A block unwrapping method for large-scale interferograms between earthquakes and related equipment

Abstract

The present invention provides a block unwrapping method for a large-scale interseismic interference pattern and related equipment, which divides the interseismic interference pattern of a study area into several sub-blocks with overlapping areas; inputs each sub-block into a trained neural network for integer ambiguity prediction to obtain the integer ambiguity and confidence of each pixel point in each sub-block, takes the average of the confidence of all pixels in each sub-block to obtain the confidence of each sub-block, and re-unwraps the corresponding sub-block or corrects the integer ambiguity of each pixel point in the corresponding sub-block according to the size of the confidence of each sub-block to obtain the updated integer ambiguity of each pixel point in each sub-block; splices the updated integer ambiguity of each pixel point in each sub-block to obtain the integer ambiguity map of the interseismic interference pattern, and obtains the unwrapping interference pattern based on the integer ambiguity map and the interseismic interference pattern, thereby greatly improving the unwrapping accuracy of the isolated island area.

Description

Technical Field

[0001] The present invention relates to the field of geodetic measurement technology, and in particular to a block unwrapping method for a large-scale interferogram between earthquakes and related equipment. Background Art

[0002] Spaceborne Synthetic Aperture Radar (InSAR) offers advantages such as wide coverage, high spatial resolution, and high-precision surface deformation monitoring. It is currently widely used for monitoring large-scale surface deformation along large fault zones during the interseismic phase. InSAR can characterize the details of strain rate fields over large areas, reveal blind faults, assess regional earthquake risk, analyze kinematic parameters along fault zones, and explore the dynamic drivers of fault zones, playing a vital role in revealing the mechanisms of crustal deformation. The satellite's wide spatial coverage and short revisit period provide a rich data source for acquiring large-scale interseismic deformation, but this also presents new challenges for InSAR data processing.

[0003] Interseismic interferograms are composed of complete image frames collected by one or two Sentinel satellites along their orbits. Because they primarily cover wilderness areas with complex landforms such as deserts, vegetation, and snow, and their long time span, they are susceptible to various incoherence factors and are prone to the presence of isolated islands. Localized islands surrounded by incoherent noise are highly susceptible to local unwrapping errors. Incoherent noise also causes unwrapping errors to propagate within the interferogram, resulting in large-scale triangulation loop errors. Therefore, low-coherence regions within large-scale interseismic interferograms are prone to unwrapping errors and their propagation, affecting the accuracy of subsequent time series solutions and inversion of fault interseismic motion parameters.

[0004] Existing technical research has categorized large-scale interseismic interferogram unwrapping into two main categories: spatial unwrapping and spatiotemporal unwrapping. For spatial unwrapping, the Minimum Cost Flow (MCF) method is currently the most commonly used robust unwrapping method. This method requires a coherence threshold mask to minimize the impact of incoherence noise, but this threshold is difficult to determine and has limited effectiveness. For spatiotemporal unwrapping, related research typically incorporates temporal information as a constraint to guide or coordinate unwrapping with the spatial dimension. Because interseismic deformation interferograms exhibit the characteristics of large scale, small gradients, multiple incoherences, and dramatic atmospheric variations, researchers have considered the triangular closure characteristics of the temporal dimension of multi-temporal interferograms and proposed an iterative weighted unwrapping method, building on traditional three-dimensional unwrapping methods, to gradually reduce large-scale unwrapping errors.

[0005] In addition to improving the unwrapping method itself, correcting unwrapping errors is also a key research focus for improving the accuracy of interseismic data processing. Unwrapping error correction can be categorized into two main categories: spatial correction and temporal correction. Spatial correction relies on interconnectedness and is not applicable to isolated regions. Temporal correction relies on redundant interferometric observations in the temporal dimension, making it difficult to implement for low-coherence points, and its effectiveness is limited by the inherent computational method.

[0006] Therefore, for large-scale, multi-phase interseismic deformation interferograms, introducing time dimension information for unwrapping and error correction may face problems such as inefficiency and ineffectiveness; and the reliance of the spatial dimension on the traditional phase continuity assumption is the root cause of the unwrapping errors in the decoherence zone. Therefore, the unwrapping accuracy of the above two existing unwrapping methods on interseismic interferograms is low. Summary of the Invention

[0007] The present invention provides a block unwrapping method for a large-scale interseismic interference pattern and related equipment, the purpose of which is to improve the unwrapping accuracy of a large-scale low-coherence interseismic interference pattern.

[0008] In order to achieve the above object, the present invention provides a block unwrapping method for large-scale interferograms between earthquakes, comprising:

[0009] Step 1: Obtain the interseismic interferogram of the study area and divide the interseismic interferogram into several sub-blocks with overlapping areas;

[0010] Step 2: For each sub-block with overlapping areas, input the sub-block into the trained neural network for integer ambiguity prediction to obtain the integer ambiguity of each pixel in each sub-block and the confidence of each pixel in each sub-block;

[0011] Step 3: Average the confidence values of all pixels in each sub-block to obtain the confidence value of each sub-block. Based on the confidence value of each sub-block, re-unwrap the corresponding sub-block or correct the integer ambiguity of each pixel in the corresponding sub-block according to the confidence value to obtain the updated integer ambiguity of each pixel in each sub-block.

[0012] Step 4: splice the updated integer ambiguity of each pixel point in each sub-block to obtain the integer ambiguity map of the interseismic interferogram, and obtain the unwrapped interferogram based on the integer ambiguity map and the interseismic interferogram.

[0013] Specifically, training a neural network involves:

[0014] Obtain several interferograms processed by the Lics platform as real samples;

[0015] Generate simulated samples and sample labels based on several real coherence maps and digital elevation models of the study area;

[0016] The neural network is trained using real samples and simulated samples to obtain a trained neural network.

[0017] Furthermore, simulated samples and sample labels are generated based on several real coherence maps and digital elevation data of the study area, including:

[0018] The incoherent noise is simulated based on several real coherence maps of the study area;

[0019] Based on the regional digital elevation model and mathematical model, the vertically stratified atmosphere, turbulent atmosphere, orbital error and interseismic deformation are simulated. The turbulent atmosphere, orbital error, interseismic deformation and vertically stratified atmosphere are added together to obtain the unwrapped true value of the simulated interseismic interferogram.

[0020] The decoherent noise is added to the unwrapped true value and then entangled and filtered to obtain the simulated interferogram and the coherence corresponding to the simulated interferogram. The coherence corresponding to the simulated interferogram is used to mask the integer ambiguity of the low-coherence pixels in the simulated interferogram to 0;

[0021] Subtract the unwrapped true value from the simulated interferogram to obtain the true value of the integer ambiguity of each pixel in the simulated interferogram;

[0022] The simulated interferogram and the true value of the integer ambiguity of each pixel in the simulated interferogram are divided into blocks to obtain simulated samples.

[0023] Furthermore, the confidence of each pixel in each sub-block is the difference between the maximum probability and the second largest probability predicted by the trained neural network for each pixel in each sub-block;

[0024] The confidence of each sub-pixel includes the average confidence of each pixel and the average confidence of non-decoherent pixels.

[0025] Furthermore, the confidence of each sub-block is obtained by averaging the confidence of all pixels in each sub-block, including:

[0026] For each sub-block, the confidence of all pixels in the sub-block is averaged to obtain the average pixel confidence of each sub-block;

[0027] For each sub-block, the confidence of all non-decoherent pixels in the sub-block is averaged to obtain the average confidence of non-decoherent pixels in each sub-block;

[0028] The confidence of each sub-block is composed of the average pixel confidence of each sub-block and the average non-decoherent pixel confidence of each sub-block.

[0029] Furthermore, for each sub-block's confidence, the corresponding sub-block is re-unwrapped or the integer ambiguity of the corresponding sub-block is corrected according to the confidence level, including:

[0030] For each sub-block, when the average pixel credibility in the confidence of the sub-block is less than a first preset threshold or the average non-decoherent pixel credibility is less than a second preset threshold, the sub-block is re-unwrapped using the MCF algorithm;

[0031] When the average pixel reliability in the confidence of the sub-block is greater than or equal to the first preset threshold and less than the third preset threshold, or the average non-decoherent pixel reliability is greater than or equal to the second preset threshold and less than the third preset threshold, the trained DeeplabV3+ model is used to correct the integer ambiguity of the sub-block.

[0032] More specifically, step 4 includes:

[0033] For each sub-block, the integer ambiguity of each pixel in the overlapping area of the sub-block is subtracted from the integer ambiguity of each pixel in the overlapping area of the adjacent sub-block to obtain the relative difference value of each pixel in the overlapping area of the two adjacent sub-blocks. The mode of the relative difference values of each pixel in the overlapping area of the two adjacent sub-blocks is used as the relative difference observation value of the overlapping area.

[0034] The overlapping area is screened based on the proportion of incoherent pixels in the overlapping area and the proportion of pixels corresponding to the mode of the relative difference value, and the relative difference observations of the corresponding overlapping area are retained or discarded according to the screening results;

[0035] The observation equation is established based on the retained relative difference observation values, and the observation equation is solved using the weighted least squares method to obtain the relative difference value of the integer ambiguity of each sub-block relative to the first sub-block.

[0036] Correcting the integer ambiguity of each pixel in each sub-block according to the relative difference value of the integer ambiguity of each sub-block relative to the integer ambiguity of the first sub-block to obtain a corrected integer ambiguity;

[0037] The corrected integer ambiguities of all pixels in the overlapping area of each sub-block are averaged to obtain the average integer ambiguity. The integer ambiguities of the interseismic interferogram are then concatenated based on the corrected integer ambiguities and the average integer ambiguity to obtain the integer ambiguity map of the interseismic interferogram.

[0038] By formula , add the integer ambiguity map of the interseismic interferogram to the interseismic interferogram to obtain the unwrapped interferogram, where represents the unwrapped phase of the pixel, represents the interference phase of the pixel point, Indicates the integer ambiguity of the pixel.

[0039] Furthermore, the observation equation is expressed as:

[0040] ;

[0041] in, represents the relative difference observation value, , Indicates the The sub-block and The relative difference observation value of the overlapping area on the sub-blocks, represents the coefficient matrix, , Indicates the relative difference value of the integer ambiguity of each sub-block relative to the first sub-block. , The interseismic interference pattern includes Sub-blocks, Indicates the The relative difference value of the integer ambiguity of the sub-block as a whole relative to the first sub-block as a whole.

[0042] The present invention also provides a terminal device, including a memory, a processor, and a computer program stored in the memory and runnable on the processor. When the processor executes the computer program, a block unwrapping method for large-scale interseismic interference patterns is implemented.

[0043] The present invention also provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, a method for block unwrapping of large-scale interferograms between earthquakes is implemented.

[0044] The above solution of the present invention has the following beneficial effects:

[0045] The present invention divides the obtained interseismic interference map of the study area into several sub-blocks with overlapping areas; for each sub-block with overlapping areas, the sub-block is input into the trained neural network for integer ambiguity prediction to obtain the integer ambiguity of each pixel in each sub-block and the confidence of each pixel in each sub-block. The confidence of all pixels in each sub-block is averaged to obtain the confidence of each sub-block, and for each sub-block, the corresponding sub-block is re-unwrapped or the integer ambiguity of each pixel in the corresponding sub-block is corrected according to the size of the confidence to obtain the updated integer ambiguity of each pixel in each sub-block; the updated integer ambiguity of each pixel in each sub-block is spliced to obtain the integer ambiguity map of the interseismic interference map, and the unwrapped interference map is obtained based on the integer ambiguity map and the interseismic interference map; compared with the prior art, the present invention obtains the integer ambiguity map of the interseismic interference map after training. The trained neural network predicts the integer ambiguity, and re-unwraps the corresponding sub-block according to the confidence level or corrects the integer ambiguity of each pixel in the corresponding sub-block to obtain the updated integer ambiguity of each pixel in each sub-block. It can consider the global information of each sub-block and accurately predict the integer ambiguity of the island area. It can also automatically identify incoherent and low-coherence areas through the trained neural network, which improves the processing efficiency while reducing the negative impact of incoherence noise on traditional spatial dimensional unwrapping, greatly improving the unwrapping accuracy of the island area.

[0046] Other beneficial effects of the present invention will be described in detail in the subsequent specific implementation section. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 A schematic diagram of a flow chart of an embodiment of the present invention;

[0048] Figure 2 A structural diagram of a neural network in an embodiment of the present invention;

[0049] Figure 3 A flowchart of simulation sample generation in an embodiment of the present invention;

[0050] Figure 4 The figure shows the comparison of the unwrapping accuracy between the traditional method and the proposed method on 25 simulated interference patterns.

[0051] Figure 5 This is a statistical comparison of the unwrapping accuracy of the traditional method and this method on the real interferogram; Figure 5 (a) is a schematic diagram of the normalized frequency distribution of closed-loop UEP under two untangling methods. Figure 5 (b) Schematic diagram of the specific performance of this method in each closed loop compared with the traditional method;

[0052] Figure 6 Schematic diagram of the structure of the terminal device in an embodiment of the present invention. DETAILED DESCRIPTION

[0053] To make the technical problems, technical solutions, and advantages to be solved by the present invention more clear, the following is a detailed description with reference to the accompanying drawings and specific embodiments. It is obvious that the embodiments described are only some of the embodiments of the present invention, not all of them. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0054] In the description of the present invention, it should be noted that the terms "first", "second" and "third" are only used for descriptive purposes and should not be understood as indicating or implying relative importance.

[0055] In the description of the present invention, it should be noted that, unless otherwise specified or limited, the term "connected" should be understood in a broad sense. For example, it can mean directly connected or indirectly connected through an intermediary. Those skilled in the art will understand the specific meaning of the above terms in the present invention in specific circumstances.

[0056] In addition, the technical features involved in the different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0057] In view of the existing problems, the present invention provides a block unwrapping method for large-scale interseismic interference patterns and related equipment.

[0058] like Figure 1 As shown, an embodiment of the present invention provides a block unwrapping method for a large-scale interferogram between earthquakes, comprising:

[0059] Step 1: Obtain the interseismic interferogram of the study area and divide the interseismic interferogram into several sub-blocks with overlapping areas;

[0060] Step 2: For each sub-block with overlapping areas, input the sub-block into the trained neural network for integer ambiguity prediction to obtain the integer ambiguity of each pixel in each sub-block and the confidence of each pixel in each sub-block;

[0061] Step 3: Average the confidence values of all pixels in each sub-block to obtain the confidence value of each sub-block. Based on the confidence value of each sub-block, re-unwrap the corresponding sub-block or correct the integer ambiguity of each pixel in the corresponding sub-block according to the confidence value to obtain the updated integer ambiguity of each pixel in each sub-block.

[0062] Step 4: splice the updated integer ambiguity of each pixel point in each sub-block to obtain the integer ambiguity map of the interseismic interferogram, and obtain the unwrapped interferogram based on the integer ambiguity map and the interseismic interferogram.

[0063] The present invention uses the western end of a fault zone in the northern part of a certain plateau as the research area. The terrain at the western end of the fault zone is highly undulating, with elevation differences of more than 3,000 meters. The interferogram phase is dominated by interseismic deformation and atmospheric noise, and contains complex landforms such as deserts, plateau permafrost, and snow. The interferogram decoherence noise is complex and severe, and there are many isolated island areas caused by decoherence in the desert. Therefore, the method provided by the present invention is specifically described using this fault zone as the research area. The process is as follows:

[0064] Step 1: Obtain an interseismic interferogram of the western end of a fault zone in the northern part of a certain plateau. The data acquisition time span is from January 2019 to June 2022. Use Gamma software for interferometry processing, and only select interferograms with a time baseline greater than 300 days and a spatial baseline less than 20 m. 763 interseismic deformation interferograms with long-term baselines are obtained. The interseismic interferogram is segmented to obtain several sub-blocks with overlapping areas. The sub-block size is set to 2048 × 2048 pixels, and the corresponding moving step size is 1600 pixels. That is, the size of the overlapping area in the interseismic interferogram is 448 × 2048 pixels.

[0065] Step 2: For each sub-block with overlapping areas, input the sub-block into the trained neural network for integer ambiguity prediction to obtain the integer ambiguity of each pixel in each sub-block and the confidence of each pixel in each sub-block.

[0066] Specifically, the SegFormer (Simple and Efficient Design for Semantic Segmentation with Transformers) neural network based on the Transformer architecture combines the advantages of the Transformer and convolutional networks, and has good robustness in semantic segmentation. Its network structure is as follows: Figure 2 Shown, including:

[0067] An encoder consisting of an embedding module, a first transformer module, a second transformer module, a third transformer module, and a fourth transformer module connected in sequence;

[0068] A decoder composed of a first multi-layer perception module and a second multi-layer perception module connected in sequence;

[0069] The output end of the first transformer module, the output end of the second transformer module, the output end of the third transformer module, and the output end of the fourth transformer module are all connected to the input end of the first multi-layer perception module;

[0070] The first transformer module, the second transformer module, the third transformer module, and the fourth transformer module each include an efficient attention layer, a hybrid feedforward network layer, and an overlapping block fusion layer connected in sequence. The input end of the efficient attention layer is the input end of the corresponding transformer module, and the output end of the overlapping block fusion layer is the output end of the corresponding transformer module.

[0071] The first multi-layer perception module and the second multi-layer perception module both include a multi-layer perceptron and an upsampling layer connected in sequence, wherein the input end of the multi-layer perceptron is the input end of the corresponding multi-layer perception module, and the output end of the upsampling layer is the output end of the corresponding multi-layer perception module.

[0072] In the encoder, the neural network in the embodiment of the present invention adopts a new overlapping patch positionless encoding, thereby avoiding the interpolation of positional encoding, increasing the flexibility of the model and its adaptability to inputs of various resolutions; introducing a mixed feedforward network layer (Mix-FFN), which can simultaneously capture local and global features; in the decoder, a lightweight multi-layer perceptron (MLP) is used to complete decoding, aggregating information from different layers, thereby combining local attention and global attention to present powerful representations.

[0073] Specifically, training the neural network includes:

[0074] Obtain several interferograms processed by the Lics platform as real samples;

[0075] Generate simulated samples and sample labels based on several real coherence maps and digital elevation models of the study area;

[0076] The neural network is trained using real samples and simulated samples to obtain a trained neural network.

[0077] To fully train the neural network, the present invention uses simulation generation and LICS (Looking into the Continents from Space) platform processing to obtain nearly 40,000 sets of data samples with varying degrees of incoherent noise, covering a rich range of pattern features such as deformation, atmospheric delay, and orbital error. Taking into account limited video memory, the spatial size of all training samples is downsampled to 512×512 pixels.

[0078] Specifically, simulated samples and sample labels are generated based on several real coherence maps and digital elevation data, including:

[0079] The incoherent noise is simulated based on several real coherence maps of the study area;

[0080] Based on the regional digital elevation model and mathematical model, the vertically stratified atmosphere, turbulent atmosphere, orbital error and interseismic deformation are simulated. The turbulent atmosphere, orbital error, interseismic deformation and vertically stratified atmosphere are added together to obtain the unwrapped true value of the simulated interseismic interferogram.

[0081] The decoherent noise is added to the unwrapped true value and then entangled and filtered to obtain the simulated interferogram and the coherence corresponding to the simulated interferogram. The coherence corresponding to the simulated interferogram is used to mask the integer ambiguity of the low-coherence pixels in the simulated interferogram to 0;

[0082] Subtract the unwrapped true value from the simulated interferogram to obtain the true value of the integer ambiguity of each pixel in the simulated interferogram;

[0083] The simulated interferogram and the true value of the integer ambiguity of each pixel in the simulated interferogram are divided into blocks to obtain simulated samples.

[0084] In this embodiment of the present invention, the ascending orbit data of the western end of a fault zone in the northern part of a certain plateau were first visually selected on the LICS platform to obtain interferograms with complex patterns without unwrapping errors. Manual corrections were then performed on some samples to obtain 13,500 sets of real samples.

[0085] Then as Figure 3 As shown, the embodiment of the present invention simulates decoherent noise based on several real coherence maps at the western end of a fault zone in the northern part of a certain plateau, simulates a vertically stratified atmosphere based on a regional digital elevation model, simulates turbulent atmosphere, orbit error, and interseismic deformation based on mathematical models such as fractal functions, quadratic polynomials, and screw dislocations, and adds the turbulent atmosphere, orbit error, interseismic deformation, and vertically stratified atmosphere to obtain the unwrapped true value of the simulated interseismic interferogram.

[0086] Next, the decoherent noise is added to the unwrapped true value and then entangled and filtered to obtain the simulated interferogram and the coherence corresponding to the simulated interferogram. The coherence corresponding to the simulated interferogram is used to mask the integer ambiguity of the low-coherence pixels in the simulated interferogram to 0.

[0087] Then, the unwrapped true value and the simulated interferogram are subtracted to obtain the true value of the integer ambiguity of each pixel in the simulated interferogram (1~6);

[0088] Finally, the simulated interferograms and the true integer ambiguity values of each pixel in the simulated interferograms were divided into blocks to obtain multiple sub-block interferogram samples and their corresponding integer ambiguity labels (0-6). After manual screening of the multiple sub-block interferogram samples, more than 26,000 groups of simulated samples were obtained.

[0089] Specifically, the confidence of each pixel in each sub-block is the difference between the maximum probability and the second largest probability predicted by the trained SegFormer network for each pixel in each sub-block;

[0090] The confidence of each sub-block includes the average confidence of each pixel and the average confidence of non-decoherent pixels.

[0091] Considering that the integer ambiguity predicted by the neural network has a certain degree of error, which is directly reflected in the unwrapped phase and affects subsequent splicing, the embodiment of the present invention calculates the confidence level of the class prediction probability output for each pixel. The confidence level is the difference between the maximum probability predicted by the neural network at each pixel and the second-highest probability. The greater the difference between the two, the more accurate the neural network's prediction and the greater the credibility of the prediction result.

[0092] Specifically, the confidence of each sub-block is obtained by averaging the confidence of all pixels in each sub-block, including:

[0093] For each sub-block, the confidence of all pixels in the sub-block is averaged to obtain the average pixel confidence of each sub-block. ;

[0094] For each sub-block, the confidence of all non-decoherent pixels in the sub-block is averaged to obtain the average confidence of non-decoherent pixels in each sub-block. ;

[0095] The average pixel credibility of each sub-block and the average credibility of non-decoherent pixels in each sub-block The confidence level of each sub-block .

[0096] Specifically, for each sub-block's confidence, the corresponding sub-block is re-unwrapped or the integer ambiguity of the corresponding sub-block is corrected according to the confidence level, including:

[0097] For each sub-block, when the average pixel credibility in the confidence of the sub-block is less than a first preset threshold or the average non-decoherent pixel credibility is less than a second preset threshold, the sub-block is re-unwrapped using the MCF algorithm;

[0098] When the average pixel reliability in the confidence of the sub-block is greater than or equal to the first preset threshold and less than the third preset threshold, or the average non-decoherent pixel reliability is greater than or equal to the second preset threshold and less than the third preset threshold, the trained DeeplabV3+ model is used to correct the integer ambiguity of the sub-block.

[0099] In an embodiment of the present invention, the first preset threshold is set to 0.8, the second preset threshold is set to 0.65, and the third preset threshold is set to 0.95. When the average pixel reliability in the confidence of the sub-block is less than 0.8 or the average reliability of the non-decoherent pixels is less than 0.65, it indicates that the sub-block may have a large-scale prediction error, and the MCF algorithm needs to be used to re-untangle the sub-block; when the average pixel reliability in the confidence of the sub-block is greater than or equal to 0.8 and less than 0.95 or the average reliability of the non-decoherent pixels is greater than or equal to 0.65 and less than 0.95, it indicates that prediction errors may occur only in part of the area. At this time, the trained DeeplabV3+ model is used to correct the integer ambiguity of the sub-block.

[0100] It should be noted that the MCF algorithm is a conventional detangling method in the art, and the specific process of detangling using the algorithm is a well-known technology. Therefore, the embodiment of the present invention will not further describe the specific process of detangling using the algorithm.

[0101] Specifically, the trained DeeplabV3+ model is used to correct the integer ambiguity of the sub-block, including:

[0102] The phase of each sub-block is added to the integer ambiguity output by the neural network in the previous step according to the basic formula of InSAR phase unwrapping to obtain the initial unwrapping result of the sub-block, and the initial unwrapping result is used as the input of the trained DeeplabV3+ model. The pixels in the sub-block are then classified into three categories: 0, 1, and 2 by the trained DeeplabV3+ model. Categories 1 and 2 respectively indicate that there are errors in the integer ambiguity output by the neural network for the pixel, and 1 needs to be subtracted or added to it to achieve correction, while category 0 indicates that there is no error in the integer ambiguity output by the neural network for the pixel and no correction is required.

[0103] The DeeplabV3+ model in the embodiment of the present invention is also trained by simulating and processing nearly 40,000 sets of data samples with varying degrees of incoherent noise using the Lics (Looking into the Continents from Space) platform.

[0104] Specifically, step 4 includes:

[0105] For each sub-block, the integer ambiguity of each pixel in the overlapping area of the sub-block is subtracted from the integer ambiguity of each pixel in the overlapping area of the adjacent sub-block to obtain the relative difference value of each pixel in the overlapping area of the two adjacent sub-blocks. The mode of the relative difference values of each pixel in the overlapping area of the two adjacent sub-blocks is used as the relative difference observation value of the overlapping area.

[0106] The overlapping area is screened based on the proportion of incoherent pixels in the overlapping area and the proportion of pixels corresponding to the mode of the relative difference value, and the relative difference observations of the corresponding overlapping area are retained or discarded according to the screening results;

[0107] The observation equation is established based on the retained relative difference observation values, and the observation equation is solved using the weighted least squares method to obtain the relative difference value of the integer ambiguity of each sub-block relative to the first sub-block.

[0108] Correcting the integer ambiguity of each pixel in each sub-block according to the relative difference value of the integer ambiguity of each sub-block relative to the first sub-block to obtain a corrected integer ambiguity;

[0109] The corrected integer ambiguities of all pixels in the overlapping area of each sub-block are averaged to obtain the average integer ambiguity. The integer ambiguities of the interseismic interferogram are then concatenated based on the corrected integer ambiguities and the average integer ambiguity to obtain the integer ambiguity map of the interseismic interferogram.

[0110] By formula , add the integer ambiguity map of the interseismic interferogram to the interseismic interferogram to obtain the unwrapped interferogram, where represents the unwrapped phase of the pixel, represents the interference phase of the pixel point, Indicates the integer ambiguity of the pixel.

[0111] The embodiment of the present invention takes into account that the difference in the overlapping area is theoretically a constant value, and therefore takes the mode of the relative difference values of each pixel point in the overlapping area as the relative difference observation value of the overlapping area; for the overlapping area where the proportion of incoherent pixels exceeds 97% and the proportion of mode pixels is less than 80%, it is considered that there is a large unwrapping error in the overlapping area, and the relative difference observation value of the overlapping area is discarded.

[0112] Specifically, the observation equation is expressed as:

[0113] ;

[0114] in, represents the relative difference observation value, , Indicates the The sub-block and The relative difference observation value of the overlapping area on the sub-blocks, represents the coefficient matrix, , Indicates the relative difference value of the integer ambiguity of each sub-block relative to the first sub-block. , The interseismic interference pattern includes Sub-blocks, Indicates the The relative difference value of the integer ambiguity of the sub-block as a whole relative to the first sub-block as a whole.

[0115] In the embodiment of the present invention, The interference pattern of the sub-blocks needs to be solved parameters, without discarding the overlapping area Overlapping observations.

[0116] It should be noted that the relative difference value of the integer ambiguity of each sub-block relative to the first sub-block is taken into account. Theoretically, it is an integer, and the solution residual should theoretically be 0. For interference patterns with a solution residual greater than 0.00001, they should be discarded, otherwise large-scale stitching errors will occur.

[0117] The unwrapping results obtained by the embodiment of the present invention and the traditional MCF method were compared with the simulated true values. The accuracy was assessed using a custom metric: the Unwrapping Error Proportion (UEP). This metric is the ratio of the number of pixels with unwrapping errors to the number of pixels unmasked in the interferogram unwrapping. In this comparative experiment, pixels where the absolute value of the difference between the unwrapping result and the simulated true value reached 5.5 radians were considered to have unwrapping errors. The UEP values of the two unwrapping methods were calculated for all simulated interferograms. The average UEP for the traditional MCF method was 2%, while the average UEP for the method provided by the embodiment of the present invention was only 0.8%. Therefore, it can be concluded that the average accuracy of the method provided by the embodiment of the present invention on this simulated dataset is 60% higher than that of the traditional method. Figure 4 The UEP comparison of 25 simulated interferograms is shown. It can be seen that the UEP of most interferograms is significantly reduced based on the method provided by the embodiment of the present invention.

[0118] The embodiment of the present invention was applied to real data unwrapping. The unwrapping results obtained by the embodiment of the present invention were compared with those obtained by the traditional MCF method using triangular phase closure error detection to determine the unwrapping error. Based on the 739 retained interferograms, more than 1,300 triangular closed loops were constructed. The accuracy assessment indicator was also the unwrapping error proportion (UEP). In the accuracy assessment of this real data, pixels with an absolute value of 5.5 radians in the triangular closed loop were considered to have unwrapping errors. The UEP values of the two unwrapping methods on all triangular closed loops were calculated separately. The average UEP of the traditional MCF method was 1.26%, while the average UEP of the method provided by the embodiment of the present invention was only 0.77%. Therefore, it can be considered that the average accuracy of the method provided by the embodiment of the present invention on this real data set is 38% higher than that of the traditional method. Figure 5 (a) shows the normalized frequency distribution of the closed-loop UEP under the two unwrapping methods. It is obvious that the closed-loop UEP of the method provided by the embodiment of the present invention is concentrated in a smaller value. Figure 5 (b) further demonstrates the specific performance of the method provided by the embodiment of the present invention on various closed-loops compared to traditional methods. Approximately 56% of closed-loops experienced no significant change in UEP. While approximately 11.7% of closed-loops experienced a significant increase in UEP, over 30% of closed-loops experienced a significant decrease. Overall, the method provided by the embodiment of the present invention effectively reduces unwrapping errors in large-scale interferograms.

[0119] The embodiment of the present invention divides the obtained interseismic interference map of the study area into several sub-blocks with overlapping areas; for each sub-block with overlapping areas, the sub-block is input into the trained neural network to perform integer ambiguity prediction, and the integer ambiguity of each pixel point in each sub-block and the confidence of each pixel point in each sub-block are obtained; the confidence of all pixels in each sub-block is averaged to obtain the confidence of each sub-block; for each confidence of the sub-block, the corresponding sub-block is re-unwrapped or the integer ambiguity of each pixel point in the corresponding sub-block is corrected according to the size of the confidence, and the updated integer ambiguity of each pixel point in each sub-block is obtained; the updated integer ambiguity of each pixel point in each sub-block is spliced to obtain the earthquake The invention obtains the integer ambiguity map of the inter-seismic interference map, and obtains the unwrapped interference map based on the integer ambiguity map and the inter-seismic interference map; compared with the prior art, the embodiment of the present invention predicts the integer ambiguity through the trained neural network, re-unwraps the corresponding sub-blocks according to the size of the confidence level or corrects the integer ambiguity of each pixel point in the corresponding sub-block, and obtains the updated integer ambiguity of each pixel point in each sub-block, which can consider the global information of each sub-block, and can more accurately predict the integer ambiguity of the island area. It can also automatically identify the incoherent and low coherent areas through the trained neural network, which improves the processing efficiency while reducing the negative impact of the incoherence noise on the traditional spatial dimensional unwrapping, and greatly improves the unwrapping accuracy of the island area.

[0120] The embodiment of the present invention further provides a terminal device, such as Figure 6 As shown, the terminal device D10 of this embodiment includes: at least one processor D100 ( Figure 6 Only one processor is shown in the figure), a memory D101, and a computer program D102 stored in the memory D101 and executable on the at least one processor D100, wherein the processor D100 implements the above-mentioned inter-seismic large-scale interference pattern block unwrapping method when executing the computer program D102.

[0121] The terminal device D10 can be a computing device such as a desktop computer, a notebook, a PDA, a server, a server cluster, a cloud server, etc. The terminal device may include, but is not limited to, a processor D100 and a memory D101. It will be understood by those skilled in the art that Figure 6 This is merely an example of the terminal device D10 and does not constitute a limitation on the terminal device D10 . The terminal device D10 may include more or fewer components than shown in the figure, or a combination of certain components, or different components. For example, the terminal device D10 may also include input and output devices, network access devices, etc.

[0122] The processor D100 may be a central processing unit (CPU), or may be another general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. A general-purpose processor may be a microprocessor or any conventional processor.

[0123] In some embodiments, the memory D101 may be an internal storage unit of the terminal device D10, such as a hard disk or memory of the terminal device D10. In other embodiments, the memory D101 may also be an external storage device of the terminal device D10, such as a plug-in hard disk, a smart memory card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the terminal device D10. Furthermore, the memory D101 may include both an internal storage unit of the terminal device D10 and an external storage device. The memory D101 is used to store an operating system, application programs, a boot loader, data, and other programs, such as the program code of the computer program. The memory D101 may also be used to temporarily store data that has been output or is about to be output.

[0124] It should be noted that the information interaction, execution process, etc. between the above-mentioned devices / units are based on the same concept as the method embodiment of this application. Their specific functions and technical effects can be found in the method embodiment section and will not be repeated here.

[0125] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0126] An embodiment of the present invention further provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, a method for block unwrapping of large-scale interferograms between earthquakes is implemented.

[0127] If the integrated unit is implemented as a software functional unit and sold or used as a standalone product, it can be stored in a computer-readable storage medium. Based on this understanding, the present application implements all or part of the process steps in the above-mentioned method embodiments by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When executed by a processor, the computer program can implement the steps of each of the above-mentioned method embodiments. The computer program includes computer program code, which can be in source code form, object code form, executable file, or some intermediate form. The computer-readable medium can include at least: any entity or device capable of carrying the computer program code to a construction device / terminal device, a recording medium, computer memory, read-only memory (ROM), random access memory (RAM), an electrical carrier signal, a telecommunications signal, and a software distribution medium. Examples include a USB flash drive, a removable hard drive, a magnetic disk, or an optical disk.

[0128] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as within the scope of protection of the present invention.

Claims

1. A block unwrapping method for large-scale interferograms between earthquakes, characterized by: include: Step 1: Obtain an interseismic interferogram of the study area and divide the interseismic interferogram into a number of sub-blocks with overlapping areas; Step 2: For each sub-block with overlapping areas, input the sub-block into the trained neural network to perform integer ambiguity prediction, and obtain the integer ambiguity of each pixel in each sub-block and the confidence of each pixel in each sub-block; Step 3: Average the confidences of all pixels in each sub-block to obtain the confidence of each sub-block. Based on the confidence of each sub-block, re-unwrap the corresponding sub-block or correct the integer ambiguity of each pixel in the corresponding sub-block according to the confidence, to obtain the updated integer ambiguity of each pixel in each sub-block. Step 4, splicing the updated integer ambiguity of each pixel point in each sub-block to obtain the integer ambiguity map of the inter-seismic interference map, and obtaining the unwrapped interference map based on the integer ambiguity map and the inter-seismic interference map.

2. The interseismic large-scale interferogram block unwrapping method according to claim 1, characterized in that: Training the neural network, including: Obtain several interferograms processed by the Lics platform as real samples; Generate simulated samples and sample labels based on several real coherence maps and digital elevation models of the study area; The neural network is trained using the real samples and the simulated samples to obtain a trained neural network.

3. The interseismic large-scale interferogram block unwrapping method according to claim 2, characterized in that: Generate simulated samples and sample labels based on several real coherence maps and digital elevation data of the study area, including: The incoherent noise is simulated based on several real coherence maps of the study area; Simulating a vertically stratified atmosphere, a turbulent atmosphere, orbital errors, and interseismic deformation based on a regional digital elevation model and a mathematical model, and adding the turbulent atmosphere, the orbital errors, the interseismic deformation, and the vertically stratified atmosphere to obtain an unwrapped true value of a simulated interseismic interferogram; Adding the decoherent noise to the unwrapped true value and then performing entanglement and filtering to obtain a simulated interferogram and coherence corresponding to the simulated interferogram, and using the coherence corresponding to the simulated interferogram to mask the integer ambiguity of low-coherence pixels in the simulated interferogram to 0; Subtracting the unwrapped true value from the simulated interferogram to obtain the true value of the integer ambiguity of each pixel in the simulated interferogram; The simulated interferogram and the integer ambiguity true value of each pixel in the simulated interferogram are divided into blocks to obtain simulated samples.

4. The interseismic large-scale interferogram block unwrapping method according to claim 2, characterized in that: The confidence of each pixel in each sub-block is the difference between the maximum probability and the second largest probability predicted by the trained neural network for each pixel in each sub-block; The confidence of each sub-block includes the average confidence of each pixel and the average confidence of non-decoherent pixels.

5. The interseismic large-scale interferogram block unwrapping method according to claim 4, characterized in that: The confidence of each sub-block is obtained by averaging the confidence of all pixels in each sub-block, including: For each sub-block, the confidence of all pixels in the sub-block is averaged to obtain the average pixel confidence of each sub-block; For each sub-block, the confidence scores of all non-decoherent pixels in the sub-block are averaged to obtain the average confidence score of the non-decoherent pixels in each sub-block; The confidence of each sub-block is composed of the average pixel confidence of each sub-block and the average non-decoherent pixel confidence of each sub-block.

6. The interseismic large-scale interferogram block unwrapping method according to claim 5, characterized in that: For each sub-block, re-unwrapping the corresponding sub-block or correcting the integer ambiguity of the corresponding sub-block according to the confidence level of the sub-block includes: For each sub-block, when the average pixel credibility in the confidence of the sub-block is less than a first preset threshold or the average non-decoherent pixel credibility is less than a second preset threshold, re-unwrapping the sub-block using the MCF algorithm; When the average pixel credibility in the confidence of the sub-block is greater than or equal to the first preset threshold and less than the third preset threshold, or the average non-decoherent pixel credibility is greater than or equal to the second preset threshold and less than the third preset threshold, the trained DeeplabV3+ model is used to correct the integer ambiguity of the sub-block.

7. The interseismic large-scale interferogram block unwrapping method according to claim 6, characterized in that: The step 4 comprises: For each sub-block, the integer ambiguity of each pixel in the overlapping area of the sub-block is subtracted from the integer ambiguity of each pixel in the overlapping area of the adjacent sub-block to obtain the relative difference value of each pixel in the overlapping area of the two adjacent sub-blocks, and the mode of the relative difference values of each pixel in the overlapping area of the two adjacent sub-blocks is used as the relative difference observation value of the overlapping area; The overlapping area is screened based on the proportion of incoherent pixels in the overlapping area and the proportion of pixels corresponding to the mode of the relative difference value, and the relative difference observations of the corresponding overlapping area are retained or discarded according to the screening results; An observation equation is established based on the retained relative difference observation value, and the observation equation is solved using the weighted least squares method to obtain the relative difference value of the integer ambiguity of each sub-block as a whole relative to the first sub-block as a whole; Correcting the integer ambiguity of each pixel in each sub-block according to the relative difference value of the integer ambiguity of each sub-block relative to the first sub-block to obtain a corrected integer ambiguity; averaging the corrected integer ambiguities of all pixels in the overlapping area of each sub-block to obtain an average integer ambiguity, and concatenating the integer ambiguities of the interseismic interferogram based on the corrected integer ambiguities and the average integer ambiguity to obtain an integer ambiguity map of the interseismic interferogram; By formula , the integer ambiguity map of the interseismic interferogram is added to the interseismic interferogram to obtain an unwrapped interferogram, where represents the unwrapped phase of the pixel, represents the interference phase of the pixel point, Indicates the integer ambiguity of the pixel.

8. The interseismic large-scale interferogram block unwrapping method according to claim 7, characterized in that: The expression of the observation equation is: in, represents the relative difference observation value, , Indicates the The sub-block and The relative difference observation value of the overlapping area on the sub-blocks, represents the coefficient matrix, , Indicates the relative difference value of the integer ambiguity of each sub-block relative to the first sub-block. , The interseismic interference pattern includes Sub-blocks, Indicates the The relative difference value of the integer ambiguity of the sub-block as a whole relative to the first sub-block as a whole.

9. A terminal device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the inter-seismic large-scale interference pattern block unwrapping method according to any one of claims 1 to 8 is implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the inter-seismic large-scale interference pattern block unwrapping method according to any one of claims 1 to 8 is implemented.

Images