Landslide sensitive area dynamic identification method and device supported by multi-source data

Abstract

The invention provides a landslide sensitive area dynamic identification method and device supported by multi-source data, and relates to the technical field of geological prospecting, and the method comprises the steps: obtaining landslide related multi-source data of a target monitoring area, the landslide related multi-source data comprising meteorological data, geological data and historical landslide data; preprocessing the landslide-related multi-source data to obtain landslide-related feature data; generating a dynamic rainfall threshold value by using a rainfall threshold value prediction model based on the landslide related characteristic data; calculating a landslide sensitivity index according to the dynamic rainfall threshold; carrying out regional division on the target monitoring region according to the landslide sensitivity index to generate a sensitive region distribution diagram; constructing a multi-feature hidden danger prediction model based on the sensitive area distribution map and the landslide related feature data; and dynamically identifying a landslide sensitive area by using the multi-feature hidden danger prediction model. According to the invention, the timeliness and accuracy of landslide risk monitoring can be significantly improved.

Description

Background Art

[0002] Landslides are a common geological disaster that are sudden, destructive, and have a wide range of impacts. The inducing factors are complex and diverse, and are closely related to geological conditions, meteorological conditions, human activities, and other factors. In landslide disaster prevention and control, accurate identification and dynamic monitoring of landslide-sensitive areas are the key to disaster warning and risk management. However, existing technologies still have many deficiencies in identifying landslide-sensitive areas.

[0003] First, existing methods mostly rely on static data analysis, such as using geological data and historical landslide data to divide fixed areas. This method fails to fully consider the dynamic impact of real-time meteorological factors such as rainfall intensity and rainfall on landslide triggering, resulting in insufficient timeliness in the division of landslide-sensitive areas. In addition, due to the limited ability to integrate static and dynamic features, existing technologies are difficult to accurately predict the real-time changes in landslide risks, especially in the case of short-term heavy rainfall or continuous rainfall, which can easily lead to a lag in landslide risk warning.

[0004] Secondly, the triggering mechanism of landslide hazards is complex, involving multiple factors such as rainfall characteristics (such as rainfall intensity, cumulative rainfall, rainfall duration), geological conditions (such as soil thickness, slope, vegetation coverage), etc. Existing technologies often use single factor analysis in feature modeling, and fail to effectively integrate multi-source data to capture the complex interactive relationship between various features, thus affecting the accuracy of hazard prediction. In addition, in traditional landslide risk assessment methods, regional division and hazard prediction are mostly separate processes, making it difficult to form a complete identification closed loop and unable to achieve dynamic adjustment of sensitive areas.

[0005] With the development of sensor technology, remote sensing technology and big data technology, the acquisition methods and accuracy of landslide-related data have been significantly improved. However, the processing and integration of these multi-source data face challenges. How to use multi-source data to dynamically assess landslide risks, accurately divide sensitive areas, and adjust key monitoring areas in real time has become a technical problem that needs to be solved in the field of landslide disaster prevention and control. Summary of the invention

[0006] In order to overcome the problems existing in the related art, the present invention provides a method and device for dynamically identifying landslide sensitive areas supported by multi-source data.

[0007] According to a first aspect of an embodiment of the present invention, a method for dynamic identification of landslide-sensitive areas supported by multi-source data is provided, comprising:

[0008] Acquire multi-source data related to landslides in the target monitoring area, wherein the multi-source data related to landslides include meteorological data, geological data, and historical landslide data;

[0009] Preprocess the multi-source data related to the landslide to obtain landslide-related feature data;

[0010] Based on the landslide-related feature data, use a rainfall threshold prediction model to generate a dynamic rainfall threshold;

[0011] Calculate the landslide sensitivity index according to the dynamic rainfall threshold;

[0012] Divide the target monitoring area according to the landslide sensitivity index to generate a sensitive area distribution map;

[0013] Based on the sensitive area distribution map and the landslide-related feature data, construct a multi-feature hidden danger prediction model;

[0014] Use the multi-feature hidden danger prediction model to dynamically identify landslide-sensitive areas.

[0015] In some exemplary embodiments of the present invention, based on the foregoing solution, the sensitive area distribution map includes a high-sensitivity area;

[0016] After dividing the target monitoring area according to the landslide sensitivity index to generate a sensitive area distribution map, the dynamic landslide-sensitive area identification method supported by the multi-source data further includes:

[0017] Based on the sensitive area distribution map, when the high-sensitivity area reaches a preset condition, dynamically adjust the sensor layout density of the high-sensitivity area to generate a new sensor network;

[0018] Use the new sensor network to obtain new geological data and new meteorological data;

[0019] Update the new geological data and the new meteorological data as the landslide-related multi-source data and perform iteration.

[0020] In some exemplary embodiments of the present invention, based on the foregoing solution, the landslide-related feature data includes cumulative rainfall and real-time rainfall intensity;

[0021] The preset conditions include:

[0022] The cumulative rainfall in the high-sensitivity area reaches a preset percentage of the dynamic rainfall threshold;

[0023] The real-time rainfall in the high-sensitivity area is greater than a preset threshold; or

[0024] The area of the high-sensitivity area is greater than 2 times the pre-adjusted sensor coverage area.

[0025] In some exemplary embodiments of the present invention, based on the foregoing solution, the sensitive area distribution map further includes a low-sensitivity area;

[0026] Dynamically adjusting the sensor layout density in the highly sensitive area includes:

[0027] Deploying the sensors in the low sensitive area to the highly sensitive area; or

[0028] Using a drone to drop new sensors into the highly sensitive area.

[0029] In some exemplary embodiments of the present invention, based on the foregoing solution, the sensitive area distribution map further includes a medium sensitive area;

[0030] Wherein, the landslide sensitivity index in the highly sensitive area is greater than a first preset value; the landslide sensitivity index in the medium sensitive area is less than the first preset value and greater than a second preset value; the landslide sensitivity index in the low sensitive area is less than the second preset value.

[0031] In some exemplary embodiments of the present invention, based on the foregoing solution, the rainfall threshold prediction model includes an input layer, a feature extraction part, a dynamic weight fusion part, a rainfall threshold prediction part, and an output layer;

[0032] The feature extraction part includes a first time series feature extraction module and a first geological feature extraction module. The first time series feature extraction module includes a time series transformer network architecture for extracting the hourly rainfall intensity, cumulative rainfall, rainfall intermittent time, temperature feature, humidity feature, and wind speed feature in the landslide-related feature data and generating time series features. The first geological feature extraction module includes a multi-layer perceptron and a graph convolutional network. The multi-layer perceptron is used to encode the landslide-related feature data into high-dimensional features and use the high-dimensional features as the input of the graph convolutional network. The graph convolutional network is used to model the high-dimensional features into a graph structure and capture the spatial relationship of the high-dimensional features to generate static features;

[0033] The dynamic weight fusion part is used to generate dynamic weights of the time series features and the static features by using an attention mechanism, and generate fusion features according to the time series features, the static features, and the dynamic weights;

[0034] The rainfall threshold prediction part includes a deep regression network for generating a dynamic rainfall threshold according to the fusion features and outputting the dynamic rainfall threshold through the output layer.

[0035] In some exemplary embodiments of the present invention, based on the foregoing solution, the multi-feature hidden danger prediction model includes: an input module, a feature extraction and analysis module, a multi-feature fusion module, a hidden danger prediction module, and an output module;

[0036] The feature extraction and analysis part includes a second time-series feature extraction module, a second geological feature extraction module, and a spatial feature extraction module. The second time-series feature extraction module includes a time-series transformer network architecture for extracting the hourly rainfall intensity, cumulative rainfall, rainfall intermittent time, temperature feature, humidity feature, and wind speed feature in the landslide-related feature data and generating time-series features. The second geological feature extraction module includes a multi-layer perceptron and a graph convolutional network. The multi-layer perceptron is used to encode the landslide-related feature data into high-dimensional features and use the high-dimensional features as the input of the graph convolutional network. The graph convolutional network is used to model the high-dimensional features as a graph structure and capture the spatial relationship of the high-dimensional features to generate static features. The spatial feature extraction module includes a convolutional neural network for extracting the spatial features of the sensitive area distribution map.

[0037] The multi-feature fusion module is used to perform feature fusion on the time-series features, the static features, and the spatial features to generate comprehensive features. The hidden danger prediction module is used to predict the landslide hidden danger level based on the comprehensive features and output through the output module.

[0038] According to the second aspect of the embodiments of the present invention, there is provided an apparatus for a landslide sensitive area dynamic recognition method based on the above multi-source data support, including:

[0039] A data acquisition module for acquiring multi-source data related to landslides in a target monitoring area, where the multi-source data related to landslides includes meteorological data, geological data, and historical landslide data;

[0040] A preprocessing module for preprocessing the multi-source data related to landslides to obtain landslide-related feature data;

[0041] A threshold generation module for generating a dynamic rainfall threshold based on the landslide-related feature data using a rainfall threshold prediction model;

[0042] An index calculation module for calculating a landslide sensitivity index according to the dynamic rainfall threshold;

[0043] A distribution map generation module for dividing the target monitoring area according to the landslide sensitivity index to generate a sensitive area distribution map;

[0044] A model construction module for constructing a multi-feature hidden danger prediction model based on the sensitive area distribution map and the landslide-related feature data;

[0045] A dynamic recognition module for dynamically recognizing landslide sensitive areas using the multi-feature hidden danger prediction model.

[0046] According to a third aspect of an embodiment of the present invention, there is provided an electronic device, including: a processor; and a memory, on which computer-readable instructions are stored, and when the computer-readable instructions are executed by the processor, the method in the first aspect is implemented.

[0047] According to a fourth aspect of an embodiment of the present invention, there is provided a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the method in the first aspect is implemented.

[0048] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects:

[0049] 1. The present invention obtains multi-source data related to landslides (meteorological data, geological data, historical landslide data) in a target monitoring area, and preprocesses them to obtain landslide-related feature data. Through the extraction and standardization of these feature data, the integrity and consistency of the model input data can be ensured, and the problems of diverse data sources and complex formats in landslide risk identification can be solved.

[0050] 2. Based on the landslide-related feature data, the present invention uses a rainfall threshold prediction model to generate a dynamic rainfall threshold. By generating a real-time dynamic rainfall threshold, the risk change of rainfall-induced landslides in the region can be accurately reflected, thus solving the problem that it is difficult to respond to rainfall changes in real time in traditional landslide warning methods, and improving the timeliness and accuracy of landslide risk warning.

[0051] 3. On the basis of generating the dynamic rainfall threshold, the present invention calculates the sensitivity index to quantitatively evaluate the landslide risk in the target monitoring area. The sensitivity index combines the dynamic rainfall threshold and geological characteristics, and describes the landslide risk through a unified quantitative index, which can effectively overcome the technical bottleneck that it is difficult to quantitatively fuse multiple influencing factors. Based on the sensitivity index, the present invention divides the target monitoring area, and generates a sensitive area distribution map. This distribution map presents the spatial distribution of landslide risks in the form of high, medium, and low sensitive areas, and can intuitively reflect the risk levels of different areas. This risk division method avoids the deficiencies of a single regional division standard and an unintuitive division result in traditional methods, and provides an important basis for the scientific allocation of landslide monitoring resources.

[0052] 4. Combining the generated sensitive area distribution map and the landslide-related feature data, the present invention further constructs a multi-feature hidden danger prediction model. By fusing static features (such as geological data), dynamic features (such as rainfall data), and spatial features (such as the sensitive area distribution map), this prediction model can deeply explore the induction mechanism of landslide hidden dangers and capture the complex interaction relationships between multiple features. Compared with the hidden danger prediction method based on a single feature, the multi-feature fusion model of the present invention can more comprehensively describe the triggering conditions of landslide hidden dangers, and significantly improve the accuracy and reliability of hidden danger prediction.

[0053] 5. The present invention realizes the dynamic identification of landslide sensitive areas by using a multi-feature hidden danger prediction model. When the risk level changes, the prediction model can dynamically adjust the sensitive area distribution map and update the landslide risk information to achieve real-time identification of the key monitoring areas. The introduction of the dynamic identification ability not only improves the accuracy and timeliness of landslide risk monitoring, but also optimizes the utilization efficiency of monitoring resources, providing scientific support for landslide disaster warning and emergency response.

[0054] It should be understood that the above general description and the following detailed description are only exemplary and explanatory, and cannot limit the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] The accompanying drawings herein are incorporated into the specification and constitute a part of the present invention, showing embodiments consistent with the present invention and, together with the specification, used to explain the principles of the present invention.

[0056] Figure 1 A schematic diagram of the system architecture showing an exemplary application environment of a method and apparatus for dynamically identifying landslide sensitive areas supported by multi-source data to which embodiments of the present invention can be applied;

[0057] Figure 2 A schematic flowchart showing the method for dynamically identifying landslide sensitive areas supported by multi-source data according to some embodiments of the present invention;

[0058] Figure 3 A schematic structural diagram showing a rainfall threshold prediction model according to some embodiments of the present invention;

[0059] Figure 4 A schematic structural diagram showing a multi-feature hidden danger prediction model according to some embodiments of the present invention;

[0060] Figure 5 A schematic diagram showing a device for dynamically identifying landslide sensitive areas supported by multi-source data according to some embodiments of the present invention;

[0061] Figure 6 A schematic structural diagram showing a computer system of an electronic device according to some embodiments of the present invention;

[0062] Figure 7 A schematic diagram showing a computer-readable storage medium according to some embodiments of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0063] Exemplary embodiments will be described in detail herein, and examples thereof are shown in the accompanying drawings. When the following description refers to the accompanying drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. On the contrary, they are merely examples of devices and methods consistent with some aspects of the present invention as detailed in the appended claims.

[0064] The terms used in the present invention are for the purpose of describing particular embodiments only and are not intended to limit the present invention. The singular forms "a", "the", and "said" used in the present invention and the appended claims are also intended to include the plural forms unless the context clearly dictates otherwise. It should also be understood that the term "and / or" as used herein refers to and encompasses any and all possible combinations of one or more of the associated listed items.

[0065] It should be understood that although the terms first, second, third, etc. may be used in the present invention to describe various information, such information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of the present invention, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the word "if" as used herein may be interpreted as "when" or "while" or "in response to determining".

[0066] Figure 1 A schematic diagram of a system architecture of an exemplary application environment of a method and apparatus for dynamically identifying landslide-sensitive areas supported by multi-source data to which embodiments of the present invention can be applied is shown.

[0067] As Figure 1 shown, the system architecture 100 may include one or more of terminal devices such as a desktop computer 101, a portable computer 102, a smart phone 103, etc., a network 104, and a server 105. The network 104 is used to provide a medium for a communication link between the terminal device and the server 105. The network 104 may include various connection types, such as wired, wireless communication links, or fiber optic cables, etc. The terminal device may be various electronic devices having data processing functions, and the electronic device has a display screen for presenting the dynamic identification result of the landslide-sensitive area supported by multi-source data to the user, including but not limited to the above-mentioned desktop computer, portable computer, smart phone, etc. It should be understood that Figure 1 the numbers of the terminal devices, the network, and the server in

[0068] The method for dynamically identifying landslide sensitive areas supported by multi-source data provided by the embodiments of the present invention can generally be executed by a terminal device. Correspondingly, the device for dynamically identifying landslide sensitive areas supported by multi-source data is generally set in the terminal device. However, those skilled in the art can easily understand that the method for dynamically identifying landslide sensitive areas supported by multi-source data provided by the embodiments of the present invention can also be executed by the server 105. Correspondingly, the device for dynamically identifying landslide sensitive areas supported by multi-source data can also be set in the server 105. No special limitation is made in this exemplary embodiment.

[0069] In addition, it should be understood that the method for dynamically identifying landslide sensitive areas supported by multi-source data in the embodiments of the present invention can be configured as a software module. In some implementation scenarios, the solution for dynamically identifying landslide sensitive areas supported by multi-source data of the present invention can be deployed independently to dynamically identify landslide sensitive areas in different target monitoring areas. In other implementation scenarios, the solution for dynamically identifying landslide sensitive areas supported by multi-source data of the present invention can be deployed in other software as a functional module of the software. For example, it can be deployed in the analysis software of underground pipelines. The present invention does not make special restrictions on the application manner of the method for dynamically identifying landslide sensitive areas supported by multi-source data.

[0070] Next, the embodiments of the present invention will be described in detail.

[0071] As Figure 2 shown, Figure 2 is a flowchart of a method for dynamically identifying landslide sensitive areas supported by multi-source data according to an exemplary embodiment of the present invention, including the following steps:

[0072] S210: Obtain multi-source data related to landslides in the target monitoring area, where the multi-source data related to landslides includes meteorological data, geological data, and historical landslide data;

[0073] S220: Preprocess the multi-source data related to landslides to obtain landslide-related feature data;

[0074] S230: Based on the landslide-related feature data, use a rainfall threshold prediction model to generate a dynamic rainfall threshold;

[0075] S240: Calculate the landslide sensitivity index according to the dynamic rainfall threshold;

[0076] S250: Divide the target monitoring area according to the landslide sensitivity index to generate a sensitive area distribution map;

[0077] S260: Based on the sensitive area distribution map and the landslide-related feature data, construct a multi-feature hidden danger prediction model;

[0078] S270: Dynamically identify landslide-sensitive areas using the multi-feature hidden danger prediction model.

[0079] In S210, obtain multi-source data related to landslides in the target monitoring area, where the multi-source data related to landslides includes meteorological data, geological data, and historical landslide data.

[0080] In some embodiments, meteorological data is used to reflect the inducing conditions of landslides, including rainfall, rainfall intensity, rainfall duration, temperature, humidity, wind speed, etc.; geological data is used to reflect the geological characteristics and stability of the landslide area, including slope gradient, soil thickness, soil type, lithology, groundwater level, etc.; historical landslide data is used to provide the historical risk characteristics of the landslide area, including the occurrence time, occurrence location, scale, inducing factors, etc. of historical landslide events.

[0081] These multi-source data related to landslides can be collected through technologies such as meteorological stations, sensor networks, and remote sensing, and transmitted to the data acquisition module of the present invention, so that the multi-source data related to landslides of the present invention has diversity and richness, and further makes the dynamic identification result of landslide-sensitive areas generated based on the multi-source data related to landslides more authentic.

[0082] In S220, preprocess the multi-source data related to landslides to obtain landslide-related feature data.

[0083] Here, the preprocessing process can be operations such as cleaning, transforming, extracting, and standardizing the multi-source data related to landslides to make it standardized and normalized.

[0084] In some embodiments, the data cleaning operation may include:

[0085] Check for missing values in the meteorological data (such as missing rainfall records), and complete them using interpolation or nearest neighbor filling; detect outliers in the geological data (such as slope exceeding the reasonable range), and remove or replace the abnormal data points; delete duplicate records, especially in historical landslide data, to avoid redundant information interfering with model calculations.

[0086] In some embodiments, the data transformation operation may include:

[0087] Standardize rainfall, rainfall intensity, etc. in the meteorological data to a unified time scale (such as hourly data); perform unit conversion on geological data (such as slope, soil thickness) to ensure that all data is used under the same measurement unit; convert historical landslide data (such as occurrence time, scale) into numerical features for convenient direct processing by the model.

[0088] In some embodiments, the feature extraction operation may include:

[0089] Meteorological data features: Extract key features such as hourly rainfall intensity, cumulative rainfall, and rainfall duration; generate time series features of sliding windows to capture the dynamic changes of rainfall.

[0090] Geological data features: Extract features that have a greater impact on landslide stability, such as slope, soil thickness, and lithology; combine geological features with spatial distribution to form spatial attribute features.

[0091] Historical landslide features: Extract features such as the occurrence frequency, scale, and distribution area of historical landslide events; use statistical methods to generate the risk probability distribution of historical landslides.

[0092] In some embodiments, the standardization operation may include:

[0093] Use a normalization method (such as Min - Max Scaling) to map the feature values to the interval [0, 1]; standardize the distribution of feature values to conform to a normal distribution to improve the model training effect.

[0094] Through cleaning, transformation, and standardization, the problems of noise, missing values, and inconsistent formats in multi - source landslide - related data can be solved; and feature extraction can capture the key influencing factors related to landslide hazards, making the data more interpretable and predictive; the generated landslide - related feature data provides high - quality input for subsequent rainfall threshold prediction models and multi - feature hazard prediction models, ensuring the computational efficiency and prediction accuracy of the models.

[0095] In S230, based on the landslide - related feature data, use a rainfall threshold prediction model to generate a dynamic rainfall threshold.

[0096] Here, the rainfall threshold prediction model can be generated based on existing network models, such as Long Short - Term Memory (LSTM), Gated Recurrent Unit (GRU), etc.

[0097] However, considering that LSTM and GRU may have limitations in dealing with complex feature interactions and long - term dependencies. Therefore, in the embodiments of the present invention, referring to Figure 3 as shown, the rainfall threshold prediction model

[0098] In some exemplary embodiments of the present invention, based on the foregoing solution, the rainfall threshold prediction model includes an input layer, a feature extraction part, a dynamic weight fusion part, a rainfall threshold prediction part, and an output layer;

[0099] The feature extraction part includes a first temporal feature extraction module and a first geological feature extraction module. The first temporal feature extraction module includes a Temporal Transformer network architecture (abbreviated as Temporal Transformer) for extracting the hourly rainfall intensity, cumulative rainfall, rainfall intermittent time, temperature feature, humidity feature, and wind speed feature in the landslide-related feature data and generating temporal features. This process can be expressed as:

[0100]

[0101] where Q, K, and V are the query, key, and value matrices of the rainfall-related features in the landslide-related feature data of rainfall features, and d k represents the dimension of the key vector.

[0102] The first geological feature extraction module includes a Multi-Layer Perceptron (MLP) and a Graph Convolutional Network (GCN). The multi-layer perceptron is used to encode the landslide-related feature data to generate high-dimensional features and use the high-dimensional features as the input of the graph convolutional network. The graph convolutional network is used to model the high-dimensional features as a graph structure and capture the spatial relationship of the high-dimensional features to generate static features. This process is expressed as:

[0103] H( l+1 ) = σ(D -1 / 2 AD -1 / 2 H l W l )

[0104] where H (l+1) represents the feature matrix of the l+1 layer, H l represents the feature matrix of the l layer, σ represents the activation function, A represents the adjacency matrix of the graph, D represents the degree matrix, D -1 / 2 AD -1 / 2 represents the normalized adjacency matrix, and W l represents the weight matrix of the l layer.

[0105] The dynamic weight fusion part is used to generate the dynamic weights of the temporal features and the static features by using the attention mechanism. This process is expressed as:

[0106] α T = softmax(W T ·T)

[0107] α S = softmax(W S ·S)

[0108] Among them, α T represents the weight of the temporal feature T, and α S represents the weight of the static feature S. W T and W S respectively represent the weight matrices of the temporal feature T and the static feature S.

[0109] After that, according to the temporal feature, the static feature, and the dynamic weight, a fusion feature F r is generated. This process is expressed as:

[0110] F r =α T ·T + α S ·S

[0111] The rainfall threshold prediction part includes a deep regression network (Fully Connected Layers) for generating a dynamic rainfall threshold according to the fusion feature, and outputs the dynamic rainfall threshold through the output layer.

[0112] The loss function L MSE of the rainfall threshold prediction model of the present invention is:

[0113]

[0114] Among them, i represents the i-th sample, n represents the number of samples, y i represents the true value of sample i, represents the predicted value of sample i.

[0115] Since the temporal Transformer, MLP, GCN, and deep regression network are all existing technologies, the present invention will not elaborate on them, and they are not shown in the figure either.

[0116] By combining the temporal feature and the static feature, the spatio-temporal correlation between rainfall data and geological data can be fully mined. By dynamically adjusting the feature importance through the attention mechanism, the adaptability of the rainfall threshold prediction model in a complex environment can be improved; by introducing the joint modeling of the temporal Transformer and GCN, the global correlation analysis ability of rainfall and spatial features can be enhanced.

[0117] Through the time series transformer network and the dynamic weight mechanism, the rainfall threshold prediction model can reflect the impact of dynamic changes such as rainfall on landslide triggering conditions in real time, combine dynamic time series features with static geological features, and capture the complex multi-dimensional influencing factors of landslide triggering. The graph convolutional network effectively models the spatial relationship of geological features, improves the adaptability of the model to regional landslide risks, and the deep regression network can effectively process complex multi-source feature relationships. The generated dynamic rainfall threshold has high accuracy, providing scientific support for landslide risk assessment. Through the collaboration of multiple modules, the spatio-temporal information of dynamic meteorological features and static geological features is successfully integrated, solving the problems of insufficient rainfall threshold prediction accuracy and poor timeliness in the existing technology.

[0118] In S240, according to the dynamic rainfall threshold, calculate the Landslide Susceptibility Index (LSI).

[0119] The calculation process of LST is as follows:

[0120] LST = ω 1 ·T n + ω 2 ·S + ω 3 ·H

[0121] Where, T n represents the dynamic rainfall threshold, S represents the geological feature index (such as slope, soil thickness, lithology, etc.), H represents the historical landslide event feature index (such as historical landslide frequency, scale), and ω 1 、ω 2 、ω 3 represent the weight parameters of the dynamic rainfall threshold, geological feature index, and historical landslide event feature index respectively.

[0122] Before calculating the LST using the dynamic rainfall threshold, the dynamic rainfall threshold can also be normalized. Then, in the above formula, T n represents the normalized dynamic rainfall threshold.

[0123] Based on the dynamically updated dynamic rainfall threshold, the landslide susceptibility index can reflect the impact of rainfall changes on landslide risks in real time, significantly improving the response ability to sudden rainfall-induced landslides; by introducing geological features and historical landslide data, the calculation results not only depend on the rainfall threshold, but also comprehensively consider the regional geological stability and historical risk characteristics, and the results are more accurate and scientific.

[0124] In S250, according to the landslide susceptibility index, divide the target monitoring area into regions and generate a sensitive area distribution map.

[0125] Here, the monitoring area can be divided into: according to the LSI value

[0126] High-sensitivity area (LSI > first preset value); medium-sensitivity area (second preset value ≤ LSI ≤ first preset value); low-sensitivity area (LSI < second preset value).

[0127] The first preset value and the second preset value can be set accordingly according to the actual situation. For example, in some embodiments, the first preset value can be 0.8, 0.7 or 0.6, etc., and the second preset value can be 0.5, 0.4 or 0.3, etc.

[0128] Then, based on the division result, a sensitive area distribution map is dynamically generated using a Geographic Information System (GIS).

[0129] In S260, based on the sensitive area distribution map and the landslide-related feature data, a multi-feature hidden danger prediction model is constructed.

[0130] In some embodiments, the multi-feature hidden danger prediction model can be constructed and generated based on machine learning (such as XGBoost) or a neural network structure.

[0131] However, considering the accuracy and prediction precision of multi-feature hidden danger prediction, as shown in Figure 4 The multi-feature hidden danger prediction model provided by the present invention includes: an input module, a feature extraction and analysis module, a multi-feature fusion module, a hidden danger prediction module, and an output module;

[0132] The feature extraction and analysis part includes a second temporal feature extraction module, a second geological feature extraction module, and a spatial feature extraction module. The second temporal feature extraction module includes a Temporal Transformer network architecture (abbreviated as Temporal Transformer) for extracting the hourly rainfall intensity, cumulative rainfall, rainfall interval time, temperature feature, humidity feature, and wind speed feature in the landslide-related feature data and generating temporal features. The second geological feature extraction module includes a Multi-Layer Perceptron (MLP) and a Graph Convolutional Network (GCN). The multi-layer perceptron is used to encode the landslide-related feature data into high-dimensional features and use the high-dimensional features as the input of the graph convolutional network. The graph convolutional network is used to model the high-dimensional features into a graph structure and capture the spatial relationship of the high-dimensional features to generate static features. The spatial feature extraction module includes a Convolutional Neural Network (CNN) for extracting the spatial features of the sensitive area distribution map;

[0133] The multi - feature fusion module is based on the attention mechanism and is used to fuse the temporal features, the static features, and the spatial features to generate a comprehensive feature F f , and this process is expressed as:

[0134] F f = ReLU(W T F T + W S F S + W G F G + W TS (F T ⊙ F S )+ W TG (F T ⊙ F G )+ W SG (F S ⊙ F G )+ b)

[0135] Among them, W T , W S and W G respectively represent the weights of the temporal feature T, the static feature S, and the spatial feature G, F T , F S and F G respectively represent the temporal feature, the static feature, and the spatial feature, W TS represents the fusion weight of the temporal feature T and the static feature S, W TG represents the fusion weight of the temporal feature T and the spatial feature G, represents the fusion weight of the static feature S and the spatial feature G, ⊙ represents the element - by - element interaction between elements. For example, (F T ⊙ F S )[i, j]= F T [i, j]· F S [i, j], where i, j represent the elements in the matrix.

[0136] The hidden - danger prediction module is used to predict the landslide hidden - danger level based on the comprehensive feature, and this process can be expressed as:

[0137] S c = σ 1 (W· F f + b 1 )

[0138] Among them, S c represents the hidden - danger score, σ 1 represents the activation function, F f represents the comprehensive feature, b 1 represents the bias term.

[0139] And output through the output module.

[0140] The multi - feature fusion module can model the non - linear relationship between temporal features, static features, and spatial features, enhancing the model's understanding of complex triggering mechanisms. Compared with simple weighted summation, the interactive fusion idea adopted in the present invention can more deeply depict the correlation between features, generating more expressive comprehensive features to provide more accurate comprehensive features for landslide hazard prediction.

[0141] Early warning rules can also be set on this basis. For example, in some embodiments:

[0142] First - level early warning (low risk): LSI > 0.5 and the cumulative rainfall is close to the threshold.

[0143] Second - level early warning (medium risk): LSI > 0.7 and the real - time rainfall intensity is close to the threshold.

[0144] Third - level early warning (high risk): LSI > 0.8 and the sensor data exceeds the threshold (such as displacement > 5mm / hour).

[0145] In S270, the multi - feature hazard prediction model is used to dynamically identify landslide - sensitive areas.

[0146] Using dynamic meteorological data, the multi - feature hazard prediction model can adjust the identification results of landslide - sensitive areas in real - time, timely reflecting the impact of environmental changes on landslide risks.

[0147] In addition, in another embodiment of the present invention, after dividing the target monitoring area according to the landslide sensitivity index to generate a sensitive area distribution map, the multi - source data - supported dynamic landslide - sensitive area identification method further includes:

[0148] Based on the sensitive area distribution map, when the high - sensitive area reaches a preset condition, dynamically adjust the sensor layout density of the high - sensitive area to generate a new sensor network;

[0149] Using the new sensor network, obtain new geological data and new meteorological data;

[0150] Update the new geological data and the new meteorological data as the landslide - related multi - source data and perform iteration.

[0151] Here, the preset condition can be that the cumulative rainfall in the high - sensitive area reaches a preset percentage of the dynamic rainfall threshold. For example, the cumulative rainfall reaches 50%, 60%, 70%, or 80% of the dynamic rainfall threshold, and the present invention does not make a limitation.

[0152] It can also be that the real-time rainfall in the highly sensitive area is greater than a preset threshold, for example, the real-time rainfall reaches 40 mm / h, 50 mm / h, 60 mm / h; or it is set that the area of the highly sensitive area is greater than 2 times the sensor coverage area after pre-adjustment.

[0153] In some embodiments, the dynamic adjustment of the sensor layout density in the highly sensitive area can be:

[0154] Deploy the sensors in the low-sensitive area to the highly sensitive area; or

[0155] Use a drone to drop new sensors into the highly sensitive area.

[0156] Deploying the sensors in the low-sensitive area to the highly sensitive area can not only increase the sensor density in the highly sensitive area, but also reduce the number of sensors in the low-sensitive area to save resources. Of course, considering that the low-sensitive area can always be in the low-sensitive area, the number of sensors originally set in it is small. At this time, a drone can be used to drop new sensors into the highly sensitive area to achieve efficient and rapid sensor setting.

[0157] In the embodiments of the present invention, the sensors include soil moisture sensors, surface displacement monitoring devices, groundwater level sensors, etc. Initially, fixed sensors can be deployed in areas with complex geology and frequent historical landslides. 3 soil moisture sensors and 2 surface displacement monitoring points are set per square kilometer, and LoRa or NB-IoT communication methods are selected to enable long-term online operation of low-power sensors.

[0158] According to the second aspect of the embodiments of the present invention, there is also provided a landslide-sensitive area dynamic identification device 500 supported by multi-source data. Refer to Figure 5 As shown, the landslide-sensitive area dynamic identification device supported by multi-source data includes:

[0159] A data acquisition module 510, configured to acquire multi-source data related to landslides in a target monitoring area, where the multi-source data related to landslides includes meteorological data, geological data, and historical landslide data;

[0160] A preprocessing module 520, configured to preprocess the multi-source data related to landslides to obtain landslide-related feature data;

[0161] A threshold generation module 530, based on the landslide-related feature data, uses a rainfall threshold prediction model to generate a dynamic rainfall threshold;

[0162] An index calculation module 540, configured to calculate a landslide sensitivity index according to the dynamic rainfall threshold;

[0163] A distribution map generation module 550 is configured to divide the target monitoring area based on the landslide sensitivity index to generate a sensitive area distribution map;

[0164] A model construction module 560 is configured to construct a multi-feature hidden danger prediction model based on the sensitive area distribution map and the landslide-related feature data;

[0165] A dynamic recognition module 570 is configured to dynamically recognize landslide-sensitive areas by using the multi-feature hidden danger prediction model.

[0166] It should be noted that although several modules of the landslide-sensitive area dynamic recognition device 500 with multi-source data support are mentioned in the above detailed description, this division is not mandatory. In fact, according to the embodiments of the present invention, the features and functions of two or more of the above-described modules can be embodied in one module or unit. Conversely, the features and functions of one module described above can be further divided and embodied by multiple modules or sub-modules.

[0167] In addition, in an exemplary embodiment of the present invention, an electronic device capable of implementing a method for dynamically recognizing landslide-sensitive areas with multi-source data support is also provided.

[0168] Those skilled in the art can understand that various aspects of the present invention can be implemented as a system, a method, or a program product. Therefore, various aspects of the present invention can be specifically implemented in the following forms, namely: a complete hardware embodiment, a complete software embodiment (including firmware, microcode, etc.), or an embodiment combining hardware and software aspects, which can be collectively referred to as "circuits", "modules", or "systems" here.

[0169] The following refers to Figure 6 to describe the electronic device 600 according to this embodiment of the present invention. Figure 6 The illustrated electronic device 600 is merely an example and should not impose any limitation on the functions and usage scope of the embodiments of the present invention.

[0170] As Figure 6 shown, the electronic device 600 is presented in the form of a general-purpose computing device. The components of the electronic device 600 may include, but are not limited to: at least one of the above-mentioned processing units 610, at least one of the above-mentioned storage units 620, a bus 630 connecting different system components (including the storage unit 620 and the processing unit 610), and a display unit 640.

[0171] Among them, the storage unit stores program code, and the program code can be executed by the processing unit 610, so that the processing unit 610 executes the steps according to various exemplary embodiments of the present invention described in the above "exemplary method" part of the present invention. For example, the processing unit 610 can execute asFigure 2 As shown in S210, obtain multi-source data related to landslides in the target monitoring area, where the multi-source data related to landslides includes meteorological data, geological data, and historical landslide data; S220, preprocess the multi-source data related to landslides to obtain landslide-related feature data; S230, based on the landslide-related feature data, use a rainfall threshold prediction model to generate a dynamic rainfall threshold; S240, calculate a landslide sensitivity index according to the dynamic rainfall threshold; S250, divide the target monitoring area according to the landslide sensitivity index to generate a sensitive area distribution map; S260, based on the sensitive area distribution map and the landslide-related feature data, construct a multi-feature hidden danger prediction model; S270, use the multi-feature hidden danger prediction model to dynamically identify landslide-sensitive areas.

[0172] The storage unit 620 may include a readable medium in the form of a volatile storage unit, such as a random access storage unit (RAM) 621 and / or a cache storage unit 622, and may further include a read-only storage unit (ROM) 623.

[0173] The storage unit 620 may further include a program / utilities 624 having a set (at least one) of program modules 625. Such program modules 625 include, but are not limited to: an operating system, one or more application programs, other program modules, and program data. Each or some combination of these examples may include an implementation of a network environment.

[0174] The bus 630 may represent one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, a graphics acceleration port, a processing unit, or a local bus using any of the various bus structures.

[0175] The electronic device 600 can also communicate with one or more external devices 670 (such as a keyboard, a pointing device, a Bluetooth device, etc.), and can also communicate with one or more devices that enable a user to interact with the electronic device 600, and / or communicate with any device (such as a router, a modem, etc.) that enables the electronic device 600 to communicate with one or more other computing devices. Such communication can be carried out through the input / output (I / O) interface 650. Moreover, the electronic device 600 can also communicate with one or more networks (such as a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) through the network adapter 660. As shown in the figure, the network adapter 660 communicates with other modules of the electronic device 600 through the bus 630. It should be understood that, although not shown in the figure, other hardware and / or software modules can be used in combination with the electronic device 600, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems, etc.

[0176] Through the description of the above embodiments, those skilled in the art can easily understand that the exemplary embodiments described herein can be implemented by software, or can be implemented by the way of software combined with necessary hardware. Therefore, the technical solution according to the embodiments of the present invention can be embodied in the form of a software product, and the software product can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, including several instructions to enable a computing device (which can be a personal computer, a server, a terminal device, or a network device, etc.) to execute the method according to the embodiments of the present invention.

[0177] In an exemplary embodiment of the present invention, there is also provided a computer-readable storage medium, on which a program product capable of implementing the above method of the present invention is stored. In some possible embodiments, various aspects of the present invention can also be implemented in the form of a program product, which includes program code. When the program product runs on a terminal device, the program code is used to enable the terminal device to execute the steps according to various exemplary embodiments of the present invention described in the above "Exemplary Method" section of the present invention.

[0178] Reference Figure 7 As shown, a program product 700 for implementing the above-mentioned multi-source data-supported landslide-sensitive area dynamic identification method according to the embodiments of the present invention is described. It can adopt a portable compact disc read-only memory (CD-ROM) and include program code, and can run on a terminal device, such as a personal computer. However, the program product of the present invention is not limited thereto. In the present invention, the readable storage medium can be any tangible medium that contains or stores a program, and the program can be used by or in combination with an instruction execution system, device, or device.

[0179] The program product can adopt any combination of one or more readable storage media. The readable storage media can be, for example, but not limited to, systems, devices or components of electricity, magnetism, light, electromagnetic, infrared ray, or semiconductor, or any combination of the above. More specific examples of the readable storage media (a non-exhaustive list) include: electrical connections with one or more wires, portable disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the above.

[0180] The program code for performing the operations of the present invention can be written in any combination of one or more programming languages, including object-oriented programming languages such as Java, C++, etc., and also including conventional procedural programming languages such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computing device, partially on the user's device, executed as an independent software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server. In the case of a remote computing device, the remote computing device can be connected to the user's computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computing device (for example, by using an Internet service provider to connect through the Internet).

[0181] In addition, the above drawings are only schematic illustrations of the processes included in the method according to the exemplary embodiments of the present invention, rather than for limiting purposes. It is easy to understand that the processes shown in the above drawings do not indicate or limit the chronological order of these processes. Additionally, it is also easy to understand that these processes can be executed synchronously or asynchronously, for example, in multiple modules.

[0182] Through the description of the above embodiments, those skilled in the art can easily understand that the exemplary embodiments described herein can be implemented by software, or can be implemented by the way of software combined with necessary hardware. Therefore, the technical solution according to the embodiments of the present invention can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, including several instructions to enable a computing device (which can be a personal computer, a server, a touch terminal, or a network device, etc.) to execute the method according to the embodiments of the present invention.

[0183] Other embodiments of the present invention will be readily apparent to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. The present invention is intended to cover any variations, uses, or adaptations of the invention following the general principles of the invention and including known common knowledge or conventional technical means in the technical field not disclosed herein. The specification and examples are only illustrative, and the true scope and spirit of the invention are pointed out by the claims.

[0184] It should be understood that the present invention is not limited to the exact structures described above and shown in the drawings, and various modifications and changes can be made without departing from its scope. The scope of the present invention is only limited by the appended claims.

Claims

1. A dynamic identification method of landslide sensitive areas supported by multi-source data, characterized in that: include: Acquire multi-source data related to landslides in the target monitoring area, wherein the multi-source data related to landslides include meteorological data, geological data, and historical landslide data; Preprocessing the multi-source data related to the landslide to obtain characteristic data related to the landslide; Based on the landslide-related characteristic data, a dynamic rainfall threshold is generated using a rainfall threshold prediction model; calculating a landslide susceptibility index according to the dynamic rainfall threshold; Divide the target monitoring area into regions according to the landslide sensitivity index and generate a sensitive area distribution map; Based on the sensitive area distribution map and the landslide-related characteristic data, a multi-characteristic hidden danger prediction model is constructed; The multi-feature hazard prediction model is used to dynamically identify landslide-sensitive areas.

2. The method for dynamic identification of landslide-sensitive areas supported by multi-source data according to claim 1 is characterized in that: The sensitive area distribution map includes highly sensitive areas; After dividing the target monitoring area into regions according to the landslide sensitivity index and generating a sensitive area distribution map, the dynamic landslide sensitive area identification method supported by multi-source data further includes: Based on the sensitive area distribution map, when the highly sensitive area reaches a preset condition, dynamically adjusting the sensor deployment density of the highly sensitive area to generate a new sensor network; Using the new sensor network, new geological data and new meteorological data are obtained; The new geological data and the new meteorological data are updated to the landslide-related multi-source data and iterated.

3. The method for dynamic identification of landslide-sensitive areas supported by multi-source data according to claim 2 is characterized in that: The landslide-related characteristic data include accumulated rainfall and real-time rainfall intensity; The preset conditions include: The accumulated rainfall in the highly sensitive area reaches a preset percentage of the dynamic rainfall threshold; The real-time rainfall in the highly sensitive area is greater than a preset threshold; or The area of ​​the high-sensitivity region is greater than twice the pre-adjusted sensor coverage area.

4. The method for dynamic identification of landslide-sensitive areas supported by multi-source data according to claim 2 is characterized in that: The sensitive area distribution map also includes low-sensitivity areas; Dynamically adjusting the density of sensor deployment in the highly sensitive area includes: Deploy the sensor of the low-sensitivity area to the high-sensitivity area; or Use drones to drop new sensors into the highly sensitive areas.

5. The method for dynamic identification of landslide-sensitive areas supported by multi-source data according to any one of claims 2 to 4, characterized in that: The sensitive area distribution map also includes medium sensitive areas; Among them, the landslide sensitivity index of the high-sensitive area is greater than a first preset value; the landslide sensitivity index of the medium-sensitive area is less than the first preset value and greater than a second preset value; and the landslide sensitivity index of the low-sensitive area is less than the second preset value.

6. The method for dynamic identification of landslide-sensitive areas supported by multi-source data according to claim 1 is characterized in that: The rainfall threshold prediction model includes an input layer, a feature extraction part, a dynamic weight fusion part, a rainfall threshold prediction part and an output layer; The feature extraction part includes a first time series feature extraction module and a first geological feature extraction module. The first time series feature extraction module includes a time series transformer network architecture, which is used to extract hourly rainfall intensity, cumulative rainfall, rainfall interval time, temperature characteristics, humidity characteristics and wind speed characteristics from the landslide related feature data, and generate time series features; the first geological feature extraction module includes a multi-layer perceptron and a graph convolution network. The multi-layer perceptron is used to encode the landslide related feature data to generate high-dimensional features, and use the high-dimensional features as inputs of the graph convolution network. The graph convolution network is used to model the high-dimensional features as a graph structure, and capture the spatial relationship of the high-dimensional features to generate static features; The dynamic weight fusion part is used to generate the dynamic weights of the temporal features and the static features by using the attention mechanism, and generate fusion features according to the temporal features, the static features and the dynamic weights; The rainfall threshold prediction part includes a deep regression network, which is used to generate a dynamic rainfall threshold according to the fusion features; and output the dynamic rainfall threshold through the output layer.

7. The method for dynamic identification of landslide-sensitive areas supported by multi-source data according to claim 1 is characterized in that: The multi-feature hidden danger prediction model includes: an input module, a feature extraction and analysis module, a multi-feature fusion module, a hidden danger prediction module and an output module; The feature extraction and analysis part includes a second time series feature extraction module, a second geological feature extraction module and a spatial feature extraction module. The second time series feature extraction module includes a time series transformer network architecture, which is used to extract hourly rainfall intensity, cumulative rainfall, rainfall interval time, temperature characteristics, humidity characteristics and wind speed characteristics from the landslide-related feature data, and generate time series features; the second geological feature extraction module includes a multi-layer perceptron and a graph convolutional network, the multi-layer perceptron is used to encode the landslide-related feature data to generate high-dimensional features, and use the high-dimensional features as inputs of the graph convolutional network, the graph convolutional network is used to model the high-dimensional features as a graph structure, and capture the spatial relationship of the high-dimensional features to generate static features; the spatial feature extraction module includes a convolutional neural network, which is used to extract the spatial features of the sensitive area distribution map; The multi-feature fusion module is used to fuse the temporal features, the static features and the spatial features to generate comprehensive features; the hidden danger prediction module is used to predict the landslide hidden danger level based on the comprehensive features and output it through the output module.

8. A device for the method for dynamic identification of landslide sensitive areas supported by multi-source data according to any one of claims 1 to 7, characterized in that: include: A data acquisition module, used to acquire multi-source data related to landslides in a target monitoring area, wherein the multi-source data related to landslides include meteorological data, geological data, and historical landslide data; A preprocessing module, used for preprocessing the multi-source data related to the landslide to obtain the characteristic data related to the landslide; A threshold generation module generates a dynamic rainfall threshold based on the landslide-related characteristic data and using a rainfall threshold prediction model; An index calculation module, used to calculate a landslide susceptibility index according to the dynamic rainfall threshold; A distribution map generating module, used for dividing the target monitoring area into regions according to the landslide sensitivity index and generating a sensitive area distribution map; A model building module, used to build a multi-feature hidden danger prediction model based on the sensitive area distribution map and the landslide-related feature data; The dynamic identification module is used to dynamically identify landslide-sensitive areas using the multi-feature hidden danger prediction model.

9. An electronic device, comprising: processor; as well as A memory having computer-readable instructions stored thereon, wherein the computer-readable instructions, when executed by the processor, implement the method for dynamic identification of landslide-sensitive areas supported by multi-source data as claimed in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the method for dynamic identification of landslide-sensitive areas supported by multi-source data as claimed in any one of claims 1 to 7 is implemented.

Images