Landslide displacement space-time prediction method, device and system and storage medium

Abstract

The invention provides a landslide displacement space-time prediction method, device and system and a storage medium, and relates to the technical field of engineering geology, and the landslide displacement space-time prediction method comprises the steps: obtaining the historical landslide multi-field time sequence monitoring data of each monitoring point of a to-be-measured region, constructing a monitoring point bitmap based on the historical space related data, constructing a time sequence characteristic matrix according to the historical displacement time sequence, the historical landslide external disturbance factor time sequence and the monitoring point bitmap; inputting the time sequence characteristic matrix and the monitoring point bitmap into a trained landslide displacement space-time prediction model to obtain a landslide displacement space-time prediction result; the landslide displacement space-time prediction model is constructed based on an improved GCN network and an improved BiLSTM network; according to the method, the space-time prediction precision of landslide displacement can be remarkably improved by combining the space correlation data and the external disturbance factors and utilizing the capability of the deep learning model.

Description

Technical Field

[0001] The present invention relates to the technical field of engineering geology, and in particular to a method, device, system and storage medium for spatiotemporal prediction of landslide displacement. Background Art

[0002] Landslides are a common geological disaster, and their internal geological environment and external environmental factors are extremely complex. Under the influence of various external disturbance factors, landslides are prone to instability and destruction, posing a major threat to the life safety and property of surrounding residents. With the advancement of technology, many large-scale landslide areas have built three-dimensional monitoring and early warning systems. The massive landslide displacement time series monitoring data obtained through these systems has become an important basis for studying landslide behavior. Landslide deformation is a key indicator for landslide safety monitoring. Real-time monitoring of the deformation process of the landslide can record the entire process information before instability and destruction, and extract the precursor characteristics before the landslide occurs, thereby providing a reliable basis for landslide disaster warning. Therefore, establishing an efficient landslide prediction model and using massive data sets to accurately predict the deformation trend of landslides is not only of great theoretical significance, but also has significant practical application value.

[0003] In recent years, with the rapid development of machine learning theory, nonlinear science and system science, the nonlinear characteristics of landslide displacement time series have been widely recognized. The machine learning model established based on nonlinear theory has gradually become a hot topic in landslide displacement prediction research due to its good applicability and generalization ability. In particular, artificial intelligence models such as artificial neural network (ANN), convolutional neural network (CNN), recurrent neural network (RNN), and long short-term memory neural network (LSTM) are widely used. However, these traditional prediction methods mainly focus on the temporal relationship of monitoring data, and analyze the data of a single monitoring point. It is difficult to fully consider the complex spatiotemporal relationship between multiple landslide monitoring points, which leads to poor accuracy of the final prediction results. That is, these methods have limitations in judging the overall deformation trend of the landslide, and they also face challenges in the prediction efficiency and time cost of multi-point monitoring time series data of landslides. Summary of the invention

[0004] The problem solved by the present invention is how to improve the accuracy of landslide displacement prediction based on the spatiotemporal relationship of multiple landslide monitoring points, thereby improving the monitoring, early warning and risk management capabilities of the overall deformation of the landslide.

[0005] In order to solve the above problems, the present invention provides a landslide displacement time-space prediction method and device.

[0006] In a first aspect, the present invention provides a method for spatiotemporal prediction of landslide displacement, comprising:

[0007] Acquire historical landslide multi-field time series monitoring data of each monitoring point in the area to be tested, wherein the historical landslide multi-field time series monitoring data includes historical spatial correlation data, historical displacement time series and historical landslide external disturbance factor time series;

[0008] Constructing a monitoring point map based on the historical spatial related data, and constructing a time series feature matrix according to the historical displacement time series, the historical landslide external disturbance factor time series and the monitoring point map;

[0009] The time series feature matrix and the monitoring point map are input into the trained landslide displacement spatiotemporal prediction model to obtain the landslide displacement spatiotemporal prediction result; the landslide displacement spatiotemporal prediction model is constructed based on the improved GCN network and the improved BiLSTM network.

[0010] Optionally, the step of inputting the time series feature matrix and the monitoring point bitmap into a trained landslide displacement spatiotemporal prediction model to obtain a landslide displacement spatiotemporal prediction result includes:

[0011] Processing the time series feature matrix and the monitoring point bitmap through the improved GCN network to obtain a spatial time series feature matrix;

[0012] The spatial time series feature matrix is ​​predicted by the improved BiLSTM network to obtain the spatiotemporal prediction result of the landslide displacement.

[0013] Optionally, the improved GCN network includes an original GCN network and a multi-head attention mechanism; the step of processing the temporal feature matrix and the monitoring point bitmap by the improved GCN network to obtain a spatial temporal feature matrix includes:

[0014] The time series feature matrix is ​​fused through the multi-head attention mechanism to obtain multi-head attention fusion output data;

[0015] The multi-head attention fusion output data and the monitoring point bitmap are reconstructed through the original GCN network to obtain the spatial temporal feature matrix.

[0016] Optionally, the reconstructing the multi-head attention fusion output data and the monitoring point bitmap through the original GCN network to obtain the spatial temporal feature matrix includes:

[0017] Calculating a weighted adjacency matrix of spatial monitoring points based on the monitoring point map;

[0018] Based on the weighted adjacency matrix, the multi-head attention fusion output data is reconstructed to obtain the spatial-temporal feature matrix.

[0019] Optionally, the improved BiLSTM network includes a temporal attention mechanism; and the improved BiLSTM network construction process includes:

[0020] Obtain an original BiLSTM network, wherein the original BiLSTM network includes a forward LSTM unit and a corresponding backward LSTM unit;

[0021] The temporal attention mechanism is added to both the forward LSTM unit and the backward LSTM unit to obtain the improved BiLSTM network, and the temporal attention mechanism is constructed based on a fully connected layer and a softmax function.

[0022] Optionally, predicting the spatial time series feature matrix through the improved BiLSTM network to obtain the spatiotemporal prediction result of the landslide displacement includes:

[0023] Performing feature extraction on the spatial temporal feature matrix through the original BiLSTM network to obtain a hidden state matrix;

[0024] Processing the hidden state matrix through the temporal attention mechanism to obtain a temporary hidden state matrix;

[0025] The spatiotemporal prediction result of the landslide displacement is obtained according to the temporary hidden state matrix.

[0026] Optionally, the processing the hidden state matrix by the temporal attention mechanism to obtain a temporary hidden state matrix includes:

[0027] Processing the hidden state matrix of the fully connected layer to obtain a temporal attention weight, and normalizing the temporal attention weight according to the softmax function to obtain an attention weight vector;

[0028] The temporary hidden state matrix is ​​calculated according to the attention weight vector and the hidden state matrix.

[0029] In a second aspect, the present invention provides a landslide displacement spatiotemporal prediction device, comprising:

[0030] An acquisition unit is used to acquire historical landslide multi-field time series monitoring data of each monitoring point in the area to be tested, wherein the historical landslide multi-field time series monitoring data includes historical spatial correlation data, historical displacement time series and historical landslide external disturbance factor time series;

[0031] A processing unit, configured to construct a monitoring point map based on the historical spatial related data, and to construct a time series feature matrix according to the historical displacement time series, the historical landslide external disturbance factor time series and the monitoring point map;

[0032] The acquisition unit is also used to input the time series feature matrix and the monitoring point bitmap into the trained landslide displacement spatiotemporal prediction model to obtain the landslide displacement spatiotemporal prediction result; the landslide displacement spatiotemporal prediction model is constructed based on the improved GCN network and the improved BiLSTM network.

[0033] In a third aspect, the present invention provides a landslide displacement spatiotemporal prediction system, including a memory and a processor;

[0034] The memory is used to store computer programs;

[0035] The processor is used to implement the landslide displacement spatiotemporal prediction method as described in the first aspect when executing the computer program.

[0036] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the method for spatiotemporal prediction of landslide displacement as described in the first aspect is implemented.

[0037] The landslide displacement time-space prediction method and device of the present invention have the following beneficial effects:

[0038] First, obtain the historical data of all monitoring points of the landslide to be tested, including historical spatial related data, that is, historical spatial coordinates, historical displacement time series, that is, for each monitoring point, record its displacement changes at different time points, including horizontal and vertical displacements. The time series of external disturbance factors of historical landslides, such as monthly displacement, monthly average rainfall, monthly average reservoir water level change values ​​and other time series data; then generate a landslide monitoring point map based on the collected historical spatial related data. Generally speaking, the monitoring point map is represented as a graph structure, in which the edges represent the spatial relationship between monitoring points and the nodes represent each monitoring point.

[0039] The historical displacement time series is combined with the historical landslide external disturbance factor time series to extract the features that affect the landslide displacement. The dimension of the feature matrix is ​​usually the product of the number of monitoring points and the number of time points, including the relevant change values ​​and timestamps.

[0040] Then, the time series feature matrix and the monitoring point map are input into the trained landslide displacement spatiotemporal prediction model to obtain the landslide displacement spatiotemporal prediction results; among them, the improved graph convolutional network (GCN) is used to process the spatial relationship and correlation between monitoring points, which can effectively capture the complex relationship between spatial features and local features. In the time series processing part of the model, the improved bidirectional long short-term memory network (BiLSTM) is used to learn the dynamic changes of the time series from the historical displacement and external disturbance factors, which can better capture the time series characteristics.

[0041] In summary, the present invention can significantly improve the spatiotemporal prediction accuracy of landslide displacement by combining spatially related data with external disturbance factors and utilizing the capabilities of deep learning models. And the method can predict new monitoring data in real time, timely feedback landslide risks, and provide important basis for emergency management. At the same time, the improved GCN and BiLSTM networks can adapt to different types of landslide events, have strong adaptability, can handle complex time series data features, and are conducive to the prediction and monitoring of landslides in different regions. And through the combination of multiple types of data, the prediction model not only focuses on a single factor, but also fully considers the spatial characteristics and associations between multiple landslide monitoring nodes, effectively improves the prediction efficiency and accuracy of landslide multi-point monitoring data, and can also comprehensively consider the influence of multiple external disturbances, providing more scientific analysis and decision-making basis. The method can be widely used in the monitoring and risk management of geological disasters such as landslides, mudslides, and collapses, and improves the scientificity and effectiveness of natural disaster prevention, assessment, and prevention. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 A schematic flow chart of a method for spatiotemporal prediction of landslide displacement according to an embodiment of the present invention;

[0043] Figure 2 Schematic diagram of the improved GCN graph convolutional network structure of an embodiment of the present invention;

[0044] Figure 3 This is a schematic diagram of the structure of a multi-head attention mechanism according to an embodiment of the present invention;

[0045] Figure 4 This is a schematic diagram of the original BiLSTM model structure of an embodiment of the present invention;

[0046] Figure 5 Schematic diagram of the structure of the temporal attention mechanism of an embodiment of the present invention;

[0047] Figure 6 A schematic diagram of the structure of a device for temporal and spatial prediction of landslide displacement according to an embodiment of the present invention;

[0048] Figure 7 The figure is a schematic diagram of the structure of a landslide displacement spatiotemporal prediction system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0049] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below in conjunction with the accompanying drawings. Although certain embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be interpreted as being limited to the embodiments described herein. On the contrary, these embodiments are provided to provide a more thorough and complete understanding of the present invention. It should be understood that the drawings and embodiments of the present invention are only for exemplary purposes and are not intended to limit the scope of protection of the present invention.

[0050] It should be understood that the various steps described in the method embodiments of the present invention may be performed in different orders and / or in parallel. In addition, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present invention is not limited in this respect.

[0051] The term "including" and its variations used in this document are open inclusions, that is, "including but not limited to"; the term "based on" means "based at least in part on"; the term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one other embodiment"; the term "some embodiments" means "at least some embodiments"; the term "optionally" means "optional embodiments". The relevant definitions of other terms will be given in the following description. It should be noted that the concepts of "first", "second", etc. mentioned in the present invention are only used to distinguish different devices, modules or units, and are not used to limit the order or interdependence of the functions performed by these devices, modules or units.

[0052] It should be noted that the modifications of "one" and "plurality" mentioned in the present invention are illustrative rather than restrictive, and those skilled in the art should understand that, unless otherwise clearly indicated in the context, it should be understood as "one or more".

[0053] The names of the messages or information exchanged between multiple devices in the embodiments of the present invention are only used for illustrative purposes, and are not used to limit the scope of these messages or information.

[0054] This embodiment provides a landslide displacement time-space prediction method, device and landslide displacement time-space prediction system.

[0055] like Figure 1 As shown, a method for spatiotemporal prediction of landslide displacement provided by an embodiment of the present invention includes:

[0056] S100, obtaining historical landslide multi-field time series monitoring data of each monitoring point in the area to be tested, wherein the historical landslide multi-field time series monitoring data includes historical spatial correlation data, historical displacement time series and historical landslide external disturbance factor time series.

[0057] Specifically, historical spatial data mainly include the spatial coordinates of multiple monitoring points in the area to be measured; historical displacement time series, which records the displacement changes of each monitoring point within a certain time range, usually with time as the horizontal coordinate and displacement as the vertical coordinate. These time series data can reflect the dynamic changes of landslides and help researchers identify the evolution law and potential instability trend of landslides.

[0058] The historical landslide external disturbance factor time series data set mainly includes external factors that affect landslides, such as rainfall, temperature changes, seismic activity, construction disturbances, etc. In some embodiments, the historical landslide external disturbance factor time series includes monitoring data such as monthly displacement at landslide monitoring points, monthly average rainfall, and monthly average reservoir water level change values. These disturbance factors have an important impact on the occurrence and development of landslides. By analyzing the historical data of these factors, their role in the stability of landslides can be better understood.

[0059] By acquiring multiple time-series monitoring data, we can conduct a comprehensive analysis of landslides from multiple dimensions, comprehensively consider spatial characteristics, historical displacements, and external disturbance factors, and provide a deeper understanding of the occurrence mechanism of landslides.

[0060] S200, constructing a monitoring point map based on the historical space-related data, and constructing a time series feature matrix according to the historical displacement time series, the historical landslide external disturbance factor time series and the monitoring point map.

[0061] Specifically, a monitoring point map is constructed based on historical spatial related data, that is, a monitoring point map G representing spatial relationships is constructed based on the spatial coordinates of multiple monitoring points, and then a time series feature matrix X is established in combination with the monitoring point map G based on time related data, namely, the historical displacement time series and the historical landslide external disturbance factor time series. The displacement data and external disturbance factor data of each monitoring point are sorted according to the time series.

[0062] By constructing a monitoring point map and a time series feature matrix, the scattered monitoring data can be systematized to facilitate comprehensive analysis in space and time. The time series feature matrix can provide rich input features for subsequent machine learning models, enhance the model's understanding of the landslide mechanism, and improve the accuracy of prediction. At the same time, the construction of the monitoring point map makes the spatial distribution of landslide monitoring points clear at a glance, which is convenient for decision makers and researchers to intuitively understand the potential risk areas of landslides. And by integrating the time series of displacement and the time series of external disturbance factors, it is possible to analyze the relationship between multiple variables and reveal the complex relationship between landslide behavior and external factors.

[0063] The constructed time series feature matrix and monitoring point map can provide a scientific basis for landslide monitoring, early warning and disaster management, and help relevant departments formulate more reasonable disaster prevention and mitigation measures.

[0064] S300, inputting the time series feature matrix and the monitoring point bitmap into a trained landslide displacement spatiotemporal prediction model to obtain a landslide displacement spatiotemporal prediction result; the landslide displacement spatiotemporal prediction model is constructed based on an improved GCN network and an improved BiLSTM network.

[0065] Specifically, the constructed time series feature matrix and monitoring point bitmap are properly formatted to ensure that they can be correctly input into the landslide displacement spatiotemporal prediction model. The time series feature matrix contains the historical displacement and external disturbance factor information of each monitoring point, while the monitoring point bitmap provides the spatial coordinate information of each monitoring point. The time series feature matrix and the monitoring point bitmap are input into the trained landslide displacement spatiotemporal prediction model to obtain the landslide displacement spatiotemporal prediction results, where the improved GCN network (graph convolutional network) is used to capture the spatial relationship between monitoring points. GCN can effectively process graph structure data and extract the spatial characteristics and mutual relationships of monitoring points. By modeling the connection relationship of monitoring points, GCN can understand the spatial distribution pattern of the landslide affected area. The improved BiLSTM network (bidirectional long short-term memory network) is designed to capture the dynamic change characteristics of time series data. By bidirectionally processing time series information, BiLSTM can learn the time dependency between the past and the future and improve the prediction ability of landslide displacement.

[0066] In the training phase, the model uses historical landslide displacement data, time series feature matrix and monitoring point map information to learn the spatiotemporal variation of landslides. In the prediction phase, the time series feature matrix and monitoring point map are input into the trained spatiotemporal prediction model of landslide displacement. The model predicts the future displacement of the landslide through the combination of GCN and BiLSTM to generate the spatiotemporal prediction results of landslide displacement.

[0067] The improved GCN and BiLSTM models that combine spatial and temporal information can more comprehensively understand landslide behavior and improve the accuracy of prediction. This is of great significance for landslide monitoring and early warning. By integrating geospatial information and time series data, the model can effectively capture the spatiotemporal characteristics of landslides and reveal the real-time dynamic changes of landslide deformation.

[0068] The improved GCN and BiLSTM networks can better handle the impact of noise and missing data, making the model more robust and adaptable when facing various complex situations in the real environment. The generated spatiotemporal prediction results of landslide displacement can provide important decision-making support for the monitoring, assessment and management of landslide disasters, and help formulate more effective prevention and emergency measures.

[0069] In this embodiment, the historical data of all monitoring points of the landslide to be tested are first obtained, including historical spatial related data, that is, historical spatial coordinates, historical displacement time series, that is, for each monitoring point, the displacement changes at different time points are recorded, including horizontal and vertical displacements. The time series of external disturbance factors of the historical landslide, such as monthly displacement, monthly average rainfall, monthly average reservoir water level change value and other time series data; then, based on the collected historical spatial related data, a landslide monitoring point bitmap is generated. Generally speaking, the monitoring point bitmap is represented as a graph structure, in which the edges represent the spatial relationship between the monitoring points and the nodes represent each monitoring point.

[0070] The historical displacement time series is combined with the historical landslide external disturbance factor time series to extract the features that affect the landslide displacement. The dimension of the feature matrix is ​​usually the product of the number of monitoring points and the number of time points, including the relevant change values ​​and timestamps.

[0071] Then, the time series feature matrix and the monitoring point map are input into the trained landslide displacement spatiotemporal prediction model to obtain the landslide displacement spatiotemporal prediction results; among them, the improved graph convolutional network (GCN) is used to process the spatial relationship and correlation between monitoring points, which can effectively capture the complex relationship between spatial features and local features. In the time series processing part of the model, the improved bidirectional long short-term memory network (BiLSTM) is used to learn the dynamic changes of the time series from the historical displacement and external disturbance factors, which can better capture the time series characteristics.

[0072] In summary, this embodiment can significantly improve the spatiotemporal prediction accuracy of landslide displacement by combining spatially related data with external disturbance factors and utilizing the capabilities of deep learning models. Moreover, this method can predict new monitoring data in real time, and timely feedback landslide risks, providing an important basis for emergency management. At the same time, the improved GCN and BiLSTM networks can adapt to different types of landslide events, have strong adaptability, can handle complex time series data features, and are conducive to the prediction and monitoring of landslides in different regions. Moreover, through the combination of multiple types of data, the prediction model not only focuses on a single factor, but also fully considers the spatial characteristics and associations between multiple landslide monitoring nodes, effectively improves the prediction efficiency and accuracy of multi-point monitoring data of landslides, and can also comprehensively consider the influence of multiple external disturbances, providing more scientific analysis and decision-making basis. This method can be widely used in the monitoring and risk management of geological disasters such as landslides, mudslides, and collapses, and improves the scientificity and effectiveness of natural disaster prevention, assessment, and prevention.

[0073] Optionally, the step of inputting the time series feature matrix and the monitoring point bitmap into a trained landslide displacement spatiotemporal prediction model to obtain a landslide displacement spatiotemporal prediction result includes:

[0074] Processing the time series feature matrix and the monitoring point bitmap through the improved GCN network to obtain a spatial time series feature matrix;

[0075] The spatial time series feature matrix is ​​predicted by the improved BiLSTM network to obtain the spatiotemporal prediction result of the landslide displacement.

[0076] Specifically, the integrated time series feature matrix and monitoring point map are used as the input of the model. These two data sets combine the spatial location and time evolution characteristics of the landslide monitoring points, providing comprehensive information for the model. The input time series feature matrix (which contains the time series of historical displacements and external disturbance factors) is combined with the monitoring point map (which shows the spatial relationship between monitoring points).

[0077] In the improved GCN network, the spatial correlation of the monitoring points is extracted by performing graph convolution operations on the connection relationship between the monitoring points. The improved GCN network can effectively capture the spatial dependency between the landslide monitoring points by aggregating the neighborhood information of the nodes. After processing, a new spatial time series feature matrix is ​​output, which contains the fusion results of spatial relationship and time series information.

[0078] The spatial time series feature matrix processed by the improved GCN is used as the input of the improved BiLSTM network. The improved BiLSTM model extracts appropriate time series features based on the matrix. It captures the previous and next relationships in the time series through bidirectional learning, so as to better understand the impact of time changes on landslide displacement. Finally, the BiLSTM network generates spatiotemporal prediction results of landslide displacement, and identifies potential landslide risks in advance by predicting displacement changes at future moments.

[0079] The improved GCN network integrates spatial information with time series data, providing a multi-dimensional information analysis method, thereby enhancing the model's ability to understand and capture landslide behavior. And by improving the spatial time series feature matrix extracted by GCN, the relationship between monitoring points and their interaction with landslide displacement are accurately reflected, which lays the foundation for BiLSTM's spatiotemporal prediction. Finally, the temporal dependency captured by the bidirectional LSTM can significantly improve the model's prediction accuracy for landslide displacement, especially for dynamic changes that are continuous in time.

[0080] The improved GCN and BiLSTM network design improves the robustness of the model in processing complex and irregular data, and adapts to landslide prediction under different geographical and environmental conditions. Through this integrated prediction method, real-time monitoring and early warning of landslide displacement can be achieved, helping relevant departments to quickly respond and implement disaster prevention and mitigation measures, thereby improving public safety.

[0081] Optionally, the improved GCN network includes an original GCN network and a multi-head attention mechanism; the step of processing the temporal feature matrix and the monitoring point bitmap by the improved GCN network to obtain a spatial temporal feature matrix includes:

[0082] The time series feature matrix is ​​fused through the multi-head attention mechanism to obtain multi-head attention fusion output data;

[0083] The multi-head attention fusion output data and the monitoring point bitmap are reconstructed through the original GCN network to obtain the spatial temporal feature matrix.

[0084] Optionally, the reconstructing the multi-head attention fusion output data and the monitoring point bitmap through the original GCN network to obtain the spatial temporal feature matrix includes:

[0085] Calculating a weighted adjacency matrix of spatial monitoring points based on the monitoring point map;

[0086] Based on the weighted adjacency matrix, the multi-head attention fusion output data is reconstructed to obtain the spatial-temporal feature matrix.

[0087] Specifically, the temporal feature matrix containing landslide historical data and external influencing factors is input into the multi-head attention mechanism. The temporal feature matrix is ​​divided into multiple heads, each of which captures a specific piece of information. The attention weight of each head is calculated, and the importance of the feature is obtained by weighted average. The results of all heads are spliced ​​to generate multi-head attention fusion output data, which carries multi-dimensional feature information. The multi-head attention fusion output data and the monitoring point bitmap are input into the original GCN network. Using the monitoring point bitmap, the weighted adjacency matrix is ​​calculated. The weighted adjacency matrix is ​​used to represent the spatial relationship between monitoring points, and the weights can be set according to distance, historical correlation or other factors. The weighted relationship counts the mutual influence of monitoring points, so that the network can be constructed more accurately.

[0088] The multi-head attention fusion output data is subjected to graph convolution operation through the calculated weighted adjacency matrix. The original GCN combines the features of each monitoring point with the features of its neighboring monitoring points and strengthens the node features through aggregation operation. After graph convolution processing, the generated spatial temporal feature matrix not only contains temporal information, but also comprehensively considers the spatial relationship between monitoring points.

[0089] The multi-head attention mechanism effectively integrates the temporal features, and by focusing on the importance of different features, the model is more accurate in information extraction. The weighted adjacency matrix can reasonably reflect the mutual influence and spatial dependence between monitoring points, enhancing the expression ability of GCN. Compared with the simple adjacency matrix, the weighted connection can better capture the complexity of the local network. The original GCN network processes the multi-head attention output and uses the weighted adjacency matrix to adaptively aggregate neighborhood information, making the obtained spatial temporal feature matrix more representative and effective, promoting the accurate prediction of landslide displacement.

[0090] That is, by focusing on and integrating multiple features, the improved GCN network has stronger robustness to the uncertainty and noise of the input data, and can continuously provide accurate predictions in complex environments. Moreover, the generation of the spatial time series feature matrix provides a dynamic decision support for landslide monitoring and early warning, improving the timeliness and accuracy of landslide risk assessment. Through such a mechanism, the performance of the improved GCN network in landslide displacement prediction is significantly enhanced, and it can better handle complex spatiotemporal data and more effectively support disaster early warning and management.

[0091] In some embodiments, Figure 2 and Figure 3 As shown, Figure 2 To improve the GCN graph convolutional network structure diagram, Figure 3 It is a schematic diagram of the multi-head attention mechanism. Based on the spatial coordinates of multiple monitoring points (monitoring points), a monitoring point map G representing the spatial relationship is constructed, and the expression is as follows:

[0092] G = (V, E, W);

[0093] Among them, V∈{v1,v2…v N} is a set of N monitoring points, each of which is connected by an adjacent edge E, which can be obtained through the spatial coordinates of the monitoring point (monitoring point), W∈[0,1] N×N Represents a weighted adjacency matrix (weighted adjacency matrix W), where the weight w ij It represents the spatial correlation between monitoring points.

[0094] Gaussian similarity function is used to calculate the weight w ij , the expression is as follows:

[0095]

[0096] Among them, dist ij The monitoring point v i With v j The Euclidean distance, σ is the standard deviation of the distance;

[0097] The expression of the weighted adjacency matrix W is as follows:

[0098]

[0099] Then, based on the time-related data, namely the historical displacement time series and the historical landslide external disturbance factor time series, combined with the monitoring point map G, the time series feature matrix X is established. The expression of the time series feature matrix X is as follows:

[0100] X∈R N×S×M ;

[0101] Among them, N is the number of monitoring points, S is the number of monitoring features, M is the length of historical time series data for each monitoring point, and X t ∈R N×S×t It is the historical time series feature matrix of each monitoring point at time t.

[0102] Then, the graph convolutional neural network GCN (original GCN network) is used, and the multi-head attention mechanism GAT is introduced to construct an improved GCN network (spatial module) to capture the spatial correlation of landslide monitoring points and reconstruct the time series feature matrix related to the landslide space.

[0103] Among them, the implementation process of the multi-head attention mechanism GAT is as follows:

[0104] X i is the time series feature matrix X, that is, the historical time series feature matrix X = {X1,X2,…,X N}, X i ∈R S×M , X j is the feature matrix of the nodes adjacent to the central node, and the neighboring node matrix X j After linear transformation, we get the query matrix Q ij , center node matrix X i After linear deformation key matrix K i , value matrix V i , the calculation process is as follows:

[0105]

[0106] in, Respectively represent Q ij , K i 、V i The weight matrix, N i It means traversing the set of nodes adjacent to node i, S represents the dimension of the weight matrix, that is, the number of monitoring features;

[0107] Using the query matrix Q ij Key matrix K iPerform dot product scaling, and then calculate the weight coefficient w through softmax normalization ij , according to the weight coefficient pair matrix V i Perform weighted summation and finally obtain the self-attention weight output of node j to node i. The calculation formula is as follows;

[0108]

[0109] in, Denotes the key matrix K i The dimension of , the softmax function can obtain the self-attention weights that sum to 1;

[0110] The multi-head attention mechanism GAT includes multiple self-attention heads, which use multiple self-attention heads to learn the correlation of different representation subspaces respectively, and concatenate and linearly transform multiple self-attention values ​​to obtain the final self-attention value. Taking the n-head self-attention mechanism as an example, the feature vector X of the input center node i is i Split into X i / / n sub-feature sequences, each of which calculates its own self-attention, and finally concatenates them to obtain the new feature matrix X after the center node i is learned by the multi-head attention mechanism i ′, the splicing process is as follows:

[0111] Multi-Head(X i )=concat(z1,z2,…z n );

[0112]

[0113] Among them, n represents the total number of multi-head attention in the multi-head attention mechanism, Indicates that the kth attention head corresponds to Q ij , K i 、V i The weight matrix, N i It means traversing the set of nodes adjacent to node i. The concat function is used to concatenate the output values ​​of each attention head. k represents the output value of the kth attention head, that is, the sub-feature sequence after learning by the attention mechanism;

[0114] Repeat the above steps N times in total, and output the historical time series feature matrix of all nodes on the monitoring point map G after learning by the multi-head attention mechanism. The multi-head attention fusion output data is X′={X1′,X2′,…,X N ′},X′∈R N×S×M .

[0115] In the original graph convolutional network (GCN) model, the composition of the input layer is the key to achieve landslide displacement prediction. The input layer mainly includes the weighted adjacency matrix W and the multi-head attention fusion output data. Together, they provide the model with the necessary spatial relationship and temporal information.

[0116] First, the weighted adjacency matrix W is used to describe the spatial relationship between monitoring points, reflecting their mutual influence in the landslide prediction process. The degree of association between different monitoring points can be expressed through the corresponding weights, which are set based on factors such as geographical location and historical data.

[0117] Secondly, the multi-head attention fusion output data contains the characteristic data collected by each monitoring point in the historical time period. These data (such as displacement, rainfall, etc.) are organized into a matrix form, where each row represents the time series characteristics of a monitoring point, and each column corresponds to the characteristic value at a specific time point. The matrix provides the model with important information in the time dimension, enabling it to analyze the changing trend of the monitoring point over time.

[0118] In the structural design, the input layer is connected to the first hidden convolutional layer through the weighted adjacency matrix W and the multi-head attention fusion output data. Each hidden convolutional layer is combined with the corresponding ReLU (rectified linear unit) activation unit to introduce nonlinear factors, thereby enhancing the expressiveness and adaptability of the model.

[0119] Finally, the last ReLU activation unit is connected to the output layer of GCN. The output layer is responsible for generating a temporal feature matrix containing spatially relevant information of landslides. This output matrix not only comprehensively reflects the temporal characteristics of the monitoring points, but also highlights the spatial correlation between them, thus providing a solid data foundation for the prediction of landslide events.

[0120] In summary, the GCN model effectively captures the spatiotemporal characteristics of landslide displacement through carefully designed input layers and hierarchical structures, providing important support for disaster prediction and management.

[0121] Among them, the propagation process between adjacent convolutional layers in the original GCN network is carried out according to the following formula:

[0122]

[0123] D ii =∑ j W ij ;

[0124] O (0) =X′;

[0125] Among them, O (l+1) represents the output of the lth convolutional layer, σ is the activation function, G (l)is the feature matrix O (l) The weight matrix for linear changes, l represents the number of convolution layers, and the matrix represents the adjacency matrix after symmetric normalization of the weighted adjacency matrix W. The degree matrix D is calculated from the weighted adjacency matrix W. (0) Represents the input of the first hidden convolutional layer, that is, the multi-head attention fusion output data.

[0126] Optionally, the improved BiLSTM network includes a temporal attention mechanism; and the improved BiLSTM network construction process includes:

[0127] Obtain an original BiLSTM network, wherein the original BiLSTM network includes a forward LSTM unit and a corresponding backward LSTM unit;

[0128] The temporal attention mechanism is added to both the forward LSTM unit and the backward LSTM unit to obtain the improved BiLSTM network, and the temporal attention mechanism is constructed based on a fully connected layer and a softmax function.

[0129] Optionally, predicting the spatial time series feature matrix through the improved BiLSTM network to obtain the spatiotemporal prediction result of the landslide displacement includes:

[0130] Performing feature extraction on the spatial temporal feature matrix through the original BiLSTM network to obtain a hidden state matrix;

[0131] Processing the hidden state matrix through the temporal attention mechanism to obtain a temporary hidden state matrix;

[0132] The spatiotemporal prediction result of the landslide displacement is obtained according to the temporary hidden state matrix.

[0133] Optionally, the processing the hidden state matrix by the temporal attention mechanism to obtain a temporary hidden state matrix includes:

[0134] Processing the hidden state matrix of the fully connected layer to obtain a temporal attention weight, and normalizing the temporal attention weight according to the softmax function to obtain an attention weight vector;

[0135] The temporary hidden state matrix is ​​calculated according to the attention weight vector and the hidden state matrix.

[0136] Specifically, Figure 4As shown in Figure 2, the original Bidirectional Long Short-Term Memory (BiLSTM) model consists of two independent sets of forward and backward LSTM units, which are designed to extract the positive and reverse features in the temporal information. Through these units, the model can fully capture the temporal dynamics of landslide displacement.

[0137] Specifically, the forward LSTM unit is responsible for processing the forward input data, while the backward LSTM unit processes the reverse input data. These two groups of LSTM units operate independently in the model and generate hidden states respectively. Subsequently, the model uses an output layer to weightedly superimpose the hidden states calculated by the forward and backward LSTM units to form a fused and spliced ​​hidden state for each time step. These spliced ​​hidden states are then input into a fully connected layer to generate the final landslide displacement prediction value. This process effectively improves the utilization of data and the accuracy of prediction.

[0138] In some embodiments, the original bidirectional long short-term memory network (BiLSTM) model includes a forward LSTM unit and a corresponding backward LSTM unit, and the internal structure of each LSTM unit includes an input gate, a forget gate, an output gate, and a cell state. These components work together to determine the flow of information in the network and the retention of importance. Specifically, the time series feature matrix related to the landslide space is passed as input to each LSTM unit, and finally the hidden state matrix of each time step is output.

[0139] The main calculation formula is as follows:

[0140]

[0141] Among them, x t is the input of time step t, representing the value vector of the time series characteristic matrix of landslide space correlation at time step t, h t-1 is the hidden state matrix at time step t-1, c t-1 is the cell state at time step t-1, i t 、f t , c t , o t 、h t Represent the input gate, forget gate, cell state update, cell state, output gate and hidden state at time step t respectively, W xi , W xf , W xo , W xc With W hi , W hf , W ho , Whc are the input matrix and hidden matrix of the above three gates and cell states, respectively. ci , W cf , W co is the weight matrix involving long-term memory in the above three gates, b i 、b f 、b o 、b c are the corresponding bias vectors, σ is the sigmoid function, and tanh is the hyperbolic tangent function.

[0142] In some preferred embodiments, Figure 5 As shown in Figure 2, the implementation process of the temporal attention mechanism is as follows:

[0143] By adding a fully connected layer and a softmax function as an attention layer in each original LSTM unit to implement the temporal attention mechanism, which is used to select the weights of the hidden states in the entire time series memory cycle, the importance of the input sequence at different time steps can be dynamically adjusted to improve the accuracy of the model.

[0144] The input data of the temporal attention mechanism is the hidden state matrix H=[h1,h2,…,h T ] T×S ,h T is the hidden state corresponding to T time steps, S is the size of each hidden state, that is, the number of monitoring features; the hidden state matrix is ​​processed by a layer of fully connected neural network to obtain the time attention weight, and the weight is normalized by the softmax function to obtain the attention weight vector α, and the attention weight vector α is multiplied by the hidden state matrix to obtain the hidden state matrix H after the time attention mechanism is allocated. at , the weight distribution formula and final output are as follows:

[0145] α=softmax(A t (H))=[α1,α2,…,α T ] 1×T ;

[0146]

[0147] Among them, A t represents a fully connected neural network (fully connected layer), α T is the attention weight corresponding to T time steps, Represents the hidden state (temporary hidden state matrix) of the final output of the LSTM unit at time T.

[0148] It should be noted that there is no limit on the number of forward LSTM units, and the specific number can be determined according to actual conditions, while the number of backward LSTM units corresponds to the number of forward LSTM units.

[0149] In some embodiments, during the training process of the landslide displacement spatiotemporal prediction model, the landslide displacement time series prediction value is often evaluated based on three error evaluation indicators: mean absolute error MAE, mean absolute proportional error MASE, and root mean square error RMSE to obtain a prediction evaluation result;

[0150] Specifically, the actual displacement monitoring sequence of the landslide (the spatiotemporal monitoring results of the landslide displacement) Y = [y1, y2, ..., y T ] 1×T The displacement time series prediction results of the model (landslide displacement time-space prediction results) are analyzed using three indicators: root mean square error (MAE), mean absolute error (MASE) and mean relative absolute error (RMSE). For accuracy test, the calculation formulas of root mean square error MAE, mean absolute error MASE and mean relative error absolute value RMSE are as follows:

[0151]

[0152] Where T is the length of the time prediction sequence, y i and are respectively the monitoring value and predicted value of landslide displacement at time i

[0153] It should be noted that all the data used in the model training and actual application process are historical data. For example, the model training process can be trained with historical data from 5 years ago, while the actual application process can use historical data from 1 year ago to predict the spatiotemporal results of future landslide displacements.

[0154] like Figure 6 As shown, an embodiment of the present invention provides a landslide displacement spatiotemporal prediction device, comprising:

[0155] An acquisition unit is used to acquire historical landslide multi-field time series monitoring data of each monitoring point in the area to be tested, wherein the historical landslide multi-field time series monitoring data includes historical spatial correlation data, historical displacement time series and historical landslide external disturbance factor time series;

[0156] A processing unit, configured to construct a monitoring point map based on the historical spatial related data, and to construct a time series feature matrix according to the historical displacement time series, the historical landslide external disturbance factor time series and the monitoring point map;

[0157] The acquisition unit is also used to input the time series feature matrix and the monitoring point bitmap into the trained landslide displacement spatiotemporal prediction model to obtain the landslide displacement spatiotemporal prediction result; the landslide displacement spatiotemporal prediction model is constructed based on the improved GCN network and the improved BiLSTM network.

[0158] The landslide displacement spatiotemporal prediction device of this embodiment is used to implement the landslide displacement spatiotemporal prediction method described above. Its advantages over the prior art are the same as the advantages of the landslide displacement spatiotemporal prediction method described above over the prior art, and will not be repeated here.

[0159] like Figure 7 As shown, a landslide displacement spatiotemporal prediction system 700 provided by an embodiment of the present invention includes a memory 710 and a processor 720; the memory 710 is used to store a computer program; the processor 720 is used to implement the landslide displacement spatiotemporal prediction method as described above when executing the computer program.

[0160] In other words, a landslide displacement spatiotemporal prediction system 700 includes a memory 710 and a processor 720 coupled to the memory 710; the memory 710 is configured to store a computer program; the processor 720 is configured to perform the following operations when executing the computer program:

[0161] Acquire historical landslide multi-field time series monitoring data of each monitoring point in the area to be tested, wherein the historical landslide multi-field time series monitoring data includes historical spatial correlation data, historical displacement time series and historical landslide external disturbance factor time series;

[0162] Constructing a monitoring point map based on the historical spatial related data, and constructing a time series feature matrix according to the historical displacement time series, the historical landslide external disturbance factor time series and the monitoring point map;

[0163] The time series feature matrix and the monitoring point map are input into the trained landslide displacement spatiotemporal prediction model to obtain the landslide displacement spatiotemporal prediction result; the landslide displacement spatiotemporal prediction model is constructed based on the improved GCN network and the improved BiLSTM network.

[0164] An embodiment of the present invention provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the above-mentioned method for spatiotemporal prediction of landslide displacement is implemented.

[0165] In other words, a non-volatile computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the processor performs the following operations:

[0166] Acquire historical landslide multi-field time series monitoring data of each monitoring point in the area to be tested, wherein the historical landslide multi-field time series monitoring data includes historical spatial correlation data, historical displacement time series and historical landslide external disturbance factor time series;

[0167] Constructing a monitoring point map based on the historical spatial related data, and constructing a time series feature matrix according to the historical displacement time series, the historical landslide external disturbance factor time series and the monitoring point map;

[0168] The time series feature matrix and the monitoring point map are input into the trained landslide displacement spatiotemporal prediction model to obtain the landslide displacement spatiotemporal prediction result; the landslide displacement spatiotemporal prediction model is constructed based on the improved GCN network and the improved BiLSTM network.

[0169] The landslide displacement spatiotemporal prediction system 700 that can be used as the server or client of the present invention will now be described, which is an example of a hardware device that can be applied to various aspects of the present invention. The landslide displacement spatiotemporal prediction system 700 is intended to represent various forms of digital electronic computer equipment, such as, laptop computers, desktop computers, workbenches, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The landslide displacement spatiotemporal prediction system 700 can also represent various forms of mobile devices, such as, personal digital processing, cellular phones, smart phones, wearable devices and other similar computing devices. The components shown herein, their connections and relationships, and their functions are only used as examples, and are not intended to limit the implementation of the present invention described herein and / or required.

[0170] The landslide displacement spatiotemporal prediction system 700 includes a computing unit, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) or a computer program loaded from a storage unit into a random access memory (RAM). In the RAM, various programs and data required for the operation of the device can also be stored. The computing unit, ROM, and RAM are connected to each other via a bus. An input / output (I / O) interface is also connected to the bus.

[0171] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, the storage medium can be a disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM), etc. In the present application, the unit described as a separate component may or may not be physically separated, and the component displayed as a unit may or may not be a physical unit, that is, it may be located in one place, or it may be distributed on multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the embodiment of the present invention. In addition, each functional unit in each embodiment of the present invention can be integrated in a processing unit, or each unit can exist physically separately, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of a software functional unit.

[0172] Although the present invention is disclosed as above, the protection scope of the present invention is not limited thereto. Those skilled in the art may make various changes and modifications without departing from the spirit and scope of the present invention, and these changes and modifications will fall within the protection scope of the present invention.

Claims

1. A method for spatiotemporal prediction of landslide displacement, characterized in that: include: Acquire historical landslide multi-field time series monitoring data of each monitoring point in the area to be tested, wherein the historical landslide multi-field time series monitoring data includes historical spatial correlation data, historical displacement time series and historical landslide external disturbance factor time series; Constructing a monitoring point map based on the historical spatial related data, and constructing a time series feature matrix according to the historical displacement time series, the historical landslide external disturbance factor time series and the monitoring point map; Inputting the time series feature matrix and the monitoring point bitmap into the trained landslide displacement spatiotemporal prediction model to obtain the landslide displacement spatiotemporal prediction result; The landslide displacement spatiotemporal prediction model is constructed based on an improved GCN network and an improved BiLSTM network.

2. The method for spatiotemporal prediction of landslide displacement according to claim 1, characterized in that: The step of inputting the time series feature matrix and the monitoring point bitmap into the trained landslide displacement spatiotemporal prediction model to obtain the landslide displacement spatiotemporal prediction result includes: Processing the time series feature matrix and the monitoring point bitmap through the improved GCN network to obtain a spatial time series feature matrix; The spatial time series feature matrix is ​​predicted by the improved BiLSTM network to obtain the spatiotemporal prediction result of the landslide displacement.

3. The method for spatiotemporal prediction of landslide displacement according to claim 2, characterized in that: The improved GCN network includes the original GCN network and a multi-head attention mechanism; The step of processing the time series feature matrix and the monitoring point bitmap through the improved GCN network to obtain a spatial time series feature matrix includes: The time series feature matrix is ​​fused through the multi-head attention mechanism to obtain multi-head attention fusion output data; The multi-head attention fusion output data and the monitoring point bitmap are reconstructed through the original GCN network to obtain the spatial temporal feature matrix.

4. The method for spatiotemporal prediction of landslide displacement according to claim 3, characterized in that: The reconstructing the multi-head attention fusion output data and the monitoring point bitmap through the original GCN network to obtain the spatial temporal feature matrix includes: Calculating a weighted adjacency matrix of spatial monitoring points based on the monitoring point map; Based on the weighted adjacency matrix, the multi-head attention fusion output data is reconstructed to obtain the spatial-temporal feature matrix.

5. The method for spatiotemporal prediction of landslide displacement according to claim 2, characterized in that: The improved BiLSTM network includes a temporal attention mechanism; the improved BiLSTM network construction process includes: Obtain an original BiLSTM network, wherein the original BiLSTM network includes a forward LSTM unit and a corresponding backward LSTM unit; The temporal attention mechanism is added to both the forward LSTM unit and the backward LSTM unit to obtain the improved BiLSTM network, and the temporal attention mechanism is constructed based on a fully connected layer and a softmax function.

6. The method for spatiotemporal prediction of landslide displacement according to claim 5, characterized in that: The method of predicting the spatial time series feature matrix through the improved BiLSTM network to obtain the spatiotemporal prediction result of the landslide displacement includes: Performing feature extraction on the spatial temporal feature matrix through the original BiLSTM network to obtain a hidden state matrix; Processing the hidden state matrix through the temporal attention mechanism to obtain a temporary hidden state matrix; The spatiotemporal prediction result of the landslide displacement is obtained according to the temporary hidden state matrix.

7. The method for spatiotemporal prediction of landslide displacement according to claim 6, characterized in that: The step of processing the hidden state matrix by the temporal attention mechanism to obtain a temporary hidden state matrix includes: Processing the hidden state matrix of the fully connected layer to obtain a temporal attention weight, and normalizing the temporal attention weight according to the softmax function to obtain an attention weight vector; The temporary hidden state matrix is ​​calculated according to the attention weight vector and the hidden state matrix.

8. A landslide displacement spatiotemporal prediction device, characterized in that: include: An acquisition unit is used to acquire historical landslide multi-field time series monitoring data of each monitoring point in the area to be tested, wherein the historical landslide multi-field time series monitoring data includes historical spatial correlation data, historical displacement time series and historical landslide external disturbance factor time series; A processing unit, configured to construct a monitoring point map based on the historical spatial related data, and to construct a time series feature matrix according to the historical displacement time series, the historical landslide external disturbance factor time series and the monitoring point map; The acquisition unit is also used to input the time series feature matrix and the monitoring point bitmap into the trained landslide displacement spatiotemporal prediction model to obtain the landslide displacement spatiotemporal prediction result; The landslide displacement spatiotemporal prediction model is constructed based on an improved GCN network and an improved BiLSTM network.

9. A landslide displacement spatiotemporal prediction system, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method for spatiotemporal prediction of landslide displacement according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by a processor, the method for spatiotemporal prediction of landslide displacement according to any one of claims 1 to 7 is implemented.

Images