A crust deformation monitoring method, electronic equipment and storage medium

Abstract

The application provides a crust deformation monitoring method, an electronic device and a storage medium, relates to the technical field of geological exploration, and obtains a crust observation data set of a target region, wherein the crust observation data set comprises at least two different types of crust observation data, the crust observation data set is input into a long short-term memory neural network model based on wavelet transform which is constructed in advance, a data fusion result is obtained, statistical analysis is performed on the data fusion result, crust anomaly information of the target region is obtained, and the crust deformation change of the target region is monitored. Since different types of crust observation data are effectively fused, the crust deformation is monitored through the data fusion result after fusion, the precision of crust deformation monitoring is improved, the three-dimensional monitoring capability of the crust deformation of the target region is expanded, the crust deformation change of a small region can be more accurately detected, and the improvement of the crust deformation monitoring precision is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of geological exploration, in particular to a crust deformation monitoring method, an electronic device and a storage medium. BACKGROUND

[0002] At present, the global navigation satellite system (GNSS) reference station network established globally provides a direct and effective observation means for crustal deformation monitoring. The crustal movement monitoring network can be divided into three scales of monitoring network according to the size of its coverage, i.e. global, regional and local. These GNSS reference stations have accumulated observation data for decades since their establishment. The coordinate time series obtained by using high-precision GNSS data software solution provides important basic data for the research in the fields of regional crustal deformation monitoring, global plate tectonic movement research, fault slip monitoring, earthquake deformation monitoring and the like.

[0003] The GNSS reference station coordinate time series contains long-term linear changes of the crust, seasonal periodic changes, earthquake displacement changes and changes affected by other factors. It is of great significance to study the regional crustal deformation law to complete the analysis of the GNSS reference station coordinate time series and find out the abnormal information therein. However, the feature analysis of the GNSS reference station coordinate time series mainly focuses on the noise model and seasonal changes of the time series. When analyzing the influence of some natural disasters on the time series, further research and analysis on non-seasonal periodic signals are still needed. However, the GNSS reference stations are distributed sparsely and it is difficult to form complete coverage of the region, thereby affecting the monitoring precision of the crustal deformation.

[0004] In summary, there is a certain demand for improving the monitoring precision of the crustal deformation monitoring at present. SUMMARY

[0005] The problem solved by the present application is how to solve the problem of low monitoring precision caused by single data.

[0006] To solve the above problems, the present application provides a crustal deformation monitoring method, an electronic device and a storage medium.

[0007] In a first aspect, the present application provides a crustal deformation monitoring method, comprising:

[0008] obtaining a crust observation data set of a target region, the crust observation data set comprising at least two different types of crust observation data, wherein at least one type of crust observation data is global navigation satellite system data;

[0009] input the crust observation data set into a pre-constructed long short-term memory neural network model based on wavelet transform to obtain a data fusion result output by the long short-term memory neural network model based on wavelet transform, the data fusion result being a result after fusion of at least two different types of crust observation data in the crust observation data set;

[0010] perform statistical analysis on the data fusion result to obtain crust abnormal information of the target area, the crust abnormal information being used to represent crust deformation change.

[0011] Optionally, inputting the crust observation data set into a pre-constructed long short-term memory neural network model based on wavelet transform to obtain a data fusion result output by the long short-term memory neural network model based on wavelet transform includes:

[0012] performing wavelet transform processing on at least two different types of crust observation data in the crust observation data set to obtain weight information of the crust observation data set;

[0013] performing feature extraction on at least two different types of crust observation data in the crust observation data set by using a convolutional neural network to obtain feature information of the crust observation data set;

[0014] input the weight information of the crust observation data set and the feature information of the crust observation data set into a long short-term memory neural network model to obtain a data fusion result output by the long short-term memory neural network model.

[0015] In a second aspect, the present application provides an electronic device, comprising a memory and a processor;

[0016] The memory is configured to store a computer program.

[0017] The processor is configured to implement the crust deformation monitoring method according to the first aspect when executing the computer program.

[0018] In a third aspect, the present application provides a computer readable storage medium, the storage medium storing a computer program, and when the computer program is executed by a processor, the crust deformation monitoring method according to the first aspect is implemented.

[0019] The crust deformation monitoring method, the electronic device and the storage medium have the following beneficial effects: a crust observation data set of a target region is obtained, the crust observation data set includes at least two types of crust observation data, at least one type of the crust observation data is global navigation satellite system data, the different types of multiple observation data are obtained, the crust observation data set is input into a long short-term memory neural network model based on wavelet transform which is constructed in advance, to obtain a data fusion result output by the long short-term memory neural network model based on wavelet transform, the obtained multi-source observation data is fused by the long short-term memory neural network model based on wavelet transform, statistical analysis is performed on the data fusion result, to obtain crust anomaly information of the target region; due to effective fusion of different types of crust observation data, the crust deformation is monitored by the fused data fusion result, the accuracy of the crust deformation monitoring is improved, the stereoscopic monitoring capability of the crust deformation of the target region is expanded, the crust deformation change of a small region can be more accurately detected, and the accuracy of the crust deformation monitoring is improved. BRIEF DESCRIPTION OF DRAWINGS

[0020] Figure 1 A flowchart of a crust deformation monitoring method according to an embodiment of the present application;

[0021] Figure 2 A flowchart of preprocessing and standardizing crust observation data;

[0022] Figure 3 A processing flowchart of inputting a crust observation data set into a long short-term memory neural network model based on wavelet transform which is constructed in advance;

[0023] Figure 4 A schematic diagram of wavelet transform processing and convolutional neural network processing of GNSS station displacement data, InSAR displacement data and remote sensing image data respectively;

[0024] Figure 5 A flowchart of wavelet transform processing of at least two types of crust observation data in a crust observation data set;

[0025] Figure 6 A structure schematic diagram of a gating convolutional layer;

[0026] Figure 7 A structure schematic diagram of a convolutional neural network;

[0027] Figure 8 A flowchart of feature extraction of at least two types of crust observation data in a crust observation data set by using a convolutional neural network;

[0028] Figure 9A processing flowchart for inputting weight information of a crust observation data set and feature information of the crust observation data set into a long short-term memory neural network model;

[0029] Figure 10 A structural schematic diagram of a cell structure in the long short-term memory neural network model;

[0030] Figure 11 A flowchart for statistical analysis of a data fusion result;

[0031] Figure 12 A structural schematic diagram of a crust deformation monitoring device according to an embodiment of the present application;

[0032] Figure 13 A structural schematic diagram of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0033] In order to make the above objectives, characteristics and advantages of the present application more obvious and easy to understand, the specific embodiments of the present application will be described in detail below with reference to the drawings. Although some embodiments of the present application are shown in the drawings, it should be understood that the present application can be implemented in various forms, and should not be interpreted as being limited to the embodiments described herein, on the contrary, these embodiments are provided to make the present application more thorough and complete. It should be understood that the drawings and embodiments of the present application are only for exemplary purposes, and are not intended to limit the scope of protection of the present application.

[0034] It should be understood that each step described in the method embodiments of the present application can be performed in different order and / or in parallel. In addition, the method embodiments can include additional steps and / or omit the steps shown. The scope of the present application is not limited in this respect.

[0035] The terms related to the present application are explained below.

[0036] GNSS (Global Navigation Satellite System) is a satellite system covering the whole world for autonomous spatial positioning, allowing small electronic receivers (stations) to determine their location (longitude, latitude and altitude) and transmit time signals with an accuracy of 10 meters along the line of sight via satellite broadcasts.

[0037] INSAR (Interferometric Synthetic Aperture Radar) is a remote sensing technology for ground deformation monitoring and topographic mapping using synthetic aperture radar images. Two or more synthetic aperture radar image maps are used to generate digital elevation models or ground deformation maps based on the phase difference of the echoes received by satellites or aircraft.

[0038] Remote sensing images are images of the earth's surface obtained by satellite, unmanned aerial vehicle, aircraft and other remote sensing platforms.

[0039] In the related art, when processing observation data for monitoring crust deformation, it is often limited by the singularity of the observation data and the complexity of the model, and it is difficult to accurately capture the subtle changes of the crust deformation. A variety of different types of observation data can provide more comprehensive and rich information, such as the fusion of a variety of different types of data such as GNSS, InSAR, etc. can reflect the crust deformation from different angles and levels, effectively improving the reliability and integrity of the data. When using a variety of different types of observation data, the plurality of data needs to be fused, and the traditional fusion method for fusing data will be limited by the influence of different resolutions between the data, greatly reducing the fusion performance and accuracy of the data.

[0040] To solve the problems in the above related art, the embodiment provides a crust deformation monitoring method, an electronic device and a storage medium.

[0041] As shown in Figure 1 The crust deformation monitoring method provided by the embodiment of the present application comprises:

[0042] S100, acquiring a crust observation data set of a target region, the crust observation data set comprising at least two different types of crust observation data, wherein at least one type of crust observation data is global navigation satellite system data.

[0043] Specifically, the different types of crust observation data are observation data obtained by different collection source methods, i.e. multi-source observation data, which can be, for example, global navigation satellite system data, interferometric synthetic aperture radar data and remote sensing image data. It should be noted that the crust observation data set of the embodiment of the present application at least includes global navigation satellite system data, and other observation data is not limited to the above-mentioned three types of data. If some of the data is missing in a certain area, the method of the embodiment of the present application can also be used for reconstruction calculation. However, in the case of less data, the accuracy will be reduced. In addition, other observation data can also be used to supplement the data amount, thereby improving the reconstruction accuracy. For example, the gravity field recovery to climate experiment gravity satellite data and global land data assimilation system data can be added. By jointly fusing these supplementary data with global navigation satellite system data, the monitoring accuracy and data integrity can be effectively improved, and the method has strong flexibility and expansibility. The problem of data loss can be solved by adding other data types, thereby realizing efficient and accurate crust deformation monitoring in a wider application scenario.

[0044] Specifically, the target region can be selected according to specific needs, and generally includes earthquake activity regions, volcanic activity regions, geological disaster prone regions, mining and oil and gas extraction regions, urban and infrastructure regions, plate boundary and active fault regions, and other special regions (such as high mountain and canyon regions, deserts, marshes, permafrost, etc.).

[0045] S200, input the crust observation data set into the long short-term memory neural network model based on wavelet transform which is constructed in advance to obtain a data fusion result output by the long short-term memory neural network model based on wavelet transform; wherein the data fusion result is the result after fusion of at least two different types of crust observation data in the crust observation data set.

[0046] Specifically, the long short-term memory neural network model based on wavelet transform can assign different weights to data according to different frequency domain dimensions, which helps to improve the accuracy of fused data, thereby enriching the observation data of crust deformation in the target area.

[0047] S300, statistically analyzing the data fusion result to obtain crust anomaly information of the target area, the crust anomaly information being used to represent the crust deformation change.

[0048] Specifically, the data fusion result is statistically analyzed, and the crust anomaly information in the target area is comprehensively counted to describe the deformation characteristics of the crust.

[0049] In this embodiment, a crust observation data set of a target area is obtained, the crust observation data set including at least two different types of crust observation data, at least one type of which is global navigation satellite system data. By obtaining multi-source observation data, the target area can be better covered. The crust observation data set is input into a long short-term memory neural network model based on wavelet transform which is constructed in advance to obtain a data fusion result output by the long short-term memory neural network model based on wavelet transform. The obtained multi-source observation data is fused by the long short-term memory neural network model based on wavelet transform. The data fusion result is statistically analyzed to obtain the crust anomaly information of the target area, so as to monitor the crust deformation change of the target area. Since the multi-source observation data is effectively fused, the crust deformation is monitored by the fused data fusion result, the precision of the crust deformation monitoring is improved, the stereoscopic monitoring capability of the target area deformation is expanded, the crust deformation change of a small area can be more accurately detected, and the precision of the crust deformation monitoring is improved.

[0050] Optionally, before the crust observation data set is input into the long short-term memory neural network model based on wavelet transform which is constructed in advance, the method further includes:

[0051] The at least two different types of crust observation data in the crust observation data set are respectively preprocessed and standardized to have consistency in the spatial coordinate system.

[0052] In some embodiments, the crust observation dataset includes: global navigation satellite system (GNSS) data, interferometric synthetic aperture radar data (InSAR image data), and remote sensing image data, as shown in Figure 2 Preprocessing and standardizing at least two different types of crust observation data in the crust observation dataset respectively includes:

[0053] S210, the GNSS data is resolved to obtain displacement data of the corresponding GNSS station in the target area.

[0054] Specifically, the displacement data is displacement data of a preset length time sequence, and the displacement data includes east direction component, north direction component and sky direction component. The preset length time sequence can be a natural day or other time length.

[0055] Specifically, the GNSS data station corresponds to the GNSS data. For example, for GNSS data A1 received by GNSS station A, the displacement data of GNSS station A can be obtained by resolving GNSS data A1.

[0056] Specifically, GNSS data processing software (such as GAMIT) can be used to resolve GNSS data. In an embodiment, TEQC software is used to check the quality of GNSS data and eliminate substandard station data. Then, the GAMIT open source data processing tool is used to resolve the GNSS data based on the satellite precise ephemeris file and other parameter files that need to be corrected, to obtain the long time sequence daily displacement data of the station in the east direction, the north direction and the sky direction. Due to weather or instrument itself, some stations may lack complete time sequence data, so GNSS missing data interpolation software is used to interpolate the missing data to obtain complete displacement data of the corresponding GNSS station in the preset length time sequence.

[0057] S220, at least two InSAR images in the InSAR image data are interfered to obtain InSAR displacement data of the target area.

[0058] Specifically, the InSAR image is radiometrically corrected to eliminate the radiation errors caused by sensor characteristics, atmospheric conditions and other factors, so that the radiation value of the image more accurately reflects the scattering characteristics of the ground object; the reference DEM data (a digital terrain model for representing the elevation of the earth's surface) is used to correct the terrain of the InSAR image, convert the InSAR image from the radar coordinate system to the geographic coordinate system, and eliminate the influence of the terrain undulation on the geometric shape of the InSAR image; the spatial resolution of the image is reduced and the signal-to-noise ratio of the scatterer is increased by processing the InSAR image in multiple views, thereby improving the quality of the interferogram; the time and spatial baselines of the multi-scene InSAR images covering different time periods are calculated, and the interferometric pairs with shorter time and spatial baselines are usually selected to extract deformation information; two or more pre-processed InSAR images are interfered to generate an interferogram, which is used to display the phase difference between different InSAR images; the coherence of the generated interferogram is calculated to obtain a coherence map, which reflects the correlation between different InSAR images, and the coherence reflects the interference quality of each pixel in the interferogram, and the interference phase in the high-coherence area is more stable and reliable; the coherence map is filtered to further enhance the signal in the high-coherence area and suppress the noise in the low-coherence area, thereby improving the overall quality of the interferogram; the interferogram usually contains a 2π periodic phase, and the least squares method is used for phase unwrapping to obtain the true displacement information of the ground surface; according to the formation of the subset of the free combination interferogram, the phase equation is formed by all the interferograms, and the least squares method is used to solve the phase equation to estimate the final InSAR displacement data of the target area.

[0059] S230, the remote sensing image data is processed to remove clouds and band fusion.

[0060] Specifically, the ENVI software is used to load the remote sensing image data, the Cloud Masking tool is selected, the exponential masking algorithm is used, the corresponding threshold parameters are set, and the remote sensing image data is processed to remove clouds. In the ENVI software, the ColorNormalized Sharpening tool is selected, the band names that need to be fused are selected, and the resampling method is set, so that the fused remote sensing image data is obtained.

[0061] S240, the displacement data of the GNSS site and the remote sensing image data are fused to obtain the displacement data of the GNSS site with unified spatial coordinates of the remote sensing image.

[0062] Specifically, according to the coordinate latitude and longitude information of the site, the displacement information of the site is added to the corresponding position of the remote sensing image, and is superimposed in the form of multi-band; then the bands of the remote sensing image are standardized, and the pixel values are set to data conforming to the normal distribution.

[0063] S250: interpolating the GNSS site displacement data to obtain the internal displacement distribution of the region.

[0064] In some embodiments, the inverse distance weighting method is used to interpolate the site displacement data. Specifically, according to the coordinate positions of the existing GNSS sites, each GNSS site is processed as follows: for each position to be interpolated, the distance between it and all known GNSS sites is calculated: wherein and denote the longitude and latitude of the position to be interpolated, respectively; and denote the longitude and latitude positions of the known samples, respectively, and the distance is converted to a weight: , p is an exponential parameter, in order to avoid division by zero, a minimum value is set for the distance, and the weight information is used to perform interpolation calculation for each grid: wherein N table denotes the number of known GNSS sites within the search radius, denotes the displacement value of the i th GNSS site, denotes element-wise multiplication. The above method is used to interpolate each grid in the target region to obtain the displacement data in the region.

[0065] S260: performing fusion processing on the InSAR displacement data and the remote sensing image data of the target region to obtain InSAR displacement data having unified spatial coordinates with the remote sensing image.

[0066] Specifically, similar to the fusion processing of the GNSS site displacement data and the remote sensing image data, the InSAR displacement data is added to the corresponding position of the remote sensing image to obtain InSAR displacement data having unified spatial coordinates with the remote sensing image.

[0067] In this optional embodiment, the different types of observation data are preprocessed and standardized to ensure that the different types of observation data are unified in space, so as to facilitate subsequent input into the neural network model for fusion processing.

[0068] Optionally, as shown in Figure 3 , the crust observation data set is input into the pre-constructed long short-term memory neural network model based on wavelet transform to obtain a data fusion result output by the long short-term memory neural network model based on wavelet transform, including:

[0069] S310: performing wavelet transform processing on at least two different types of crust observation data in the crust observation data set to obtain weight information of the crust observation data set.

[0070] S320, feature extraction is performed on at least two different types of crust observation data in the crust observation data set by using a convolutional neural network to obtain feature information of the crust observation data set.

[0071] S330, the weight information of the crust observation data set and the feature information of the crust observation data set are input into a long short-term memory neural network model to obtain a data fusion result output by the long short-term memory neural network model.

[0072] In some embodiments, the crust observation data set includes GNSS data, InSAR data and remote sensing image data, and after preprocessing and standardization processing, GNSS station displacement data, InSAR displacement data and remote sensing image data are obtained. As shown in Figure 4 The GNSS station displacement data, InSAR displacement data and remote sensing image data are respectively processed by wavelet transform and convolutional neural network to obtain weight information and feature information, and then the weight information and feature information are input into a long short-term memory neural network model (LSTM neural network model) for data fusion to obtain a data fusion result.

[0073] In this optional embodiment, compared with the traditional fusion method, the long short-term memory neural network model based on wavelet transform can give different weights to data according to different frequency domain dimensions, which helps to improve the accuracy of the data fusion result, thereby enriching the observation data of the crust deformation in the target area.

[0074] Optionally, as shown in Figure 5 The wavelet transform processing is performed on at least two different types of crust observation data in the crust observation data set to obtain weight information of the crust observation data set, which includes:

[0075] S510, wavelet analysis is performed on at least two different types of crust observation data in the crust observation data set to obtain high-frequency feature data and low-frequency feature data, the high-frequency feature data being used to represent short-term changes and noise of the crust observation data set, and the low-frequency feature data being used to represent long-term trends and low-frequency information of the crust observation data set.

[0076] Specifically, the multiple observation data are first decomposed by using wavelet analysis, and the data are usually decomposed into three layers by using wavelet analysis, and the core of the decomposition is to decompose the data into different scale approximation coefficients (low-frequency feature data) and detail coefficients (high-frequency feature data). The expression of data decomposition by using wavelet analysis is as follows:

[0077] ;

[0078] wherein, represents the decomposition coefficient of the i-th layer, j and h and gcoefficients representing a filter, j and n respectively represent the number of decomposition layers and node numbers, l representing a discrete index related to a time axis, k representing a position index in signal processing.

[0079] In this embodiment, different types of multiple observation data are decomposed into 3 layers, and the approximate coefficients (low-frequency feature data) of each layer mainly include long-term trends and low-frequency information of the information, and the detail coefficients (high-frequency feature data) mainly include short-term changes and noise of the information.

[0080] S520, the high-frequency feature data and the low-frequency feature data are input into the first double convolution layer to obtain a first high-frequency data feature map and a first low-frequency data feature map output by the first double convolution layer.

[0081] Specifically, the high-frequency feature data and the low-frequency feature data are input into the first double convolution layer. In an embodiment, the first double convolution layer is composed of two convolution kernels with a size of 3, the step is set to 1, the padding is set to 1, and the output channel is set to 32. Through the first double convolution layer, the size of the data can be kept unchanged, more feature data can be extracted, and the first high-frequency data feature map and the first low-frequency data feature map can be obtained.

[0082] S530, the first high-frequency data feature map and the first low-frequency data feature map are input into the maximum pooling layer to obtain a second high-frequency data feature map and a second low-frequency data feature map output by the maximum pooling layer.

[0083] Specifically, after the data is extracted by S520, its size needs to be changed to half of the original size. In an embodiment, the convolution kernel size of the maximum pooling layer is set to 2, the step is set to 2, and the padding is set to 0. Therefore, the size of the first high-frequency data feature map and the first low-frequency data feature map will be reduced by half after passing through the maximum pooling layer, so that the global related features can be extracted by the next convolution operation.

[0084] S540, the second high-frequency data feature map and the second low-frequency data feature map are input into the first gated convolution layer to obtain a third high-frequency data feature map and a third low-frequency data feature map output by the first gated convolution layer.

[0085] Specifically, the convolution kernel size of the first gated convolution layer is set to 3, the stride is set to 1, the padding is set to 1, and the output channel is set to 64. When performing gated convolution, similar to standard convolution, the output channel is first set to 128. After the first convolution of S520, the size of the data remains unchanged, but the output channel becomes four times the input channel. Then, through S530, the convolved data is evenly divided into two parts in the channel dimension, one part of which is input into the sigmoid activation function and then multiplied with the remaining half of the data to obtain the output result of the last first gated convolution layer, that is, the third high-frequency data feature map and the third low-frequency data feature map are obtained.

[0086] Specifically, the high-frequency feature data and low-frequency feature data after S510 wavelet analysis are respectively input into the encoding layer for feature extraction, wherein the first gated convolution layer is used to further extract key features. Figure 6 As shown, the gated convolution structure mainly includes a convolution layer 610 and a gating layer 620. The convolution expression of the convolution layer 610 is: A=E*W+b, B=E*V+c, which is an expression for double convolution of the input data, where A and B are the results of convolution, E represents the data input to the convolution layer 610, W and V represent the weights of the convolution, b and c represent the convolution bias, and the convolution layer 610 applies the filter to the tensor to generate the output tensor. The gated convolution modifies the ordinary convolution process by introducing a gating mechanism, so that the output is calculated as a combination of the convolution output and the gating signal. The processing expression of the gated layer 620 is: ,in Represents the result after gated convolution, represents element-wise multiplication, Represents the sigmoid activation function.

[0087] Convolutional layer 610 convolves the input high-frequency feature data with the low-frequency feature data to obtain twice the feature data of the output channel. The feature data is evenly divided into two parts, one of which is input into the sigmoid activation function, and the other is multiplied by the result of the input activation function to obtain the final result. Gated layer 620 is an advanced convolution operation that introduces a gating mechanism to enhance the model's ability to capture complex dependencies in the data. Compared with ordinary convolutional layers, gated convolution is particularly effective in capturing long-range dependencies in sequences and provides flexibility in processing information, enabling the model to adaptively learn which features are important to the current task, improving the robustness of the model. Using an encoding layer with gated convolution can effectively extract important feature data.

[0088] S550 , upsampling the third high-frequency data feature map and the third low-frequency data feature map and inputting the upsampled data into the second double convolutional layer to obtain a fourth high-frequency data feature map and a fourth low-frequency data feature map output by the second double convolutional layer.

[0089] Specifically, the third high-frequency data feature map and the third low-frequency data feature map are input into an up-sampling function, the size is set to be twice the current size, the sampling method is set to be nearest-neighbor interpolation, and the size is restored to the original size; then the feature map with the restored size is input into the second double convolution layer, the convolution kernel size of the second double convolution layer is set to be 3, the step size is set to be 1, the padding is set to be 1, and the output channel number is set to be 64, and the fourth high-frequency data feature map and the fourth low-frequency data feature map are obtained through the second double convolution layer.

[0090] S560, the fourth high-frequency data feature map is spliced with the first high-frequency data feature map to obtain a first high-frequency spliced feature map; the fourth low-frequency data feature map is spliced with the first low-frequency data feature map to obtain a first low-frequency spliced feature map; and the first high-frequency spliced feature map and the first low-frequency spliced feature map are input into the third double convolution layer to obtain a fifth high-frequency data feature map and a fifth low-frequency data feature map output by the third double convolution layer.

[0091] Specifically, since the fourth high-frequency data feature map and the first high-frequency data feature map, and the fourth low-frequency data feature map and the first low-frequency data feature map are inconsistent in the channel dimension, the embodiment merges the two feature maps with inconsistent channel dimensions to obtain a feature map with consistent channel dimensions in the dimension scale through splicing, and then inputs the spliced feature map into the third double convolution layer. In an embodiment, if the channel number of the spliced feature map is 96, the convolution kernel of the third double convolution layer is set to be 3, the step size is 1, the padding is 1, and the output channel is 48, and finally the fifth high-frequency data feature map and the fifth low-frequency data feature map are obtained.

[0092] S570, the fifth high-frequency data feature map and the fifth low-frequency data feature map are input into the second gating convolution layer to obtain a sixth high-frequency data feature map and a sixth low-frequency data feature map output by the second gating convolution layer; wherein the sixth high-frequency data feature map and the sixth low-frequency data feature map are used to represent the weight information of the crust observation data set.

[0093] Specifically, the convolution kernel size of the second gating convolution layer is set to be 3, the step size is 1, the padding is 1, the output channel is set to be 1, and the sixth high-frequency data feature map and the sixth low-frequency data feature map are obtained through the second gating convolution layer.

[0094] In the optional embodiment, the weight information of the different types of multiple data under the wavelet transform is obtained after the above convolution.

[0095] Optionally, as Figure 7As shown, the convolutional neural network comprises an encoding module, a standard convolutional layer, a decoding module and a jump connection, wherein the encoding module is composed of a wavelet down-sampling layer and a double convolutional layer (including a gated convolution), and the decoding module is composed of an inverse convolutional layer and a double convolutional layer.

[0096] Optionally, as shown, feature extraction is performed on at least two different types of crustal observation data in the crustal observation data set by using a convolutional neural network to obtain feature information of the crustal observation data set, which comprises: Figure 8

[0097] S810, input at least two different types of crustal observation data in the crustal observation data set into the fourth double convolutional layer to obtain the first fusion feature map output by the fourth double convolutional layer.

[0098] In an embodiment, the convolution kernel size of the fourth double convolutional layer is set to 3, the step is 1, the padding is 1, and the output channel number is 32, and the first fusion feature map can be obtained through the fourth double convolutional layer.

[0099] S820, input the first fusion feature map into the first wavelet down-sampling layer, and input the output result of the first wavelet down-sampling layer into the fifth double convolutional layer to obtain the second fusion feature map output by the fifth double convolutional layer.

[0100] In an embodiment, the channel number of the first wavelet down-sampling layer is set to 32, the convolution kernel size of the fifth double convolutional layer is set to 3, the step is 1, the padding is 1, and the output channel number is 64.

[0101] S830, input the second fusion feature map into the second wavelet down-sampling layer, and input the output result of the second wavelet down-sampling layer into the sixth double convolutional layer to obtain the third fusion feature map output by the sixth double convolutional layer.

[0102] In an embodiment, the channel number of the second wavelet down-sampling layer is set to 64, the convolution kernel size of the sixth double convolutional layer is set to 3, the step is 1, the padding is 1, and the output channel number is 128.

[0103] S840, input the third fusion feature map into the third wavelet down-sampling layer, and input the output result of the third wavelet down-sampling layer into the seventh double convolutional layer to obtain the fourth fusion feature map output by the seventh double convolutional layer.

[0104] In an embodiment, the convolution kernel size of the seventh double convolutional layer is set to 3, the step is 1, the padding is 1, and the output channel number is 256.

[0105] S850, input the fourth fusion feature map into the eighth double convolutional layer to obtain the fifth fusion feature map output by the eighth double convolutional layer.

[0106] ​In an embodiment, the convolution kernel size of the eighth double convolution layer is set to 3, the step is 1, the padding is 1, and the number of output channels is 512.

[0107] S860, input the fifth fusion feature map into the first deconvolution layer, perform dimension concatenation on the output result of the first deconvolution layer and the third fusion feature map, and input the dimension concatenation result into the ninth double convolution layer to obtain a sixth fusion feature map output by the ninth double convolution layer.

[0108] In an embodiment, the convolution kernel size of the first deconvolution layer is 3, the step is 1, and the padding is 2.

[0109] S870, input the sixth fusion feature map into the second deconvolution layer, perform dimension concatenation on the output result of the second deconvolution layer and the second fusion feature map, and input the dimension concatenation result into the tenth double convolution layer to obtain a seventh fusion feature map output by the tenth double convolution layer.

[0110] In an embodiment, the convolution kernel size of the second deconvolution layer is 3, the step is 1, and the padding is 2. The output channel of the tenth double convolution layer is 32.

[0111] S880, input the seventh fusion feature map into the third deconvolution layer, perform dimension concatenation on the output result of the third deconvolution layer and the first fusion feature map, and input the dimension concatenation result into the eleventh double convolution layer to obtain an eighth fusion feature map output by the eleventh double convolution layer, which is used to represent the feature information of the crust observation data set.

[0112] In an embodiment, the convolution kernel size of the third deconvolution layer is 3, the step is 1, and the padding is 2. The output channel of the eleventh double convolution layer is 32.

[0113] In this optional embodiment, using the wavelet downsampling layer can reduce the dimension and compress the data through discrete wavelet transform, reduce the data volume while retaining key information, thereby improving storage efficiency and analysis speed. In addition, wavelet transform can effectively separate noise and useful signals, and by selectively retaining low-frequency approximation coefficients or threshold processing high-frequency detail coefficients, noise filtering can be achieved to improve data quality. At the same time, the multi-resolution analysis characteristic of wavelet allows observation of time series data at different time scales, both capturing long-term trends and analyzing short-term fluctuations, providing a more comprehensive perspective for complex data. The global and local features extracted can be used for classification and pattern recognition, enhancing the discrimination ability of the model, and can also be used for anomaly detection, thereby not only improving the analysis efficiency, but also enhancing the interpretability and accuracy of time series data.

[0114] Optionally, as Figure 9As shown, the weight information of the crust observation data set and the feature information of the crust observation data set are input into the long short-term memory neural network model to obtain a data fusion result output by the long short-term memory neural network model.

[0115] S910, the weight information of the crust observation data set and the feature information of the crust observation data set are merged to obtain merged data.

[0116] Specifically, the sixth high-frequency data feature map and the sixth low-frequency data feature map are respectively multiplied with the eighth fusion feature map pixel by pixel, and then the results of the pixel-by-pixel multiplication are spliced in the channel dimension to obtain a weighted fusion feature map.

[0117] S920, the merged data is input into the third gate convolution layer to obtain merged feature data output by the third gate convolution layer.

[0118] Specifically, the weighted fusion feature map is input into the third gate convolution layer, and in an embodiment, the convolution kernel size of the third gate convolution layer is 3, the step is 1, the padding is 1, and the output channel is 3; the feature data of the weighted fusion feature map is extracted by the third gate convolution layer to obtain merged feature data input into the long short-term memory neural network model, and the merged feature data is long time series data.

[0119] S930, the merged feature data is input into the long short-term memory neural network model (LSTM) to obtain a data fusion result output by the long short-term memory neural network model.

[0120] Specifically, the structure of the long short-term memory neural network model is composed of a cell sequence, and the cell structure is as shown in Figure 10 The structure mainly includes three parts: a forgetting gate, an input gate and an output gate, in this cell structure, is the input of the time sequence, represents the output of the time sequence, t represents time; the forgetting gate is used to control the discard (forget) or retention of some information in the memory unit, and its calculation formula is: The input gate decides to update the information of the memory unit, including Sigmoid and Tanh two parts, wherein the formula of Sigmoid is: The formula of Tanh is: , wherein is a Sigmoid function, x represents the input data, and Tanh ( x ) is a Tanh function; both Sigmoid and Tanh two parts contain the input of the current moment and the output of the last moment , which is: , The two formulas determine that the function of the output gate is to read the just updated neural network state and output the memory cell, and the specific information that can be output is controlled by the output gate, and the main calculation formula is: , , ; in the above formula, , , , are internal functions of the long short-term memory cell, , and are memory vectors, , , , are weight parameters, , , , are convolution biases. It should be noted that the weight parameters are obtained through the training of the neural network.

[0121] In the optional embodiment, the long short-term memory neural network model is widely used in time series modeling as a kind of recurrent neural network architecture, and can also be applied to deformation monitoring based on monitoring data. By constructing a multi-layer stacked long short-term memory neural network model, displacement transformation monitoring data is established through data training, data fusion and deformation reconstruction are realized, and the long short-term memory neural network model has the characteristics of long-time memory information. Not only can information be extracted from a single data point, but also can information be extracted from the entire series of data. The long short-term memory neural network model allows information to persist and circulate.

[0122] Optionally, as shown in Figure 11 , the statistical analysis of the data fusion result is performed to obtain the crustal anomaly information of the target area, including:

[0123] S1110, the statistical analysis of the data fusion result is performed to obtain the mean and standard deviation of the data fusion result.

[0124] S1120, the normal distribution test is performed on the data fusion result, and the data in the data fusion result that does not conform to the normal distribution is converted to normal distribution to obtain the data fusion result that conforms to the normal distribution.

[0125] Specifically, the normal distribution test is performed on the data fusion result, and if the data fusion result does not conform to the normal distribution, the data that does not conform to the normal distribution is converted using the Q statistic method. Q statistic is usually used in the process of normality conversion, which can effectively process the variables that do not conform to the normal distribution, so that they are more consistent with the assumption of normal distribution.

[0126] S1130, the mean and standard deviation of the data fusion result are calculated by using the cumulative sum and control chart formula, the test data are obtained, and the test data are converted into time type sequence to obtain time sequence test data.

[0127] Specifically, the mean and standard deviation of the data fusion result in S1110 are input into the cumulative sum and control chart (CUSUM) formula for calculation to obtain the upper offset statistic and the lower offset statistic in CUSUM, and the upper offset statistic and the lower offset statistic in CUSUM are converted into time type sequence as test data.

[0128] S1140, the least square method is used to linearly fit the time sequence test data to obtain the linear change trend of the target region.

[0129] S1150, the linear change trend is removed from the time sequence test data, and the periodic signal in the removed data is fitted to obtain the periodic motion trend of the target region in one or more predetermined time periods.

[0130] Specifically, the predetermined time period can be an annual period and / or a semi-annual period.

[0131] S1160, according to the linear change trend of the crust of the target region and the periodic crust motion trend, the crust anomaly information of the target region is obtained.

[0132] Specifically, the abnormal points of the linear change trend are detected, for each grid point, the difference between the actual displacement and the fitted linear change trend is calculated, if the difference exceeds a certain set threshold (for example, 3 times the standard deviation), it is considered that the linear trend of the point is abnormal; the abnormal points of the periodic crust motion trend are detected, by comparing the actual periodic crust motion trend of each grid point with the fitted periodic crust motion trend, if the periodic amplitude and frequency exceed the preset range, it may indicate that the periodic change of the region is abnormal. Using standardized residual detection, for each point, the standardized residual between the actual value and the fitted value is calculated, if the residual is too large, it may mean that the point is abnormal. For the obtained abnormal points, clustering analysis is used to identify the abnormal regions that may exist in the target region, these abnormal regions may have abnormal motion trend or periodic fluctuation, and the distribution of the abnormal points, such as the number of abnormal grid points, the spatial distribution characteristics of the abnormal points, etc., are helpful to understand whether the abnormal phenomenon is concentrated in certain regions.

[0133] In the optional embodiment, the fused data fusion result is statistically analyzed, the CUBFASET method is used to effectively detect the abnormal information in the data, and then the least square method is used to fit the fused data to obtain linear velocity and periodic change data of crust deformation, and the linear change trend and periodic motion trend in the target region are comprehensively counted to describe the deformation characteristics of the crust.

[0134] As shown in Figure 12 The embodiment of the present application provides a crust deformation monitoring device 1200, which comprises:

[0135] The acquisition module 1210 is configured to acquire a crust observation data set of a target region, wherein the crust observation data set comprises at least two types of crust observation data, and at least one type of crust observation data is global navigation satellite system data.

[0136] The fusion module 1220 is configured to input the crust observation data set into a long short-term memory neural network model based on wavelet transform which is constructed in advance, to obtain a data fusion result output by the long short-term memory neural network model based on wavelet transform, and the data fusion result is a result after fusion of at least two types of crust observation data in the crust observation data set.

[0137] The statistical analysis module 1230 is configured to statistically analyze the data fusion result to obtain crust abnormal information of the target region, and the crust abnormal information is used to represent crust deformation change.

[0138] Optionally, inputting the crust observation data set into the long short-term memory neural network model based on wavelet transform which is constructed in advance to obtain the data fusion result output by the long short-term memory neural network model based on wavelet transform comprises:

[0139] Wavelet transform is performed on at least two types of crust observation data in the crust observation data set to obtain weight information of the crust observation data set;

[0140] Feature extraction is performed on at least two types of crust observation data in the crust observation data set by using a convolutional neural network to obtain feature information of the crust observation data set;

[0141] The weight information of the crust observation data set and the feature information of the crust observation data set are input into a long short-term memory neural network model to obtain a data fusion result output by the long short-term memory neural network model.

[0142] Optionally, the wavelet transform is performed on at least two types of crust observation data in the crust observation data set to obtain the weight information of the crust observation data set, comprising:

[0143] performing wavelet analysis on at least two different types of crust observation data in the crust observation data set to obtain high-frequency feature data and low-frequency feature data, the high-frequency feature data being used to represent short-term changes and noise of the crust observation data set, and the low-frequency feature data being used to represent long-term trends and low-frequency information of the crust observation data set;

[0144] inputting the high-frequency feature data and the low-frequency feature data into a first double convolution layer to obtain a first high-frequency data feature map and a first low-frequency data feature map output by the first double convolution layer;

[0145] inputting the first high-frequency data feature map and the first low-frequency data feature map into a maximum pooling layer to obtain a second high-frequency data feature map and a second low-frequency data feature map output by the maximum pooling layer;

[0146] inputting the second high-frequency data feature map and the second low-frequency data feature map into a first gated convolution layer to obtain a third high-frequency data feature map and a third low-frequency data feature map output by the first gated convolution layer;

[0147] inputting the third high-frequency data feature map and the third low-frequency data feature map into a second double convolution layer after upsampling to obtain a fourth high-frequency data feature map and a fourth low-frequency data feature map output by the second double convolution layer;

[0148] splicing the fourth high-frequency data feature map with the first high-frequency data feature map to obtain a first high-frequency splicing feature map, splicing the fourth low-frequency data feature map with the first low-frequency data feature map to obtain a first low-frequency splicing feature map, and inputting the first high-frequency splicing feature map and the first low-frequency splicing feature map into a third double convolution layer to obtain a fifth high-frequency data feature map and a fifth low-frequency data feature map output by the third double convolution layer;

[0149] inputting the fifth high-frequency data feature map and the fifth low-frequency data feature map into a second gated convolution layer to obtain a sixth high-frequency data feature map and a sixth low-frequency data feature map output by the second gated convolution layer, wherein the sixth high-frequency data feature map and the sixth low-frequency data feature map are used to represent weight information of the crust observation data set.

[0150] Optionally, the feature extraction on at least two different types of crust observation data in the crust observation data set by using the convolutional neural network to obtain the feature information of the crust observation data set comprises:

[0151] inputting at least two different types of crust observation data in the crust observation data set into a fourth double convolution layer to obtain a first fusion feature map output by the fourth double convolution layer;

[0152] input the first fusion feature map into a first wavelet downsampling layer, and input an output result of the first wavelet downsampling layer into a fifth double convolution layer to obtain a second fusion feature map output by the fifth double convolution layer;

[0153] input the second fusion feature map into a second wavelet downsampling layer, and input an output result of the second wavelet downsampling layer into a sixth double convolution layer to obtain a third fusion feature map output by the sixth double convolution layer;

[0154] input the third fusion feature map into a third wavelet downsampling layer, and input an output result of the third wavelet downsampling layer into a seventh double convolution layer to obtain a fourth fusion feature map output by the seventh double convolution layer;

[0155] input the fourth fusion feature map into an eighth double convolution layer to obtain a fifth fusion feature map output by the eighth double convolution layer;

[0156] input the fifth fusion feature map into a first deconvolution layer, perform dimension splicing on an output result of the first deconvolution layer and the third fusion feature map, input a dimension splicing result into a ninth double convolution layer to obtain a sixth fusion feature map output by the ninth double convolution layer;

[0157] input the sixth fusion feature map into a second deconvolution layer, perform dimension splicing on an output result of the second deconvolution layer and the second fusion feature map, input a dimension splicing result into a tenth double convolution layer to obtain a seventh fusion feature map output by the tenth double convolution layer;

[0158] input the seventh fusion feature map into a third deconvolution layer, perform dimension splicing on an output result of the third deconvolution layer and the first fusion feature map, input a dimension splicing result into an eleventh double convolution layer to obtain an eighth fusion feature map output by the eleventh double convolution layer, wherein the eighth fusion feature map is used to represent feature information of the crust observation data set.

[0159] Optionally, the weight information of the crust observation data set and the feature information of the crust observation data set are input into a long short-term memory neural network model to obtain a data fusion result output by the long short-term memory neural network model, and the data fusion result includes:

[0160] merge the weight information of the crust observation data set and the feature information of the crust observation data set to obtain merged data;

[0161] input the merged data into a third gated convolution layer to obtain merged feature data output by the third gated convolution layer;

[0162] Input the combined feature data into the long short-term memory neural network model to obtain a data fusion result output by the long short-term memory neural network model.

[0163] Optionally, the statistical analysis on the data fusion result comprises:

[0164] The statistical analysis on the data fusion result obtains a mean value and a standard deviation of the data fusion result.

[0165] The data fusion result is subjected to normal distribution test, and data in the data fusion result that does not obey normal distribution is subjected to normality conversion to obtain data fusion result in normal distribution.

[0166] The mean value and the standard deviation of the data fusion result are calculated by using a cumulative sum control chart formula to obtain test data, and the test data is converted into a time type sequence to obtain time sequence test data.

[0167] The time sequence test data is subjected to linear trend fitting by using a least square method to obtain a linear change trend of the target region.

[0168] The linear change trend is removed from the time sequence test data, and a periodic signal in the removed data is fitted to obtain a periodic motion trend of the target region in one or more predetermined time periods.

[0169] According to the linear change trend and the periodic motion trend of the target region, the crustal abnormal information of the target region is obtained.

[0170] Optionally, the crust observation data set at least comprises global navigation satellite system data, interferometric synthetic aperture radar data and remote sensing image data.

[0171] Optionally, the method further comprises a standardization processing module configured to pre-process and standardize at least two different types of crust observation data in the crust observation data set to make the at least two different types of crust observation data consistent in a spatial coordinate system.

[0172] As shown in Figure 13 The electronic device 1300 provided by the embodiment of the present application comprises a memory 1310 and a processor 1320; the memory 1310 is configured to store a computer program; and the processor 1320 is configured to implement the crust deformation monitoring method as described above when executing the computer program.

[0173] The embodiment of the present application provides a computer readable storage medium, and the storage medium stores a computer program. When the computer program is executed by a processor, the computer program implements the crust deformation monitoring method.

[0174] An electronic device 1300, which can be a server or a client of the present application, will now be described, which is an example of a hardware device that can be applied to aspects of the present application. The electronic device 1300 is intended to represent various forms of digital electronic computing devices such as laptops, desktops, workstations, personal digital assistants, servers, blade servers, mainframes, and other appropriate computing devices. The electronic device 1300 can also represent various forms of mobile devices such as personal digital processing, cellular telephones, smart phones, wearable devices, and other similar computing devices. The components shown here, their connections and relationships, and their functions, are meant to be examples only, and are not meant to limit the implementations of the present application described and / or claimed in this document.

[0175] Although the present application discloses the above, the protection scope of the present application is not limited to this. Those skilled in the art can make various changes and modifications without departing from the spirit and scope of the present application, and these changes and modifications shall fall within the protection scope of the present application.

Claims

1. A method for monitoring crustal deformation, characterized in that: include: Acquiring a crustal observation dataset of a target area, wherein the crustal observation dataset includes at least two different types of crustal observation data, wherein at least one type of crustal observation data is global navigation satellite system data; Inputting the crustal observation data set into a pre-built long short-term memory neural network model based on wavelet transform to obtain a data fusion result output by the long short-term memory neural network model based on wavelet transform, wherein the data fusion result is a result of fusing at least two different types of crustal observation data in the crustal observation data set; Statistical analysis is performed on the data fusion results to obtain crustal anomaly information of the target area, where the crustal anomaly information is used to characterize crustal deformation changes.

2. The crustal deformation monitoring method according to claim 1, characterized in that: Inputting the crust observation data set into a pre-built long short-term memory neural network model based on wavelet transform to obtain a data fusion result output by the long short-term memory neural network model based on wavelet transform includes: performing wavelet transform processing on at least two different types of crustal observation data in the crustal observation data set to obtain weight information of the crustal observation data set; performing feature extraction on at least two different types of crustal observation data in the crustal observation data set using a convolutional neural network to obtain feature information of the crustal observation data set; The weight information of the crust observation data set and the feature information of the crust observation data set are input into a long short-term memory neural network model to obtain a data fusion result output by the long short-term memory neural network model.

3. The crustal deformation monitoring method according to claim 2, characterized in that: The step of performing wavelet transform processing on at least two different types of crust observation data in the crust observation data set to obtain weight information of the crust observation data set includes: performing wavelet analysis on at least two different types of crustal observation data in the crustal observation data set to obtain high-frequency feature data and low-frequency feature data, wherein the high-frequency feature data is used to characterize short-term changes and noise in the crustal observation data set, and the low-frequency feature data is used to characterize long-term trends and low-frequency information in the crustal observation data set; Inputting the high-frequency feature data and the low-frequency feature data into a first double convolutional layer to obtain a first high-frequency data feature map and a first low-frequency data feature map output by the first double convolutional layer; Inputting the first high-frequency data feature map and the first low-frequency data feature map into a maximum pooling layer to obtain a second high-frequency data feature map and a second low-frequency data feature map output by the maximum pooling layer; Inputting the second high-frequency data feature map and the second low-frequency data feature map into the first gated convolutional layer to obtain a third high-frequency data feature map and a third low-frequency data feature map output by the first gated convolutional layer; Upsampling the third high-frequency data feature map and the third low-frequency data feature map and inputting the resulting data into a second double convolutional layer to obtain a fourth high-frequency data feature map and a fourth low-frequency data feature map output by the second double convolutional layer; Splicing the fourth high-frequency data feature map with the first high-frequency data feature map to obtain a first high-frequency splicing feature map; splicing the fourth low-frequency data feature map with the first low-frequency data feature map to obtain a first low-frequency splicing feature map; inputting the first high-frequency splicing feature map and the first low-frequency splicing feature map into a third double convolutional layer to obtain a fifth high-frequency data feature map and a fifth low-frequency data feature map output by the third double convolutional layer; The fifth high-frequency data feature map and the fifth low-frequency data feature map are input into a second gated convolutional layer to obtain a sixth high-frequency data feature map and a sixth low-frequency data feature map output by the second gated convolutional layer; wherein the sixth high-frequency data feature map and the sixth low-frequency data feature map are used to characterize weight information of the crust observation dataset.

4. The crustal deformation monitoring method according to claim 2, characterized in that: The step of extracting features from at least two different types of crustal observation data in the crustal observation data set using a convolutional neural network to obtain feature information of the crustal observation data set includes: Inputting at least two different types of crustal observation data in the crustal observation dataset into a fourth double convolutional layer to obtain a first fusion feature map output by the fourth double convolutional layer; Inputting the first fused feature map into a first wavelet downsampling layer, and inputting the output result of the first wavelet downsampling layer into a fifth double convolutional layer to obtain a second fused feature map output by the fifth double convolutional layer; Inputting the second fused feature map into the second wavelet downsampling layer, and inputting the output result of the second wavelet downsampling layer into the sixth double convolution layer to obtain a third fused feature map output by the sixth double convolution layer; Inputting the third fused feature map into a third wavelet downsampling layer, and inputting the output result of the third wavelet downsampling layer into a seventh double convolutional layer to obtain a fourth fused feature map output by the seventh double convolutional layer; Inputting the fourth fused feature map into the eighth double convolutional layer to obtain a fifth fused feature map output by the eighth double convolutional layer; Inputting the fifth fused feature map into the first deconvolution layer, performing dimension splicing on the output result of the first deconvolution layer and the third fused feature map, and inputting the dimension splicing result into the ninth double convolution layer to obtain the sixth fused feature map output by the ninth double convolution layer; Inputting the sixth fused feature map into the second deconvolution layer, dimensionally concatenating the output result of the second deconvolution layer and the second fused feature map, and inputting the dimensionally concatenated result into the tenth biconvolution layer to obtain the seventh fused feature map output by the tenth biconvolution layer; The seventh fused feature map is input into the third deconvolution layer, the output result of the third deconvolution layer is dimensionally spliced ​​with the first fused feature map, and the dimension splicing result is input into the eleventh double convolution layer to obtain the eighth fused feature map output by the eleventh double convolution layer, and the eighth fused feature map is used to characterize the characteristic information of the crust observation dataset.

5. The crustal deformation monitoring method according to claim 2, characterized in that: Inputting the weight information of the crust observation dataset and the feature information of the crust observation dataset into a long short-term memory neural network model to obtain a data fusion result output by the long short-term memory neural network model includes: Merging the weight information of the crust observation dataset and the feature information of the crust observation dataset to obtain merged data; Inputting the combined data into a third gated convolutional layer to obtain combined feature data output by the third gated convolutional layer; The combined feature data is input into the long short-term memory neural network model to obtain a data fusion result output by the long short-term memory neural network model.

6. The crustal deformation monitoring method according to claim 1, characterized in that: The performing statistical analysis on the data fusion result to obtain crustal anomaly information of the target area includes: Performing statistical analysis on the data fusion results to obtain the mean and standard deviation of the data fusion results; Performing a normal distribution test on the data fusion result, and performing normality transformation on the data in the data fusion result that does not obey the normal distribution, to obtain a data fusion result with a normal distribution; The mean and standard deviation of the data fusion result are calculated using a cumulative sum control chart formula to obtain test data, and the test data is converted into a time series to obtain time series test data; Performing linear trend fitting on the time series test data using the least squares method to obtain a linear change trend of the target area; removing the linear variation trend from the time series test data, and fitting the periodic signal in the removed data to obtain the periodic motion trend of the target area in one or more predetermined time periods; According to the linear change trend and the periodic movement trend of the target area, crustal anomaly information of the target area is obtained.

7. The crustal deformation monitoring method according to any one of claims 1 to 6, characterized in that: The crust observation data set includes at least: global navigation satellite system data, interferometric synthetic aperture radar data and remote sensing image data.

8. The crustal deformation monitoring method according to any one of claims 1 to 6, characterized in that: Before inputting the crust observation data set into the pre-built long short-term memory neural network model based on wavelet transform, the method further includes: Preprocessing and standardization are performed on at least two different types of crustal observation data in the crustal observation data set, respectively, so that the at least two different types of crustal observation data have consistency in a spatial coordinate system.

9. An electronic device, characterized in that: including memory and processor; The memory is used to store computer programs; The processor is configured to implement the crust deformation monitoring method according to any one of claims 1 to 8 when executing the computer program.

10. A computer-readable storage medium, characterized in that The storage medium stores a computer program, and when the computer program is executed by the processor, the crust deformation monitoring method according to any one of claims 1 to 8 is implemented.

Images