Reservoir bank deformation analysis and landslide identification method and device based on InSAR technology

Abstract

The invention provides a reservoir bank deformation analysis and landslide identification method and device based on an InSAR technology, and the method comprises the steps: collecting multiple groups of SAR images of a reservoir bank of a hydropower station reservoir region through a satellite remote sensing technology, and preprocessing the collected multiple groups of SAR images; multiple groups of SAR images are shot based on different time and different angles; collecting a multi-temporal SAR image of a reservoir bank of the same hydropower station, optimizing the multi-temporal SAR image, and extracting surface deformation data from the optimized multi-temporal SAR image; through data fitting and analysis, a deformation rate and a change trend in the earth surface deformation data are extracted, a landslide area is identified and positioned, and meanwhile, landslide information of a reservoir area is updated in real time; the future landslide trend of the identified landslide area is predicted, and real-time early warning information is provided for hydropower station operation. The deformation process of the reservoir area can be accurately monitored, the landslide area can be identified and predicted, and technical support is provided for current hydropower station geological disaster monitoring.

Description

Technical Field

[0001] The present application relates to the deep learning technology in the field of landslide monitoring technology and artificial intelligence technology, and in particular to a reservoir bank deformation analysis and landslide identification method and device based on InSAR technology. Background Art

[0002] As critical energy facilities, the safety of hydropower stations is directly linked to the stable operation of society, the economy, and the environment. As hydropower station construction continues to advance, the geological safety of reservoir areas is receiving increasing attention. During reservoir impoundment, bank deformation, particularly geological hazards such as landslides and subsidence, can pose a serious threat to hydropower station operations. Traditional monitoring methods rely primarily on ground-based measurements or manual observations, which suffer from long data collection cycles, high monitoring costs, and limited coverage. Summary of the Invention

[0003] The embodiments of the present application provide a reservoir bank deformation analysis and landslide identification method and device based on InSAR technology.

[0004] In a first aspect, the present invention provides a method for reservoir bank deformation analysis and landslide identification based on InSAR technology, including:

[0005] Collecting multiple sets of synthetic aperture radar (SAR) images of the reservoir area and banks of the hydropower station using satellite remote sensing technology, and preprocessing the collected multiple sets of SAR images; the multiple sets of SAR images are taken at different times and different angles;

[0006] Collecting multi-temporal SAR images of the reservoir area and banks of the same hydropower station, optimizing the multi-temporal SAR images, and extracting surface deformation data from the optimized multi-temporal SAR images;

[0007] Through data fitting and analysis, the deformation rate and change trend in the surface deformation data are extracted, and the landslide area is identified and located, while the landslide information of the reservoir area is updated in real time;

[0008] The future landslide trend of the identified landslide area is predicted to provide real-time early warning information for hydropower station operation.

[0009] In a second aspect, an embodiment of the present application provides a reservoir bank deformation analysis and landslide identification device based on InSAR technology, comprising:

[0010] An acquisition and preprocessing module is used to acquire multiple sets of synthetic aperture radar (SAR) images of the reservoir area and banks of the hydropower station using satellite remote sensing technology, and to preprocess the acquired multiple sets of SAR images; the multiple sets of SAR images are taken at different times and angles;

[0011] A data extraction module is used to collect multi-temporal SAR images of the reservoir area and banks of the same hydropower station, optimize the multi-temporal SAR images, and extract surface deformation data from the optimized multi-temporal SAR images;

[0012] An identification and positioning module is used to extract the deformation rate and change trend from the surface deformation data through data fitting and analysis, and to identify and locate the landslide area, while updating the landslide information of the reservoir area in real time;

[0013] The prediction module is used to predict the future landslide trend of the identified landslide area and provide real-time warning information for the operation of the hydropower station.

[0014] In a third aspect, an embodiment of the present application provides an electronic device, including:

[0015] one or more processors;

[0016] The processor is used to call instructions to enable the electronic device to execute the method described in the first aspect.

[0017] In a fourth aspect, an embodiment of the present application provides a storage medium storing instructions. When the instructions are executed on an electronic device, the electronic device executes the method described in the first aspect.

[0018] In a fifth aspect, an embodiment of the present application provides a program product, comprising at least one of a program and an instruction, wherein when the at least one of the program and the instruction is executed by an electronic device, the steps of the method described in the first aspect are implemented.

[0019] According to the technical solution of this application, high-precision surface deformation data of the hydropower station reservoir area and reservoir bank can be extracted through multi-phase SAR imagery. Through data fitting and analysis, the deformation rate and change trend in the surface deformation data can be extracted, and landslide areas can be identified and located. The future landslide trend of the identified landslide area can be predicted, providing real-time early warning information for hydropower station operations. This application can accurately monitor the deformation process of the reservoir area and identify and predict landslide areas, providing technical support for current geological disaster monitoring of hydropower stations. It can solve the problems of traditional monitoring methods that mainly rely on ground measurements or manual observations, resulting in long data collection cycles, high monitoring costs, and limited coverage.

[0020] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0022] Figure 1 A schematic diagram of the process of reservoir bank deformation analysis and landslide identification method based on InSAR technology provided in an embodiment of the present application;

[0023] Figure 2 A schematic diagram of the process of reservoir bank deformation analysis and landslide identification method based on InSAR technology provided in an embodiment of the present application;

[0024] Figure 3 A schematic diagram of the process of reservoir bank deformation analysis and landslide identification method based on InSAR technology provided in an embodiment of the present application;

[0025] Figure 4 A block diagram of the reservoir bank deformation analysis and landslide identification device based on InSAR technology provided in an embodiment of the present application;

[0026] Figure 5 is a block diagram of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0027] The following describes in detail embodiments of the present application, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present application, and should not be construed as limiting the present application.

[0028] The embodiments of the present application are not exhaustive, but are merely illustrative of some embodiments and are not intended to be a specific limitation on the scope of protection of the present application. In the absence of contradiction, each step in a certain embodiment can be implemented as an independent embodiment, and the steps can be arbitrarily combined. For example, a solution after removing some steps in a certain embodiment can also be implemented as an independent embodiment, and the order of the steps in a certain embodiment can be arbitrarily exchanged. In addition, the optional implementations in a certain embodiment can be arbitrarily combined; in addition, the embodiments can be arbitrarily combined. For example, some or all of the steps in different embodiments can be arbitrarily combined, and a certain embodiment can be arbitrarily combined with the optional implementations of other embodiments.

[0029] In each embodiment of the present application, unless otherwise specified or there is any logical conflict, the terms and / or descriptions between the embodiments are consistent and can be referenced by each other. The technical features in different embodiments can be combined to form a new embodiment based on their inherent logical relationships.

[0030] It is worth noting that in the embodiments of this application, certain software, components, models, etc. that already exist in the industry may be mentioned. They should be considered as exemplary. Their purpose is only to illustrate the feasibility of implementing the technical solution of this application, but it does not mean that the applicant has or will necessarily use the solution.

[0031] The following describes a reservoir bank deformation analysis and landslide identification method and device based on InSAR (Synthetic Aperture Radar Interferometry) technology according to an embodiment of the present application with reference to the accompanying drawings.

[0032] It should be noted that the execution entity of the reservoir bank deformation analysis and landslide identification method based on InSAR technology in the embodiment of the present application can be a reservoir bank deformation analysis and landslide identification device based on InSAR technology, which can be implemented by software and / or hardware. The device can be configured in an electronic device, which can include but is not limited to a terminal, a server, etc.

[0033] Figure 1 This is a flow chart of the reservoir bank deformation analysis and landslide identification method based on InSAR technology provided in the embodiment of this application. Figure 1 As shown, the reservoir bank deformation analysis and landslide identification method based on InSAR technology may include but is not limited to the following steps.

[0034] In step 101, multiple sets of SAR (Synthetic Aperture Radar) images of the reservoir area and banks of the hydropower station are collected by satellite remote sensing technology, and the collected multiple sets of SAR images are preprocessed; the multiple sets of SAR images are taken at different times and different angles.

[0035] In some embodiments, a plurality of SAR images captured at different times and angles may be selected from the plurality of SAR images, one of which is used as a reference image. Based on the feature points of the reference image, the remaining SAR images in the plurality of SAR images are registered to the reference image using a pixel-level error minimization method. The feature points may include, but are not limited to, one or more of corners, edges, and objects. For example, corners, edges, and objects of the reference image may be selected as feature points, and the remaining SAR images are registered to the reference image using a pixel-level error minimization method based on the selected feature points. Based on the different sizes of the SAR images, a set of N×N sliding windows is generated, the position of the sliding window corresponding to each SAR image is initialized, and the sliding window is moved according to a preset path and step size. During each movement, the pixel values ​​in the sliding window are smoothed. For example, the pixel values ​​in the sliding window may be smoothed using a Gaussian filter to remove high-frequency noise. The geometric distortion information in each group of SAR images is detected based on factors such as perspective differences, platform height changes, and terrain undulations. The DEM (Digital Elevation Model) data and satellite orbit information of the target area are obtained. A regional geometric model of the target area is constructed based on the DEM data and satellite orbit information. The radar line of sight corresponding to each pixel in the SAR image is converted into a position in the ground coordinate system to correct the SAR image. The SAR image is reconstructed pixel by pixel using the bilinear interpolation method, and the spatial resolution of each SAR image is adjusted according to the resampling results.

[0036] In step 102, multi-temporal SAR images of the reservoir area and banks of the same hydropower station are collected, and the multi-temporal SAR images are optimized, and surface deformation data are extracted from the optimized multi-temporal SAR images.

[0037] In some embodiments, as Figure 2 As shown, an optional implementation method of the above steps of collecting multi-temporal SAR images of the reservoir area and bank of the same hydropower station and optimizing the multi-temporal SAR images may include the following steps:

[0038] In step 201, multi-temporal SAR images of the reservoir area and bank of the same hydropower station can be collected by various satellites, and a corresponding objective function can be constructed based on the error between the orbit data and the known true orbit.

[0039] Exemplarily, the calculation formula of the objective function is expressed as follows:

[0040]

[0041] Among them, f Te (x Te ) represents the sum of squares of orbit errors; Represents the i-th Te The orbital position of the observation; Represents the i Te The actual orbital position of the observation; N Te Represents the total number of orbital observation points.

[0042] In step 202, a corresponding parameter multidimensional space is established according to the orbital parameters, and a population consisting of multiple orbital parameter combinations is initialized, and the fitness value of each individual in the population is calculated through the objective function.

[0043] In the embodiment of the present application, a corresponding parameter multidimensional space can be established based on the orbital parameters, wherein each dimension in the parameter multidimensional space can represent a parameter of the orbital data, and then a population consisting of multiple orbital parameter combinations can be initialized (e.g., using Indicates that P Orb Represents the size of the population), and the fitness value of each individual in the population is calculated through the objective function.

[0044] In step 203, individuals with fitness values ​​higher than a preset threshold are selected as parent individuals, and mutation operations are performed on each parent individual based on the differences between the selected parent individuals to generate new individuals. The mutated new individuals are mixed with the parent individuals through a crossover operation to generate a new population.

[0045] As an example, the calculation formula of the mutation operation can be expressed as follows:

[0046]

[0047] in, Represents the i DE A new individual after mutation; as well as represents three individuals randomly selected from the population; F represents the variation scaling factor, which is used to control the variation amplitude.

[0048] As an example, the judgment conditions for the above cross operation may be as follows:

[0049]

[0050] in, Represents the i DE candidate solutions; Represents the i DE parent individuals; rand represents a random number used to control the crossover ratio; C r Represents the preset crossover probability.

[0051] In step 204, the fitness value of each individual in the new population is calculated, and each individual in the new population is regarded as a path composed of multiple trajectory parameter nodes. The pheromone concentration between each node at the current moment is then initialized to 0. According to the number of individuals in the new population, the same number of exploration bodies as the number of individuals is generated, and the position of each exploration body on each population individual is initialized.

[0052] As an example, the position of each explorer on each group of individuals can be initialized by the following formula:

[0053]

[0054] in, Represents the i Orb The initial solution of the exploration body, that is, the orbital parameter combination, x Orb,0 represents the global initial solution, represents a random perturbation.

[0055] In step 205, each exploration body calculates the selection probability of the next orbital parameter node based on the current orbital parameter node, and randomly selects nodes that meet the preset threshold value for movement based on the calculated selection probabilities of each group, and repeats the node selection until a complete path (i.e., a complete orbital parameter combination) is constructed, and the fitness value of the constructed path is calculated, and the pheromone concentration between each node is updated according to the calculated fitness value.

[0056] As an example, the calculation formula for the above selection probability is expressed as follows:

[0057]

[0058] in, Represents the exploration body from node i ACO to j ACO The probability of selection; represents the pheromone intensity, indicating the quality of the path; Represents the heuristic factor, which is the inverse of the fitness value; α ACO and β ACO Represent the weights of pheromone and heuristic factors respectively; k ACO represents any parameter value in the neighborhood solution set; Represents the current node i ACO The neighborhood solution set of .

[0059] As an example, the formula used to update the pheromone concentration between nodes is as follows:

[0060]

[0061] in, Represents the increment generated by the current exploration path; Q ACO Represents the total amount of pheromone; Represents the i ACO The fitness value of the individuals in the group; Represents t ACO +1 from node i ACO to j ACO Pheromone update value; ρ ACO It stands for pheromone volatile factor.

[0062] In step 206, the path construction and pheromone concentration update are repeated until the fitness value changes of each exploration body after multiple rounds of iteration converge to the preset range, and then the iteration is stopped. The new individuals generated based on each individual in the population are compared, and the new individual with the highest fitness value is selected to replace the original individual, and then the selection threshold of the parent individual is updated.

[0063] Repeat steps 203 to 206 until the fitness values ​​of the individuals in the new population converge to within a preset range, then stop the iteration, calculate the fitness values ​​of the individuals in each group of the final population, and select the individual with the highest fitness value as the optimal orbit parameter combination. Based on the optimal orbit parameter combination, orbit refinement is performed on the multi-temporal SAR image to reduce the orbit error in the multi-temporal SAR image, thereby improving orbit accuracy, reducing noise interference caused by orbit error, and avoiding false deformation information caused by orbit error, thereby improving the reliability of landslide identification and risk assessment, and enabling more rapid and accurate calculation of surface deformation and landslide prediction.

[0064] In some embodiments, an optional implementation of the above-mentioned step of extracting surface deformation data from the optimized multi-temporal SAR image may include: arranging the optimized multi-temporal SAR images in chronological order (e.g., in chronological order from new to old), selecting multiple groups of SAR images from the multi-temporal SAR images as main images according to a preset spatial baseline threshold and a temporal baseline threshold; aligning the remaining SAR images in the multi-temporal SAR image to the corresponding main image according to the spatial baseline threshold and the temporal baseline threshold to generate multiple groups of small baseline sets, and performing differential interferometry processing on each pair of images to generate a corresponding interferogram, using a high-pass filter to remove the low-frequency phase components caused by atmospheric effects and noise in each interferogram; calculating the corresponding phase difference in each group of interferograms, combining the phase difference data formed by the multi-temporal SAR image through differential interferometry technology into deformation time series data, and converting the wrapped phase into actual surface deformation data through a phase unwrapping algorithm. As an example, the calculation formula of the phase difference is expressed as follows:

[0065]

[0066] Among them, Δφ SARRepresents the phase difference between two observed SAR images; ΔR SAR represents the relative displacement of the two observed SAR images; λ SAR represents the wavelength of the SAR signal; Δφ SAR represents the total phase difference of the small baseline set; Representative i SBAS Phase difference of the interference pattern; N SBAS Represents the number of interferogram pairs selected.

[0067] In step 103, the deformation rate and change trend in the surface deformation data are extracted through data fitting and analysis, and the landslide area is identified and located, while the landslide information of the reservoir area is updated in real time.

[0068] In some embodiments, as Figure 3 As shown, the optional implementation of the above steps of extracting the deformation rate and change trend from the surface deformation data through data fitting and analysis, and identifying and locating the landslide area may include the following steps.

[0069] In step 301, the surface deformation data is preprocessed to obtain the corresponding deformation data matrix. The data of each time step in the deformation data matrix represents a deformation image. The deformation time series data at different time points are combined to form an input data set with a time dimension, and a landslide recognition model is established. The landslide recognition model includes an input layer, a YOLO layer, an LSTM layer, and an output layer.

[0070] In step 302, the input data set is divided into a training set and a test set. The training set data is divided into multiple batches of small training groups. The data of each training group is then passed into the landslide recognition model as input data. The input layer normalizes the image data in the input data and then performs forward propagation. The YOLO layer performs convolution operations of different scales on each group of input data through convolution units at different levels to extract feature maps of different scales for each data.

[0071] In step 303, each group of feature maps is pooled by a pooling unit to perform feature dimensionality reduction on each group of extracted feature maps, and then the feature maps of different levels after convolution and pooling are fused through jump connections, and the processed feature maps are activated using the ReLU activation function, and the activated feature maps with nonlinear characteristics and multiple scales are output.

[0072] In step 304, each set of feature maps is divided into an S×S grid, and then the inclusion category probability, bounding box coordinates and confidence of each network are calculated. All bounding boxes are sorted according to the confidence, and the box with the highest confidence is selected. Other bounding boxes that overlap with the box by more than a set threshold are removed.

[0073] In step 305, the LSTM layer receives the landslide area bounding box output by the YOLO layer. The LSTM layer identifies the temporal characteristics of the landslide deformation and the evolution trend of the landslide through the back-propagation mechanism. After performing nonlinear activation on the identification results through the fully connected unit, the LSTM layer outputs the probability of occurrence of the landslide area and the landslide development trend.

[0074] In step 306, the loss values ​​of the YOLO layer and the LSTM layer are calculated using the regression loss function and the MSE function, respectively. The calculated loss values ​​are then back-propagated from the output layer of the landslide recognition model using the chain rule. The gradient values ​​of the loss values ​​for each layer of the landslide recognition model are then calculated, and the parameters of the landslide recognition model are optimized based on the gradient values. For example, the parameters of the landslide recognition model can be optimized using the Adam optimizer based on the gradient values.

[0075] In step 307, the landslide recognition model is iteratively trained until the model loss value converges to a preset range, and then the training is stopped. The performance of the landslide recognition model on unknown data is evaluated using a test set. When the performance reaches the preset index, the trained landslide recognition model is used to predict the real-time multi-temporal SAR image to obtain the bounding box of the landslide area, the temporal change of the landslide, and the predicted landslide extension range.

[0076] That is to say, the landslide recognition model is iteratively trained until the model loss value converges to a preset range and then the training is stopped. The test set is used to evaluate the performance of the landslide recognition model on unknown data. If the performance does not meet the preset indicators, the model is retrained. If the performance meets the preset indicators, the trained landslide recognition model is deployed to the analysis and recognition platform, and the real-time multi-phase SAR image is input into the landslide recognition model. Through forward propagation, the model outputs the bounding box of the landslide area, the time change of the landslide, and the predicted landslide expansion range.

[0077] In step 104, the future landslide trend of the identified landslide area is predicted to provide real-time warning information for the operation of the hydropower station.

[0078] In an embodiment of the present application, the future landslide trend of the identified landslide area can be predicted based on the trained landslide recognition model, such as the boundary box of the landslide area, the time change of the landslide occurrence, and the predicted landslide expansion range. Based on the future landslide trend of the identified landslide area, real-time warning information can be provided for the operation of the hydropower station.

[0079] In the above embodiment, high-precision surface deformation data of the hydropower station reservoir area and reservoir bank can be extracted through multi-phase SAR images. Through data fitting and analysis, the deformation rate and change trend in the surface deformation data can be extracted, and the landslide area can be identified and located. The future landslide trend of the identified landslide area can be predicted, providing real-time early warning information for the operation of the hydropower station. This application can accurately monitor the deformation process of the reservoir area and identify and predict the landslide area, providing technical support for the current geological disaster monitoring of hydropower stations. It can solve the problems of traditional monitoring methods that mainly rely on ground measurement or manual observation, resulting in long data collection cycles, high monitoring costs, and limited coverage.

[0080] Optionally, in some embodiments, the probability of landslide occurrence and its influencing factors are analyzed, the landslide risk is quantitatively assessed, and an early warning of the potential hazards of the landslide is issued. Exemplarily, deformation data and environmental factors of the reservoir bank of the hydropower station can be collected and preprocessed, and the corresponding standardized deformation gradients are calculated based on each set of collected data. The weight values ​​of each environmental factor are then calculated through principal component analysis. The multi-factor coupling correlation matrix of the corresponding landslide area is constructed according to the calculated deformation rate and weight value. According to the standardized deformation gradient, the state space of the reservoir bank area of ​​the hydropower station is divided into three discrete states: stable, critical and sliding. The transition frequency of each state is counted based on historical landslide events, and the statistical results are smoothed and corrected. A transition probability matrix is ​​constructed. The transition frequency of each state is updated according to the latest monitoring data at preset periodic intervals, and the transition probability matrix is ​​recalculated. When the state is sliding, the transition probability is fixed, and the corresponding area is marked as a high-risk area for landslide. According to the current state of the area, the prior probability of each state of the current area is calculated, and then the GMM model is used to fit the prior probability distribution under each state. The evidence factor is calculated using importance sampling, and the corresponding risk index of the reservoir bank area of ​​the hydropower station is calculated based on the evidence factor and the fitting result.

[0081] For example, the deformation data and environmental factors of the reservoir bank of the hydropower station are collected and preprocessed, and the corresponding standardized deformation gradients are calculated for each set of collected data. The weight values ​​of each environmental factor are then calculated through principal component analysis. The multi-factor coupling correlation matrix of the corresponding landslide area is constructed based on the calculated deformation rate and weight value. According to the standardized deformation gradient, the state space of the reservoir bank area of ​​the hydropower station is divided into three discrete states: stable, critical and sliding. The frequency of state transitions is counted based on historical landslide events, and the statistical results are smoothed and corrected, and a transition probability matrix is ​​constructed. After that, the frequency of state transitions is updated according to the latest monitoring data at preset period lengths (such as every 30 days). If the pixel (i La ,j LaIf no state transition occurs, the state transition frequency remains unchanged. Otherwise, the state transition frequency is updated and the transition probability matrix is ​​recalculated. When the state is sliding, the transition probability is fixed, and the area is marked as a high-risk landslide area. Based on the current regional state, the prior probability of each state in the current region is calculated. The GMM model is then used to fit the prior probability distribution under each state. Importance sampling is used to calculate the evidence factor. Then, based on the evidence factor and the fitting results, the corresponding risk index of the reservoir area and the bank area of ​​the hydropower station is calculated. Based on the calculated risk index, the corresponding warning information is issued.

[0082] As an example, the multi-factor coupling correlation matrix of the landslide area constructed above may be a correlation matrix of a nonlinear coupling relationship, and the expression of the nonlinear coupling relationship is as follows:

[0083]

[0084] in, represents the normalized deformation gradient; Represents a pixel (i La ,j La ) at time t La The calibrated deformation rate of Represents a pixel (i La ,j La ) within a preset range of the spatial region deformation rate; Represents a pixel (i La ,j La ) within a preset range of spatial deformation rate standard deviation; α La represents the time sensitivity coefficient; Represents a pixel (i La ,j La ) at time t La -Δt La The calibrated deformation rate; Δt La represents a time interval; Represents the kth La The weight of each environmental factor; Represents the kth La The eigenvalues ​​of the principal components; represents the sum of all principal component eigenvalues; K La represents the total number of environmental factors; Represents the kth La The coupling effect value of an environmental factor and deformation; exp() represents the exponential function; Represents the kth La Sensitivity parameters of the factors; Represents the kth La Pixels (i La ,jLa ) and time t La Observed value of .

[0085] As an example, the optional implementation method of the above step of issuing corresponding warning information based on the calculated risk index is as follows: if the risk index is ≥0.8, a red warning is triggered, and the emergency plan is immediately activated to evacuate downstream residents. If the risk index is 0.6≤<0.8, an orange warning is triggered, and the monitoring frequency is increased to once a day, and the activities of people in the reservoir area are restricted. If the risk index is 0.3≤<0.6, a yellow warning is triggered, a risk warning is issued, and drainage facilities are inspected.

[0086] Therefore, the weight value of each environmental factor can be calculated through principal component analysis, and the multi-factor coupling correlation matrix of the corresponding landslide area can be constructed according to the calculated deformation rate and weight value. According to the standardized deformation gradient, the state space of the reservoir area and the bank area of ​​the hydropower station is divided into three discrete states: stable, critical and sliding. The transition frequency of each state is statistically analyzed based on historical landslide events, and the statistical results are smoothed and corrected, and a transition probability matrix is ​​constructed. After that, the transition frequency of each state is updated according to the latest monitoring data at regular intervals, and the transition probability matrix is ​​recalculated. When the state is sliding, the transition probability is fixed, and the state is converted to sliding state. The area is marked as a high-risk area for landslides. Based on the current state of the area, the prior probabilities of each state in the current area are calculated. The GMM model is used to fit the prior probability distribution under each state, and the evidence factor is calculated using importance sampling. Then, based on the evidence factor and the fitting results, the corresponding risk index of the reservoir and bank area of ​​the hydropower station is calculated. This can improve the spatiotemporal continuity of the landslide evolution process modeling, solve the problem of high false alarm rate caused by ignoring time series correlation in traditional methods, and is more in line with actual geological responses. It can adapt to changes in different operation stages of the reservoir area without relying on high-performance computing platforms, and realize real-time deployment in remote hydropower stations.

[0087] Figure 4 This is a block diagram of the reservoir bank deformation analysis and landslide identification device based on InSAR technology provided in the embodiment of this application. Figure 4 As shown, the reservoir bank deformation analysis and landslide identification device based on InSAR technology may include an acquisition and preprocessing module 401 , a data extraction module 402 , an identification and positioning module 403 and a prediction module 404 .

[0088] Among them, the acquisition and preprocessing module 401 is used to acquire multiple sets of synthetic aperture radar (SAR) images of the reservoir area and banks of the hydropower station through satellite remote sensing technology, and preprocess the acquired multiple sets of SAR images; the multiple sets of SAR images are taken at different times and different angles.

[0089] The data extraction module 402 is used to collect multi-temporal SAR images of the reservoir area and banks of the same hydropower station, optimize the multi-temporal SAR images, and extract surface deformation data from the optimized multi-temporal SAR images.

[0090] The identification and positioning module 403 is used to extract the deformation rate and change trend in the surface deformation data through data fitting and analysis, and to identify and locate the landslide area, while updating the landslide information of the reservoir area in real time.

[0091] The prediction module 404 is used to predict the future landslide trend of the identified landslide area and provide real-time warning information for the operation of the hydropower station.

[0092] In some embodiments, the acquisition and preprocessing module 401 is configured to: select a group of SAR images from the multiple groups of SAR images as a reference image, and based on the feature points of the reference image, register the remaining groups of SAR images in the multiple groups of SAR images to the reference image by minimizing pixel-level error; generate a group of N×N sliding windows according to the different sizes of the SAR images in each group, initialize the position of the sliding window corresponding to each group of SAR images, move the sliding window according to a preset path and step size, and smooth the pixel values ​​in the sliding window during each movement; detect geometric distortion information in each group of SAR images based on factors such as viewing angle difference, platform height change, and terrain undulation, obtain DEM data and satellite orbit information of the target area, construct a regional geometric model of the target area based on the DEM data and satellite orbit information, convert the radar line of sight corresponding to each pixel in the SAR image into a position in the ground coordinate system to correct the SAR image, reconstruct the SAR image pixels by bilinear interpolation, and adjust the spatial resolution of each SAR image based on the resampling result.

[0093] In some embodiments, the data extraction module 402 is used to perform the following steps: S1, collect multi-phase SAR images of the reservoir area and bank of the same hydropower station through a satellite, and construct a corresponding objective function based on the error between the orbit data and the known true orbit; S2, establish a corresponding parameter multidimensional space based on the orbit parameters, and initialize a population composed of multiple orbit parameter combinations, and calculate the fitness value of each individual in the population through the objective function; S3, select individuals with fitness values ​​higher than a preset threshold as parent individuals, and perform mutation operations on each parent individual according to the differences between the selected parent individuals to generate new individuals, and mix the mutated new individuals with the parent individuals through a crossover operation to generate a new population; S4, calculate the fitness value of each individual in the new population, and use each individual in the new population as a path composed of multiple orbit parameter nodes, and then initialize the pheromone concentration between each node at the current moment to 0, generate the same number of exploration bodies as the number of individuals in the new population according to the number of individuals, and initialize the exploration bodies on the individuals of each population. Position; S5, each of the exploration bodies calculates the selection probability of the next orbital parameter node according to the current orbital parameter node, and randomly selects nodes that meet the preset threshold value to move according to the calculated selection probability of each group, and repeats the node selection until a complete path is constructed, and calculates the fitness value of the constructed path, and updates the pheromone concentration between each node according to the calculated fitness value; S6, repeats the path construction and pheromone concentration update until the fitness value change of each of the exploration bodies after multiple rounds of iteration converges to a preset range, then stops the iteration, compares the new individuals generated based on each individual in the population, and selects the new individual with the highest fitness value to replace the original individual, and then updates the selection threshold of the parent individual; S7, repeatedly executes the above steps S3 to S6 until the fitness value of each individual in the new population converges to a preset range, then stops the iteration, calculates the fitness value of each group of individuals in the final population, and selects the individual with the highest fitness value as the optimal orbital parameter combination, and refines the orbit of the multi-temporal SAR image based on the optimal orbital parameter combination to reduce the orbit error in the multi-temporal SAR image.

[0094] In some embodiments, the data extraction module 402 is used to: arrange the optimized multi-temporal SAR images in chronological order, and select multiple groups of SAR images from the multi-temporal SAR images as main images according to preset spatial baseline thresholds and temporal baseline thresholds; align the remaining SAR images in the multi-temporal SAR images to the corresponding main images according to the spatial baseline thresholds and temporal baseline thresholds to generate multiple groups of small baseline sets, and perform differential interferometry processing on each pair of images to generate corresponding interferograms, and remove low-frequency phase components caused by atmospheric effects and noise in each interferogram; calculate the corresponding phase difference in each group of the interferograms, combine the phase difference data formed by the differential interferometry technology of the multi-temporal SAR images into deformation time series data, and convert the wrapped phase into actual surface deformation data through a phase unwrapping algorithm.

[0095] In some embodiments, the identification and positioning module 403 is used to: pre-process the surface deformation data to obtain the corresponding deformation data matrix, the data of each time step of the deformation data matrix represents a deformation image, and the deformation time series data at different time points are combined to form an input data set with a time dimension to establish a landslide identification model; the landslide identification model includes an input layer, a YOLO layer, an LSTM layer and an output layer; the input data set is divided into a training set and a test set, the training set data is divided into multiple batches of small training groups, and then the data of each training group is passed into the landslide identification model as input data, and the input layer performs image processing on the input data. After the image data is normalized, it is forward propagated. The YOLO layer performs convolution operations of different scales on each group of input data through convolution units of different levels to extract feature maps of different scales of each data; each group of feature maps is pooled through the pooling unit to reduce the dimension of each group of extracted feature maps, and then the feature maps of different levels after convolution and pooling are fused through jump connections, and the ReLU activation function is used to activate each group of feature maps after processing, and outputs a multi-scale feature map with nonlinear characteristics after activation; each group of feature maps is divided into S×S grids, and then the grid is calculated. The inclusion category probability, bounding box coordinates and confidence of the network are calculated, and all bounding boxes are sorted according to the confidence, the box with the highest confidence is selected, and other bounding boxes that overlap with the box by more than a set threshold are removed; the LSTM layer receives the landslide area bounding box output by the YOLO layer, and identifies the temporal characteristics of the landslide deformation through the back propagation mechanism, identifies the evolution trend of the landslide, and outputs the occurrence probability of the landslide area and the landslide development trend after nonlinear activation of the recognition result through the fully connected unit; the loss values ​​of the YOLO layer and the LSTM layer are calculated by the regression loss function and the MSE function respectively, and the calculated The loss value starts from the output layer of the landslide identification model, performs back propagation based on the chain rule, and calculates the gradient value of the loss value for each layer of the landslide identification model, and optimizes the parameters of the landslide identification model based on the gradient value; the landslide identification model is iteratively trained repeatedly until the model loss value converges to a preset range, and the training is stopped. The performance of the landslide identification model on unknown data is evaluated using a test set. When the performance reaches the preset index, the real-time multi-phase SAR image is predicted by the trained landslide identification model to obtain the bounding box of the landslide area, the time change of the landslide occurrence, and the predicted landslide expansion range.

[0096] In some embodiments, the device may further include an evaluation module, wherein the evaluation module is used to analyze the probability of landslide occurrence and its influencing factors, perform quantitative evaluation of landslide risks, and issue early warnings for potential landslide hazards. Optionally, in a possible implementation, the evaluation module is used to: collect and pre-process deformation data and environmental factors of the reservoir bank of the hydropower station, and calculate the corresponding standardized deformation gradient based on each set of collected data, and then calculate the weight value of each environmental factor through principal component analysis, and construct a multi-factor coupling correlation matrix corresponding to the landslide area according to the calculated deformation rate and weight value, and divide the state space of the reservoir bank area of ​​the hydropower station into three discrete states of stable, critical and sliding according to the standardized deformation gradient; based on historical landslide events, the transition frequency of each state is counted, and the statistical results are smoothed and corrected, and a transition probability matrix is ​​constructed, and the transition frequency of each state is updated according to the latest monitoring data at preset periodic intervals, and the transition probability matrix is ​​recalculated. When the state is sliding, the transition probability is fixed, and the corresponding area is marked as a high-risk area for landslide; according to the current area state, the prior probability of each state of the current area is calculated, and then the prior probability distribution under each state is fitted by the GMM model, and the evidence factor is calculated by importance sampling, and the corresponding risk index of the reservoir bank area of ​​the hydropower station is calculated according to the evidence factor and the fitting result.

[0097] It should be noted that the above explanation of the embodiment of the reservoir bank deformation analysis and landslide identification method based on InSAR technology is also applicable to the reservoir bank deformation analysis and landslide identification device based on InSAR technology in this embodiment, and will not be repeated here.

[0098] According to an embodiment of the present application, the present application also provides an electronic device and a readable storage medium.

[0099] like Figure 5 , is a block diagram of an electronic device according to an embodiment of the present application. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present application described and / or required herein.

[0100] like Figure 5As shown, the electronic device includes: one or more processors 501, a memory 502, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. The various components are connected to each other using different buses and can be installed on a common mainboard or installed in other ways as needed. The processor can process instructions executed in the electronic device, including instructions stored in or on the memory to display graphical information of a GUI on an external input / output device (such as a display device coupled to the interface). In other embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Similarly, multiple electronic devices can be connected, and each device provides some necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system). Figure 5 A processor 501 is taken as an example.

[0101] Memory 502 is the non-transitory computer-readable storage medium provided in this application. The memory stores instructions executable by at least one processor, causing the at least one processor to execute the InSAR-based reservoir bank deformation analysis and landslide identification method provided in this application. The non-transitory computer-readable storage medium of this application stores computer instructions for causing a computer to execute the InSAR-based reservoir bank deformation analysis and landslide identification method provided in this application.

[0102] The memory 502 is a non-transient computer-readable storage medium that can be used to store non-transient software programs, non-transient computer executable programs and modules, such as the program instructions / modules corresponding to the reservoir bank deformation analysis and landslide identification method based on InSAR technology in the embodiment of the present application (for example, the attached Figure 4 The processor 501 executes the non-transient software programs, instructions, and modules stored in the memory 502 to execute various functional applications and data processing of the server, thereby implementing the reservoir bank deformation analysis and landslide identification method based on InSAR technology in the above method embodiment.

[0103] The memory 502 may include a program storage area and a data storage area, wherein the program storage area may store an operating system and applications required for at least one function; the data storage area may store data created based on the use of the electronic device, etc. In addition, the memory 502 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory 502 may optionally include a memory remotely located relative to the processor 501, and these remote memories may be connected to the electronic device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0104] The electronic device may further include: an input device 503 and an output device 504. The processor 501, the memory 502, the input device 503 and the output device 504 may be connected via a bus or other means. Figure 5 The bus connection is taken as an example.

[0105] The input device 503 can receive input digital or character information and generate key signal input related to user settings and function control of the electronic device, such as input devices such as a touch screen, a keypad, a mouse, a trackpad, a touch pad, an indicator stick, one or more mouse buttons, a trackball, and a joystick. The output device 504 can include a display device, an auxiliary lighting device (e.g., an LED), and a tactile feedback device (e.g., a vibration motor). The display device can include, but is not limited to, a liquid crystal display (LCD), a light emitting diode (LED) display, and a plasma display. In some embodiments, the display device can be a touch screen.

[0106] Various implementations of the systems and techniques described herein can be realized in digital electronic circuit systems, integrated circuit systems, dedicated ASICs (application specific integrated circuits), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system that includes at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0107] These computer programs (also referred to as programs, software, software applications, or code) include machine instructions for a programmable processor and can be implemented using high-level procedural and / or object-oriented programming languages, and / or assembly / machine languages. As used herein, the terms "machine-readable medium" and "computer-readable medium" refer to any computer program product, apparatus, and / or device (e.g., a magnetic disk, an optical disk, a memory, a programmable logic device (PLD)) for providing machine instructions and / or data to a programmable processor, including a machine-readable medium that receives machine instructions as a machine-readable signal. The term "machine-readable signal" refers to any signal for providing machine instructions and / or data to a programmable processor.

[0108] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0109] The systems and techniques described herein can be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), the Internet, and a blockchain network.

[0110] A computer system may include a client and a server. The client and server are generally remote from each other and typically interact via a communication network. This client-server relationship is established by computer programs running on the respective computers, establishing a client-server relationship. The server may be a cloud server, also known as a cloud computing server or cloud host, a host product within the cloud computing service ecosystem that addresses the management difficulties and limited scalability of traditional physical hosts and VPS services ("Virtual Private Servers" or simply "VPS"). The server may also be a server in a distributed system or a server integrated with blockchain.

[0111] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art can combine and combine different embodiments or examples described in this specification and features of different embodiments or examples without contradiction.

[0112] In the description of the present application, “multiple groups” means at least two groups, such as two groups, three groups, etc., unless otherwise clearly defined.

[0113] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code comprising one or more executable instructions for implementing the steps of a custom logical function or process, and the scope of the preferred embodiments of the present application includes alternative implementations in which functions may be performed out of the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by those skilled in the art to which the embodiments of the present application belong.

[0114] The logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing the logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device). For purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include the following: an electrical connection with one or more wires (electronic devices), a portable computer disk cartridge (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and programmable read-only memory (EPROM or flash memory), fiber optic devices, and a portable compact disc read-only memory (CDROM). Furthermore, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting or processing it in another suitable manner if necessary, and then storing it in a computer memory.

[0115] It should be understood that various parts of the present application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used to implement: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, an application-specific integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0116] Those skilled in the art will understand that all or part of the steps in the method of the above embodiment can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium. When the program is executed, it includes one or a combination of the steps of the method embodiment.

[0117] In addition, the functional units in the various embodiments of the present application may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into a module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.

[0118] The storage medium mentioned above may be a read-only memory, a magnetic disk, or an optical disk, etc. Although the embodiments of the present application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present application. Persons skilled in the art may make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present application.

Claims

1. A reservoir bank deformation analysis and landslide identification method based on InSAR technology, characterized in that: include: Collecting multiple sets of synthetic aperture radar (SAR) images of the reservoir area and banks of the hydropower station using satellite remote sensing technology, and preprocessing the collected multiple sets of SAR images; the multiple sets of SAR images are taken at different times and different angles; Collecting multi-temporal SAR images of the reservoir area and banks of the same hydropower station, optimizing the multi-temporal SAR images, and extracting surface deformation data from the optimized multi-temporal SAR images; Through data fitting and analysis, the deformation rate and change trend in the surface deformation data are extracted, and the landslide area is identified and located, while the landslide information of the reservoir area is updated in real time; The future landslide trend of the identified landslide area is predicted to provide real-time early warning information for hydropower station operation.

2. The method according to claim 1, wherein The pre-processing of the plurality of groups of SAR images collected comprises: Selecting one set of SAR images from the multiple sets of SAR images as a reference image, and registering the remaining sets of SAR images in the multiple sets of SAR images to the reference image by minimizing pixel-level error based on feature points of the reference image; generating a set of N×N sliding windows according to the different sizes of the groups of SAR images, initializing the position of the sliding windows corresponding to the groups of SAR images, moving the sliding windows according to a preset path and step size, and smoothing the pixel values ​​in the sliding windows during each movement; The geometric distortion information in each group of SAR images is detected based on factors such as viewing angle difference, platform height change, and terrain undulation, and the DEM data and satellite orbit information of the target area are obtained. A regional geometric model of the target area is constructed based on the DEM data and satellite orbit information. The radar line of sight corresponding to each pixel in the SAR image is converted into a position in the ground coordinate system to correct the SAR image. The SAR image is pixel-reconstructed by bilinear interpolation, and the spatial resolution of each SAR image is adjusted according to the resampling result.

3. The method according to claim 1, wherein The collecting of multi-temporal SAR images of the reservoir area and bank of the same hydropower station and optimizing the multi-temporal SAR images include: S1, collect multi-temporal SAR images of the reservoir area and bank of the same hydropower station through satellite, and construct the corresponding objective function based on the error between the orbit data and the known true orbit; S2, establishing a corresponding parameter multidimensional space based on the orbital parameters, and initializing a population consisting of multiple orbital parameter combinations, and calculating the fitness value of each individual in the population using the objective function; S3, selecting individuals with fitness values ​​higher than a preset threshold as parent individuals, performing mutation operations on the parent individuals according to the differences between the selected parent individuals to generate new individuals, and mixing the mutated new individuals with the parent individuals through a crossover operation to generate a new population; S4, calculating the fitness value of each individual in the new population, and treating each individual in the new population as a path consisting of multiple trajectory parameter nodes, then initializing the pheromone concentration between each node at the current moment to 0, generating the same number of explorers as the number of individuals in the new population, and initializing the position of each explorer on each individual in the population; S5, each of the exploration bodies calculates the selection probability of the next trajectory parameter node based on the current trajectory parameter node, and randomly selects a node that meets a preset threshold to move based on the calculated selection probability of each group, repeating the node selection until a complete path is constructed, and calculating the fitness value of the constructed path, and updating the pheromone concentration between each node based on the calculated fitness value; S6, repeating the path construction and pheromone concentration update until the fitness value of each explorer after multiple rounds of iteration converges to a preset range, then stopping the iteration, comparing the new individuals generated based on each individual in the population, and selecting the new individual with the highest fitness value to replace the original individual, and then updating the selection threshold of the parent individual; S7, repeatedly executing steps S3 to S6 above until the fitness values ​​of each individual in the new population converge to a preset range, then stopping the iteration, calculating the fitness values ​​of each group of individuals in the final population, and selecting the individual with the highest fitness value as the optimal orbit parameter combination. Based on the optimal orbit parameter combination, orbit refinement is performed on the multi-temporal SAR image to reduce the orbit error in the multi-temporal SAR image.

4. The method according to claim 1, wherein The extracting of surface deformation data from the optimized multi-temporal SAR image comprises: Arranging the optimized multi-temporal SAR images in chronological order, and selecting multiple groups of SAR images as main images from the multi-temporal SAR images according to a preset spatial baseline threshold and a temporal baseline threshold; registering the remaining SAR images in the multi-temporal SAR image to the corresponding main image according to the spatial baseline threshold and the temporal baseline threshold to generate multiple groups of small baseline sets, performing differential interferometry on each pair of images to generate a corresponding interferogram, and removing low-frequency phase components caused by atmospheric effects and noise in each interferogram; The corresponding phase difference in each group of the interference patterns is calculated, the phase difference data formed by the differential interferometry technology of the multi-phase SAR images are combined into deformation time series data, and the entangled phases are converted into actual surface deformation data through a phase unwrapping algorithm.

5. The method according to claim 1, wherein The method of extracting deformation rate and change trend from the surface deformation data through data fitting and analysis, and identifying and locating landslide areas, includes: Preprocessing the surface deformation data to obtain a corresponding deformation data matrix, wherein the data of each time step of the deformation data matrix represents a deformation image, combining the deformation time series data at different time points to form an input data set with a time dimension, and establishing a landslide recognition model; the landslide recognition model includes an input layer, a YOLO layer, an LSTM layer, and an output layer; The input data set is divided into a training set and a test set, and the training set data is divided into multiple batches of small training groups. The data of each training group is then passed into the landslide recognition model as input data. The input layer normalizes the image data in the input data and then performs forward propagation. The YOLO layer performs convolution operations of different scales on each group of input data through convolution units at different levels to extract feature maps of different scales for each data; The pooling unit is used to perform pooling processing on each group of feature maps to reduce the dimension of each group of extracted feature maps. Then, the feature maps of different levels after convolution and pooling are fused through jump connections, and the ReLU activation function is used to activate each group of feature maps after processing, and the activated feature maps with nonlinear characteristics and multiple scales are output; Divide each set of feature maps into an S×S grid, then calculate the inclusion category probability, bounding box coordinates and confidence of each network, sort all bounding boxes according to the confidence, select the box with the highest confidence, and remove other bounding boxes that overlap with it by more than a set threshold; The LSTM layer receives the landslide area bounding box output by the YOLO layer, identifies the temporal characteristics of the landslide deformation through the back-propagation mechanism, identifies the evolution trend of the landslide, and outputs the probability of occurrence of the landslide area and the landslide development trend after performing nonlinear activation on the identification results through the fully connected unit; Calculating the loss values ​​of the YOLO layer and the LSTM layer respectively by using a regression loss function and an MSE function, performing backpropagation based on the chain rule starting from the output layer of the landslide identification model, and calculating the gradient values ​​of the loss values ​​for each layer of the landslide identification model, and optimizing the parameters of the landslide identification model based on the gradient values; The landslide recognition model is iteratively trained until the model loss value converges to a preset range and then the training is stopped. The performance of the landslide recognition model on unknown data is evaluated using a test set. When the performance reaches the preset index, the trained landslide recognition model is used to predict the real-time multi-temporal SAR image to obtain the bounding box of the landslide area, the time change of the landslide occurrence, and the predicted landslide extension range.

6. The method according to any one of claims 1 to 5, wherein The method further comprises: Analyze the probability of landslides and their influencing factors, conduct quantitative assessment of landslide risks, and provide early warning of potential landslide hazards.

7. The method according to claim 6, wherein The analysis of the probability of landslide occurrence and its influencing factors, and the quantitative assessment of landslide risk, include: Deformation data and environmental factors of the reservoir bank of the hydropower station are collected and preprocessed. Based on each set of collected data, the corresponding standardized deformation gradient is calculated. The weight value of each environmental factor is then calculated through principal component analysis. A multi-factor coupling correlation matrix corresponding to the landslide area is constructed based on the calculated deformation rate and weight value. Based on the standardized deformation gradient, the state space of the reservoir bank of the hydropower station is divided into three discrete states: stable, critical, and sliding. Based on historical landslide events, the frequency of each state transition is counted, the statistical results are smoothed and corrected, and a transition probability matrix is ​​constructed. The frequency of each state transition is updated according to the latest monitoring data every preset period, and the transition probability matrix is ​​recalculated. When the state is sliding, the transition probability is fixed, and the corresponding area is marked as a high-risk area for landslide; According to the current regional status, the prior probability of each state in the current region is calculated, and then the GMM model is used to fit the prior probability distribution under each state. The evidence factor is calculated using importance sampling, and the corresponding risk index of the reservoir area and bank area of ​​the hydropower station is calculated based on the evidence factor and the fitting results.

8. A reservoir bank deformation analysis and landslide identification device based on InSAR technology, characterized in that: include: An acquisition and preprocessing module is used to acquire multiple sets of synthetic aperture radar (SAR) images of the reservoir area and banks of the hydropower station using satellite remote sensing technology, and to preprocess the acquired multiple sets of SAR images; the multiple sets of SAR images are taken at different times and angles; A data extraction module is used to collect multi-temporal SAR images of the reservoir area and banks of the same hydropower station, optimize the multi-temporal SAR images, and extract surface deformation data from the optimized multi-temporal SAR images; An identification and positioning module is used to extract the deformation rate and change trend from the surface deformation data through data fitting and analysis, and to identify and locate the landslide area, while updating the landslide information of the reservoir area in real time; The prediction module is used to predict the future landslide trend of the identified landslide area and provide real-time warning information for the operation of the hydropower station.

9. An electronic device, characterized in that: include: one or more processors; The processor is configured to call instructions so that the electronic device executes the method according to any one of claims 1 to 7.

10. A storage medium storing instructions, characterized in that: When the instructions are executed on an electronic device, the electronic device is caused to execute the method according to any one of claims 1 to 7.

Images