Landslide deformation zone monitoring method and system based on surface and deep data combination

Abstract

The present disclosure provides a landslide deformation zone monitoring method and system based on the combination of surface and deep data, relating to the technical field of deformation measurement. The method comprises: collecting surface deformation data and deep monitoring data of the landslide deformation zone region at multiple observation time periods, and based on the surface deformation data and the deep monitoring data, constructing a time series monitoring sequence corresponding to each monitoring point in the landslide deformation zone region; taking the monitoring point as a graph node, and constructing a graph structure according to the spatial distribution relationship of the monitoring point; adding the time series monitoring sequence to the corresponding node in the graph structure as a node attribute to obtain a multi-dimensional graph model; predicting the multi-dimensional graph model using a graph neural network to obtain predicted deformation data of each monitoring point at a future observation time period; and combining the predicted deformation data with a current landslide deformation zone image to determine the expansion area of the landslide deformation zone. The technical solution in the present disclosure can realize accurate prediction of the deformation trend and expansion area of the landslide deformation zone.

Description

Technical Field

[0001] The present disclosure relates to the field of deformation measurement technology, and in particular to a landslide deformation zone monitoring method and system based on the combination of surface and deep data. Background Art

[0002] Slope stability is crucial for the safe operation of large-scale projects such as reservoirs and hydropower stations. Landslide deformation zones are a key factor in slope instability, and their dynamic changes often reflect potential slope risks. Due to the complex topography, geological conditions, and water effects in the slope areas of reservoirs and hydropower stations, the deformation process of landslide deformation zones is highly dynamic, nonlinear, and hidden. Especially under the influence of external factors such as rainfall, reservoir water level fluctuations, and earthquakes, the state and expansion of landslide deformation zones are difficult to predict, seriously affecting the safety of engineering facilities and personnel.

[0003] In related technologies, monitoring of landslide deformation zones primarily relies on observing surface data of the zones, using techniques such as GNSS and surface crack monitoring equipment to obtain surface displacement, settlement, and other data. However, this type of monitoring method is limited to static observation of the surface of the landslide deformation zone and is unable to determine the deformation trend of the landslide deformation zone. Therefore, some related technologies use surface deformation data to predict the deformation trend of the landslide deformation zone, thereby better understanding the direction of change in landslide deformation. However, prediction methods that only use single surface deformation data are insufficiently accurate. Therefore, there is still room for improvement in the comprehensiveness and accuracy of existing landslide deformation zone monitoring technologies.

[0004] It should be noted that the information disclosed in the above background technology section is only used to enhance the understanding of the background of the present disclosure, and therefore may include information that does not constitute prior art known to ordinary technicians in the field. Summary of the Invention

[0005] The purpose of the embodiments of the present disclosure is to provide a landslide deformation zone monitoring method based on the combination of surface and deep data, a landslide deformation zone monitoring system based on the combination of surface and deep data, an electronic device and a computer-readable storage medium, which can improve the comprehensiveness and accuracy of landslide deformation zone monitoring at least to a certain extent by fusing surface deformation data with deep monitoring data, constructing a graph structure and introducing a graph neural network for deformation prediction.

[0006] Other features and advantages of the present disclosure will become apparent from the following detailed description, or may be learned in part by practice of the present disclosure.

[0007] According to a first aspect of an embodiment of the present disclosure, a landslide deformation zone monitoring method based on a combination of surface and deep data is provided, the method comprising: collecting surface deformation data and deep monitoring data of a landslide deformation zone area in multiple observation periods, and constructing a time series monitoring sequence corresponding to each monitoring point in the landslide deformation zone area based on the surface deformation data and the deep monitoring data; taking the monitoring points as graph nodes, and constructing a graph structure according to the spatial distribution relationship of the monitoring points; adding the time series monitoring sequence as a node attribute to the corresponding node in the graph structure to obtain a multidimensional graph model; using a graph neural network to predict the multidimensional graph model to obtain predicted deformation data of each monitoring point in future observation periods; and determining the expansion area of ​​the landslide deformation zone by combining the predicted deformation data with the current landslide deformation zone image.

[0008] In some example embodiments of the present disclosure, based on the aforementioned scheme, the above-mentioned collection of surface deformation data and deep monitoring data in the landslide deformation zone area in multiple observation time periods includes: using GNSS antenna devices and tilt sensors deployed at multiple target monitoring points in the landslide deformation zone area to collect the surface deformation data and the deep monitoring data in multiple observation time periods; the surface deformation data include horizontal displacement and vertical settlement, and the deep monitoring data include tilt angle change data and stress data.

[0009] In some example embodiments of the present disclosure, based on the aforementioned scheme, the above-mentioned method constructs a time series monitoring sequence corresponding to each monitoring point in the landslide deformation zone area based on the surface deformation data and the deep monitoring data, including: determining the weight data corresponding to each monitoring point according to the terrain data corresponding to each monitoring point in the landslide deformation zone area, wherein each parameter in the weight data corresponds to a different type of data in the time series monitoring sequence; constructing an initial time series monitoring sequence corresponding to each monitoring point in the landslide deformation zone area based on the surface deformation data and the deep monitoring data; and performing weighted processing on the initial time series monitoring sequence according to the weight data to obtain the time series monitoring sequence.

[0010] In some example embodiments of the present disclosure, based on the aforementioned scheme, the above-mentioned method of determining weight data corresponding to each monitoring point according to the terrain data corresponding to each monitoring point in the landslide deformation zone area includes: obtaining slope data, elevation data, terrain curvature and slope foot distance corresponding to each monitoring point in the landslide deformation zone area; determining the weight data according to the slope data, elevation data, terrain curvature and slope foot distance; wherein the slope data is used to determine a first weight corresponding to the horizontal displacement, the elevation data is used to determine a second weight corresponding to the vertical settlement, the terrain curvature is used to determine a third weight corresponding to the inclination angle change data, and the slope foot distance is used to determine a fourth weight corresponding to the stress data.

[0011] In some example embodiments of the present disclosure, based on the aforementioned scheme, the monitoring points are used as graph nodes, and a graph structure is constructed according to the spatial distribution relationship of the monitoring points, including: determining the Euclidean distance, elevation difference and slope angle between any two monitoring points according to the position coordinates, elevation data and slope information corresponding to the monitoring points; establishing a graph connection edge between the corresponding two monitoring points in response to the Euclidean distance being less than a first threshold, the elevation difference being less than a second threshold and the slope angle being less than a third threshold; and constructing the graph structure that satisfies terrain constraints based on multiple graph connection edges.

[0012] In some example embodiments of the present disclosure, based on the aforementioned scheme, the above-mentioned use of a graph neural network to predict the multidimensional graph model and obtain the predicted deformation data of each monitoring point in the future observation period includes: segmenting the time series monitoring sequence corresponding to each node in the multidimensional graph model according to a preset time step to obtain a set of node feature sequences divided based on time steps; inputting the node features corresponding to each time step in the node feature sequence set into the graph neural network in turn, and performing feature propagation processing in each time step based on the edge connection relationship in the multidimensional graph model; in each time step, updating the graph representation state of the corresponding node according to the result of the feature propagation processing, and passing the graph representation state to the next time step as the state input of the subsequent node feature; in response to the node feature completing the iterative operation on all time steps and forming the final graph representation state, generating the predicted deformation data of each monitoring point in the target prediction period based on the final graph representation state.

[0013] In some example embodiments of the present disclosure, based on the aforementioned scheme, the above-mentioned segmentation processing of the timing monitoring sequence corresponding to each node in the multidimensional graph model according to a preset time step includes: normalizing the timing monitoring sequence corresponding to each node in the multidimensional graph model based on a linear normalization method to obtain a normalized monitoring sequence; and segmenting the normalized monitoring sequence corresponding to each node according to a preset time step.

[0014] In some example embodiments of the present disclosure, based on the aforementioned scheme, the above-mentioned feature propagation processing based on the edge connection relationship in the multidimensional graph model includes: determining the adjacent nodes corresponding to each node based on the edge connection relationship in the multidimensional graph model; using edge weights to weight the node feature sequence corresponding to the adjacent nodes at the current time step to obtain the aggregated features of each node at the current time step; fusing the aggregated features with the node feature sequence of the current node to generate the graph representation state of the current node at the current time step.

[0015] In some example embodiments of the present disclosure, based on the aforementioned scheme, the above-mentioned combination of the predicted deformation data and the current landslide deformation band image to determine the extended area of ​​the landslide deformation band includes: determining the deformation gradient vector of each of the monitoring points based on the predicted deformation data; extracting the boundary contour of the landslide area in the current landslide deformation band image, and constructing a mask image based on the boundary contour; determining the extension direction of the deformation gradient vector relative to the boundary of the mask image, and performing an expansion operation in the corresponding area of ​​the extension direction to generate the extended area of ​​the landslide deformation band.

[0016] According to a second aspect of an embodiment of the present disclosure, a landslide deformation zone monitoring system based on a combination of surface and deep data is provided, the system comprising: a sequence construction module for collecting surface deformation data and deep monitoring data of a landslide deformation zone area in multiple observation periods, and constructing a time series monitoring sequence corresponding to each monitoring point in the landslide deformation zone area based on the surface deformation data and the deep monitoring data; a graph structure establishment module for taking the monitoring points as graph nodes and constructing a graph structure according to the spatial distribution relationship of the monitoring points; a graph model construction module for adding the time series monitoring sequence as a node attribute to the corresponding node in the graph structure to obtain a multidimensional graph model; a deformation prediction module for predicting the multidimensional graph model using a graph neural network to obtain predicted deformation data of each monitoring point in future observation periods; an extension area module for combining the predicted deformation data with the current landslide deformation zone image to determine the extension area of ​​the landslide deformation zone.

[0017] According to a third aspect of an embodiment of the present disclosure, an electronic device is provided, comprising: a processor; and a memory, wherein the memory stores computer-readable instructions, and when the computer-readable instructions are executed by the processor, a landslide deformation zone monitoring method based on the combination of surface and deep data as in the first aspect is implemented.

[0018] According to a fourth aspect of an embodiment of the present disclosure, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the landslide deformation zone monitoring method based on the combination of surface and deep data as in the first aspect is implemented.

[0019] The technical solutions provided by the embodiments of the present disclosure may have the following beneficial effects:

[0020] In the embodiment of the present disclosure, a landslide deformation zone monitoring method based on the combination of surface and deep data is proposed. On the one hand, a time series monitoring sequence is constructed by combining surface deformation data and deep monitoring data, so that the monitoring data can simultaneously cover the deformation characteristics of the shallow and deep layers of the surface, thereby improving the representativeness of the monitoring data. On the other hand, a graph structure is constructed by taking the monitoring points as graph nodes, so that the geographical relationship between the monitoring points is structured. On the other hand, a multidimensional graph model is formed by adding the time series monitoring sequence as a node attribute to the graph structure, so that each node not only contains local time series information, but also receives information input from adjacent nodes through edge connection relationships. Based on the multidimensional graph model, the graph neural network can perform feature propagation and state update at each time step, and gradually generate predicted deformation data for each monitoring point in the future observation period. Compared with the traditional time series prediction model, it can better capture local trends and improve prediction accuracy. Finally, by combining the predicted deformation data with the current landslide deformation zone image, the current landslide boundary can be updated along the deformation direction, thereby obtaining the expansion area of ​​the landslide deformation zone.

[0021] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The accompanying drawings are incorporated into and constitute a part of the specification, illustrate embodiments consistent with the present disclosure, and together with the specification, are used to explain the principles of the present disclosure. Obviously, the drawings described below are only some embodiments of the present disclosure, and those skilled in the art can derive other drawings based on these drawings without inventive effort.

[0023] Figure 1 The following schematically illustrates a flow chart of a landslide deformation zone monitoring method based on the combination of surface and deep data according to some embodiments of the present disclosure.

[0024] Figure 2 The following schematically illustrates a flow chart of obtaining predicted deformation data according to some embodiments of the present disclosure.

[0025] Figure 3 The following schematically illustrates a flow chart of determining an extension area of ​​a landslide deformation zone according to some embodiments of the present disclosure.

[0026] Figure 4 The following schematically illustrates the composition of a landslide deformation zone monitoring system based on the combination of surface and deep data according to some embodiments of the present disclosure.

[0027] Figure 5 A schematic structural diagram of a computer system of an electronic device according to some embodiments of the present disclosure is schematically shown.

[0028] Figure 6 A schematic diagram of a computer-readable storage medium according to some embodiments of the present disclosure is schematically shown.

[0029] In the drawings, the same or corresponding reference numerals denote the same or corresponding parts. DETAILED DESCRIPTION

[0030] Exemplary embodiments will be described in detail herein, with examples illustrated in the accompanying drawings. In the following description, when referring to the drawings, identical numerals in different figures represent identical or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments are not intended to represent all embodiments consistent with this specification. Rather, they are merely examples of apparatus and methods consistent with certain aspects of this specification, as detailed in the appended claims.

[0031] The terms used in this specification are for the purpose of describing specific embodiments only and are not intended to limit this specification. As used in this specification and the appended claims, the singular forms "a," "an," "the," and "the" are intended to include the plural forms as well, unless the context clearly indicates 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.

[0032] It should be understood that although the terms first, second, third, etc. may be used in this specification to describe various information, such information should not be limited to these terms. These terms are merely used to distinguish information of the same type from one another. For example, first information may also be referred to as second information, and similarly, second information may also be referred to as first information without departing from the scope of this specification. Depending on the context, the term "if" as used herein may be interpreted as "when," "when," or "in response to determining."

[0033] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that this disclosure will be thorough and complete and will fully convey the concepts of the example embodiments to those skilled in the art.

[0034] In addition, the described features, structures or characteristics may be combined in any suitable manner in one or more embodiments. In the following description, many specific details are provided to provide a full understanding of the embodiments of the present disclosure. However, those skilled in the art will appreciate that the technical solutions of the present disclosure can be practiced without one or more of the specific details, or other methods, components, devices, steps, etc. can be adopted. In other cases, well-known methods, devices, implementations or operations are not shown or described in detail to avoid blurring various aspects of the present disclosure.

[0035] Furthermore, the drawings are schematic illustrations only and are not necessarily drawn to scale. The block diagrams shown in the drawings are merely functional entities and do not necessarily correspond to physically separate entities. In other words, these functional entities may be implemented in software, in one or more hardware modules or integrated circuits, or in different networks and / or processor devices and / or microcontroller devices.

[0036] In this exemplary embodiment, a landslide deformation zone monitoring method based on the combination of surface and deep data is first provided. Figure 1 The following schematically illustrates a flow chart of a landslide deformation zone monitoring method based on the combination of surface and deep data according to some embodiments of the present disclosure. Figure 1 As shown, the landslide deformation zone monitoring method based on the combination of surface and deep data may include the following steps:

[0037] Step S110, collecting surface deformation data and deep monitoring data of the landslide deformation zone area during multiple observation periods, and constructing a time series monitoring sequence corresponding to each monitoring point in the landslide deformation zone area based on the surface deformation data and the deep monitoring data;

[0038] Step S120: Using the monitoring points as graph nodes, a graph structure is constructed based on the spatial distribution relationship of the monitoring points;

[0039] Step S130 , adding the time series monitoring sequence as a node attribute to the corresponding node in the graph structure to obtain a multi-dimensional graph model;

[0040] Step S140: Using a graph neural network to predict the multidimensional graph model, obtain the predicted deformation data of each monitoring point in the future observation period;

[0041] Step S150 : Determine the extension area of ​​the landslide deformation zone by combining the predicted deformation data with the current landslide deformation zone image.

[0042] According to the landslide deformation zone monitoring method based on the combination of surface and deep data in this example embodiment, on the one hand, a time series monitoring sequence is constructed by surface deformation data and deep monitoring data, so that the monitoring data can simultaneously cover the deformation characteristics of the shallow and deep layers of the surface, thereby improving the representativeness of the monitoring data. On the other hand, the monitoring points are used as graph nodes to construct a graph structure, so that the geographical relationship between the monitoring points is structured. On the other hand, by adding the time series monitoring sequence as a node attribute to the graph structure, a multidimensional graph model is formed, so that each node not only contains local time series information, but also receives information input from adjacent nodes through edge connection relationships. Based on the multidimensional graph model, the graph neural network can perform feature propagation and state update at each time step, and gradually generate predicted deformation data for each monitoring point in the future observation period. Compared with the traditional time series prediction model, it can better capture local trends and improve prediction accuracy. Finally, by combining the predicted deformation data with the current landslide deformation zone image, the current landslide boundary can be updated along the deformation direction, thereby obtaining the expansion area of ​​the landslide deformation zone.

[0043] The landslide deformation zone monitoring method based on the combination of surface and deep data in this exemplary embodiment will be further described below.

[0044] Step S110 , collecting surface deformation data and deep monitoring data of the landslide deformation zone area in multiple observation periods, and constructing a time series monitoring sequence corresponding to each monitoring point in the landslide deformation zone area based on the surface deformation data and the deep monitoring data.

[0045] Among them, the landslide deformation zone can represent the spatial range of significant deformation activity in the landslide area, which can be located in areas prone to landslides, such as mountainous areas, reservoir slopes, road slopes, and mining slopes. The observation period can represent the set of time periods used to collect surface deformation data and deep monitoring data, which has a time sequence, and each observation period corresponds to a complete data collection cycle. The surface deformation data can represent the surface position change data obtained by GNSS, InSAR, optical remote sensing, etc. during the observation period. The deep monitoring data can represent the underground structure state change information obtained by sensors deployed underground in the landslide deformation zone. The monitoring point can represent the fixed spatial location in the landslide deformation zone area for deploying sensors or collecting data. The time series monitoring sequence can represent the data sequence of surface deformation data and deep monitoring data corresponding to a monitoring point in multiple observation periods, arranged in time sequence.

[0046] Further, the ground deformation data can reflect the displacement trend of the landslide deformation zone in the shallow range, and the deep monitoring data can record the stress evolution of the underground structure in the sliding process. After combining the two types of data, the shortcomings of single monitoring method in data representation ability can be effectively made up, and more comprehensive input data is provided for subsequent graph modeling and trend prediction.

[0047] In step S120, the monitoring points are taken as graph nodes, and a graph structure is constructed according to the spatial distribution relationship of the monitoring points.

[0048] The graph nodes can represent the corresponding nodes of the monitoring points in the graph structure, and are used to carry the time series monitoring sequence and other attribute information corresponding to the monitoring points in the graph structure. The graph structure can represent a topological network model composed of graph nodes and the connection relationship between the edges, and is used to simulate the spatial distribution relationship and interaction between the monitoring points in the landslide deformation zone. By constructing the graph structure, the geographical proximity and spatial correlation between the monitoring points in the landslide deformation zone are structurally expressed, so that the spatial coordination features of the landslide deformation zone are more effectively obtained, and the prediction accuracy of the overall deformation trend of the landslide deformation zone is improved.

[0049] In step S130, the time series monitoring sequence is added to the corresponding node in the graph structure as a node attribute, and a multi-dimensional graph model is obtained.

[0050] The multi-dimensional graph model can represent a graph data structure in which the time series monitoring sequence associated with each graph node is embedded as a node attribute on the basis of the graph structure. The model has both spatial topological relationship and time series features. The node attribute is used to record the ground deformation data and deep monitoring data of the corresponding monitoring point at multiple observation periods. By constructing the multi-dimensional graph model, not only the spatial distribution relationship between the monitoring points can be expressed, but also the continuous monitoring data in the time dimension can be associated through the graph nodes.

[0051] In step S140, the multi-dimensional graph model is predicted by using a graph neural network, and the predicted deformation data of each monitoring point at a future observation period is obtained.

[0052] The graph neural network can represent a deep learning model for processing graph structure data, which performs feature propagation and state update operations through edge connections between nodes to extract the mutual dependency between spatial topological features and node attributes in the graph structure. The predicted deformation data can represent the result obtained by processing the multi-dimensional graph model based on the graph neural network, which is used to reflect the deformation trend of displacement, settlement, tilt or stress change of each monitoring point in the future observation period. By using the graph neural network to predict the multi-dimensional graph model, the spatial connection relationship between the monitoring points and the time series monitoring features of each node can be used to accurately predict the landslide deformation trend in the future observation period.

[0053] In step S150, the predicted deformation data and the current landslide deformation zone image are combined to determine the expansion area of the landslide deformation zone.

[0054] The current landslide deformation zone image can represent image data reflecting the surface morphology and boundary contour of the landslide deformation zone in the current observation period. The image can be obtained by remote sensing image, unmanned aerial vehicle aerial photography, optical photography, etc. The expansion area can represent a newly added spatial area with potential deformation risk inferred in the extension direction outside the landslide boundary according to the predicted deformation data.

[0055] Next, the technical content in the above embodiments will be described in detail.

[0056] In some embodiments, the ground surface deformation data and deep monitoring data of the landslide deformation zone in multiple observation periods are collected, specifically including: collecting ground surface deformation data and deep monitoring data in multiple observation periods by using GNSS antenna devices and tilt sensors arranged at multiple target monitoring points in the landslide deformation zone; the ground surface deformation data includes horizontal displacement and vertical settlement, and the deep monitoring data includes tilt angle change data and stress data.

[0057] The GNSS antenna device can represent a positioning device based on the Global Navigation Satellite System (GNSS). It is deployed on the surface at monitoring points in the landslide deformation zone and is used to collect surface deformation data at the monitoring points over multiple observation periods. The tilt sensor can represent a sensor device used to measure the angular change of the target monitoring point relative to its initial position. It can be installed in shallow or deep boreholes and deployed along a vertical profile to collect deep monitoring data. Horizontal displacement and vertical settlement can represent the spatial offset of the monitoring point relative to its initial position in the horizontal and vertical directions, respectively, over the observation period. Tilt angle change data can represent the tilt angle change data of the monitoring point in the vertical profile, recorded by the tilt sensor over multiple observation periods. Stress data can represent the stress change data on the underground structure at the monitoring point. By deploying the GNSS antenna device and tilt sensor in the landslide deformation zone to collect surface deformation data and deep monitoring data, deformation information at different depths can be simultaneously acquired, improving the comprehensiveness of landslide deformation data.

[0058] In some embodiments, based on surface deformation data and deep monitoring data, a time series monitoring sequence corresponding to each monitoring point in the landslide deformation zone area is constructed, specifically including: determining weight data corresponding to each monitoring point according to the terrain data corresponding to each monitoring point in the landslide deformation zone area, wherein each parameter in the weight data corresponds to a different type of data in the time series monitoring sequence; based on surface deformation data and deep monitoring data, an initial time series monitoring sequence corresponding to each monitoring point in the landslide deformation zone area is constructed; and weighted processing is performed on the initial time series monitoring sequence according to the weight data to obtain a time series monitoring sequence.

[0059] Among them, the weight data can represent a set of parameters calculated based on the terrain data of the monitoring point, which is used to perform weighted processing on different types of data in the time series monitoring sequence. The time series monitoring initial sequence can represent a data sequence composed of surface deformation data and deep monitoring data under multiple observation periods arranged in chronological order before weighted processing. In this embodiment, the surface deformation data and deep monitoring data are differentially weighted by introducing terrain data, so that when constructing the time series monitoring sequence, different importance can be given to each type of data according to the geomorphological characteristics of the monitoring point, thereby enhancing the main features and reducing the secondary features.

[0060] In some embodiments, weight data corresponding to each monitoring point is determined based on the terrain data corresponding to each monitoring point in the landslide deformation zone, specifically including: obtaining slope data, elevation data, terrain curvature and slope foot distance corresponding to each monitoring point in the landslide deformation zone; determining weight data based on the slope data, elevation data, terrain curvature and slope foot distance; wherein the slope data is used to determine a first weight corresponding to the horizontal displacement, the elevation data is used to determine a second weight corresponding to the vertical settlement, the terrain curvature is used to determine a third weight corresponding to the inclination angle change data, and the slope foot distance is used to determine a fourth weight corresponding to the stress data.

[0061] Slope data represents the inclination angle of the ground surface at the monitoring point relative to the horizontal. Elevation data represents the vertical height of the monitoring point relative to a uniform datum. Terrain curvature represents the degree of change in the horizontal and vertical curvature of the terrain surface. Toe distance represents the horizontal distance from the monitoring point to the lower edge of the landslide, i.e., the toe of the slope.

[0062] For example, define The slope value of each monitoring point is , the elevation value is , the terrain curvature value is , the distance to the foot of the slope is The first weight corresponding to the horizontal displacement can be determined by the following formula:

[0063]

[0064] in, Indicates the The first weight of each monitoring point, Indicates the The slope value of the point, Indicates the maximum slope value of all monitoring points in the landslide deformation zone. Represents a small positive number used to prevent the denominator from being zero.

[0065] The second weight corresponding to the vertical settlement can be determined by the following formula:

[0066]

[0067] in, Indicates the The second weight of each monitoring point, Indicates the elevation of the monitoring point. Indicates the elevation of the highest point in the monitoring area.

[0068] The third weight corresponding to the tilt angle change data can be determined by the following formula:

[0069]

[0070] in, Indicates the The third weight of each monitoring point, represents the terrain curvature value, Indicates the maximum absolute value of the curvature of all monitoring points.

[0071] The fourth weight corresponding to the stress data can be determined by the following formula:

[0072]

[0073] in, Indicates the The fourth weight of each monitoring point, Indicates the horizontal distance from the monitoring point to the slope foot. represents the control coefficient.

[0074] After obtaining the weight coefficients, the various types of data in the initial time series monitoring sequence are multiplied by the corresponding weight coefficients to obtain a weighted time series monitoring sequence. Of course, in other embodiments of the present disclosure, the time series monitoring sequence can also be constructed directly using the collected surface deformation data and deep monitoring data without weighting.

[0075] In some embodiments, the monitoring points are used as graph nodes, and a graph structure is constructed according to the spatial distribution relationship of the monitoring points, specifically including: determining the Euclidean distance, elevation difference and aspect angle between any two monitoring points according to the position coordinates, elevation data and aspect information corresponding to the monitoring points; establishing a graph connection edge between the corresponding two monitoring points in response to the Euclidean distance being less than a first threshold, the elevation difference being less than a second threshold and the aspect angle being less than a third threshold; and constructing a graph structure that satisfies terrain constraints based on multiple graph connection edges.

[0076] Among them, the elevation difference can represent the height difference between any two monitoring points in the vertical direction. The aspect angle can represent the directional difference between the surface slopes at two monitoring points, that is, the minimum angle difference between the aspect angle values ​​of the two points. For example, the first threshold value can be any value between 20 meters and 30 meters, the second threshold value can be any value between 5 meters and 15 meters, and the third threshold value can be any value between 20 degrees and 45 degrees. Of course, in other embodiments of the present disclosure, the sizes of the first threshold value, the second threshold value, and the third threshold value can be specifically set according to the actual application scenario.

[0077] In this example, by combining the spatial location, elevation data, and aspect information of monitoring points, a triple terrain constraint—Euclidean distance, elevation difference, and aspect angle—is set to select edges that demonstrate both geomorphic continuity and physical accessibility, thereby constructing a graph structure that conforms to the actual landslide path. The Euclidean distance constraint limits the spatial extent of graph edges, preventing long-distance connections without physical connections. The elevation difference constraint eliminates edge connections in areas with sudden elevation changes, helping to maintain vertical connectivity between nodes in the graph. Finally, the aspect angle constraint ensures that adjacent nodes lie on slopes with consistent terrain trends.

[0078] In some embodiments, reference Figure 2 As shown in the figure, a graph neural network is used to predict the multidimensional graph model to obtain the predicted deformation data of each monitoring point in the future observation period, including:

[0079] In step S210 , the time series monitoring sequence corresponding to each node in the multi-dimensional graph model is segmented according to a preset time step to obtain a node feature sequence set divided based on the time step.

[0080] The time step can represent the time interval parameter used to divide the time series monitoring sequence into multiple equally spaced subsequences according to the time dimension, which can be other appropriate time intervals such as 10 days, months, quarters, etc. The node feature sequence set can represent the structured input data set formed by dividing the time series monitoring sequence of each node in the multidimensional graph model by time steps, where each node corresponds to a set of node feature vector sequences obtained by segmenting according to the time step, which is used to reflect the state change characteristics of the node at each time step.

[0081] In some embodiments, the timing monitoring sequence corresponding to each node in the multidimensional graph model is segmented according to a preset time step, specifically including: normalizing the timing monitoring sequence corresponding to each node in the multidimensional graph model based on a linear normalization method to obtain a normalized monitoring sequence; segmenting the normalized monitoring sequence corresponding to each node according to a preset time step.

[0082] Linear normalization refers to a method for normalizing raw numerical data using a linear mapping function. This method scales the processed data by the ratio between its minimum and maximum values, linearly mapping the raw data to a specified interval, which can be [0, 1]. This normalization ensures consistency in the numerical scale of different types of monitoring data, preventing the impact of dimensional differences or different value ranges on subsequent graph neural network predictions.

[0083] In step S220, the node features corresponding to each time step in the node feature sequence set are sequentially input into the graph neural network, and feature propagation processing is performed based on the edge connection relationship in the multidimensional graph model in each time step.

[0084] A node feature represents a data vector corresponding to a node in a graph structure. It consists of normalized monitoring data at the monitoring point at the current time step and serves as input to the graph neural network at the current time step. Feature propagation processing is the process of aggregating and updating the node features of each node and its adjacent nodes in a graph neural network based on the edge connections in the graph structure.

[0085] In some embodiments, the node features corresponding to each time step in the node feature sequence set are sequentially input into the graph neural network, and feature propagation processing is performed in each time step based on the edge connection relationship in the multidimensional graph model, specifically including: determining the adjacent nodes corresponding to each node based on the edge connection relationship in the multidimensional graph model; using edge weights to weight the node feature sequences corresponding to the adjacent nodes at the current time step to obtain the aggregated features of each node at the current time step; and fusing the aggregated features with the node feature sequence of the current node to generate the graph representation state of the current node at the current time step.

[0086] Edge connections can represent the spatial adjacency or structural correlation between any two nodes in a graph. Edge weights can represent numerical weight parameters attached to edge connections in a graph, reflecting the degree of influence between connected nodes in the feature propagation process. These weights can be calculated based on spatial distance, terrain similarity, or correlation with historical monitoring data. Aggregate features can represent the feature results of a node calculated based on the node feature vectors of its neighboring nodes and the corresponding edge weights at the current time step.

[0087] In the specific implementation, firstly, based on the edge connection relationship in the multi-dimensional graph model, for each node , determine its adjacent node set, recorded as ,in Representation and Node There are all nodes with directly connected edges. At the current time step Next, let each adjacent node The node feature vector of ,in Representation node At time step The corresponding normalized monitoring data constitutes the feature vector. With node The edge weight between .

[0088] Secondly, perform edge weighted aggregation on all adjacent node feature vectors to obtain node At time step The aggregate feature vector of , the calculation formula is:

[0089]

[0090] in, Representation node At time step The adjacent feature vectors obtained by aggregation, represents the edge connection weight, Represents adjacent nodes The node feature vector of .

[0091] Then, the node The original node feature vector at the current time step and aggregated feature vectors Perform fusion processing to generate its The graph represents the state The fusion process can be achieved through weighted linear combination, and the calculation formula is:

[0092]

[0093] in, Representation node At time step The diagram represents the state, represents the transformation weight matrix of the original eigenvector, represents the transformation weight matrix of the aggregated feature vector, represents the bias vector, Represents the activation function, which can be a ReLU activation function.

[0094] In step S230, at each time step, the graph representation state of the corresponding node is updated based on the results of the feature propagation process, and the graph representation state is passed to the next time step as the state input for subsequent node features. The graph representation state can represent the expression vector formed by fusing the node's own features with the features of adjacent nodes in the graph neural network at a specific time step.

[0095] Step S240 , in response to the node features completing the iterative operation at all time steps and forming a final graph representation state, generates predicted deformation data of each monitoring point in the target prediction period based on the final graph representation state.

[0096] Specifically, after the node features complete the iterative calculation of the graph neural network at all time steps, the graph representation state corresponding to each node at the last time step is extracted. The final graph representation state contains the deformation change information of the node itself during the entire historical observation period, as well as the feature propagation results between it and the adjacent nodes at each time step, which can comprehensively reflect the structural response characteristics of the node in the dual dimensions of time and space. Subsequently, the above-mentioned graph representation state is input into the prediction module for processing. The prediction module can be a neural network structure with nonlinear mapping capabilities, or it can be a preset regression analysis model, which is used to convert the graph representation state into the deformation estimation result of the node in the target prediction period. The prediction result includes deformation data such as future horizontal displacement, vertical settlement, inclination angle change and stress change corresponding to the monitoring point, which is used to reflect the structural deformation that may occur at the monitoring point in the future period.

[0097] In some embodiments, the predicted deformation data for each monitoring point in the target prediction period is generated based on the final graphical representation state, specifically including the following process: for each monitoring point, the corresponding final graphical representation state is input into a prediction network, which is a neural network structure with nonlinear mapping capabilities. In the prediction network, a feature transformation operation is performed on the final graphical representation state to extract a predicted feature expression related to the deformation trend. Based on the predicted feature expression, the predicted deformation data corresponding to the monitoring point in the target prediction period is output. The predicted deformation data may include horizontal displacement, vertical settlement, tilt angle change data, and stress change data related to the monitoring point.

[0098] In some embodiments, reference Figure 3 As shown in the figure, the predicted deformation data and the current landslide deformation zone image are combined to determine the extension area of ​​the landslide deformation zone, which specifically includes the following technical steps:

[0099] Step S310: determining the deformation gradient vector of each monitoring point based on the predicted deformation data.

[0100] Among them, the deformation gradient vector can represent a vector calculated by the change rate of the predicted deformation data in the spatial distribution, reflecting the direction and amplitude of the deformation change of each monitoring point relative to its surrounding points during the prediction period. In the specific implementation, the predicted deformation data of multiple monitoring points in the landslide deformation zone area during the target prediction period are first obtained. The predicted deformation data is the spatial deformation information corresponding to each monitoring point output based on the graph neural network. For each monitoring point, its spatial coordinate information within the landslide deformation zone area is determined, and a deformation gradient vector field is constructed based on the spatial difference between the corresponding predicted deformation data and the predicted deformation data of the adjacent monitoring points. According to the deformation gradient vector field, the deformation gradient vector of each monitoring point is extracted.

[0101] Step S320 : extracting the boundary contour of the landslide area from the current landslide deformation zone image, and constructing a mask image based on the boundary contour.

[0102] The boundary contour can represent the set of edge lines between the identified landslide area and the surrounding non-sliding area in the current landslide deformation zone image, which is used to define the existing spatial extent of the landslide area. The boundary contour can be obtained through image recognition techniques such as image edge extraction or segmentation algorithms. The mask image can represent a binary layer constructed based on the boundary contour in the landslide deformation zone image, in which the image pixels corresponding to the landslide area are set as the foreground and other areas as the background. By extracting the boundary contour, the actual distribution range of the landslide can be accurately located. By constructing the mask image, the relevant areas in the image can be effectively separated from the non-relevant areas, providing constraints for the subsequent deformation direction identification and image area expansion processing.

[0103] Step S330 : determining the extension direction of the deformation gradient vector relative to the boundary of the mask image, and performing a dilation operation in the area corresponding to the extension direction to generate an extended area of ​​the landslide deformation zone.

[0104] The extension direction can represent the directional component of the deformation gradient vector pointing outside the mask near the boundary contour, which is used to indicate the boundary expansion direction of the predicted deformation trend. The dilation operation can be used to represent the image area expansion processing along the extension direction area based on the mask image boundary during the image processing process, which is used to extend the current landslide area by a specific pixel unit in the predicted direction.

[0105] Specifically, for the boundary pixel points in the mask image, the image normal vector direction corresponding to each boundary point is determined, and the normal vector direction is used to characterize the local external orientation of the boundary. For each boundary point, its corresponding deformation gradient vector is obtained, and the deformation gradient vector is compared with the corresponding image normal vector. If the angle between the deformation gradient vector and the normal vector is less than the preset angle threshold, the direction is determined to be the boundary extension orientation. According to the determined extension orientation, a direction-guided pixel expansion operation is performed in the corresponding extension area of ​​the mask image. The pixel expansion operation includes expanding the pixel points of a specific step size in the image coordinates according to the extension direction, and setting its mask value to the landslide area mark value to complete the boundary expansion of the mask image. The mask image updated after the expansion operation is output as the extension area layer of the landslide deformation zone. The extension area layer can be used to represent the possible extension range of the landslide area within the prediction period.

[0106] In addition, in other embodiments of the present disclosure, after obtaining the extended area of ​​the landslide deformation zone, landslide warning can be further performed based on the extended area. Specifically, the extended area can be spatially superimposed and analyzed with the existing basic geographic information data to identify target objects such as residential areas, roads, power transmission lines, and rivers covered by the extended area. Subsequently, multiple risk levels are set in combination with the predicted deformation amplitude of each monitoring point in the extended area. For example, based on the maximum horizontal displacement prediction value, maximum settlement or stress concentration degree in the extended area, it can be divided into multiple warning levels according to preset thresholds. After completing the risk classification, a warning map with spatial boundaries, risk level identifiers and key point annotations can be generated on the landslide area layer, and the warning map can be pushed to the monitoring terminal as the basis for landslide trend analysis and emergency response, thereby achieving early identification and linkage response of landslide risks.

[0107] It should be noted that although the steps of the method disclosed herein are depicted in a particular order in the accompanying drawings, this does not require or imply that the steps must be performed in that particular order, or that all steps must be performed to achieve the desired result. Additionally or alternatively, certain steps may be omitted, multiple steps may be combined into one, and / or one step may be decomposed into multiple steps.

[0108] In addition, in this exemplary embodiment, a landslide deformation zone monitoring system based on the combination of surface and deep data is also provided. Figure 4 As shown, the landslide deformation zone monitoring system 400 based on the combination of surface and deep data may include: a sequence construction module 410, a graph structure establishment module 420, a graph model construction module 430, a deformation prediction module 440 and an expansion area module 450. Among them:

[0109] The sequence construction module 410 may be used to collect surface deformation data and deep monitoring data of the landslide deformation zone area during multiple observation periods, and construct a time series monitoring sequence corresponding to each monitoring point in the landslide deformation zone area based on the surface deformation data and the deep monitoring data;

[0110] The graph structure building module 420 may be used to use monitoring points as graph nodes and build a graph structure according to the spatial distribution relationship of the monitoring points;

[0111] The graph model construction module 430 may be used to add the time series monitoring sequence as a node attribute to the corresponding node in the graph structure to obtain a multi-dimensional graph model;

[0112] The deformation prediction module 440 may be configured to use a graph neural network to predict the multidimensional graph model and obtain predicted deformation data of each monitoring point in a future observation period;

[0113] The expansion area module 450 may be configured to combine the predicted deformation data with the current landslide deformation zone image to determine the expansion area of ​​the landslide deformation zone.

[0114] The specific details of each module in the above landslide deformation zone monitoring system based on the combination of surface and deep data have been described in detail in the corresponding landslide deformation zone monitoring method based on the combination of surface and deep data, so they will not be repeated here.

[0115] It should be noted that although the detailed description above mentions several modules or units of the landslide deformation zone monitoring system based on the combination of surface and deep data, this division is not mandatory. In fact, according to the embodiments of the present disclosure, the features and functions of two or more modules or units described above can be embodied in a single module or unit. Conversely, the features and functions of a single module or unit described above can be further divided and embodied by multiple modules or units.

[0116] In addition, in an exemplary embodiment of the present disclosure, an electronic device capable of implementing the above-mentioned landslide deformation zone monitoring method based on the combination of surface and deep data is also provided.

[0117] Those skilled in the art will appreciate that various aspects of the present disclosure may be implemented as systems, methods, or program products. Therefore, various aspects of the present disclosure may be implemented in the following forms: a complete hardware embodiment, a complete software embodiment (including firmware, microcode, etc.), or an embodiment combining hardware and software aspects, which may be collectively referred to herein as a "circuit," "module," or "system."

[0118] Refer to the following Figure 5 hereinafter, an electronic device 500 according to such an embodiment of the present disclosure is described. Figure 5 The electronic device 500 shown is merely an example and should not limit the functions and scope of use of the embodiments of the present disclosure.

[0119] like Figure 5 As shown, electronic device 500 is implemented as a general-purpose computing device. Components of electronic device 500 may include, but are not limited to, the aforementioned at least one processing unit 510, the aforementioned at least one storage unit 520, a bus 530 connecting various system components (including storage unit 520 and processing unit 510), and a display unit 540.

[0120] The storage unit stores program code, which can be executed by the processing unit 510, causing the processing unit 510 to perform the steps described in the "Exemplary Methods" section above according to various exemplary embodiments of the present disclosure. The storage unit 520 may include a readable medium in the form of a volatile memory unit, such as a random access memory unit (RAM) 521 and / or a cache memory unit 522, and may further include a read-only memory unit (ROM) 523.

[0121] The storage unit 520 may also include a program / utility 524 having a set (at least one) of program modules 525, such program modules 525 including but not limited to: an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment.

[0122] Bus 530 may represent one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, an accelerated graphics port, a processing unit, or a local bus using any of a variety of bus architectures.

[0123] The electronic device 500 can also communicate with one or more external devices 570 (e.g., a keyboard, pointing device, Bluetooth device, etc.), one or more devices that enable a user to interact with the electronic device 500, and / or any device that enables the electronic device 500 to communicate with one or more other computing devices (e.g., a router, modem, etc.). This communication can occur via an input / output (I / O) interface 550. Furthermore, the electronic device 500 can communicate with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network such as the Internet) via a network adapter 560. As shown, the network adapter 560 communicates with other modules of the electronic device 500 via a bus 530. It should be understood that, although not shown in the figure, other hardware and / or software modules can be used in conjunction with the electronic device 500, including but not limited to microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0124] Through the description of the above embodiments, those skilled in the art will readily understand that the exemplary embodiments described herein may be implemented by software, or by combining software with necessary hardware.

[0125] In exemplary embodiments of the present disclosure, a computer-readable storage medium is also provided, on which is stored a program product capable of implementing the aforementioned methods of this specification. In some possible embodiments, various aspects of the present disclosure may also be implemented in the form of a program product, which includes program code. When the program product is executed on a terminal device, the program code is used to cause the terminal device to execute the steps according to various exemplary embodiments of the present disclosure described in the "Exemplary Methods" section of this specification.

[0126] refer to Figure 6 As shown, a program product 600 for implementing the above-described landslide deformation zone monitoring method based on the integration of surface and deep data according to an embodiment of the present disclosure is described. This program product 600 can be implemented in a portable compact disc read-only memory (CD-ROM) and include program code, and can be executed on a terminal device, such as a personal computer. However, the program product of the present disclosure is not limited thereto. In this document, a readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0127] The program product may utilize any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. The readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or component, or any combination thereof. More specific examples (a non-exhaustive list) of readable storage media include: an electrical connection having one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof.

[0128] A computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, which carries readable program code. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A readable signal medium may also be any readable medium other than a readable storage medium that can transmit, propagate, or transfer a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0129] Furthermore, the figures above are merely illustrative of the processes included in the methods according to exemplary embodiments of the present disclosure and are not intended to be limiting. It is readily understood that the processes illustrated in the figures above do not indicate or limit the temporal order of these processes. Furthermore, it is readily understood that these processes may be executed synchronously or asynchronously, for example, in multiple modules.

[0130] It should be understood that the present disclosure is not limited to the exact structures that have been described above and shown in the drawings, and that various modifications and changes can be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

Claims

1. A landslide deformation zone monitoring method based on the combination of surface and deep data, characterized in that: include: Collecting surface deformation data and deep monitoring data in the landslide deformation zone during multiple observation periods, and constructing a time series monitoring sequence corresponding to each monitoring point in the landslide deformation zone based on the surface deformation data and the deep monitoring data; The monitoring points are used as graph nodes, and a graph structure is constructed according to the spatial distribution relationship of the monitoring points; Adding the time series monitoring sequence as a node attribute to a corresponding node in the graph structure to obtain a multi-dimensional graph model; Using a graph neural network to predict the multidimensional graph model, obtaining predicted deformation data of each monitoring point in a future observation period; The predicted deformation data is combined with the current landslide deformation zone image to determine the extension area of ​​the landslide deformation zone.

2. The landslide deformation zone monitoring method based on the combination of surface and deep data according to claim 1 is characterized in that: The acquisition of surface deformation data and deep layer monitoring data in the landslide deformation zone during multiple observation periods includes: Using GNSS antenna devices and tilt sensors deployed at multiple target monitoring points in the landslide deformation zone, collecting the surface deformation data and the deep monitoring data for multiple observation periods; The surface deformation data includes horizontal displacement and vertical settlement, and the deep layer monitoring data includes tilt angle change data and stress data.

3. The landslide deformation zone monitoring method based on the combination of surface and deep data according to claim 1 is characterized in that: The step of constructing a time series monitoring sequence corresponding to each monitoring point in the landslide deformation zone based on the surface deformation data and the deep layer monitoring data includes: Determining weight data corresponding to each monitoring point according to the terrain data corresponding to each monitoring point in the landslide deformation zone, wherein each parameter in the weight data corresponds to a different type of data in the time series monitoring sequence; Based on the surface deformation data and the deep layer monitoring data, constructing an initial time series monitoring sequence corresponding to each monitoring point in the landslide deformation zone; The initial sequence of time series monitoring is weighted according to the weight data to obtain the time series monitoring sequence.

4. The landslide deformation zone monitoring method based on the combination of surface and deep data according to claim 3 is characterized in that: Determining weight data corresponding to each monitoring point according to the terrain data corresponding to each monitoring point in the landslide deformation zone includes: Obtaining slope data, elevation data, terrain curvature, and distance to the slope foot corresponding to each monitoring point in the landslide deformation zone; Determining the weight data according to the slope data, elevation data, terrain curvature and distance to the foot of the slope; Among them, the slope data is used to determine the first weight corresponding to the horizontal displacement, the elevation data is used to determine the second weight corresponding to the vertical settlement, the terrain curvature is used to determine the third weight corresponding to the inclination angle change data, and the slope foot distance is used to determine the fourth weight corresponding to the stress data.

5. The landslide deformation zone monitoring method based on the combination of surface and deep data according to claim 1 is characterized in that: The monitoring points are used as graph nodes, and a graph structure is constructed according to the spatial distribution relationship of the monitoring points, including: Determine the Euclidean distance, elevation difference, and slope angle between any two monitoring points based on the location coordinates, elevation data, and slope information corresponding to the monitoring points; In response to the Euclidean distance being less than a first threshold, the elevation difference being less than a second threshold, and the aspect angle being less than a third threshold, establishing a graph connection edge between the corresponding two monitoring points; The graph structure satisfying the terrain constraint condition is constructed according to a plurality of graph connection edges.

6. The landslide deformation zone monitoring method based on the combination of surface and deep data according to claim 1 is characterized in that: The method of using a graph neural network to predict the multidimensional graph model and obtain predicted deformation data of each monitoring point in a future observation period includes: Segmenting the time series monitoring sequence corresponding to each node in the multidimensional graph model according to a preset time step to obtain a node feature sequence set based on time step division; The node features corresponding to each time step in the node feature sequence set are sequentially input into the graph neural network, and feature propagation processing is performed based on the edge connection relationship in the multidimensional graph model in each time step; In each time step, the graph representation state of the corresponding node is updated according to the result of the feature propagation process, and the graph representation state is passed to the next time step as the state input of the subsequent node feature; In response to the node features completing the iterative operation at all time steps and forming a final graph representation state, predicted deformation data of each monitoring point in a target prediction period is generated based on the final graph representation state.

7. The landslide deformation zone monitoring method based on the combination of surface and deep data according to claim 6 is characterized in that: The segmenting of the time series monitoring sequence corresponding to each node in the multi-dimensional graph model according to a preset time step includes: Normalizing the time series monitoring sequence corresponding to each node in the multidimensional graph model based on a linear normalization method to obtain a normalized monitoring sequence; The normalized monitoring sequence corresponding to each node is segmented according to the preset time step.

8. The landslide deformation zone monitoring method based on the combination of surface and deep data according to claim 6 is characterized in that: The feature propagation processing based on the edge connection relationship in the multi-dimensional graph model includes: Determining adjacent nodes corresponding to each node based on edge connection relationships in the multidimensional graph model; The node feature sequence corresponding to the adjacent nodes at the current time step is weighted using the edge weight to obtain the aggregated feature of each node at the current time step; The aggregated features are fused with the node feature sequence of the current node to generate a graph representation state of the current node at the current time step.

9. The landslide deformation zone monitoring method based on the combination of surface and deep data according to claim 1 is characterized in that: The step of combining the predicted deformation data with the current landslide deformation zone image to determine the extension area of ​​the landslide deformation zone includes: Determining the deformation gradient vector of each of the monitoring points based on the predicted deformation data; Extracting a boundary contour of the landslide area from the current landslide deformation zone image, and constructing a mask image based on the boundary contour; The extension direction of the deformation gradient vector relative to the boundary of the mask image is determined, and a dilation operation is performed in a region corresponding to the extension direction to generate an extended region of the landslide deformation zone.

10. A landslide deformation zone monitoring system based on the combination of surface and deep data, characterized in that: include: A sequence construction module is used to collect surface deformation data and deep monitoring data in the landslide deformation zone area during multiple observation periods, and to construct a time series monitoring sequence corresponding to each monitoring point in the landslide deformation zone area based on the surface deformation data and the deep monitoring data; A graph structure building module is used to use monitoring points as graph nodes and build a graph structure according to the spatial distribution relationship of the monitoring points; A graph model construction module, configured to add the time series monitoring sequence as a node attribute to a corresponding node in the graph structure to obtain a multi-dimensional graph model; A deformation prediction module is used to predict the multidimensional graph model using a graph neural network to obtain predicted deformation data of each monitoring point in a future observation period; The expansion area module is used to combine the predicted deformation data with the current landslide deformation zone image to determine the expansion area of ​​the landslide deformation zone.

Images