A method for solving SAR time series displacement synchronous observation values ​​considering environmental factors

Abstract

The present application provides a method for solving SAR time-series displacement synchronous observation values ​​considering environmental factors, which relates to the field of landslide disaster monitoring. The method includes: obtaining SAR image data and boundary environmental factor data; processing SAR image data based on short baseline set technology to obtain ascending and descending landslide displacement time-series data; constructing a synchronous observation value solving model based on CNN model and LSTM model; training the synchronous observation value solving model through ascending displacement observation data and external environmental factor data to obtain synchronous observation values ​​in ascending mode; training the synchronous observation value solving model through descending displacement observation data and external environmental factor data to obtain synchronous observation values ​​in descending mode; and realizing three-dimensional displacement solution of landslide through synchronous observation values. The synchronous observation value solving model constructed based on CNN model and LSTM model can well capture and model complex nonlinear dynamics and improve the accuracy of model solving.

Description

Technical Field

[0001] The present application relates to the field of landslide disaster monitoring, and in particular to a method for solving SAR time-series displacement synchronous observation values ​​taking environmental factors into consideration. Background Art

[0002] Landslides are a common geological disaster that often causes significant casualties and economic losses. Accurately grasping the deformation of landslides is a key issue in landslide monitoring systems and is essential for predicting landslide disasters and issuing early warnings. InSAR technology can cover a wide area, provide continuous spatial coverage and high-precision deformation measurement, and is not restricted by weather conditions. It can achieve all-weather monitoring and can monitor small deformations of the surface with millimeter-level accuracy. Therefore, InSAR technology occupies an important position in landslide monitoring. However, due to the revisit period and orbit design of SAR satellites, direct observations in some areas may have time gaps. And the data acquisition frequency of a single orbit may not be sufficient to meet the needs of deformation monitoring. Therefore, it is necessary to monitor displacement through multi-track SAR time series. Data from different orbits can provide different observation perspectives, help identify and correct terrain-related errors, provide more comprehensive deformation information, and thus more accurately analyze the surface deformation process.

[0003] However, since the ascending and descending orbits of SAR data represent the situation where the satellite images the same area from two opposite directions, the imaging dates of the ascending and descending orbits are often obtained at different times. Therefore, it is necessary to solve the synchronous observation values ​​of the multi-track SAR time series displacement.

[0004] At present, the commonly used solution methods include linear interpolation, polynomial interpolation, spline interpolation and other conventional solution methods. However, landslide is essentially a dissipative dynamic system that is controlled by rock and soil conditions and develops and evolves under the influence of various influencing factors such as rainfall, reservoir water level fluctuations, earthquakes, etc. Its change is a nonlinear operation process. The traditional solution model does not consider the response relationship between the landslide displacement sequence and environmental factors, and often cannot fit more accurate nonlinear results, especially in the landslide mutation stage. Summary of the invention

[0005] The purpose of the present invention is to provide a SAR time series displacement synchronous observation value solution method taking environmental factors into account in order to solve the problem that the traditional solution model does not consider the response relationship between the landslide displacement sequence and the environmental factors and often cannot fit a more accurate nonlinear result.

[0006] The above-mentioned purpose of the present application is achieved through the following technical solutions:

[0007] Step S1: Acquire the ascending and descending SAR image data of the target landslide and the external environmental factor data affecting the deformation of the landslide;

[0008] Step S2: Processing SAR image data based on the short baseline set technology to obtain the time series data of the landslide displacement on the ascending track and the time series data of the landslide displacement on the descending track;

[0009] Step S3: Based on the CNN model and the LSTM model, a synchronous observation value solution model is constructed;

[0010] Step S4: training the synchronous observation value solution model through the orbit raising displacement observation data and the external environmental factor data to obtain the synchronous observation value in the orbit raising mode; training the synchronous observation value solution model through the orbit lowering displacement observation data and the external environmental factor data to obtain the synchronous observation value in the orbit lowering mode;

[0011] Step S6: The three-dimensional displacement of the landslide is solved by synchronous observation values ​​in the orbit ascending mode and the orbit descending mode.

[0012] Optionally, step S1 includes: the external environmental factor data affecting landslide deformation are rainfall and reservoir water level data.

[0013] Optionally, step S2 includes:

[0014] Step S21: Perform InSAR processing on the SAR image data to obtain an interferometric data set;

[0015] Step S22: Perform time series analysis on the interference pair data set by using the short baseline set technical method to obtain the landslide deformation information of the ascending and descending rails, namely, the ascending rail landslide displacement time series data and the descending rail landslide displacement time series data.

[0016] Optionally, step S4 includes:

[0017] Step S41: merging the imaging time T1 of the ascending SAR image data and the imaging time T2 of the descending SAR image data to determine the imaging time of all SAR image data for three-dimensional displacement solution;

[0018] Step S42: applying the grey correlation analysis method to calculate the correlation between the landslide deformation and the external environmental factor data; selecting the external environmental factor data with a correlation greater than a preset threshold as the input of the synchronous observation value solution model;

[0019] Step S43: training the synchronous observation value solution model through the orbit raising displacement observation data and the external environmental factor data to obtain the function model of the landslide environment and the orbit raising deformation time series;

[0020] Step S44: training the synchronous observation value solution model through the down-track displacement observation data and the external environmental factor data to obtain the function model of the landslide environment and the down-track deformation time series;

[0021] Step S45: Based on the function model of the landslide environment and the orbit raising deformation time series, the external environment factor data corresponding to the time point T2 is input to solve the synchronous observation value in the orbit raising mode at the time point T2;

[0022] Based on the function model of landslide environment and orbit-dropping deformation time series, the external environmental factor data corresponding to time point T1 are input to solve the synchronous observation value under the orbit-dropping mode at time point T1.

[0023] Optionally, step S43 includes:

[0024] S43a: extracting the environmental factor sequence of the selected external environmental factor data through the CNN model of the synchronous observation value solution model;

[0025] S43b: Input the environmental factor sequence into the LSTM model of the synchronous observation value solution model to obtain the observation time series displacement of the ascending SAR image data, that is, the synchronous observation value.

[0026] Optionally, the synchronous observation value solution model uses root mean square error MSE as a loss function; and uses Adam optimizer for parameter optimization.

[0027] An electronic device comprises a processor, a memory, a user interface and a network interface, wherein the memory is used to store instructions, the user interface and the network interface are used to communicate with other devices, and the processor is used to execute the instructions stored in the memory so that the electronic device executes a method for solving SAR time series displacement synchronization observation values ​​taking environmental factors into consideration.

[0028] A computer-readable storage medium stores instructions. When the instructions are executed, a method for solving SAR time series displacement synchronization observation values ​​taking environmental factors into consideration is executed.

[0029] The beneficial effects of the technical solution provided by this application are:

[0030] 1. Based on the CNN model and the LSTM model, a synchronous observation value solution model is constructed, which can well capture and model complex nonlinear dynamics. This model combines the advantages of convolutional neural network (CNN) and long short-term memory network (LSTM). CNN can automatically learn the key features of inducing factors from time series data and has strong learning ability. The LSTM model is a time-based recurrent neural network used to memorize long-term dependencies, which can handle long-term trends and periodic changes in time series, and achieve more accurate synchronous observation value solutions for ascending and descending orbits.

[0031] 2. The key to solving the three-dimensional deformation of the landslide surface using synthetic aperture radar (SAR) satellites is to solve the synchronous observation values ​​of SAR time series deformation data on different orbits. The short baseline set (SBAS) technology is used to process SAR image data, and the time series data of landslide displacement on ascending and descending orbits are obtained respectively. The CNN and LSTM deep learning methods are combined to solve the synchronous observation values ​​of SAR time series deformation data on different orbits. The displacements of ascending and descending orbits at all imaging time points can be used to solve the three-dimensional displacement of the landslide. Considering the external environmental factors, the synchronous observation values ​​of multi-orbit SAR time series displacement of the landslide can be solved more accurately. SAR synchronous observation values ​​of different orbits can be solved, especially in the landslide mutation stage, to provide more reliable data support for the three-dimensional deformation of the landslide surface, which is of great significance for landslide surface deformation monitoring and geological disaster risk prevention and control. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] The present application will be further described below with reference to the accompanying drawings and embodiments, in which:

[0033] Figure 1 is a step diagram in an embodiment of the present application;

[0034] Figure 2 It is a schematic diagram of SAR data lifting orbit monitoring in an embodiment of the present application;

[0035] Figure 3 is a flowchart of SBAS-InSAR processing in an embodiment of the present application;

[0036] Figure 4 It is a structural diagram of the CNN-LSTM model in the embodiment of the present application;

[0037] Figure 5 It is a schematic diagram of the LSTM principle in the embodiment of the present application;

[0038] Figure 6 It is a schematic diagram of the structure of an electronic device in an embodiment of the present application. DETAILED DESCRIPTION

[0039] In order to have a clearer understanding of the technical features, purposes and effects of the present application, the specific implementation methods of the present application are now described in detail with reference to the accompanying drawings.

[0040] The embodiments of the present application provide a method for solving SAR time series displacement synchronization observation values ​​taking environmental factors into consideration.

[0041] Please refer to Figure 1 , Figure 1 This is a step diagram of a method for solving SAR time series displacement synchronization observation values ​​considering environmental factors in an embodiment of the present application, including:

[0042] Step S1: Acquire the ascending and descending SAR image data of the target landslide and the external environmental factor data affecting the deformation of the landslide;

[0043] Step S1 includes: the external environmental factor data affecting landslide deformation are rainfall and reservoir water level data.

[0044] Specifically, based on the ASF data download platform, the ascending and descending SAR image data and elevation data of the target area are obtained; data information such as rainfall and reservoir water level are obtained through monitoring station data download.

[0045] Step S2: Processing SAR image data based on the short baseline set technology to obtain the time series data of the landslide displacement on the ascending track and the time series data of the landslide displacement on the descending track;

[0046] Step S2 includes:

[0047] Step S21: Perform InSAR processing on the SAR image data to obtain an interferometric data set;

[0048] Step S22: Perform time series analysis on the interference pair data set by using the short baseline set technical method to obtain the landslide deformation information of the ascending and descending rails, namely, the ascending rail landslide displacement time series data and the descending rail landslide displacement time series data.

[0049] Specifically, Figure 3 As shown in the figure, the entire SAR image data is mainly divided into three steps, namely, interference pair data set generation, differential interference processing and StaMPS SBAS (short baseline set technology) processing. First, the precise orbit parameters and digital elevation model are used to align and crop the image. After obtaining high-quality interference pairs, the StaMPS SBAS method is used to perform time series InSAR analysis to obtain the surface deformation rate information of the working area. The processing process of StaMPS mainly includes phase noise evaluation, distributed target selection and phase unwrapping. Finally, the interference pairs with large residuals are deleted, and the deformation information of the ascending and descending orbits is solved. Figure 2 It is a schematic diagram of SAR data lifting and lowering orbit monitoring in an embodiment of the present application.

[0050] Step S3: Based on the CNN model and the LSTM model, a synchronous observation value solution model is constructed;

[0051] Step S4: training the synchronous observation value solution model through the orbit raising displacement observation data and the external environmental factor data to obtain the synchronous observation value in the orbit raising mode; training the synchronous observation value solution model through the orbit lowering displacement observation data and the external environmental factor data to obtain the synchronous observation value in the orbit lowering mode;

[0052] Step S4 includes:

[0053] Step S41: merging the imaging time T1 of the ascending SAR image data and the imaging time T2 of the descending SAR image data to determine the imaging time of all SAR image data for three-dimensional displacement solution;

[0054] Step S42: applying the grey correlation analysis method to calculate the correlation between the landslide deformation and the external environmental factor data; selecting the external environmental factor data with a correlation greater than a preset threshold as the input of the synchronous observation value solution model;

[0055] Specifically, the higher the correlation between environmental factors and landslide deformation, the closer the correlation with landslide deformation. The collected rainfall series and reservoir water level series were divided into five inducing factor series, namely, 12-day, 24-day, and 36-day cumulative rainfall and 12-day average reservoir water level and its changes. Grey correlation was performed to obtain correlation results, and factors with correlation greater than 0.7 were selected as inputs to the synchronous observation value solution model.

[0056] Step S43: training the synchronous observation value solution model through the orbit raising displacement observation data and the external environmental factor data to obtain the function model of the landslide environment and the orbit raising deformation time series;

[0057] Step S43 includes:

[0058] S43a: extracting the environmental factor sequence of the selected external environmental factor data through the CNN model of the synchronous observation value solution model;

[0059] Specifically, Figure 4 As shown in the figure, the CNN model includes input layer, convolution layer, pooling layer, fully connected layer and output layer; in the convolution layer, a large number of redundant parameters are reduced by sharing the convolution kernel parameters. Assuming that the data input data is x, the calculation formula is as follows:

[0060]

[0061] The ReLU function is used as the activation function. ReLU can add some nonlinear factors to the network and make the output of some neurons equal to 0, reducing the interdependence between parameters. Therefore, relevant features and more suitable training data can be better mined, thereby alleviating the occurrence of overfitting problems. The expression of ReLU is as follows:

[0062]

[0063] The pooling layer samples the output of the convolutional layer, retaining important features and reducing the number of features.

[0064] Three convolutional layers are added to the CNN neural network. The first layer has 2 input channels, 16 output channels, and a convolution kernel size of 3. The second layer has 16 input channels, a convolution kernel size of 3, and an output channel of 32. The third layer has 32 input channels, a convolution kernel size of 3, and an output channel of 2. Each convolutional layer is followed by a ReLU activation function. The landslide environmental factors selected by the grey correlation method are used as the input of the CNN model, and a stronger feature time series is extracted through the multi-layer convolution of the CNN model.

[0065] S43b: Input the environmental factor sequence into the LSTM model of the synchronous observation value solution model to obtain the observation time series displacement of the ascending SAR image data, that is, the synchronous observation value.

[0066] Specifically, the LSTM model for synchronous observations in ascending orbit mode is trained. The LSTM network model is a type of recurrent neural network (RNN). Unlike traditional RNNs with the "vanishing gradient" problem, the LSTM network uses "storage units" and "gates" to selectively retain or discard information from previous time steps, which enables the LSTM network to capture long-term correlations in time series data. The neuron structure of the LSTM network is mainly composed of a forget gate, an input gate, an output gate, and a memory unit. The input gate determines how much of the latest information will be added to the unit at the current time point. The forget gate determines how much information should be discarded from the previous unit and how much information should be retained, which prevents the internal unit from growing without limit. The output gate filters the new state and outputs the filtered information.

[0067] like Figure 4 , Figure 5 As shown in the figure, in this case, in the LSTM model, the input feature dimension is set to 2, the time step is 1, the number of hidden layers is 30, and the output dimension is 1. In the process of training the CNN-LSTM network model, 70% of them are used as training sets and the remaining 30% are used as test sets. During the training process, the root mean square error MSE is used as the loss function, and the Adam optimizer is used for parameter optimization. It is a kind of adaptive estimation gradient algorithm, and the learning rate is 0.001 to achieve the update of the initial model parameters. Set the number of training times to 500. If the error does not decrease for at least 50 consecutive training times, stop training.

[0068] Step S44: training the synchronous observation value solution model through the down-track displacement observation data and the external environmental factor data to obtain the function model of the landslide environment and the down-track deformation time series;

[0069] Specifically, in the process of training the CNN-LSTM network model, 70% of them are used as training sets and the remaining 30% are used as test sets.

[0070] Step S45: Based on the function model of the landslide environment and the orbit raising deformation time series, the external environment factor data corresponding to the time point T2 is input to solve the synchronous observation value in the orbit raising mode at the time point T2;

[0071] Based on the function model of landslide environment and orbit-dropping deformation time series, the external environmental factor data corresponding to time point T1 are input to solve the synchronous observation value under the orbit-dropping mode at time point T1.

[0072] Step S6: The three-dimensional displacement of the landslide is solved by synchronous observation values ​​in the orbit ascending mode and the orbit descending mode.

[0073] The synchronous observation value solution model adopts the root mean square error MSE as the loss function; and uses the Adam optimizer for parameter optimization.

[0074] Finally, the accuracy results of the synchronous observation values ​​of the SAR data set in this study area are shown in the following table:

[0075] <![CDATA[R 2 ]]> MAE RMSE Linear solution model 0.874 6.373 5.942 CNN Model 0.889 5.814 5.431 LSTM Model 0.907 5.129 4.902 CNN-LSTM Model 0.925 4.755 4.508

[0076] The present application also discloses an electronic device. Figure 6 , Figure 6 The electronic device 500 may include: at least one processor 501 , at least one network interface 504 , a user interface 503 , a memory 505 , and at least one communication bus 502 .

[0077] The communication bus 502 is used to realize the connection and communication between these components.

[0078] The user interface 503 may include a display screen, and the optional user interface 503 may also include a standard wired interface or a wireless interface.

[0079] The network interface 504 may optionally include a standard wired interface or a wireless interface (such as a WI-FI interface).

[0080] The present application also discloses a computer-readable storage medium storing a plurality of instructions suitable for a processor to load in order to execute the above-mentioned method for solving SAR time series displacement synchronization observation values ​​taking environmental factors into consideration.

[0081] The above are only exemplary embodiments of the present disclosure and cannot be used to limit the scope of the present disclosure. That is, any equivalent changes and modifications made according to the teachings of the present disclosure are still within the scope of the present disclosure. After considering the disclosure of the specification and the truth of practice, those skilled in the art will easily think of other embodiments of the present disclosure.

[0082] This application is intended to cover any variation, use or adaptation of the present disclosure, which follows the general principles of the present disclosure and includes common knowledge or customary technical means in the art not described in the present disclosure. The description and examples are to be regarded as exemplary only, and the scope and spirit of the present disclosure are defined by the claims.

Claims

1. A method for solving SAR time series displacement synchronous observation values ​​considering environmental factors, characterized in that: The method comprises the following steps: Step S1: Acquire the ascending and descending SAR image data of the target landslide and the external environmental factor data affecting the deformation of the landslide; Step S1 includes: the external environmental factor data affecting landslide deformation are rainfall and reservoir water level data; Step S2: Processing SAR image data based on the short baseline set technology to obtain the time series data of the landslide displacement on the ascending track and the time series data of the landslide displacement on the descending track; Step S2 includes: Step S21: Perform InSAR processing on the SAR image data to obtain an interferometric data set; Step S22: Performing time series analysis on the interference pair data set by using the short baseline set technical method to obtain the landslide deformation information of the ascending and descending rails, namely, the ascending rail landslide displacement time series data and the descending rail landslide displacement time series data; Step S3: Based on the CNN model and the LSTM model, a synchronous observation value solution model is constructed; Step S4: training the synchronous observation value solution model through the orbit raising displacement observation data and the external environmental factor data to obtain the synchronous observation value in the orbit raising mode; training the synchronous observation value solution model through the orbit lowering displacement observation data and the external environmental factor data to obtain the synchronous observation value in the orbit lowering mode; Step S4 includes: Step S41: merging the imaging time T1 of the ascending SAR image data and the imaging time T2 of the descending SAR image data to determine the imaging time of all SAR image data for three-dimensional displacement solution; Step S42: applying the grey correlation analysis method to calculate the correlation between the landslide deformation and the external environmental factor data; selecting the external environmental factor data with a correlation greater than a preset threshold as the input of the synchronous observation value solution model; Step S43: training the synchronous observation value solution model through the orbit raising displacement observation data and the external environmental factor data to obtain the function model of the landslide environment and the orbit raising deformation time series; Step S44: training the synchronous observation value solution model through the down-track displacement observation data and the external environmental factor data to obtain the function model of the landslide environment and the down-track deformation time series; Step S45: Based on the function model of the landslide environment and the orbit raising deformation time series, the external environment factor data corresponding to the time point T2 is input to solve the synchronous observation value in the orbit raising mode at the time point T2; Based on the function model of landslide environment and orbit-dropping deformation time series, the external environmental factor data corresponding to the T1 time point is input to solve the synchronous observation value under the orbit-dropping mode at the T1 time point; Step S5: The three-dimensional displacement of the landslide is solved by synchronous observation values ​​in the orbit ascending mode and the orbit descending mode.

2. The method for solving SAR time series displacement synchronous observation values ​​considering environmental factors as claimed in claim 1, characterized in that: Step S43 includes: S43a: extracting the environmental factor sequence of the selected external environmental factor data through the CNN model of the synchronous observation value solution model; S43b: Input the environmental factor sequence into the LSTM model of the synchronous observation value solution model to obtain the observation time series displacement of the ascending SAR image data, that is, the synchronous observation value.

3. The method for solving SAR time series displacement synchronous observation values ​​considering environmental factors as claimed in claim 2, characterized in that: The synchronous observation value solution model adopts the root mean square error MSE as the loss function; and uses the Adam optimizer for parameter optimization.

4. An electronic device, characterized in that: The electronic device comprises a processor (501), a memory (505), a user interface (503) and a network interface (504), wherein the memory (505) is used to store instructions, the user interface (503) and the network interface (504) are used to communicate with other devices, and the processor (501) is used to execute the instructions stored in the memory (505) so that the electronic device executes the method according to any one of claims 1 to 3.

5. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores instructions, and when the instructions are executed by a computer, the method according to any one of claims 1 to 3 is executed.

Images