Landslide Susceptibility Assessment Method and System Based on Multi-Source Data and Deep Learning

Abstract

The present invention discloses a landslide susceptibility assessment method and system based on multi-source data and deep learning, which combines the landslide susceptibility assessment method of SBAS-InSAR technology, remote sensing image analysis, deep learning algorithm, TRIGRS-SCOOPS3D coupling model and MASSFLOW simulation. This method is used for high-precision identification of landslide-prone areas, prediction of potential landslide locations and three-dimensional landslide process simulation, and is especially suitable for landslide disaster risk management and emergency response in mountainous areas.

Description

Technical Field

[0001] The present invention belongs to the field of landslide disaster assessment, and particularly relates to a landslide susceptibility assessment method and system based on multi-source data and deep learning. Background Art

[0002] Existing landslide monitoring technologies can effectively obtain large-scale surface deformation information, but traditional methods have limitations in accurately identifying potential landslide areas. Especially in complex terrains, there is still a lack of efficient automated methods for identifying potential landslide points through surface deformation data. Existing technologies generally rely on manual selection and manual annotation of potential landslide points, and cannot accurately and efficiently extract potential landslide information from a large amount of remote sensing data.

[0003] Moreover, existing landslide process deduction technologies mainly focus on the volume, path, and stability analysis of landslide bodies, but there is still insufficient accuracy in simulating the distribution of dangerous areas after landslides occur at all potential landslide points. Existing technologies usually can only simulate a single landslide event and it is difficult to comprehensively consider the risk distribution after landslides occur at multiple potential landslide points under different environmental conditions. Especially in complex terrains with multiple influencing factors intertwined, existing models cannot accurately predict the landslide consequences and their expansion areas for each potential landslide point. Summary of the Invention

[0004] To solve the above technical problems, the present invention proposes a landslide susceptibility assessment method and system based on multi-source data and deep learning to solve the problems existing in the above-mentioned prior art.

[0005] To achieve the above object, the present invention provides a landslide susceptibility assessment method based on multi-source data and deep learning, including the following steps:

[0006] Obtain high-resolution remote sensing images and digital elevation model data of the study area, and obtain surface deformation features and potential landslide point information through SBAS-InSAR technology, deep convolutional neural network, and YOLO algorithm;

[0007] Obtain the geological environment data of potential landslide points, and screen out the features of the geological environment data that have landslide influencing factors. Based on the surface deformation features and the screened features, construct a landslide susceptibility assessment model through quantum Bayesian optimization;

[0008] Use the TRIGRS-SCOOPS3D coupling model to analyze the characteristics of rainfall-induced landslides, and obtain the stability of landslides and the spatial distribution of potential landslides under different rainfall conditions;

[0009] According to the stability of landslides and the spatial distribution of potential landslides under different rainfall conditions, the three-dimensional sliding surface of the potential landslide area is simulated by the MASSFLOW model, the landslide volume and the potential landslide path are calculated, and the landslide susceptibility assessment results are obtained.

[0010] Preferably, obtaining the surface deformation characteristics and potential landslide point information includes:

[0011] Calculating the surface deformation of the study area in the satellite image by the SBAS-InSAR technology to obtain the surface deformation information of the study area;

[0012] Obtaining the high-resolution remote sensing image and digital elevation model data of the study area, and extracting features through a pre-trained deep convolutional neural network to obtain the convolution extraction features;

[0013] According to the surface deformation information and the convolution extraction features, detecting and locating potential landslide points by the YOLO algorithm.

[0014] Preferably, constructing a landslide susceptibility assessment model includes:

[0015] Obtaining the geological environment data of potential landslide points, where the geological environment data includes slope, aspect, soil type, vegetation coverage, seismic intensity, rainfall, and groundwater level;

[0016] Screening out the features with landslide influence factors from the geological environment data through the Elastic Net regression model to obtain the screening features;

[0017] According to the surface deformation characteristics and the screening features, constructing a landslide susceptibility assessment model through quantum Bayesian optimization and adaptive deep reinforcement learning.

[0018] Preferably, the method further includes: converting the landslide susceptibility assessment results into a spatial distribution map, presenting the potential landslide area and the high-risk area, and sending out a warning message when the landslide risk is detected to increase.

[0019] The present invention also provides a landslide susceptibility assessment system based on multi-source data and deep learning, including:

[0020] A potential landslide point information acquisition module, configured to obtain the high-resolution remote sensing image and digital elevation model data of the study area, and obtain the surface deformation characteristics and potential landslide point information through the SBAS-InSAR technology, deep convolutional neural network, and YOLO algorithm;

[0021] A landslide susceptibility assessment model construction module, configured to obtain the geological environment data of potential landslide points, screen out the features with landslide influence factors in the geological environment data, and construct a landslide susceptibility assessment model through quantum Bayesian optimization based on the surface deformation characteristics and the screened features;

[0022] A rainfall factor analysis module, which is used to analyze the characteristics of rainfall-induced landslides using the TRIGRS-SCOOPS3D coupling model, and obtain the stability of landslides and the spatial distribution of potential landslides under different rainfall conditions;

[0023] An evaluation module, which is used to simulate the three-dimensional sliding surface of the potential landslide area through the MASSFLOW model according to the stability of the landslide and the spatial distribution of potential landslides under different rainfall conditions, calculate the landslide volume and the potential landslide path, and obtain the landslide susceptibility evaluation result.

[0024] Preferably, the potential landslide point information acquisition module includes:

[0025] A surface deformation information calculation unit, which is used to calculate the surface deformation of the study area in the satellite image through the SBAS-InSAR technology, and obtain the surface deformation information of the study area;

[0026] A convolutional feature extraction unit, which is used to obtain the high-resolution remote sensing image and digital elevation model data of the study area, and perform feature extraction through a pre-trained deep convolutional neural network to obtain the convolutional extraction features;

[0027] A positioning unit, which is used to detect and locate potential landslide points through the YOLO algorithm according to the surface deformation information and the convolutional extraction features.

[0028] Preferably, the landslide susceptibility evaluation model construction module includes:

[0029] A geological environment data acquisition unit, which is used to acquire the geological environment data of potential landslide points, where the geological environment data includes slope, aspect, soil type, vegetation coverage, seismic intensity, rainfall, and groundwater level;

[0030] A screening unit, which is used to screen out the features with landslide influence factors from the geological environment data through the Elastic Net regression model to obtain the screening features;

[0031] A construction unit, which is used to construct a landslide susceptibility evaluation model through quantum Bayesian optimization and adaptive deep reinforcement learning according to the surface deformation characteristics and the screening features.

[0032] Preferably, the system further includes: a display platform, which is used to convert the landslide susceptibility evaluation result into a spatial distribution map, present the potential landslide area and the high-risk area, and send out a warning message when the landslide risk is detected to increase.

[0033] The present invention also provides a computer device, including a memory, a processor, and a computer program stored on the memory, and the processor executes the computer program to implement the steps of the method.

[0034] The present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the above method are implemented.

[0035] Compared with the prior art, the present invention has the following advantages and technical effects:

[0036] The present invention provides a landslide susceptibility assessment method and system based on multi-source data and deep learning, including: obtaining high-resolution remote sensing images and digital elevation model data of the study area, and obtaining surface deformation features and potential landslide point information through SBAS-InSAR technology, deep convolutional neural network and YOLO algorithm; obtaining geological environment data of potential landslide points, and screening out the features of the geological environment data with landslide influence factors. Based on the surface deformation features and the screened features, a landslide susceptibility assessment model is constructed through quantum Bayesian optimization; using the TRIGRS-SCOOPS3D coupling model to analyze the characteristics of rainfall-induced landslides, and obtaining the stability of landslides and the spatial distribution of potential landslides under different rainfall conditions; according to the stability of landslides and the spatial distribution of potential landslides under different rainfall conditions, the three-dimensional sliding surface of the potential landslide area is simulated through the MASSFLOW model, and the landslide volume and potential landslide path are calculated to obtain the landslide susceptibility assessment result. The present invention integrates technologies such as SBAS-InSAR, satellite imagery, deep learning, TRIGRS-SCOOPS3D and MASSFLOW, and can realize the comprehensive analysis of global susceptibility assessment, local stability analysis and landslide process deduction of landslide disasters, with great innovation and application prospects. The present invention provides a new technical means and theoretical support for the prediction, assessment and prevention of landslide disasters. This method can be widely applied to the risk management of mountain landslide disasters, and provides an important reference for promoting the progress and development of disaster prevention technologies. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] The drawings constituting a part of this application are used to provide a further understanding of this application. The schematic embodiments of this application and their descriptions are used to explain this application and do not constitute an improper limitation of this application. In the drawings:

[0038] Figure 1 is a flowchart of a landslide susceptibility assessment method based on multi-source data and deep learning according to an embodiment of the present invention;

[0039] Figure 2 is a flowchart of area recognition according to an embodiment of the present invention;

[0040] Figure 3 is a flowchart of susceptibility assessment according to an embodiment of the present invention;

[0041] Figure 4Flowchart for simulating the landslide process according to an embodiment of the present invention. Detailed implementation manners

[0042] It should be noted that, without conflict, the embodiments in this application and the features in the embodiments may be combined with each other. The following will describe this application in detail with reference to the drawings and in combination with the embodiments.

[0043] It should be noted that the steps shown in the flowchart of the drawings may be executed in a computer system such as a set of computer-executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than here. Embodiment 1

[0044] As Figures 1-4 shown, this embodiment provides a landslide susceptibility assessment method based on multi-source data and deep learning, including the following steps:

[0045] Obtain high-resolution remote sensing images and digital elevation model data of the study area, and obtain surface deformation features and potential landslide point information through SBAS-InSAR technology, deep convolutional neural network, and YOLO algorithm;

[0046] Obtain the geological environment data of potential landslide points, and screen out the features of the geological environment data that have landslide influence factors. Based on the surface deformation features and the screened features, construct a landslide susceptibility assessment model through quantum Bayesian optimization;

[0047] Use the TRIGRS-SCOOPS3D coupling model to analyze the characteristics of rainfall-induced landslides, and obtain the stability of landslides and the spatial distribution of potential landslides under different rainfall conditions;

[0048] According to the stability of landslides and the spatial distribution of potential landslides under different rainfall conditions, simulate the three-dimensional sliding surface of the potential landslide area through the MASSFLOW model, calculate the landslide volume and potential landslide path, and obtain the landslide susceptibility assessment result.

[0049] Specifically, it includes the following steps:

[0050] I. Data preprocessing and landslide area identification:

[0051] SBAS-InSAR data processing:

[0052] Through SBAS-InSAR technology, use Sentinel-1 or other high-resolution satellite data to extract the surface deformation information of the study area. Through time-series image analysis, calculate the surface deformation velocity and its change trend of each pixel.

[0053] The data preprocessing steps include image registration, removing atmospheric effects, differential interferometric processing of time series, and phase unwrapping. The SBAS-InSAR method is used to calculate the surface deformation amount within consecutive time periods, with special attention paid to the high-deformation areas where landslides may exist.

[0054] Combination of remote sensing images and DEM data:

[0055] High-resolution remote sensing images (such as Google Earth images) and digital elevation model (DEM) data of the study area are obtained. Combining the surface features of the remote sensing images, potential landslide areas are evaluated.

[0056] Deep learning and visual interpretation:

[0057] A pre-trained deep convolutional neural network (ResNet-34) is used to extract features from remote sensing images, and the YOLO algorithm is used to detect and locate potential landslide points.

[0058] The data of SAR images and remote sensing images are fused, and a deep learning model is used to finely interpret the surface deformation area, so as to accurately identify potential landslide points.

[0059] (1) CNN architecture: A ResNet deep convolutional neural network architecture is selected for feature extraction. High-order features in the image, such as the texture, shape, color, etc. of the landslide area, are extracted through multiple convolutional layers and pooling layers.

[0060] Input layer: Input remote sensing optical images.

[0061] Convolutional layer: Multiple convolutional operations gradually extract features from low-level features (edges, textures) to high-level features (landslide morphology, deformation areas, etc.).

[0062] Pooling layer: The size of the feature map is reduced through pooling operations to reduce the computational complexity while retaining key features.

[0063] Fully connected layer: The features extracted by the convolutional and pooling layers are fully connected to generate landslide prediction results.

[0064] (2) YOLO architecture: YOLO is a deep learning-based object detection algorithm that can predict the position (bounding box) and category of objects in an image. This method is applicable to the positioning and detection of landslide areas.

[0065] ① Input layer: The features extracted by the CNN and the SBAS-InSAR surface deformation image are used as inputs.

[0066] ② Feature map generation: YOLO concatenates multiple grid cells in the image, and each cell is responsible for predicting the object bounding box and category within that area.

[0067] ③ Landslide detection: The model predicts the location of the landslide area (through bounding boxes) and the type of landslide (such as shallow landslide, deep landslide, etc.) in the image. YOLO can simultaneously perform object localization and classification in a single forward pass, with high efficiency.

[0068] SAR and optical images can provide different information. SAR images mainly provide surface deformation information, while optical images can provide richer surface texture and color information. Through deep model learning, these two types of information can be fused to improve the accuracy of landslide detection.

[0069] II. Screening of landslide susceptibility assessment factors and model establishment:

[0070] Selection of landslide susceptibility factors:

[0071] Combined with the geological environment of the study area, factors related to landslide susceptibility are selected, including: slope, aspect, soil type, vegetation coverage, seismic intensity, rainfall, groundwater level, etc. According to existing literature and expert opinions, other factors closely related to regional landslide occurrence are further supplemented, such as the impact of human activities.

[0072] Factor screening and correlation analysis:

[0073] Data preprocessing and model training: Standardize the data, use the Elastic Net regression model, set the alpha and l1_ratio parameters to balance L1 and L2 regularization, and train the model.

[0074] Factor screening and result analysis: Select the most influential features based on the model coefficients and evaluate the relationship between each feature and the target variable.

[0075] Construction of the landslide susceptibility assessment model:

[0076] Factor screening and model construction: Based on the screened factors, a landslide susceptibility assessment model is established using quantum Bayesian optimization and adaptive deep reinforcement learning (Deep Q Network, DQN). Through deep neural network learning of complex non-linear relationships, the strategy is adaptively adjusted to optimize the landslide prediction results.

[0077] DQN framework: Input layer: The number of nodes in the input layer is equal to the number of screened landslide influencing factors. Output layer: The nodes in the output layer represent the deformation rate of the landslide points, serving as the prediction results of the model.

[0078] State Space and Action Space: State Space (State): The input landslide characteristic factors (e.g., slope, precipitation, etc.), which reflect the current geological environment and external factors. Action Space (Action): Learn a policy through reinforcement learning to adjust the weights of the input characteristic factors, ultimately optimizing the prediction results.

[0079] Reward Function: Set the reward function based on the error between the prediction result and the actual observation value. If the predicted deformation rate is close to the actual observed value, a positive reward is given; if the error is large, a negative reward is given.

[0080] Training the Deep Neural Network: Use DQN to update the weights of the neural network through the backpropagation algorithm. During the training process, by interacting with the environment, the network learns the complex non-linear relationship between the characteristic factors and the deformation rate.

[0081] Experience Replay and Target Network: During the training process, use the experience replay mechanism to store the past state-action pairs (S, A) to avoid overfitting, and use the target network to stabilize the training process.

[0082] Policy Update: Continuously adjust the policy according to the reward function, and improve the accuracy of the prediction result through multiple trainings.

[0083] Weighted Decision and Optimization: Introduce quantum Bayesian optimization to automatically adjust the hyperparameters of the model (such as the learning rate, discount factor (gamma), number of network layers, number of nodes, exploration rate (epsilon) of DQN). According to the weights and influence degrees of different factors, optimize the model parameters and improve the accuracy of the evaluation results. Quantum computing accelerates the parameter optimization process and improves the efficiency of high-dimensional space search.

[0084] Initializing the Optimization Space: Define the optimization objective and hyperparameter space, such as the learning rate, number of layers, number of nodes in each layer, etc. in the deep neural network (DNN).

[0085] Using Quantum Computing to Accelerate Bayesian Optimization: Use quantum computing methods (such as quantum convolutional neural network or quantum support vector machine) to accelerate the hyperparameter tuning process, thereby optimizing the prediction accuracy of the landslide susceptibility assessment model.

[0086] Model Stability and Performance Evaluation: To ensure the stability of the model, use cross-validation and AUC (Area Under Curve) to evaluate the model performance. Generate a landslide susceptibility zoning map based on the model training results.

[0087] III. Analysis of Rainfall-Induced Landslide Characteristics and Simulation of Landslide Process:

[0088] TRIGRS-SCOOPS3D Coupled Model:

[0089] The TRIGRS model is used to analyze the stability of rainfall-induced landslides in the study area. The rainfall forecast information, soil type, groundwater level and other data of the area are input, and the stability of the slope under different rainfall conditions is calculated through the model.

[0090] The SCOOPS3D module was used to search for three-dimensional landslides, study the changing process of rainfall-induced landslides, and obtain the spatial distribution of potential landslides in the region.

[0091] MASSFLOW model applications:

[0092] According to the output results of the TRIGRS-SCOOPS3D model, the MASSFLOW model is used to simulate the three-dimensional sliding surface of the potential landslide area in the region. The MASSFLOW model can simulate the flow process of mass landslides under the action of rainfall, and calculate the volume, flow velocity, landslide path, and dangerous areas of the landslide.

[0093] The dynamic process of rainfall-induced mass landslides can be predicted through model simulation, providing early warning data for subsequent disaster prevention and control.

[0094] Model parameter optimization and simulation verification:

[0095] The simulation results of the model are verified by using field observation data and historical landslide events. The reliability of the model prediction results is ensured by adjusting the parameters.

[0096] Combined with real-time rainfall forecasts and meteorological data, the model is dynamically adjusted to achieve real-time prediction of the landslide process.

[0097] 4. Real-time presentation of landslide risk areas and disaster warning:

[0098] Spatial Visualization of Landslide Prone Areas:

[0099] The results of the landslide susceptibility assessment model are converted into a spatial distribution map through the ArcGIS platform, showing potential landslide areas and high-risk areas. The spatial accuracy of the risk area is further improved by combining geological data and remote sensing images.

[0100] Dynamically update and display the distribution of landslide-prone areas to support regional landslide risk management and emergency response.

[0101] Real-time landslide disaster warning system:

[0102] Based on deep learning models and big data analysis, a real-time disaster warning system is established, which can predict the occurrence of potential landslides based on real-time meteorological data and surface deformation information.

[0103] When the early warning system detects an increased landslide risk, it will promptly issue warning information and provide real-time risk assessment to relevant departments and the public to facilitate the adoption of disaster prevention measures.

[0104] 5. System architecture and implementation environment:

[0105] Data platform and computing environment:

[0106] The implementation of the method of the present invention requires the combination of remote sensing image data, geographic information system (GIS), meteorological data and large-scale computing platform. The data collection, preprocessing and analysis process requires the use of cloud computing platform and high-performance computing environment to handle large data volumes and complex computing models.

[0107] Establish a unified data storage and management system to support real-time updating and automated processing of data.

[0108] Software tools and technical support:

[0109] The deep learning model and landslide assessment model used in this method can be implemented using programming languages ​​such as Python. The deep learning model can be trained based on frameworks such as TensorFlow and PyTorch, and ArcGIS can be used for spatial data processing and visualization.

[0110] By integrating the calculation results of TRIGRS-SCOOPS3D and MASSFLOW models, a landslide risk assessment and disaster warning system is established to ensure the real-time and high efficiency of the model.

[0111] This embodiment further provides a computer device, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method.

[0112] This embodiment also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the method are implemented. Embodiment 2

[0113] This embodiment provides a landslide susceptibility assessment system based on multi-source data and deep learning, including:

[0114] Potential landslide point information acquisition module, which is used to obtain high-resolution remote sensing images and digital elevation model data of the study area, and obtain surface deformation characteristics and potential landslide point information through SBAS-InSAR technology, deep convolutional neural network and YOLO algorithm;

[0115] The landslide susceptibility assessment model construction module is used to obtain the geological environment data of potential landslide points, screen out the characteristics of the geological environment data with landslide influence factors, and construct a landslide susceptibility assessment model through quantum Bayesian optimization based on the surface deformation characteristics and the screened-out characteristics;

[0116] The rainfall factor analysis module is used to analyze the characteristics of rainfall-induced landslides using the TRIGRS-SCOOPS3D coupling model, and obtain the stability of landslides and the spatial distribution of potential landslides under different rainfall conditions;

[0117] The evaluation module is used to simulate the three-dimensional slip surface of the potential landslide area through the MASSFLOW model according to the stability of landslides and the spatial distribution of potential landslides under different rainfall conditions, calculate the landslide volume and the potential landslide path, and obtain the landslide susceptibility assessment result.

[0118] Furthermore, the potential landslide point information acquisition module includes:

[0119] The surface deformation information calculation unit is used to calculate the surface deformation of the study area in the satellite image through the SBAS-InSAR technology, and obtain the surface deformation information of the study area;

[0120] The convolutional feature extraction and acquisition unit is used to obtain the high-resolution remote sensing image and digital elevation model data of the study area, and perform feature extraction through a pre-trained deep convolutional neural network to obtain the convolutional extracted features;

[0121] The positioning unit is used to detect and locate potential landslide points through the YOLO algorithm according to the surface deformation information and the convolutional extracted features.

[0122] Furthermore, the landslide susceptibility assessment model construction module includes:

[0123] The geological environment data acquisition unit is used to obtain the geological environment data of potential landslide points, where the geological environment data includes slope, aspect, soil type, vegetation coverage, seismic intensity, rainfall, and groundwater level;

[0124] The screening unit is used to screen out the characteristics with landslide influence factors from the geological environment data through the Elastic Net regression model to obtain the screened features;

[0125] The construction unit is used to construct a landslide susceptibility assessment model through quantum Bayesian optimization and adaptive deep reinforcement learning according to the surface deformation characteristics and the screened features.

[0126] Furthermore, the system also includes: a display platform, which is used to convert the landslide susceptibility assessment result into a spatial distribution map, present the potential landslide area and the high-risk area, and send out a warning message when the landslide risk is detected to increase.

[0127] The above are only the preferred specific embodiments of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by those skilled in the art within the technical scope disclosed in the present application should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A landslide susceptibility assessment method based on multi-source data and deep learning, characterized in that, It includes the following steps: Obtain high-resolution remote sensing images and digital elevation model data of the study area, and acquire surface deformation characteristics and potential landslide point information through SBAS-InSAR technology, deep convolutional neural network, and YOLO algorithm; Obtain the geological environment data of potential landslide points, and screen out the characteristics of the geological environment data with landslide influence factors. Based on the surface deformation characteristics and the screened-out characteristics, construct a landslide susceptibility assessment model through quantum Bayesian optimization; Use the TRIGRS-SCOOPS3D coupling model to analyze the characteristics of rainfall-induced landslides, and obtain the stability of landslides and the spatial distribution of potential landslides under different rainfall conditions; According to the stability of landslides and the spatial distribution of potential landslides under different rainfall conditions, simulate the three-dimensional slip surface of the potential landslide area through the MASSFLOW model, calculate the landslide volume and potential landslide path, and obtain the landslide susceptibility assessment result; Obtaining surface deformation characteristics and potential landslide point information includes: Calculate the surface deformation of the study area in the satellite image through SBAS-InSAR technology to obtain the surface deformation information of the study area; Obtain high-resolution remote sensing images and digital elevation model data of the study area, and perform feature extraction through a pre-trained deep convolutional neural network to obtain convolutionally extracted features; Detect and locate potential landslide points through the YOLO algorithm according to the surface deformation information and the convolutionally extracted features; Constructing a landslide susceptibility assessment model includes: Obtain the geological environment data of potential landslide points, where the geological environment data includes slope, aspect, soil type, vegetation coverage, seismic intensity, rainfall, and groundwater level; Screen out the characteristics with landslide influence factors from the geological environment data through the Elastic Net regression model to obtain the screened-out characteristics; Construct a landslide susceptibility assessment model through quantum Bayesian optimization and adaptive deep reinforcement learning according to the surface deformation characteristics and the screened-out characteristics.

2. The method according to claim 1, wherein The method further includes: converting the landslide susceptibility assessment result into a spatial distribution map, presenting potential landslide areas and high-risk areas, and sending out warning information when an increase in landslide risk is detected.

3. A landslide susceptibility assessment system based on multi-source data and deep learning, characterized in that, It includes: A potential landslide point information acquisition module, which is used to obtain high-resolution remote sensing images and digital elevation model data of the study area, and acquire surface deformation characteristics and potential landslide point information through SBAS-InSAR technology, deep convolutional neural network, and YOLO algorithm; A landslide susceptibility assessment model construction module, which is used to obtain the geological environment data of potential landslide points, and screen out the characteristics of the geological environment data with landslide influence factors. Based on the surface deformation characteristics and the screened-out characteristics, construct a landslide susceptibility assessment model through quantum Bayesian optimization; A rainfall factor analysis module, which is used to use the TRIGRS-SCOOPS3D coupling model to analyze the characteristics of rainfall-induced landslides, and obtain the stability of landslides and the spatial distribution of potential landslides under different rainfall conditions; An evaluation module, configured to simulate the three-dimensional slip surface of the potential landslide area through the MASSFLOW model according to the stability of the landslide and the spatial distribution of potential landslides under different rainfall conditions, calculate the landslide volume and the potential landslide path, and obtain the landslide susceptibility evaluation result; The potential landslide point information acquisition module includes: A surface deformation information calculation unit, configured to calculate the surface deformation of the study area in the satellite image through the SBAS-InSAR technology, and obtain the surface deformation information of the study area; A convolutional extraction feature acquisition unit, configured to acquire the high-resolution remote sensing image and digital elevation model data of the study area, and perform feature extraction through a pre-trained deep convolutional neural network to obtain the convolutional extraction feature; A positioning unit, configured to detect and locate potential landslide points through the YOLO algorithm according to the surface deformation information and the convolutional extraction feature; The landslide susceptibility evaluation model construction module includes: A geological environment data acquisition unit, configured to acquire the geological environment data of the potential landslide point, where the geological environment data includes slope, aspect, soil type, vegetation coverage, seismic intensity, rainfall, and groundwater level; A screening unit, configured to screen out the features with landslide influence factors from the geological environment data through the Elastic Net regression model to obtain the screened features; A construction unit, configured to construct a landslide susceptibility evaluation model through quantum Bayesian optimization and adaptive deep reinforcement learning according to the surface deformation features and the screened features.

4. The system according to claim 3, characterized in that, The system further includes: a display platform, configured to convert the landslide susceptibility evaluation result into a spatial distribution map, present the potential landslide area and the high-risk area, and send out a warning message when the detected landslide risk increases.

5. A computer device, comprising a memory, a processor, and a computer program stored on the memory, characterized in that, The processor executes the computer program to implement the steps of the method according to any one of claims 1-2.

6. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, the steps of the method according to any one of claims 1-2 are implemented.

Images