Ground subsidence monitoring method, device, equipment, storage medium and program product

Abstract

The application discloses a kind of ground surface settlement monitoring method, device, equipment, storage medium and program product, it is related to the technical field of ground surface settlement monitoring.Determine the data acquisition interval time of adjacent two InSAR settlement data in time series InSAR settlement data set, time coding is carried out to data acquisition interval time, obtain time interval matrix, time interval matrix and time series InSAR settlement data set are used to train ground surface settlement prediction model, obtain target model, target model is used to carry out ground surface settlement monitoring to target area to be monitored.By time coding to data acquisition interval time and using the time interval matrix and time series InSAR settlement data set obtained by coding to train ground surface settlement prediction model, to reduce the precision error brought by non-uniform time series data, improve the adaptability of target model to irregular sampling data, to improve the prediction accuracy of target model, finally accurately monitor ground surface settlement.

Description

Technical Field

[0001] This application relates to the technical field of surface subsidence monitoring, and more particularly to surface subsidence monitoring methods, surface subsidence monitoring devices, surface subsidence monitoring equipment, storage media, and computer program products. Background Technology

[0002] Traditional methods of land subsidence monitoring primarily utilize instruments such as total stations, high-precision levels, and Global Navigation Satellite Systems (GNSS) to establish monitoring stations on the plane of the deformed area. These stations measure subsidence values ​​and rates at different time intervals, providing real-time feedback of ground subsidence data and issuing warnings for anomalies exceeding thresholds. However, these methods mainly deploy monitoring points in a point-to-line manner, acquiring only discrete elevation change data for these points and failing to reflect the deformation trend and spatial distribution patterns of the entire monitored area. In response, the development of Synthetic Aperture Radar Interferometry (InSAR) has brought a new direction to land subsidence monitoring. This technology is unaffected by environmental conditions and enables all-weather, all-day, large-scale, and high-precision Earth observation, without requiring the establishment of ground control points or physical contact with the monitoring area.

[0003] In recent years, with the development of artificial intelligence, deep learning-based methods have shone brightly in the field of remote sensing. This technology has strong learning capabilities and can handle complex nonlinear deformation prediction problems. It does not rely on manually defined physical parameters (such as lithology and hydrological characteristics) and can automatically extract subsidence features from time-series InSAR data through neural networks, reducing the dependence of traditional physical models on parameter acquisition. Through automatic feature extraction, multi-source data fusion, and an efficient computing architecture, it solves the bottlenecks of traditional remote sensing methods in complex scene interpretation, large-scale data processing, and dynamic monitoring.

[0004] However, during the processing of InSAR data, when remote sensing images are affected by atmospheric interference (such as clouds, ionospheric disturbances, etc.) or noise, interferograms may become incoherent, and such low-quality data will be automatically discarded. Additionally, factors such as snow cover or seasonal vegetation changes can also reduce radar signal coherence. To balance accuracy and data availability, some time-point data needs to be discarded. Due to varying degrees of missing time points, the non-uniform periodicity of subsidence data (i.e., InSAR data) may not accurately reflect the continuous trend of subsidence changes, leading to low accuracy in subsidence prediction and ultimately, inaccurate surface subsidence monitoring.

[0005] The above content is only used to help understand the technical solution of this application and does not represent an admission that the above content is prior art. Summary of the Invention

[0006] The main objective of this application is to provide a method, device, equipment, storage medium, and computer program for monitoring land subsidence, aiming to solve the technical problem that the lack of InSAR data cannot truly reflect the continuous trend of subsidence changes, resulting in low accuracy of subsidence prediction and ultimately inaccurate land subsidence monitoring.

[0007] To achieve the above objectives, this application proposes a method for monitoring land subsidence, the method comprising:

[0008] Acquire a time-series InSAR subsidence dataset and determine the data acquisition interval between two adjacent InSAR subsidence data periods in the time-series InSAR subsidence dataset;

[0009] The data acquisition interval is time-encoded to obtain a time interval matrix;

[0010] A surface subsidence prediction model is constructed, and the surface subsidence prediction model is trained using the time interval matrix and the time-series InSAR subsidence dataset to obtain a target model, wherein the target model is used to monitor surface subsidence in the target area to be monitored.

[0011] In one embodiment, the step of training the surface subsidence prediction model using the time interval matrix and the time-series InSAR subsidence dataset includes:

[0012] Determine the time interval matrix and the temporal and spatial attention of the fused data corresponding to the temporal InSAR settlement dataset;

[0013] By fusing the temporal attention and the spatial attention, a spatiotemporal joint attention is obtained;

[0014] The land subsidence prediction model is trained based on the spatiotemporal joint attention.

[0015] In one embodiment, the step of constructing the surface subsidence prediction model includes:

[0016] A bidirectional spatiotemporal subsidence feature extraction network combining convolutional neural networks and long short-term memory neural networks was used to construct a surface subsidence prediction model.

[0017] In one embodiment, the step of training the surface subsidence prediction model includes:

[0018] Construct a spatiotemporal statistical weighted loss for settlement;

[0019] The surface subsidence prediction model is trained using the spatiotemporal statistical weighted loss of subsidence.

[0020] In one embodiment, the step of constructing the spatiotemporal statistical weighted loss of settlement includes:

[0021] Spatiotemporal attention loss and statistical loss are constructed separately, wherein the statistical loss includes correlation loss, variance loss and mean loss;

[0022] The balance coefficients of the spatiotemporal attention loss and the statistical loss are obtained. Based on the spatiotemporal attention loss, the statistical loss, and the balance coefficients, a spatiotemporal statistical weighted loss for settlement is constructed.

[0023] In one embodiment, the step of training the surface subsidence prediction model using the time interval matrix and the time-series InSAR subsidence dataset to obtain the target model includes:

[0024] Identify the target area to be monitored;

[0025] The target model is used to monitor surface subsidence in the target area to be monitored.

[0026] Furthermore, to achieve the above objectives, this application also proposes a land subsidence monitoring device, which includes:

[0027] The preparation module is used to acquire the time-series InSAR subsidence dataset and determine the data acquisition interval between two adjacent InSAR subsidence data periods in the time-series InSAR subsidence dataset.

[0028] The encoding module is used to perform time encoding on the data acquisition interval to obtain a time interval matrix;

[0029] The application module is used to construct a land subsidence prediction model and train the land subsidence prediction model using the time interval matrix and the time-series InSAR subsidence dataset to obtain a target model, wherein the target model is used to monitor land subsidence in the target area to be monitored.

[0030] In addition, to achieve the above objectives, this application also proposes a land subsidence monitoring device, the device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the land subsidence monitoring method as described above.

[0031] In addition, to achieve the above objectives, this application also proposes a storage medium, which is a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements the steps of the land subsidence monitoring method described above.

[0032] In addition, to achieve the above objectives, this application also provides a computer program product, which includes a computer program that, when executed by a processor, implements the steps of the land subsidence monitoring method described above.

[0033] One or more technical solutions proposed in this application have at least the following technical effects:

[0034] In this application, firstly, the data acquisition interval between two adjacent InSAR settlement data periods in the time-series InSAR settlement dataset is determined. Then, the data acquisition interval is time-coded to obtain a time interval matrix. Next, the time interval matrix and the time-series InSAR settlement dataset are used to train the surface settlement prediction model to obtain the target model. Finally, the target model is used to monitor surface settlement in the target area to be monitored.

[0035] By time-coding the data acquisition interval and using the time interval matrix and time-series InSAR subsidence dataset obtained from the encoding to train the surface subsidence prediction model, the accuracy error caused by non-uniform time-series data is reduced, the adaptability of the target model to irregular sampling data is improved, and the prediction accuracy of the target model is improved, ultimately enabling accurate monitoring of surface subsidence. Attached Figure Description

[0036] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0037] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0038] Figure 1 A flowchart illustrating the first embodiment of the surface subsidence monitoring method of this application;

[0039] Figure 2 A schematic diagram of spatiotemporal joint attention provided for the first embodiment of the surface subsidence monitoring method of this application;

[0040] Figure 3 A schematic diagram of a two-way spatiotemporal settlement feature extraction network provided for the first embodiment of the surface settlement monitoring method of this application;

[0041] Figure 4 A schematic diagram of the Bi-LSTM structure provided in the first embodiment of the surface subsidence monitoring method of this application;

[0042] Figure 5 A schematic diagram of the surface subsidence prediction model structure provided in the first embodiment of the surface subsidence monitoring method of this application;

[0043] Figure 6 A schematic diagram of the first inference result provided for the surface subsidence monitoring method of this application;

[0044] Figure 7 A schematic diagram of the second inference result provided for the surface subsidence monitoring method of this application;

[0045] Figure 8 A schematic diagram of the third inference result provided for the surface subsidence monitoring method of this application;

[0046] Figure 9 This is a schematic diagram of the module structure of the surface subsidence monitoring device according to an embodiment of this application;

[0047] Figure 10 This is a schematic diagram of the equipment structure of the hardware operating environment involved in the surface subsidence monitoring method in this application embodiment.

[0048] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0049] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of this application and are not intended to limit this application.

[0050] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.

[0051] It should be noted that the executing entity in this embodiment can be a computing service device with data processing, network communication, and program execution functions, such as a tablet computer, personal computer, or mobile phone, or an electronic device or land subsidence monitoring device capable of performing the above functions. The following description uses a land subsidence monitoring device as an example to illustrate this embodiment and the subsequent embodiments.

[0052] Based on this, the embodiments of this application provide a method for monitoring land subsidence, referring to... Figure 1 , Figure 1 This is a flowchart illustrating the first embodiment of the surface subsidence monitoring method of this application.

[0053] In this embodiment, the surface subsidence monitoring method includes steps S10 to S30:

[0054] Step S10: Obtain the time-series InSAR settlement dataset and determine the data acquisition interval between two adjacent InSAR settlement data periods in the time-series InSAR settlement dataset.

[0055] The surface subsidence monitoring method proposed in this embodiment can be applied to different monitoring areas, including but not limited to: urban land subsidence, subsidence / uplift of reservoir basins and banks after impoundment of large reservoirs, deformation of dam bodies and foundations, ground compaction and surface subsidence caused by oilfield development (especially water injection development) or natural gas field development, or surface subsidence, deformation, and even collapse caused by mining activities. Taking mining as an example, most of my country's coal resources are located underground, requiring underground mining for development and utilization. This method disrupts the original stress balance of rock strata, leading to movement and deformation of rock strata and the surface in underground mining areas, thereby causing geological disasters such as collapses, cracks, and landslides, severely damaging the ecological environment of mining areas. In this case, the surface subsidence monitoring method proposed in this embodiment can be used to accurately monitor surface subsidence in mining areas, precisely and efficiently predict deformation trends in mining areas, issue early warnings of geological disasters, and provide a theoretical basis for more rational coal resource mining, protection of the fragile ecological environment of mining areas, and safe production.

[0056] In one embodiment, interferometric data from the synthetic aperture radar (SAR) of Sentinel-1 satellite are used to process InSAR data using SBAS (Small Baseline Subset) to obtain SBAS-InSAR cumulative deformation time series data. Further processing of the SBAS-InSAR cumulative deformation time series data involves overlapping and cropping to a size of 512*512, and performing normalization and missing value handling to form a time-series InSAR settlement dataset.

[0057] In another embodiment, the data acquisition interval between two adjacent InSAR settlement data periods is obtained by calculating the difference based on the data acquisition time points of two adjacent InSAR settlement data periods in the time-series InSAR settlement dataset. In a specific implementation step, suppose the time-series InSAR settlement dataset contains N data periods, and its time point ascending set is... Then the corresponding characteristics of the data collection interval between two adjacent periods. It can be represented as: The time interval for each data collection is as follows: .

[0058] Step S20: Time-encode the data acquisition interval to obtain a time interval matrix;

[0059] In one embodiment, DSTE (Dynamic Subsidence-related Time-position Embedding) can be used to time-encode the data acquisition interval to obtain a time interval matrix. In a specific implementation step, the following formula is used in DSTE. The data collection interval is time-coded, where... , The weight matrix is ​​a learnable matrix. Representing a moment, is the bias value, and GeLU is the Gaussian error linear unit activation function.

[0060] Step S30: Construct a surface subsidence prediction model and train it using a time interval matrix and a time-series InSAR subsidence dataset to obtain a target model. The target model is used to monitor surface subsidence in the target area to be monitored.

[0061] In one feasible implementation, step S30 may include steps A11 to A13:

[0062] Step A11: Determine the time interval matrix and the temporal and spatial attention of the fused data corresponding to the time-series InSAR settlement dataset;

[0063] Step A12: Merge temporal attention and spatial attention to obtain spatiotemporal joint attention;

[0064] Step A13: Train the land subsidence prediction model based on spatiotemporal joint attention.

[0065] Subsidence processes exhibit high spatiotemporal heterogeneity, with subsidence rates distributed unevenly across space; subsidence rates within the same region vary significantly over time; and the current subsidence state is closely related to historical spatiotemporal patterns, yet traditional surface subsidence prediction models struggle to capture these complex spatiotemporal dependencies. Specifically, traditional spatiotemporal modeling methods employ fixed weights to handle all spatiotemporal locations, failing to adapt to the dynamic changes in subsidence processes and exhibiting limited ability to capture long-distance spatiotemporal dependencies, resulting in delayed predictions of subsidence trends.

[0066] In this embodiment, by introducing weight calculations at three levels—spatial attention, temporal attention, and joint attention—the key spatiotemporal regions in the settlement process are adaptively focused on. Spatial attention identifies settlement-sensitive areas, temporal attention distinguishes the importance of different time steps, and joint attention captures the interaction patterns of specific spatiotemporal regions.

[0067] In a specific implementation step, such as Figure 2 As shown, temporal attention and spatial attention are applied to the fused data of the time interval matrix and the temporal InSAR settlement dataset, respectively. Specifically, the temporal and spatial attention of the time interval matrix and the corresponding fused data of the temporal InSAR settlement dataset are determined, and then the temporal and spatial attention are fused (jointly) to obtain spatiotemporal joint attention. The fusion of temporal and spatial attention can be achieved by concatenating the data along a specified dimension.

[0068] In another feasible implementation, step S30 may include:

[0069] A bidirectional spatiotemporal subsidence feature extraction network combining convolutional neural networks and long short-term memory neural networks was used to construct a surface subsidence prediction model.

[0070] The temporal InSAR settlement dataset contains both temporal and spatial features. Spatial features indicate that the settlement at a certain point in time is highly coherent with that of neighboring pixels. Temporal features indicate the trend of settlement over time at the same location.

[0071] Most traditional methods currently only utilize recurrent neural networks (RNNs) or long short-term memory neural networks (LSTMs), fitting the settlement change trend curve of sampling points in the study area only in the time dimension, without simultaneously considering time and spatial characteristics.

[0072] In this embodiment, a bidirectional spatiotemporal subsidence feature extraction network (ConvBi-LSTM) combining convolutional neural networks (CNN) and long short-term memory neural networks (LSTM) is used to construct a land subsidence prediction model. The bidirectional spatiotemporal subsidence feature extraction network (ConvBi-LSTM) can capture the spatial correlation of land surface deformation through a sliding window, while simultaneously learning forward and reverse time image sequences, thereby improving the accuracy of subsidence prediction.

[0073] ConvLSTM is a variant of LSTM neural networks, extending the core idea of ​​LSTM. Its structure is similar to LSTM, but the fully connected layers from the input layer to hidden layers and from hidden layers to hidden layers in LSTM are replaced with convolutions. This allows ConvLSTM to have convolutional structures in different states. It not only possesses the ability of LSTM to handle temporal problems but also the ability of CNNs to learn local spatial features. Furthermore, as... Figure 3 As shown, ConvLSTM mainly has input gates. Forgotten Gate and output gate The network consists of several gates. The input gate determines how much of the current input data needs to be saved to the cell state; the forget gate determines how much of the cell state from the previous time step needs to be retained; and the output gate controls how much of the current cell state needs to be output to the current output value, as shown in the following formula:

[0074] ;

[0075] ;

[0076] ;

[0077] ;

[0078] ;

[0079] in, Representing a moment; , and These represent the forget gate, input gate, and output gate, respectively. and These represent the memory cell state and the hidden state, respectively. This is the weight matrix. This is the bias value. This represents the convolution operation. This represents the Hadamard product operation. Here, is the Sigmoid function, and tanh is the hyperbolic tangent function.

[0080] Bi-LSTM, through parallel processing of forward and backward LSTM layers, can simultaneously capture correlation information from historical and future time steps. This bidirectional information flow enables the model to identify implicit correlations in time-series data. The Bi-LSTM structure is as follows: Figure 4 As shown, where Indicates positive propagation. This indicates backpropagation. At each time step, BiLSTM concatenates or performs other fusion operations on the forward and backward hidden states to obtain the final hidden state representation for that time step, thereby capturing the features and contextual relationships of the time series more comprehensively.

[0081] In other words, the LSTM in ConvLSTM is unidirectional and can only capture the contextual information of past time steps, but cannot utilize the features of future time steps. The ConvBi-LSTM in this embodiment can capture the contextual information of the past and future by processing forward and reverse time series at the same time, which enhances the model's global understanding of time dimension features. The unidirectional LSTM layer in the original ConvLSTM is replaced with a LSTM layer wrapped by a Bidirectional layer, while retaining the convolution operation.

[0082] In another feasible implementation, step S30 may include steps B11-B12:

[0083] Step B11: Construct the spatiotemporal statistical weighted loss of settlement;

[0084] Step B12: Train the surface subsidence prediction model using the spatiotemporal statistical weighted loss of subsidence.

[0085] Most current prediction models use MSE (Mean Squared Error) as the loss function. MSE measures the difference between the predicted result and the true value by calculating the average error of the predicted result on a sample point basis. However, InSAR data has characteristics such as non-periodicity and non-linearity. If MSE is used, it will be unable to fit the true distribution of the samples, resulting in an inability to accurately capture the dynamic changes of time series data.

[0086] In this embodiment, a spatiotemporal statistical weighted loss for settlement is constructed and used as the optimization direction of the prediction model to further strengthen the attention mechanism's constraint on error and improve the accuracy of settlement prediction.

[0087] In one feasible implementation, step B11 may include steps B110 to B111:

[0088] Step B110: Construct the spatiotemporal attention loss and statistical loss respectively. The statistical loss includes correlation loss, variance loss and mean loss.

[0089] Step B111: Obtain the balance coefficient of spatiotemporal attention loss and statistical loss. Based on the spatiotemporal attention loss, statistical loss and balance coefficient, construct the spatiotemporal statistical weighted loss of settlement.

[0090] In this embodiment, statistical losses corresponding to statistical indicators such as mean, variance, and correlation coefficient are combined with spatio-temporal attention loss to impose a higher penalty on errors in areas with large subsidence change trends, thereby improving the model's ability to focus on key spatio-temporal features.

[0091] In one specific implementation step, as shown in the following formula,

[0092]

[0093] Among them, L Statistics For the spatiotemporal statistical weighted loss of settlement, L Corr For the correlation coefficient loss, L Var For variance loss, L MeanFor mean loss, L ST-Attn For spatiotemporal attention loss, y i Here, N represents the true value of the label, and N is the number of samples. , , , Represents the balance coefficient. , This represents the mean and variance of the i-th sample in the prediction results. , Let represent the mean and variance of the i-th sample in the true values. Denotes KL divergence, This represents the softmax function. Let be the spatiotemporal attention weight for sample i at time t during model adaptive learning, with a value range of [0, 1].

[0094] In one embodiment, the root mean square error (RMSE) and the coefficient of determination R can also be used. 2 In addition, indicators such as the Pearson Correlation Coefficient are used to quantitatively analyze the performance of the surface subsidence prediction model and evaluate the model's prediction accuracy.

[0095] Reference Figure 5 In this embodiment, the surface subsidence prediction model is constructed using an encoder-decoder architecture, which connects a portion of the encoder's output with a portion of the decoder's output to fuse multi-scale spatiotemporal information of subsidence features, thereby enhancing the model's ability to extract multi-scale spatiotemporal features. The model input data dimension is (T, 1, H, W), where T represents the InSAR subsidence data period, H and W represent the width and height of the InSAR subsidence map, and the value represents the cumulative subsidence at the current time point.

[0096] The encoder module mainly consists of a spatiotemporal joint attention module and a bidirectional spatiotemporal settlement feature extraction network module. The input data dimension is (B, T, 1, H, W), where B represents the batch size and the time interval matrix dimension is (B, T-1). After dynamic settlement-related temporal encoding, the data is concatenated with the input data. First, the temporal and spatial information of the settlement data is fused through a spatiotemporal attention mechanism. Then, the feature map after passing through 8 3*3 convolutional kernels and LeakyReLU operation is input into the ConvBi-LSTM. In the ConvBi-LSTM cell, the feature map passes through 128 5*5 convolutional kernels with a padding of 2. The output dimension is (B, 128, 512, 512). After passing through 32 3x3 convolutions with a stride of 2 and LeakyReLU, the input is fed into the second ConvBi-LSTM. After passing through 256 5x5 convolution kernels and convolution operations with padding of 2, it becomes the encoder output with the output dimension (B, 256, 256, 256).

[0097] The decoder module also contains two convolutional kernels and two ConvBi-LSTM layers. First, it goes through the first ConvBi-LSTM layer with 256 5*5 convolutional kernels. After the convolution operation with padding of 2, the feature map size is restored to 512*512 using transpose convolution. Then, it goes through the second ConvBi-LSTM layer with 256 5*5 convolutional kernels. After the convolution operation with padding of 2, it goes through two convolutional layers and LeakyReLU to output the prediction result with dimensions (T`, 1, H, W), where T` is the set output time step.

[0098] In one possible implementation, steps C11-C12 may be included after step S30:

[0099] Step C11: Determine the target area to be monitored;

[0100] Step C12: Using the target model, monitor the surface subsidence of the target area to be monitored.

[0101] In one embodiment, SBAS-InSAR time series data on surface subsidence in mining areas can be acquired, and a target model can be used to achieve high-precision, large-scale, high-efficiency, and low-cost prediction of surface subsidence in mining areas, effectively preventing the occurrence of subsidence disasters and providing auxiliary reference for mine ecological governance, safe production, and the formulation of reasonable sustainable development decisions.

[0102] Specifically, a coal mine was selected as the target area to be detected. InSAR data was acquired using the Sentinel-1 satellite and processed by SBAS-InSAR to obtain time-series InSAR cumulative surface subsidence data, which was then used as the dataset for model training.

[0103] Taking 10 periods of SBAS-InSAR data from January 18, 2024 to May 5, 2024 as an example, the subsidence prediction data represents the cumulative change in subsidence in the region, with data dimensions of (T, H, W), where T represents the time period, and H and W are the width and height of the InSAR image. Due to the large spatial resolution of the InSAR image, the data needs to be overlapped and cropped to a size of 512*512 with a step size of 128 to form the dataset. The last 3 periods are used as labels for calculating the accuracy of the index. To adapt to the model training data, the NoData value is set to 0, and the entire dataset is normalized. 80% of the dataset is used as the training set, and 20% as the validation set. The model is trained for 500 epochs and the optimal model is saved.

[0104] A network framework was built using PyTorch, and the model was trained with an initial learning rate of 1e-3. A cosine annealing strategy was used to dynamically adjust the learning rate, and the Adam optimizer was employed to optimize the model. , An early stopping mechanism is employed to monitor validation set performance and dynamically terminate training, preventing the model from overfitting to noisy or invalid features on the training set and improving generalization ability. A spatiotemporal statistical weighted loss of settlement is used as the loss function to optimize the model. The early stopping mechanism saves the optimal model, which is then used to predict settlement on test data. The prediction results are then compared with InSAR settlement results.

[0105] Using data from January 18, 2024 to March 30, 2024 as input, we predicted settlement data for the next three periods. The predicted results were compared with the actual values, and the predicted results for the three periods showed a high degree of similarity to the actual values. We then calculated R0 using the prediction results and Label. 2 The analysis used Pearson's Correlation and RMSE metrics, sampling 100,000 samples from the prediction results. (Refer to...) Figures 6-8 As shown in the results for the three periods of 2024-04-11, 2024-04-23, and 2024-05-05, R... 2 Both the Pearson correlation coefficient and the linear fitting results are above 0.99, indicating that the model prediction results fit the label well.

[0106] Based on SBAS-InSAR cumulative subsidence data, this method combines an ST-ConvBi-LSTM neural network model to predict subsidence trends in target areas. The model was trained using historical InSAR data and tested with existing data. The test results showed good agreement with the label's performance, validating the feasibility of this method for subsidence prediction. This approach achieves high-precision, large-scale, high-efficiency, and low-cost surface subsidence prediction in mining areas. By combining subsidence monitoring results obtained from SBAS-InSAR technology with a deep learning prediction model, leveraging the all-weather, all-time, and large-scale characteristics of InSAR technology and the automatic feature extraction capabilities of deep learning, the deformation trends in mining areas can be effectively monitored and predicted. This avoids the potential safety risks of manually setting up monitoring points, helping mine management departments to make timely and feasible decisions, and ensuring the safety and sustainable development of mining projects.

[0107] It should be noted that the above examples are only for understanding this application and do not constitute a limitation on the surface subsidence monitoring method of this application. Any simple modifications based on this technical concept are within the protection scope of this application.

[0108] This application also provides a surface subsidence monitoring device; please refer to... Figure 9 The surface subsidence monitoring device includes:

[0109] Preparation module 10 is used to acquire time-series InSAR settlement dataset and determine the data acquisition interval between two adjacent InSAR settlement data periods in the time-series InSAR settlement dataset.

[0110] Encoding module 20 is used to time-encode the data acquisition interval to obtain a time interval matrix;

[0111] Application module 30 is used to construct a surface subsidence prediction model and train the surface subsidence prediction model using a time interval matrix and a time-series InSAR subsidence dataset to obtain a target model. The target model is used to monitor surface subsidence in the target area to be monitored.

[0112] In one embodiment, the application module 30 is further configured to:

[0113] Determine the time interval matrix and the temporal and spatial attention of the fused data corresponding to the temporal InSAR settlement dataset;

[0114] By integrating temporal attention and spatial attention, we obtain spatiotemporal joint attention;

[0115] A land subsidence prediction model is trained based on spatiotemporal joint attention.

[0116] In one embodiment, the application module 30 is further configured to:

[0117] A bidirectional spatiotemporal subsidence feature extraction network combining convolutional neural networks and long short-term memory neural networks was used to construct a surface subsidence prediction model.

[0118] In one embodiment, the application module 30 is further configured to:

[0119] Construct a spatiotemporal statistical weighted loss for settlement;

[0120] The surface subsidence prediction model was trained by using spatiotemporal statistical weighted loss.

[0121] In one embodiment, the application module 30 is further configured to:

[0122] Spatiotemporal attention loss and statistical loss are constructed separately. The statistical loss includes correlation loss, variance loss and mean loss.

[0123] The balance coefficients of spatiotemporal attention loss and statistical loss are obtained. Based on the spatiotemporal attention loss, statistical loss and balance coefficients, the spatiotemporal statistical weighted loss of settlement is constructed.

[0124] In one embodiment, the application module 30 is further configured to:

[0125] After the step of training the surface subsidence prediction model using the time interval matrix and the time-series InSAR subsidence dataset to obtain the target model:

[0126] Identify the target area to be monitored;

[0127] Surface subsidence monitoring is conducted in the target area using a target model.

[0128] The surface subsidence monitoring device provided in this application, employing the surface subsidence monitoring method described in the above embodiments, can solve the technical problem that the lack of InSAR data fails to accurately reflect the continuous trend of subsidence changes, resulting in low accuracy of subsidence prediction and ultimately inaccurate surface subsidence monitoring. Compared with the prior art, the beneficial effects of the surface subsidence monitoring device provided in this application are the same as those of the surface subsidence monitoring method provided in the above embodiments, and other technical features of the surface subsidence monitoring device are the same as those disclosed in the methods of the above embodiments, and will not be repeated here.

[0129] This application provides a land subsidence monitoring device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the land subsidence monitoring method in the first embodiment described above.

[0130] The following is for reference. Figure 10The diagram illustrates a structural schematic of a land subsidence monitoring device suitable for implementing embodiments of this application. The land subsidence monitoring device in the embodiments of this application may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital radio receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Description), PMPs (Portable Media Players), vehicle terminals (e.g., vehicle navigation terminals), and fixed terminals such as digital TVs and desktop computers. Figure 10 The land subsidence monitoring device shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this application.

[0131] like Figure 10 As shown, the land subsidence monitoring device may include a processing unit 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory 1002 or a program loaded from a storage device 1003 into a random access memory 1004. The random access memory 1004 also stores various programs and data required for the operation of the land subsidence monitoring device. The processing unit 1001, the read-only memory 1002, and the random access memory 1004 are interconnected via a bus 1005. An input / output interface 1006 is also connected to the bus. Typically, the following systems can be connected to the input / output interface 1006: input devices 1007 including, for example, a touch screen, touchpad, keyboard, mouse, image sensor, microphone, accelerometer, gyroscope, etc.; output devices 1008 including, for example, a liquid crystal display (LCD), speaker, vibrator, etc.; storage devices 1003 including, for example, magnetic tape, hard disk, etc.; and communication devices 1009. Communication device 1009 allows the land subsidence monitoring equipment to communicate wirelessly or wiredly with other equipment to exchange data. Although the figure shows land subsidence monitoring equipment with various systems, it should be understood that implementation or possession of all the systems shown is not required. More or fewer systems may be implemented alternatively.

[0132] Specifically, according to the embodiments disclosed in this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from storage device 1003, or installed from read-only memory 1002. When the computer program is executed by processing device 1001, it performs the functions defined in the methods of the embodiments disclosed in this application.

[0133] The surface subsidence monitoring equipment provided in this application, employing the surface subsidence monitoring method described in the above embodiments, can solve the technical problem that the lack of InSAR data cannot accurately reflect the continuous trend of subsidence changes, resulting in low accuracy of subsidence prediction and ultimately inaccurate surface subsidence monitoring. Compared with the prior art, the beneficial effects of the surface subsidence monitoring equipment provided in this application are the same as those of the surface subsidence monitoring method provided in the above embodiments, and other technical features of this surface subsidence monitoring equipment are the same as those disclosed in the previous embodiment method, and will not be repeated here.

[0134] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0135] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0136] This application provides a computer-readable storage medium having computer-readable program instructions (i.e., a computer program) stored thereon, which are used to execute the land subsidence monitoring method in the above embodiments.

[0137] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems or devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0138] The aforementioned computer-readable storage medium may be included in the land subsidence monitoring equipment; or it may exist independently and not be assembled into the land subsidence monitoring equipment.

[0139] The aforementioned computer-readable storage medium carries one or more programs. When the aforementioned one or more programs are executed by the land subsidence monitoring device, the land subsidence monitoring device: acquires a time-series InSAR subsidence dataset and determines the data acquisition interval between two adjacent periods of InSAR subsidence data in the time-series InSAR subsidence dataset; performs time encoding on the data acquisition interval to obtain a time interval matrix; constructs a land subsidence prediction model and trains the land subsidence prediction model using the time interval matrix and the time-series InSAR subsidence dataset to obtain a target model, wherein the target model is used to monitor land subsidence in the target area to be monitored.

[0140] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a Local Area Network (LAN) or a Wide Area Network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0141] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0142] The modules described in the embodiments of this application can be implemented in software or hardware. The names of the modules do not necessarily limit the functionality of the unit itself.

[0143] The readable storage medium provided in this application is a computer-readable storage medium that stores computer-readable program instructions (i.e., a computer program) for executing the above-described surface subsidence monitoring method. This solves the technical problem that the lack of sufficient InSAR data to accurately reflect the continuous trend of subsidence changes leads to low accuracy in subsidence prediction, ultimately resulting in inaccurate surface subsidence monitoring. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as those of the surface subsidence monitoring method provided in the above embodiments, and will not be repeated here.

[0144] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the land subsidence monitoring method described above.

[0145] The computer program product provided in this application can solve the technical problem that the lack of InSAR data cannot truly reflect the continuous trend of subsidence changes, resulting in low accuracy of subsidence prediction and ultimately inaccurate surface subsidence monitoring. Compared with the prior art, the beneficial effects of the computer program product provided in this application are the same as those of the surface subsidence monitoring method provided in the above embodiments, and will not be repeated here.

[0146] The above description is only a part of the embodiments of this application and does not limit the patent scope of this application. All equivalent structural transformations made under the technical concept of this application and using the contents of the specification and drawings of this application, or direct / indirect applications in other related technical fields, are included in the patent protection scope of this application.

Claims

1. A method for monitoring land subsidence, characterized in that, The surface subsidence monitoring method includes: Acquire a time-series InSAR subsidence dataset and determine the data acquisition interval between two adjacent InSAR subsidence data periods in the time-series InSAR subsidence dataset; The data acquisition interval is time-encoded to obtain a time interval matrix; the following formula is used. The data collection interval is time-coded, where... , The weight matrix is ​​a learnable matrix. Representing a moment, Here is the bias value, and GeLU is the activation function of the Gaussian error linear unit; A surface subsidence prediction model is constructed, and the surface subsidence prediction model is trained using the time interval matrix and the time-series InSAR subsidence dataset to obtain a target model, wherein the target model is used to monitor surface subsidence in the target area to be monitored.

2. The surface subsidence monitoring method as described in claim 1, characterized in that, The steps of training the surface subsidence prediction model using the time interval matrix and the time-series InSAR subsidence dataset include: Determine the time interval matrix and the temporal and spatial attention of the fused data corresponding to the temporal InSAR settlement dataset; By fusing the temporal attention and the spatial attention, a spatiotemporal joint attention is obtained; The land subsidence prediction model is trained based on the spatiotemporal joint attention.

3. The surface subsidence monitoring method as described in claim 1, characterized in that, The steps for constructing the land subsidence prediction model include: A bidirectional spatiotemporal subsidence feature extraction network combining convolutional neural networks and long short-term memory neural networks was used to construct a surface subsidence prediction model.

4. The surface subsidence monitoring method as described in claim 1, characterized in that, The steps for training the surface subsidence prediction model include: Construct a spatiotemporal statistical weighted loss for settlement; The surface subsidence prediction model is trained using the spatiotemporal statistical weighted loss of subsidence.

5. The surface subsidence monitoring method as described in claim 4, characterized in that, The steps for constructing the spatiotemporal statistical weighted loss for settlement include: Spatiotemporal attention loss and statistical loss are constructed separately, wherein the statistical loss includes correlation loss, variance loss and mean loss; The balance coefficients of the spatiotemporal attention loss and the statistical loss are obtained. Based on the spatiotemporal attention loss, the statistical loss, and the balance coefficients, a spatiotemporal statistical weighted loss of settlement is constructed.

6. The surface subsidence monitoring method as described in claim 1, characterized in that, The step of training the surface subsidence prediction model using the time interval matrix and the time-series InSAR subsidence dataset to obtain the target model includes: Identify the target area to be monitored; The target model is used to monitor surface subsidence in the target area to be monitored.

7. A surface subsidence monitoring device, characterized in that, The surface subsidence monitoring device includes: The preparation module is used to acquire the time-series InSAR subsidence dataset and determine the data acquisition interval between two adjacent InSAR subsidence data periods in the time-series InSAR subsidence dataset. The encoding module is used to time-encode the data acquisition interval to obtain a time interval matrix; the following formula is used. The data collection interval is time-coded, where... , The weight matrix is ​​a learnable matrix. Representing a moment, Here is the bias value, and GeLU is the activation function of the Gaussian error linear unit; The application module is used to construct a land subsidence prediction model and train the land subsidence prediction model using the time interval matrix and the time-series InSAR subsidence dataset to obtain a target model, wherein the target model is used to monitor land subsidence in the target area to be monitored.

8. A surface subsidence monitoring device, characterized in that, The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the land subsidence monitoring method as described in any one of claims 1 to 6.

9. A storage medium, characterized in that, The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, it implements the steps of the land subsidence monitoring method as described in any one of claims 1 to 6.

10. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the steps of the land subsidence monitoring method as described in any one of claims 1 to 6.

Images