The results of both are quite consistent. Figure 7As shown, a) is the final elevation residual obtained using this method, and b) is the result of processing the entire image using the traditional method. The results of both are quite consistent. In a) there are 3.4 million PS points, and in b) there are 700,000 PS points. Depending on the workstation configuration used, the traditional method requires sparse sampling with a regular grid after point selection due to the large number of PS points before subsequent PSInSAR processing. The method in this paper does not require consideration of excessive memory usage. Furthermore, in b) the upper left and lower right corners of the image are not covered by PS points due to connectivity issues. This highlights the advantage of this method. Figure 8 As shown, a scatter plot of the deformation rate of the example data splicing and the traditional method is presented. Figure 9 As shown, a scatter plot of the elevation residuals from the example data stitching method and the traditional method is presented. The results of the two methods are quite consistent, with a correlation of over 0.95.

[0106] In summary, compared with existing technologies, the invention has the following technical advantages:

[0107] (1) A block-segmentation strategy suitable for large-scale PSInSAR processing is proposed. It does not require the influence of other prior information such as the distribution characteristics of PS points. It sets the sub-block size and overlap parameters according to the configuration of its own hardware, especially the computer memory size, and performs the operation of first dividing into blocks, then reading and processing, and finally stitching.

[0108] (2) A PSInSAR processing method based on the aforementioned block strategy is proposed, including methods for point selection, network construction and parameter calculation. The processing of each sub-block can also be performed asynchronously in parallel, in multi-threaded processing or in a distributed system to improve efficiency.

[0109] (3) A method for stitching sub-block PSInSAR processing results based on the aforementioned block strategy is proposed. The method stitches the results according to the information of the same points in the overlapping area and the connectivity of each sub-block, so as to ensure the consistency and integrity of the overall results.

[0110] Accordingly, such as Figure 10 As shown, this embodiment also provides a large-area PSInSAR processing device based on a block-based strategy, including:

[0111] The data block partitioning module is used to automatically partition the data into several sub-blocks based on the number of rows and columns of the original data and the sub-block parameters using a regular grid partitioning method with overlap.

[0112] The block processing module is used to perform block PS point selection, block mesh construction, and block parameter calculation for each sub-block.

[0113] The block splicing module is used to adjust the parameters between sub-blocks by utilizing the overlap information and connectivity between sub-blocks, so as to achieve smooth and accurate splicing of the results.

[0114] It should be noted that the large-area PSInSAR processing device based on the block strategy provided in the embodiments of this application can execute the large-area PSInSAR processing method based on the block strategy provided in any embodiment of this application, and has the corresponding functions and beneficial effects of the method.

[0115] like Figure 11 As shown in the embodiments of this application, an electronic device is also provided, including a processor 111, a communication interface 112, a memory 113, and a communication bus 114. The processor 111, communication interface 112, and memory 113 communicate with each other via the communication bus 114. The memory 113 stores computer programs. When the processor 111 executes the program stored in the memory 113, it implements the steps of the large-area PSInSAR processing method based on a block-based strategy provided in any of the aforementioned method embodiments. For example, the steps of the large-area PSInSAR processing method based on a block-based strategy may include the following steps: automatically dividing the data into several sub-blocks using a regular grid partitioning method with overlap, based on the number of rows and columns and sub-block parameters of the original data; performing block PS point selection, block mesh construction, and block parameter calculation processing on each sub-block; and adjusting the parameters between sub-blocks using the overlap information and connectivity relationships between sub-blocks to achieve smooth and accurate stitching of the results.

[0116] This application also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the large-area PSInSAR processing method based on a block-based strategy as provided in any of the foregoing method embodiments.

[0117] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0118] The above description is merely a specific embodiment of this application, enabling those skilled in the art to understand or implement this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the claims herein.

[0119] The above embodiments are merely illustrative of the technical concept and features of the present invention, and are intended to enable those skilled in the art to understand the content of the present invention and implement it accordingly. They should not be construed as limiting the scope of protection of the present invention. All equivalent changes or modifications made based on the essence of the content of the present invention should be covered within the scope of protection of the present invention.

Claims

1. A large-area PSInSAR processing method based on a block-based strategy, characterized in that, include: Based on the number of rows and columns and the sub-block parameters of the original data, the data is automatically divided into several sub-blocks using a regular grid partitioning method with overlap. Asynchronous parallel computing, multi-threaded processing, or distributed systems are used to perform block PS point selection, block network construction, and block parameter calculation for each sub-block. By utilizing the overlap information and connectivity between sub-blocks, the parameters between sub-blocks are adjusted to achieve smooth and accurate splicing of the results; The selection of block PS points includes: In each sub-block, the intensity value sequence of each pixel is read sequentially, relative radiometric correction is performed, and the amplitude deviation value is calculated. The coherence value sequence of each pixel is read sequentially to calculate the average coherence. The PS point of the sub-block is selected using the amplitude deviation threshold and the average coherence threshold. Take the union of the PS points of the overlapping region and retain all PS points of the overlapping region; then update the PS points of each sub-block. The adjustment of parameters between sub-blocks using inter-block overlap information and connectivity includes: Based on the block-based strategy, find the corresponding points in the overlapping areas of each pair of sub-blocks in turn. If the number of corresponding points is greater than a specified threshold, calculate the difference between the deformation rate and the elevation residual of each corresponding point. Outliers are removed using three times the mean square error. The average difference between the deformation rate and elevation residual of the remaining corresponding points after outlier removal is calculated and used as the adjustment value between sub-blocks. Based on the calculated adjustment values ​​between sub-blocks and the corresponding sub-block relationships, a coefficient matrix is ​​established. and observation matrix m represents the number of adjustment values, and n represents the total number of sub-blocks, where the coefficient matrix... There are only three elements: 1, -1, and 0. 1 and -1 represent the positions of the two sub-blocks of the overlapping block with adjustment values. This indicates the adjustment value for the elevation residual or deformation rate of the corresponding overlapping block; Find and delete columns in the coefficient matrix where all elements are 0, to obtain the updated coefficient matrix. For the updated coefficient matrix The connectivity relationships of overlapping sub-blocks are found using connected component analysis. Based on the connectivity, from the coefficient matrix and observation matrix Extract the coefficient matrix of each connected component subset and observation matrix ,in and Indicates the first The total number of overlapping sub-blocks and the total number of sub-blocks in each connected component; Least square adjustment is performed in each connected component subset, where the block with the most overlapping sub-blocks with other sub-blocks is selected as the reference block with zero adjustment value, and the other sub-blocks are adjusted relatively.

2. The large-area PSInSAR processing method based on a block-based strategy as described in claim 1, characterized in that, The method for dividing a regular grid with overlap is as follows: Set the size and overlap of the sub-blocks according to the configuration of the computer used, and calculate the step size for moving each sub-block; By moving the sub-blocks within the original data size matrix with a step size, the position and boundary range of each block are determined, and the size of the sub-blocks is adjusted at the last row and last column to fit the boundaries of the matrix, ensuring that all sub-blocks are divided according to the specified size and overlap requirements, and that the last sub-block fits the true size of the matrix.

3. The large-area PSInSAR processing method based on a block-based strategy as described in claim 1, characterized in that, The segmented network structure includes: Determine the number of PS points in the sub-block. If it is less than the specified number, skip the sub-block. Otherwise, read the position information, phase sequence, time and space baseline, incident angle and slant range information of the PS point set in the sub-block, construct a local Delaunay triangulation, and use the distance threshold to delete arc segments with excessive distance. Perform arc segment parameter calculation and retain arc segments that are greater than the overall coherence threshold; Remove isolated arcs and points to generate the maximum connected network.

4. The large-area PSInSAR processing method based on a block-based strategy as described in claim 3, characterized in that, The overall coherence Calculation formula: in, Phase difference between the two endpoints of the arc segment It is the vertical baseline. Time baseline, wavelength, Slope distance Angle of incidence It is the total number of images. and It is the elevation residual increment and deformation rate increment of the two endpoints of the arc segment that are being substituted.

5. The large-area PSInSAR processing method based on a block-based strategy as described in claim 3, characterized in that, The block parameter calculation includes: Based on the arc segment parameter solution results and the maximum connected network in the sub-block, the point with the smallest amplitude deviation value in the sub-block is selected as the reference point. The elevation residual and deformation rate parameter of the PS point are calculated and nonlinear deformation decomposition is performed using the weighted least squares adjustment method.

6. A large-area PSInSAR processing device based on a block-based strategy, characterized in that, include: The data block partitioning module is used to automatically partition the data into several sub-blocks based on the number of rows and columns of the original data and the sub-block parameters using a regular grid partitioning method with overlap. The block processing module is used to perform block PS point selection, block network construction, and block parameter calculation for each sub-block using asynchronous parallel computing, multi-threaded processing, or a distributed system. The block splicing module is used to adjust the parameters between sub-blocks by utilizing the overlap information and connectivity between sub-blocks, so as to achieve smooth and accurate splicing of the results. The selection of block PS points includes: In each sub-block, the intensity value sequence of each pixel is read sequentially, relative radiometric correction is performed, and the amplitude deviation value is calculated. The coherence value sequence of each pixel is read sequentially to calculate the average coherence. The PS point of the sub-block is selected using the amplitude deviation threshold and the average coherence threshold. Take the union of the PS points of the overlapping region and retain all PS points of the overlapping region; then update the PS points of each sub-block. The adjustment of parameters between sub-blocks using inter-block overlap information and connectivity includes: Based on the block-based strategy, find the corresponding points in the overlapping areas of each pair of sub-blocks in turn. If the number of corresponding points is greater than a specified threshold, calculate the difference between the deformation rate and the elevation residual of each corresponding point. Outliers are removed using three times the mean square error. The average difference between the deformation rate and elevation residual of the remaining corresponding points after outlier removal is calculated and used as the adjustment value between sub-blocks. Based on the calculated adjustment values ​​between sub-blocks and the corresponding sub-block relationships, a coefficient matrix is ​​established. and observation matrix m represents the number of adjustment values, and n represents the total number of sub-blocks, where the coefficient matrix... There are only three elements: 1, -1, and 0. 1 and -1 represent the positions of the two sub-blocks of the overlapping block with adjustment values. This indicates the adjustment value for the elevation residual or deformation rate of the corresponding overlapping block; Find and delete columns in the coefficient matrix where all elements are 0, to obtain the updated coefficient matrix. For the updated coefficient matrix The connectivity relationships of overlapping sub-blocks are found using connected component analysis. Based on the connectivity, from the coefficient matrix and observation matrix Extract the coefficient matrix of each connected component subset and observation matrix ,in and Indicates the first The total number of overlapping sub-blocks and the total number of sub-blocks in each connected component; Least square adjustment is performed in each connected component subset, where the block with the most overlapping sub-blocks with other sub-blocks is selected as the reference block with zero adjustment value, and the other sub-blocks are adjusted relatively.

7. An electronic device, characterized in that, It includes a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; Memory, used to store computer programs; When a processor executes a program stored in memory, it implements the steps of the large-area PSInSAR processing method based on a block-based strategy as described in any one of claims 1-6.

Images