A method and device for early warning of an accelerated deformation stage of a slope, a medium and a product

Abstract

A slope accelerated deformation stage early warning method, device, medium and product are related to the field of slope early warning. In the method, first deformation data is extracted based on an optical image sequence, and second deformation data is extracted based on radar data; the first deformation data and the second deformation data are time-space registered to form corresponding joint deformation time series on each monitoring point; the joint deformation time series of each monitoring point is time-space mode decomposed to obtain multiple intrinsic mode function components of different frequency scales; the intrinsic mode function components are classified into slow trend items, periodic fluctuation items and sudden change components; the mode energy and energy proportion of the slow trend items, the periodic fluctuation items and the sudden change components are calculated; based on the mode energy and the energy proportion, it is determined that each monitoring point is in a deformation stage; based on the deformation stage of each monitoring point, a slope risk level and early warning information are generated and output. The application is implemented, and the reliability of the slope deformation early warning is improved.

Description

Technical Field

[0001] This application relates to the field of slope early warning, specifically to a method, device, medium, and product for early warning of the accelerated deformation stage of a slope. Background Technology

[0002] With the rapid development of global infrastructure construction, especially mountain highways, railways, large-scale water conservancy and hydropower projects, and open-pit mines, various steep slope engineering projects are increasing. The long-term stability of slopes is directly related to the safe operation of engineering projects, the safety of life and property, and even social stability. Catastrophic accidents caused by geological disasters such as landslides and collapses are common, causing huge economic losses and casualties. Therefore, long-term, real-time, and high-precision monitoring of slope deformation, and the ability to accurately identify its evolutionary stages from stability to instability, and to detect the turning point of accelerated deformation in advance, has become a key technical challenge and core requirement in the field of disaster prevention and mitigation.

[0003] Currently, the identification and early warning technologies for slope deformation stages suffer from the following shortcomings that urgently need to be addressed: First, the monitoring information from a single sensor is insufficient. For example, while ground-based radar can provide high-precision line-of-sight one-dimensional deformation, it cannot capture the two-dimensional planar motion characteristics of the landslide surface, leading to misjudgments of the overall slope instability mode. While optical image measurement can acquire two-dimensional displacement fields, it is easily affected by lighting and weather conditions and struggles to achieve millimeter-level monitoring accuracy, failing to effectively capture early, minute deformations. Second, existing stage identification methods largely rely on empirical time-domain thresholds. These methods trigger alarms by calculating whether the deformation rate or acceleration exceeds a certain fixed threshold. This is not only extremely sensitive to data noise but also fails to effectively distinguish, from a physical perspective, periodic fluctuations caused by external environmental factors such as rainfall and temperature changes from the irreversible accelerated creep trend truly caused by the accumulation of damage to the slope's internal structure, thus frequently leading to false alarms or missed alarms. Finally, existing technologies lack analytical methods for deeply fusing multi-source data in the time and frequency domains. Even when studies combine multiple data sources, they often remain at the level of simple superposition and comparison of results, failing to perform joint mode decomposition of the two-dimensional planar motion information provided by optical images and the high-precision deformation data from radar. In summary, existing technologies have shortcomings in data dimensionality, analysis methods, and fusion depth, resulting in low reliability of early warning systems. Summary of the Invention

[0004] This application provides a method, device, medium, and product for early warning of accelerated deformation stages of slopes, which improves the reliability of slope deformation early warning.

[0005] A first aspect of this application provides a method for early warning of accelerated deformation stage of a slope. The method includes: acquiring an optical image sequence and radar data of a target area of ​​the slope; extracting first deformation data based on the optical image sequence and extracting second deformation data based on the radar data, wherein the first deformation data is two-dimensional planar displacement field data of multiple monitoring points in the target area of ​​the slope, and the second deformation data is line-of-sight deformation field data of the multiple monitoring points; performing spatiotemporal registration of the first deformation data and the second deformation data to form a corresponding joint deformation time sequence at each monitoring point; and processing the joint deformation time sequence at each monitoring point. The sequence undergoes spatiotemporal modal decomposition to obtain multiple intrinsic mode function (IMF) components at different frequency scales. Based on the center frequencies of the IMF components, they are classified into slow trend terms, periodic fluctuation terms, and abrupt change components. The modal energy and energy percentage of the slow trend terms, periodic fluctuation terms, and abrupt change components are calculated. Based on the modal energy and energy percentage, the deformation stage of each monitoring point is determined, including the initial creep stage, constant creep stage, accelerated creep stage, and imminent slip stage. Based on the deformation stage of each monitoring point, slope risk level and early warning information are generated and output.

[0006] By adopting the above technical solutions and constructing a complete early warning method for accelerated slope deformation stages, high-precision slope deformation monitoring and early warning through multi-source data fusion were achieved. Optical image sequences and radar data of the target slope area were acquired, fully utilizing the complementary advantages of the two data sources. Optical images provide high spatial resolution two-dimensional planar displacement field information, while radar data provides high-precision line-of-sight deformation measurement capabilities. By extracting the first and second deformation data, comprehensive observation of slope deformation was achieved. Spatiotemporal registration formed a joint deformation time series, effectively integrating deformation information from heterogeneous data sources and improving the integrity and reliability of the monitoring data. Spatiotemporal modal decomposition yielded intrinsic mode function components at different frequency scales, achieving refined separation of complex deformation signals and accurately identifying deformation characteristics at different time scales. The intrinsic mode function components were classified into slow trend terms, periodic fluctuation terms, and abrupt change components, clearly distinguishing various deformation components and facilitating subsequent quantitative analysis. By calculating modal energy and energy proportion, a quantitative assessment of the contribution of various deformation components was achieved. Based on modal energy and energy ratio to determine deformation stages, a complete evolutionary sequence identification mechanism from the initial creep stage to the pre-slip stage was established, providing a scientific basis for slope instability early warning. Finally, slope risk levels and early warning information are generated and output, achieving fully automated processing from data acquisition to risk assessment, providing timely and reliable decision support for slope disaster prevention and control.

[0007] Optionally, the step of performing spatiotemporal registration of the first deformation data and the second deformation data to form a corresponding joint deformation time series at each of the monitoring points specifically includes: constructing a three-dimensional spatial grid coordinate system based on the digital elevation model of the slope target area, wherein the three-dimensional spatial grid coordinate system contains the three-dimensional coordinates of each of the monitoring points; mapping the two-dimensional pixel coordinates of each of the monitoring points in the first deformation data to the three-dimensional spatial grid coordinate system through camera model parameters to obtain first registration data; registering the radar coordinates of each of the monitoring points in the second deformation data to the three-dimensional spatial grid coordinate system to obtain second registration data; and interpolating and aligning the first registration data and the second registration data in the time dimension to form the joint deformation time series at each of the monitoring points.

[0008] By adopting the above technical solution and employing a precise spatiotemporal registration method, effective fusion of heterogeneous deformation data was achieved. A three-dimensional spatial grid coordinate system was constructed based on a digital elevation model, providing a unified reference framework for spatial registration of different data sources and ensuring the accuracy and consistency of the three-dimensional coordinates of each monitoring point. The two-dimensional pixel coordinates in the first deformation data were mapped to the three-dimensional spatial grid coordinate system through camera model parameters, achieving a precise conversion of optical image data from two-dimensional image space to three-dimensional geographic space, eliminating the effects of perspective distortion and terrain undulation, and obtaining the first registration data. The radar coordinates in the second deformation data were registered to the three-dimensional spatial grid coordinate system to obtain the second registration data, achieving an accurate conversion of radar slant range measurement to the geographic coordinate system, overcoming the geometric distortion problem of radar side-view imaging. Interpolation and alignment processing in the time dimension solved the problem of asynchronous acquisition time between optical and radar data, ensuring that the joint deformation time series formed at each monitoring point has a unified time reference. This refined spatiotemporal registration process effectively improves the accuracy of multi-source data fusion, provides high-quality input data for subsequent deformation analysis, avoids false deformation signals caused by registration errors, and enhances the reliability of slope deformation monitoring results.

[0009] Optionally, the step of performing spatiotemporal modal decomposition on the joint deformation time series of each monitoring point to obtain multiple intrinsic mode function (IMF) components at different frequency scales specifically includes: using the joint deformation time series as a multivariate input signal and setting the number of IMF components to be decomposed; constructing a constrained variational optimization model with the objective of minimizing the sum of the bandwidths of each IMF component, wherein the constraint is that the sum of each IMF component can reconstruct the joint deformation time series; iteratively solving the variational optimization model using an alternating direction multiplier method that introduces center frequency and Lagrange multipliers, alternately updating each IMF component and its center frequency in each iteration until the convergence condition is met; and outputting the converged IMF components and their corresponding center frequencies, wherein each IMF component is a multidimensional vector with the same dimension as the joint deformation time series.

[0010] By employing the above technical solution and an optimized spatiotemporal mode decomposition method, accurate separation of complex deformation signals is achieved. Using the joint deformation time series as a multivariate input signal and setting the number of intrinsic mode function (EMF) components fully considers the characteristics of multidimensional deformation data, avoiding information loss common in traditional univariate decomposition methods. A constrained variational optimization model with the objective of minimizing the sum of bandwidths is constructed, ensuring that each EMF component obtained from the decomposition has compact frequency support and effectively suppressing mode aliasing. Constraints guarantee that the sum of each EMF component can completely reconstruct the original signal, maintaining the completeness and accuracy of the signal decomposition. An iterative solution using the alternating direction multiplier method, by introducing the center frequency and Lagrange multipliers, achieves efficient solution to complex optimization problems. Alternating updates of the EMF components and center frequency in each iteration ensure the convergence and stability of the algorithm. Each output EMF component is a multidimensional vector with the same dimension as the joint deformation time series, fully preserving the multidimensional characteristics of the deformation signal and providing rich information for subsequent deformation feature analysis. This decomposition method can accurately capture the dynamic characteristics of slope deformation at different time scales, laying a solid foundation for identifying precursor signals of slope instability.

[0011] Optionally, classifying the intrinsic mode function (IMF) components into slow trend terms, periodic fluctuation terms, and abrupt change components based on their center frequencies specifically includes: using a multivariate variational mode decomposition algorithm to decompose the joint deformation time series into a predetermined number of IMF components; calculating the center frequency of each IMF component and matching the center frequency with a predetermined frequency feature library; classifying IMF components with center frequencies lower than a first predetermined frequency threshold as slow trend terms; classifying IMF components with center frequencies matching a predetermined environmental period as periodic fluctuation terms; and classifying the remaining IMF components as abrupt change components.

[0012] By employing the above technical solution and a scientific frequency characteristic classification method, accurate identification of different deformation components was achieved. A multivariate variational mode decomposition algorithm was used to decompose the joint deformation time series into a predetermined number of intrinsic mode function (EMF) components, ensuring the systematic nature and completeness of the decomposition process. The center frequency of each EMF component was calculated, obtaining the frequency characteristic quantification index of each component, providing an objective basis for subsequent classification. Matching the center frequency with a predetermined frequency feature library utilized prior knowledge of slope deformation, improving the accuracy and reliability of classification. Components with center frequencies below a first predetermined frequency threshold were classified as slow trend terms, accurately identifying low-frequency deformation characteristics caused by long-term slope creep and reflecting the cumulative process of internal stress adjustment and plastic deformation in the slope. Components with center frequencies matching a predetermined environmental period were classified as periodic fluctuation terms, effectively separating periodic elastic deformation caused by environmental factors such as temperature changes and seasonal rainfall, avoiding interference from environmental noise on the real deformation signal. The remaining components were classified as abrupt change components, successfully capturing sudden deformation events such as local slope failure and crack propagation; these high-frequency signals are often important precursors to slope instability. This frequency-based intelligent classification enables the interpretation of the physical meaning of complex deformation signals, providing clear characteristic indicators for subsequent deformation stage determination.

[0013] Optionally, the calculation of the modal energy and energy percentage of the slow trend term, the periodic fluctuation term, and the abrupt change component specifically includes: for each intrinsic modal function component, accumulating the square of the vector L2 norm of the intrinsic modal function component at each sampling time of the joint deformation time series to obtain the modal energy of the intrinsic modal function component; summing the modal energies of all intrinsic modal function components to obtain the total modal energy of the monitoring point; summing the modal energies of each intrinsic modal function component classified as the slow trend term, the periodic fluctuation term, and the abrupt change component to obtain the trend term energy, the periodic fluctuation term energy, and the abrupt change component energy; dividing the trend term energy, the periodic fluctuation term energy, and the abrupt change component energy by the total modal energy to obtain the corresponding trend term energy percentage, periodic fluctuation term energy percentage, and abrupt change component energy percentage.

[0014] By adopting the above technical solution and employing a quantitative energy calculation method, a precise assessment of the contribution of various deformation components was achieved. The modal energy is obtained by accumulating the squared vector L2 norm of each intrinsic modal function component over time at each sampling moment. This calculation method fully considers the vector characteristics of multidimensional deformation data and accurately reflects the energy accumulation of each component throughout the monitoring period. The total modal energy is obtained by summing the modal energies of all intrinsic modal function components, establishing an energy normalization benchmark and ensuring the rationality of subsequent energy proportion calculations. The energies of each category of components are summed separately to obtain the trend term energy, periodic fluctuation term energy, and abrupt component energy, achieving energy aggregation from individual components to category levels and improving the stability and representativeness of energy analysis. The energy proportion is calculated by dividing each type of energy by the total modal energy, obtaining dimensionless relative indicators, eliminating the influence of differences in deformation amplitude at different monitoring points, and making the deformation characteristics at different locations comparable. The trend term energy proportion reflects the contribution of long-term creep to the total deformation, the periodic fluctuation term energy proportion reflects the intensity of environmental factors, and the abrupt component energy proportion quantifies the severity of sudden deformation events. This energy-based quantitative analysis method provides a reliable numerical basis for the objective determination of slope deformation stages.

[0015] Optionally, determining the deformation stage of each monitoring point based on the modal energy and the energy percentage specifically includes: when the energy percentage of the periodic fluctuation term is the highest, the current stage is determined as the initial creep stage; when the energy percentage of the slow trend term replaces the periodic fluctuation term as the energy component with the highest percentage, and the energy percentage of the abrupt change component does not show a continuous increasing trend, the current stage is determined as the constant-rate creep stage; when the energy percentage of the slow trend term remains the highest and the energy percentage of the abrupt change component shows the continuous increasing trend, the current stage is determined as the accelerated creep stage; when the energy of the abrupt change component is the energy component with the highest percentage or an intrinsic modal function component with a center frequency higher than a preset frequency threshold is identified, the current stage is determined as the slippage stage.

[0016] By adopting the above technical solution and using a stage-based determination rule based on energy proportion, the evolution of slope deformation was accurately identified. When the energy proportion of the periodic fluctuation term is highest, it is determined to be the initial creep stage, accurately capturing the characteristic that early slope deformation is mainly controlled by environmental factors. At this stage, the slope is generally stable, and deformation is mainly elastic adjustment. When the energy proportion of the slow trend term replaces the periodic fluctuation term as the highest and the energy proportion of the abrupt change component does not show a continuous increasing trend, it is determined to be the isochronous creep stage, accurately identifying the slope entering a stable plastic deformation state. At this stage, the internal stress redistribution is basically completed, and the deformation rate is relatively constant. When the energy proportion of the slow trend term remains highest and the energy proportion of the abrupt change component shows a continuous increasing trend, it is determined to be the accelerated creep stage, timely identifying the critical turning point of accelerated internal damage accumulation in the slope, indicating that stability is rapidly deteriorating. When the energy proportion of the abrupt change component is highest or high-frequency intrinsic mode function components are identified, it is determined to be the pre-slip stage, accurately capturing the danger signal that the slope is about to become unstable, providing early warning time for emergency evacuation. This stage-based method of energy evolution establishes a scientific conversion mechanism from quantitative indicators to qualitative states, enabling dynamic monitoring of the deformation process of slopes throughout their entire life cycle, and providing clear guidance for the formulation of graded early warning and differentiated prevention and control measures.

[0017] Optionally, generating and outputting slope risk level and early warning information based on the deformation stage of each monitoring point specifically includes: generating a spatial distribution map of slope deformation stages based on the deformation stages corresponding to each monitoring point; identifying and aggregating adjacent monitoring points in a preset high-level deformation stage into deformation region clusters on the spatial distribution map using a preset clustering algorithm, wherein the preset high-level deformation stages include accelerated creep stage and near-slip stage; calculating a risk index for each deformation region cluster, wherein the risk index is calculated by weighted summation based on the number of monitoring points in the deformation region cluster, the geometric area of ​​the deformation region cluster, and the average energy of the mutation components of each monitoring point in the deformation region cluster; determining the slope risk level of the target slope area based on the highest risk index of each deformation region cluster, and determining the corresponding early warning information based on the slope risk level.

[0018] By adopting the above technical solutions and employing spatial clustering and risk assessment methods, an upgrade from point-based monitoring to regional early warning has been achieved. Spatial distribution maps are generated based on the deformation stages of each monitoring point, intuitively displaying the differences in deformation states across different areas of the slope and providing spatial visualization support for identifying potential hazardous areas. A pre-defined clustering algorithm aggregates adjacent monitoring points at higher deformation stages into deformation region clusters, achieving information integration from discrete points to a continuous surface, accurately delineating the main deformation areas of the slope, and avoiding misjudgments caused by single-point anomalies. The risk index of deformation region clusters comprehensively considers the number of monitoring points, geometric area, and average energy of abrupt changes. The number of monitoring points reflects the spatial breadth of deformation, the geometric area reflects the potential impact range, and the average energy of abrupt changes quantifies the severity of deformation. A comprehensive and objective risk assessment index is obtained through weighted summation. The slope risk level is determined based on the highest risk index, employing a conservative principle to ensure no potential hazards are overlooked. Corresponding early warning information is generated based on the risk level, achieving effective connection between technical analysis and management decision-making, and providing clear and explicit action guidelines for emergency response departments. This regionalized risk assessment approach improves the spatial targeting and practicality of early warning systems, and helps optimize the allocation of disaster prevention resources and the implementation of emergency measures.

[0019] In a second aspect, embodiments of this application provide an early warning device for the accelerated deformation stage of a slope. The early warning device for the accelerated deformation stage of a slope includes: one or more processors and a memory; the memory is coupled to the one or more processors and is used to store computer program code, the computer program code including computer instructions, and the one or more processors call the computer instructions to cause the early warning device for the accelerated deformation stage of a slope to perform the method described in the first aspect and any possible implementation thereof.

[0020] Thirdly, embodiments of this application provide a computer-readable storage medium including instructions that, when executed on a slope accelerated deformation stage early warning device, cause the slope accelerated deformation stage early warning device to perform the method described in the first aspect and any possible implementation thereof.

[0021] Fourthly, embodiments of this application provide a computer program product containing instructions that, when the computer program product is run on a slope accelerated deformation stage early warning device, cause the slope accelerated deformation stage early warning device to perform the method described in the first aspect and any possible implementation thereof.

[0022] In summary, one or more technical solutions provided in this application have at least the following technical effects or advantages: 1. By combining optical images and radar data, multi-dimensional slope deformation information is acquired, and the data is spatiotemporally registered in a unified three-dimensional spatial grid coordinate system to generate a joint deformation time series. By fusing deformation information from different data sources, the accuracy and comprehensiveness of slope deformation monitoring are improved, enabling more accurate capture of the deformation characteristics of the target slope area.

[0023] 2. Spatiotemporal modal decomposition (SMD) technology was used to decompose the joint deformation time series, extracting deformation information at multiple frequency scales (slow trend term, periodic fluctuation term, and abrupt change component). By calculating modal energy and energy proportion, the criteria for different creep stages of the slope were clarified. A rigorous and scientific analytical framework was provided to address the differences in monitoring points and deformation characteristics, effectively improving the sensitivity and accuracy of slope deformation stage determination.

[0024] 3. Based on the energy distribution characteristics and deformation stage determination results of monitoring points, combined with spatial distribution and a preset clustering algorithm, a spatial distribution map of slope deformation stages is generated, and high-risk area clusters are identified. Furthermore, the risk level of target areas of the slope is precisely classified through quantitative assessment using risk indices, and early warning information is output. This significantly improves the accuracy and reliability of slope risk assessment and early warning, providing strong technical support for the early prevention and control of slope disasters. Attached Figure Description

[0025] Figure 1 This is a flowchart illustrating an early warning method for the accelerated deformation stage of a slope disclosed in an embodiment of this application. Figure 2 This is another schematic diagram of a slope accelerated deformation stage early warning method disclosed in the embodiments of this application; Figure 3 This is a schematic diagram of the structure of an early warning device for the accelerated deformation stage of a slope provided in an embodiment of this application.

[0026] Explanation of reference numerals in the attached drawings: 301, Central Processing Unit; 302, Read-Only Memory; 303, Random Access Memory; 304, Bus; 305, Input / Output Interface; 306, Input Section; 307, Output Section; 308, Storage Section; 309, Communication Section; 310, Driver; 311, Removable Media. Detailed Implementation

[0027] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments.

[0028] In the description of the embodiments of this application, the words "for example" or "for instance" are used to indicate examples, illustrations, or explanations. Any embodiment or design that is described as "for example" or "for instance" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design options. Rather, the use of the words "for example" or "for instance" is intended to present the relevant concepts in a specific manner.

[0029] In the description of the embodiments of this application, the term "multiple" means two or more. For example, multiple system devices refer to two or more system devices, and multiple screen terminals refer to two or more screen terminals. Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the indicated technical features. Thus, a feature defined with "first" or "second" may explicitly or implicitly include one or more of that feature. The terms "comprising," "including," "having," and variations thereof all mean "including but not limited to," unless otherwise specifically emphasized.

[0030] This application provides a method for early warning of accelerated deformation stages of slopes, referring to... Figure 1 , Figure 1 This is a flowchart illustrating a method for early warning of accelerated slope deformation stages provided in an embodiment of this application. The method is applied to a server, which can execute an early warning program for accelerated slope deformation stages. The server can be a single server, a server cluster consisting of multiple servers, or a cloud computing service center. The method includes steps S101 to S108, as follows: Step S101: Acquire optical image sequences and radar data of the target area of ​​the slope.

[0031] In step S101, the target area of ​​the slope refers to a specific mountain or slope range that requires deformation monitoring and risk warning. The optical image sequence refers to a series of static images of the target area of ​​the slope taken by a high-resolution camera at consecutive time points. Radar data refers to the raw measurement values ​​related to the surface deformation of the target area of ​​the slope, obtained after processing by a ground-based radar system that transmits and receives electromagnetic waves.

[0032] Specifically, the server performs data acquisition by controlling monitoring equipment deployed on-site. The server instructs a high-resolution optical camera to automatically capture a digital image of the target slope area at preset time intervals, such as every 10 minutes, and organizes these images chronologically to form an optical image sequence. Simultaneously, the server instructs a distributed MIMO ground-based radar system to scan the target slope area at a preset scanning cycle, such as every 30 minutes, acquiring phase information reflecting minute surface displacements and storing this information as radar data.

[0033] Step S102: Extract the first deformation data based on the optical image sequence and extract the second deformation data based on the radar data. The first deformation data is the two-dimensional planar displacement field data of multiple monitoring points in the target area of ​​the slope, and the second deformation data is the line-of-sight deformation field data of multiple monitoring points.

[0034] In step S102, the first deformation data refers to the movement of the ground surface on a plane calculated by analyzing optical image sequences. The second deformation data refers to the movement of the ground surface along the radar line-of-sight direction calculated by analyzing radar data. A monitoring point refers to a virtual or physical point set on the surface of the target slope area for discretizing deformation analysis. Two-dimensional planar displacement field data is used to represent the displacement of each monitoring point in the horizontal and vertical directions. Line-of-sight deformation field data is used to represent the displacement of each monitoring point along the radar beam direction.

[0035] Specifically, the server first processes the acquired optical image sequence. Using sub-pixel image registration algorithms, such as phase correlation or optical flow, the server compares and analyzes consecutive images in the sequence to accurately calculate the horizontal and vertical displacements of multiple monitoring points on the gridded surface of the slope target area in the image coordinate system, forming two-dimensional planar displacement field data with sub-centimeter accuracy. Next, the server performs interferometric processing on the acquired radar data, calculating the deformation time series of each monitoring point in different radar line-of-sight directions, forming line-of-sight deformation field data with millimeter accuracy.

[0036] Step S103: Perform spatiotemporal registration of the first deformation data and the second deformation data to form a corresponding joint deformation time series at each monitoring point.

[0037] In step S103, spatiotemporal registration refers to the process of unifying deformation data from different sources, coordinate systems, and time bases onto the same spatial coordinate frame and time axis. The joint deformation time series refers to a time series vector containing multi-dimensional deformation information that each monitoring point possesses after spatiotemporal registration is completed.

[0038] Specifically, the server calls a pre-established digital elevation model (DEM) and, using the known location information from radar and optical sensors, transforms and projects the two-dimensional planar displacement field data (represented by the first deformation data) and the line-of-sight deformation field data (represented by the second deformation data) from their respective sensor coordinate systems to a unified geographic coordinate system or slope local coordinate system. Simultaneously, the server uses methods such as time interpolation to align deformation data from different sampling frequencies to a unified time point. After registration, for each monitoring point, the server generates a multi-dimensional joint deformation time series, where each moment of the series contains components such as horizontal displacement, vertical displacement, and one or more line-of-sight deformation components.

[0039] In one possible implementation, the first deformation data and the second deformation data are spatiotemporally registered to form a corresponding joint deformation time series at each monitoring point, specifically including steps S1031-S1034, as follows: Step S1031: Based on the digital elevation model of the target area of ​​the slope, construct a three-dimensional spatial grid coordinate system, which contains the three-dimensional coordinates of each monitoring point.

[0040] In step S1031, the digital elevation model (DEM) refers to a model that digitally represents the topographic relief of the land surface, recording the planar coordinates and elevation information of each point within the target area of ​​the slope. The three-dimensional spatial grid coordinate system refers to a reference frame based on real-world geographic coordinates established to unify subsequent data processing; this frame covers the entire target area of ​​the slope in a grid format. Three-dimensional coordinates refer to the unique location identifier of each monitoring point in the three-dimensional spatial grid coordinate system, typically composed of eastward coordinates, northward coordinates, and elevation coordinates.

[0041] Specifically, the server first loads a pre-acquired high-precision digital elevation model (DEM) of the target slope area. This model can be point cloud or raster data generated by LiDAR scanning or UAV oblique photogrammetry. Then, based on this DEM, the server defines a unified Cartesian coordinate system, such as the Gauss-Kruger projection coordinate system or a local independent coordinate system. Within this coordinate system, the server generates a regular two-dimensional grid with set horizontal and vertical spacing, for example, 2 meters by 2 meters, and projects this grid onto the slope surface represented by the DEM. The intersections of the grid lines and the slope surface are defined as monitoring points. The server extracts and stores the three-dimensional coordinates of each monitoring point, thus completing the construction of the three-dimensional spatial grid coordinate system.

[0042] Step S1032: Map the two-dimensional pixel coordinates of each monitoring point in the first deformation data to the three-dimensional spatial grid coordinate system through the camera model parameters to obtain the first registration data.

[0043] In step S1032, the two-dimensional pixel coordinates refer to the position of the first deformation data on the optical image plane, represented by the row and column pixel indices. Camera model parameters refer to a set of parameters describing the camera's imaging geometry, including internal parameters such as the camera's focal length and principal point coordinates, as well as external parameters such as the camera's position and orientation in the three-dimensional spatial grid coordinate system. The first registration data refers to the two-dimensional planar displacement field data that has been transformed from the two-dimensional image plane to the three-dimensional spatial grid coordinate system.

[0044] Specifically, the server first utilizes pre-calibrated camera model parameters. For each monitoring point in the 3D spatial grid coordinate system, the server transforms the point's 3D coordinates from the world coordinate system to the camera coordinate system using the extrinsic parameters in the camera model parameters. Then, the server uses the intrinsic parameters in the camera model parameters to project the 3D coordinates in the camera coordinate system onto the 2D image plane, obtaining the precise 2D pixel coordinates corresponding to the monitoring point. Through this mapping relationship, the server establishes a one-to-one correspondence between the first deformation data, i.e., the 2D planar displacement field data, calculated based on the sub-pixel image registration algorithm, and the monitoring points in the 3D spatial grid coordinate system, forming the first registration data.

[0045] Step S1033: Register the radar coordinates of each monitoring point in the second deformation data to the three-dimensional spatial grid coordinate system to obtain the second registration data.

[0046] In step S1033, radar coordinates refer to the position of the second deformation data in the radar's own coordinate system, typically represented by range, azimuth, and elevation angles. The second registration data refers to the line-of-sight deformation field data that has been transformed from the radar coordinate system to a three-dimensional spatial grid coordinate system.

[0047] Specifically, the server uses the pre-measured three-dimensional coordinates of the radar antenna phase center in a three-dimensional spatial grid coordinate system and the antenna's orientation information. For each monitoring point in the three-dimensional spatial grid coordinate system, the server calculates the distance, azimuth, and elevation angles of that point relative to the radar antenna phase center. Then, the server matches these calculated radar coordinates with the radar coordinates in the second deformation data. By finding the nearest matching point, the server assigns the original line-of-sight deformation value measured by the radar to the corresponding monitoring point in the three-dimensional spatial grid coordinate system. This process completes the spatial alignment of the radar measurement data to a unified three-dimensional grid, and the resulting data is the second registration data.

[0048] Step S1034: Interpolate and align the first and second registration data in the time dimension to form a joint deformation time series at each monitoring point.

[0049] In step S1034, interpolation and alignment refer to the process of resampling on a unified time axis using mathematical methods to generate a time-synchronized data sequence in order to solve the problem of inconsistent sampling frequencies of different sensors. The joint deformation time series refers to a multidimensional data sequence formed by arranging the time-synchronized multi-source deformation data in chronological order at each monitoring point.

[0050] Specifically, the server first establishes a unified time reference and sampling interval, for example, using the sampling time of radar data as the reference, recording once every 30 minutes. Since the sampling frequency of the first registration data, i.e., optical image data, may be higher, such as once every 10 minutes, the server selects the closest optical measurement result to the unified time point from multiple results, or averages several measurements from neighboring time points to achieve downsampling and alignment. For scenarios where sampling frequencies may be missing, or where low-frequency data needs to be matched to a high-frequency time axis, the server uses time interpolation algorithms, such as linear interpolation or cubic spline interpolation, to estimate the deformation value at a specified time point based on the deformation values ​​at previous and subsequent times. After alignment, at each unified time point, each monitoring point simultaneously possesses a two-dimensional planar displacement component from the first registration data and a line-of-sight deformation component from the second registration data. The server combines these components into a vector, arranging them in chronological order, thus forming the final joint deformation time series for each monitoring point.

[0051] Step S104: Perform spatiotemporal mode decomposition on the joint deformation time series of each monitoring point to obtain multiple eigenmode function components at different frequency scales.

[0052] In step S104, spatiotemporal mode decomposition refers to a signal processing technique used to decompose complex multidimensional time series signals into a series of simpler oscillation modes with specific physical meanings. Intrinsic mode function components (IMFs) refer to each independent oscillation mode obtained after decomposition. Each IMF is a multidimensional time series with the same dimensions as the original sequence and is concentrated within a specific frequency range.

[0053] Specifically, the server applies a multivariate variational mode decomposition algorithm to the joint deformation time series for each monitoring point. The server sets a preset number of modes K, and then decomposes the input joint deformation time series into K eigenmode function components by solving a constrained variational problem. This solution process is iteratively optimized using the alternating direction multiplier method based on the center frequency, with the goal of minimizing the sum of the bandwidths of all decomposed eigenmode function components, while ensuring that the sum of all eigenmode function components and a residual term can accurately reconstruct the original joint deformation time series.

[0054] In one possible implementation, spatiotemporal mode decomposition is performed on the joint deformation time series of each monitoring point to obtain multiple eigenmode function components at different frequency scales, specifically including steps S1041-S1044, as follows: Step S1041: Use the joint deformation time series as the multivariate input signal and set the number of intrinsic mode function components to be decomposed.

[0055] In step S1041, the multivariate input signal refers to a multidimensional time series formed by fusing data from multiple sensors; in this paper, it is the joint deformation time series for each monitoring point. The number of intrinsic mode function components refers to the pre-defined number of basic oscillation modes that are expected to be decomposed from the multivariate input signal.

[0056] Specifically, the server processes a joint deformation time series for a single monitoring point. This joint deformation time series is a multi-channel signal, such as a five-dimensional time series containing five components: horizontal displacement, vertical displacement, and three line-of-sight deformations. Before performing the decomposition, the server needs to set a key parameter, denoted by K, the number of intrinsic mode function components to be decomposed. This number can be determined based on prior knowledge or signal complexity analysis; for example, in this embodiment, it is set to 5 to separate long-term trends, periodic terms, and several abrupt changes.

[0057] Step S1042: With the goal of minimizing the sum of the bandwidths of each intrinsic mode function component, construct a constrained variational optimization model, whereby the sum of the bandwidths of each intrinsic mode function component can reconstruct the joint deformation time series.

[0058] In step S1042, minimizing the sum of bandwidths is the core optimization objective of the entire decomposition process, which requires that each decomposed eigenmode function component be as compact as possible in the frequency domain. The variational optimization model is a mathematical model used to solve for the extrema of a functional; here, it is used to find a set of functions that satisfy specific conditions. Constraints refer to the conditions that must be satisfied during the optimization process; here, it requires that the sum of all decomposed components accurately reconstructs the original signal. Reconstruction refers to summing the decomposed eigenmode function components to recover the original joint deformation time series.

[0059] Specifically, to achieve signal decomposition, the server establishes a rigorous variational optimization model. The objective function of this model is the sum of the estimated bandwidths of all K eigenmode function components to be determined. By minimizing this objective function, it is ensured that each decomposed component is tightly clustered around a center frequency, possessing a clear physical frequency characteristic. Simultaneously, the server imposes a strict equality constraint on the model: the vector sum of all K eigenmode function components at any given time must equal the vector value of the original joint deformation time series at that time. This constraint guarantees the completeness of the decomposition process, ensuring that no information is lost during the decomposition.

[0060] Step S1043: The variational optimization model is solved iteratively by introducing the alternating direction multiplier method with the center frequency and Lagrange multipliers. In each iteration, the eigenmode function components and the center frequency of the eigenmode function components are updated alternately until the convergence condition is met.

[0061] In step S1043, the center frequency refers to the energy center of each eigenmode function component in the frequency domain. Lagrange multipliers are auxiliary variables introduced when dealing with constrained optimization problems to incorporate constraints into the objective function. The alternating direction multiplier method is an efficient optimization algorithm that decomposes the original problem into multiple simpler subproblems and solves them alternately. Iterative solution refers to the process of gradually approximating the optimal solution through repeated calculations. Alternating update refers to fixing some variables and optimizing another set of variables in one iteration, and then reversing the process. The convergence condition is the criterion for determining whether the iterative process can terminate.

[0062] Specifically, the server employs the alternating direction multiplier method to solve the aforementioned variational optimization model. The server first initializes the K intrinsic mode function (IMF) components and the K corresponding center frequencies. Then, the server enters an iterative loop. In each iteration, the server first fixes the current center frequency values ​​and then solves a subproblem to update the K IMF components. Next, the server fixes the just-updated IMF components and solves another subproblem to update the K center frequencies. Simultaneously, the Lagrange multipliers are updated in each iteration to ensure that the reconstruction constraints are gradually satisfied. After each iteration, the server checks the convergence condition, for example, whether the sum of the changes in all IMF components between two consecutive iterations is less than a preset minimum threshold. If the convergence condition is met, the iteration process stops.

[0063] Step S1044: Output the converged intrinsic mode function components and the center frequencies corresponding to the converged intrinsic mode function components, wherein each intrinsic mode function component is a multidimensional vector with the same dimension as the joint deformation time series.

[0064] In step S1044, the converged intrinsic mode function components refer to the final decomposition results obtained after the iterative process satisfies the convergence condition. The multidimensional vector refers to the value of each intrinsic mode function component at each time point, and the dimension of this value is exactly the same as the dimension of the original input signal.

[0065] Specifically, after the iterative solution process is completed, the server obtains stable and optimal decomposition results. The server outputs these results, which mainly consist of two parts: The first part is the K converged intrinsic mode function (EMF) components. Each component is a multidimensional time series with the same length and dimension as the original joint deformation time series, representing an oscillation mode at a specific frequency scale. The second part is the center frequencies corresponding to the K converged EMF components, each of which is a scalar value indicating the core oscillation frequency of the corresponding component.

[0066] To illustrate steps S1041-S1044 with a complete example: The server processes a five-dimensional joint deformation time series of a monitoring point over the past 30 days, with 1440 sampling points. The server first uses this series as a multivariate input signal and sets the number of intrinsic mode function (EMF) components to be decomposed to 5. Then, the server constructs a variational optimization model, aiming to find the EMF components with the smallest sum of bandwidths, while constraining that the sum of these 5 components must completely reconstruct the original five-dimensional time series. To solve this model, the server initiates iterative calculations using the alternating direction multiplier method, alternately updating the 5 EMF components and their respective center frequencies in the loop until the difference between the results of two consecutive iterations is sufficiently small, satisfying the convergence condition. After the iteration stops, the server finally outputs five five-dimensional intrinsic mode function component time series with a length of 1440, as well as five center frequency values ​​corresponding to these five components. For example, the center frequency of mode 1 is 0.003 Hz, the center frequency of mode 2 is 0.0000116 Hz, etc. These decomposed components and frequencies will be directly used for subsequent modal classification and slope deformation stage identification.

[0067] Step S105: Based on the center frequency of the intrinsic mode function components, classify the intrinsic mode function components into slow trend terms, periodic fluctuation terms, and abrupt change components.

[0068] In step S105, the center frequency refers to the energy center position of each intrinsic mode function component on the spectrum, representing the dominant oscillation frequency of that component. The slow trend term refers to the intrinsic mode function component with the lowest center frequency, reflecting long-term, unidirectional creep deformation of the slope. The periodic fluctuation term refers to the intrinsic mode function component whose center frequency matches a known natural period, such as sunshine, rainfall, or seasonal changes, reflecting elastic deformation caused by environmental factors. The abrupt change component refers to the intrinsic mode function component with a high center frequency and discontinuous energy over time, reflecting sudden events such as local cracking or instability.

[0069] Specifically, after obtaining all intrinsic mode function (IMF) components, the server calculates the center frequency of each IMF component. Then, the server classifies them according to a preset frequency threshold: one or two IMF components with a center frequency lower than, for example, 0.01 Hz are identified as slow trend terms; IMF components with a center frequency matching a daily periodic frequency of, for example, 0.0000116 Hz are identified as periodic fluctuation terms; and the remaining IMF components with higher center frequencies are uniformly classified as abrupt change components.

[0070] In one possible implementation, the intrinsic mode function components are classified into slow trend terms, periodic fluctuation terms, and abrupt change components based on their center frequencies. This specifically includes steps S1051-S1055, as follows: Step S1051: Use a multivariate variational mode decomposition algorithm to decompose the joint deformation time series into a preset number of eigenmode function components.

[0071] In step S1051, the multivariate variational mode decomposition algorithm refers to a signal processing method that can decompose a multi-channel signal into multiple eigenmode function components with specific frequency characteristics. The preset number refers to the number of eigenmode function components to be decomposed, pre-set according to analysis requirements or experience before executing the decomposition algorithm.

[0072] Specifically, the server first acquires the joint deformation time series of specific monitoring points and uses this series as input to a multivariate variational mode decomposition (MMD) algorithm. The server executes the decomposition algorithm, and through an iterative optimization process, decomposes the complex original multidimensional signal into a set of simpler intrinsic mode function (EMF) components of a predetermined number. This step is fundamental to all subsequent classification work, and its output is a series of EMF components and their corresponding center frequencies.

[0073] Step S1052: Calculate the center frequency of each intrinsic mode function component and match the center frequency with a preset frequency feature library.

[0074] In step S1052, the preset frequency feature library refers to a pre-established database or list that stores known frequency values ​​of typical periodic fluctuations caused by environmental factors, such as the diurnal cycle frequency corresponding to changes in solar radiation temperature or the seasonal cycle frequency corresponding to seasonal rainfall. Matching refers to comparing the calculated center frequency with the frequency values ​​in the preset frequency feature library to determine whether there are any equal or approximately equal values.

[0075] Specifically, after completing the decomposition in step S1051, the server has obtained the center frequency of each intrinsic mode function component. The server accesses a preset frequency feature library stored internally, which records frequency values ​​such as 0.0000116 Hz, corresponding to a 24-hour daily cycle. The server checks the center frequency of each intrinsic mode function component one by one to determine whether the frequency matches any frequency value in the preset frequency feature library within the allowable error range.

[0076] Step S1053: Classify the intrinsic mode function components with center frequencies lower than the first preset frequency threshold as slow trend terms.

[0077] In step S1053, the first preset frequency threshold refers to a pre-set extremely low frequency value used to distinguish long-term trends in the signal from higher-frequency fluctuations. The slow trend term refers to the part of the signal that characterizes the long-term, unidirectional creep deformation of the slope; this part typically has the lowest frequency and the most gradual changes.

[0078] Specifically, the server iterates through all intrinsic mode function (IMF) components obtained through decomposition. For each IMF component, the server extracts its corresponding center frequency and compares it with a first preset frequency threshold, such as 0.01 Hz. If the center frequency of an IMF component is less than 0.01 Hz, the server marks or classifies it as a slow trend term. Typically, the server identifies one or two IMF components that meet this condition.

[0079] Step S1054: Classify the intrinsic mode function components whose center frequency matches the preset environmental period as periodic fluctuation terms.

[0080] In step S1054, the preset environmental period refers to a specific frequency defined in the preset frequency feature library that is associated with changes in the natural environment (such as diurnal temperature range, tides, and seasonal changes). The periodic fluctuation term refers to the elastic deformation component caused by these environmental factors that exhibits periodic reciprocating changes in slope deformation.

[0081] Specifically, the server continues to process intrinsic mode function (IMF) components that have not yet been classified. The server utilizes the matching results from step S1052. If the center frequency of an IMF component is determined to match a preset environmental periodic frequency in a preset frequency feature library, for example, a high degree of match with the daily periodic frequency of 0.0000116 Hz, the server marks or classifies the IMF component as a periodic fluctuation term.

[0082] Step S1055: Classify the remaining intrinsic mode function components as mutation components.

[0083] In step S1055, the remaining intrinsic mode function components refer to all intrinsic mode function components that have not been assigned a category after the classification of slow trend terms and periodic fluctuation terms. Abrupt components refer to those non-periodic oscillations in the signal that have a high frequency and short duration, which usually correspond to sudden physical events such as local cracking of slopes and small-scale instability.

[0084] Specifically, after identifying and classifying the slow trend term and the periodic fluctuation term, the server summarizes all remaining intrinsic mode function (IMF) components that have not been assigned any category. The server then uniformly labels or classifies all these remaining IMF components as abrupt change components. These components typically have a higher center frequency than the slow trend term and the periodic fluctuation term, and do not possess a clear periodicity.

[0085] Step S106: Calculate the modal energy and energy percentage of the slow trend term, periodic fluctuation term, and abrupt change component.

[0086] In step S106, modal energy refers to the total energy contained in a certain modal component throughout the entire observation period, quantifying the contribution intensity of that mode to the overall deformation. Energy percentage refers to the percentage of energy of a single modal category, such as the slow trend term, in the total modal energy.

[0087] Specifically, the server first calculates the square of the vector L2 norm at all time points for each intrinsic mode function component, and then accumulates these squared values ​​over time to obtain the modal energy of that component. Next, the server sums the modal energies of all intrinsic mode function components to obtain the total modal energy. Then, the server sums the modal energies of all components classified as slow trend terms to obtain the total energy of trend terms, and divides the total energy of trend terms by the total modal energy to obtain the energy percentage of trend terms. The server calculates the energy percentage of periodic fluctuation terms and abrupt change components in the exact same way.

[0088] In one possible implementation, the modal energy and energy percentage of the slow trend term, periodic fluctuation term, and abrupt component are calculated, specifically including steps S1061-S1064, as follows: Step S1061: For each intrinsic mode function component, accumulate the square of the vector L2 norm of the intrinsic mode function component at each sampling time of the joint deformation time series to obtain the modal energy of the intrinsic mode function component.

[0089] In step S1061, the squared value of the vector 2 norm refers to the sum of the squares of each element in the vector of the intrinsic mode function component at a specific sampling time, which is used to quantify the signal strength at that time. Modal energy refers to the total energy contained in an intrinsic mode function component throughout the entire observation period, reflecting the strength of that component in the overall signal.

[0090] Specifically, the server processes each of the decomposed intrinsic mode function (IMF) components. An IMF component is a multidimensional time series, with a corresponding vector at each sampling time. The server iterates through all sampling times of the time series, calculating the square of the L2 norm of the IMF component vector at each time step. After calculation, the server sums the squares of the L2 norms obtained from all sampling times; the final sum is the modal energy of that IMF component. The server repeats this process for all IMF components, calculating the corresponding modal energy for each component.

[0091] Step S1062: Sum the modal energies of all intrinsic modal function components to obtain the total modal energy of the monitoring point.

[0092] In step S1062, the total modal energy refers to the sum of the modal energies of all intrinsic modal function components, representing the total energy of the original joint deformation time series over the entire observation period.

[0093] Specifically, after completing step S1061, the server has obtained the modal energy value of each intrinsic modal function component. The server sums up all these calculated modal energy values ​​to obtain a total. This total is the total modal energy of the monitoring point, and it will serve as the benchmark for subsequent calculations of energy proportions.

[0094] Step S1063: Sum the modal energies of each intrinsic mode function component classified as slow trend term, periodic fluctuation term, and abrupt change component to obtain the trend term energy, periodic fluctuation term energy, and abrupt change component energy.

[0095] In step S1063, the trend term energy refers to the sum of the modal energies of all intrinsic modal function components classified as slow trend terms. The periodic fluctuation term energy refers to the sum of the modal energies of all intrinsic modal function components classified as periodic fluctuation terms. The abrupt change component energy refers to the sum of the modal energies of all intrinsic modal function components classified as abrupt changes.

[0096] Specifically, the server first obtains the classification results of each intrinsic mode function component from the previous steps. Then, the server sums the modal energies of all intrinsic mode function components classified as slow trend terms to obtain the trend term energy. Similarly, the server sums the modal energies of all intrinsic mode function components classified as periodic fluctuation terms to obtain the periodic fluctuation term energy. Finally, the server sums the modal energies of all intrinsic mode function components classified as abrupt change components to obtain the abrupt change component energy.

[0097] Step S1064: Divide the trend energy, periodic fluctuation energy, and mutation component energy by the total mode energy to obtain the corresponding trend energy ratio, periodic fluctuation energy ratio, and mutation component energy ratio.

[0098] In step S1064, the trend term energy percentage refers to the proportion of trend term energy in the total modal energy. The periodic fluctuation term energy percentage refers to the proportion of periodic fluctuation term energy in the total modal energy. The mutation component energy percentage refers to the proportion of mutation component energy in the total modal energy.

[0099] Specifically, the server uses the calculation results from steps S1062 and S1063. The server divides the trend term energy by the total modal energy to obtain the trend term energy ratio. The server then divides the periodic fluctuation term energy by the total modal energy to obtain the periodic fluctuation term energy ratio. Finally, the server divides the abrupt change component energy by the total modal energy to obtain the abrupt change component energy ratio. These three energy ratio values ​​are the core quantitative basis for determining the slope deformation stage.

[0100] For example, suppose a server has decomposed the joint deformation time series of a monitoring point into five intrinsic mode function (EMF) components, where component 1 is classified as a slow trend term, component 2 as a periodic fluctuation term, and components 3, 4, and 5 as abrupt changes. The server first calculates the modal energy of each component, obtaining 85 units for component 1, 10 units for component 2, 2 units for component 3, 2 units for component 4, and 1 unit for component 5. Then, the server sums the modal energies of all components, obtaining a total modal energy of 100 units. Next, the server aggregates the energy based on the classification results, obtaining a trend term energy of 85 units, a periodic fluctuation term energy of 10 units, and an abrupt change component energy of 5 units. Finally, the server divides each of these three energy categories by the total modal energy, calculating that the trend term energy accounts for 85%, the periodic fluctuation term energy accounts for 10%, and the abrupt change component energy accounts for 5%. These energy percentage data will be used for subsequent slope deformation stage identification and early warning.

[0101] Step S107: Based on modal energy and energy ratio, determine the deformation stage of each monitoring point. The deformation stages include the initial creep stage, the constant creep stage, the accelerated creep stage, and the slip stage.

[0102] In step S107, the deformation stage refers to the different physical states experienced by the slope during its evolution from stability to instability. The initial creep stage indicates minimal deformation, primarily characterized by elastic adjustment. The isochronous creep stage indicates a stable deformation rate, with the slope in a long-term stable creep state. The accelerated creep stage indicates a significantly faster deformation rate, serving as a critical early warning window before instability. The imminent sliding stage indicates that the slope is about to or is currently undergoing sliding failure.

[0103] Specifically, the server compares the calculated energy percentage with a pre-defined rule base to determine the current deformation stage of each monitoring point. For example, if the energy percentage of periodic fluctuations exceeds 40% and the energy percentage of slow trends is below 50%, the server determines it to be in the initial creep stage. If the energy percentage of slow trends rises to 50% to 80% and the energy percentage of abrupt changes is below 5%, the server determines it to be in the constant-rate creep stage. When the energy percentage of slow trends exceeds 80% and the energy percentage of abrupt changes rises significantly from below 5% to above 15% in a short period of time, the server determines it to be in the accelerated creep stage. When the energy percentage of abrupt changes exceeds 50%, the server determines it to be in the pre-slippery stage.

[0104] In one possible implementation, the deformation stage of each monitoring point is determined based on modal energy and energy ratio, specifically including steps S1071-S1074, as follows: Step S1071: When the energy proportion of the periodic fluctuation term is the highest, the current stage is determined as the initial creep stage.

[0105] In step S1071, the initial creep stage refers to the first stage of slope deformation evolution. The deformation in this stage is mainly dominated by periodic and elastic adjustments caused by environmental factors such as temperature and rainfall, and has not yet formed a continuous plastic creep.

[0106] Specifically, the server obtains the energy percentages of the trend term, the periodic fluctuation term, and the mutation component calculated in the aforementioned steps. The server compares the values ​​of these three energy percentages. If the value of the periodic fluctuation term energy percentage is greater than the values ​​of the trend term energy percentage and the mutation component energy percentage, the server determines that the current deformation stage of the monitoring point is the initial creep stage.

[0107] Step S1072: When the energy proportion of the slow trend term replaces the periodic fluctuation term as the energy component with the highest proportion, and the energy proportion of the mutation component does not show a continuous growth trend, the current stage is determined to be the constant-rate creep stage.

[0108] In step S1072, the constant-rate creep stage refers to the stage where, under long-term gravity, the internal stress redistribution of the slope is basically completed, and it begins to undergo continuous plastic deformation at a relatively stable rate. A continuous growth trend refers to a situation where, within a certain time window, the value of a certain variable exhibits a continuous or monotonically increasing pattern.

[0109] Specifically, the server continuously monitors the temporal evolution of each energy percentage. When the server detects that the energy percentage of the slow trend term has exceeded that of the periodic fluctuation term, becoming the largest among the three energy percentages (e.g., the slow trend term energy percentage rises from below 50% to the range of 50% to 80%), and simultaneously, the server analyzes the recent time series changes of the mutation component energy percentage, confirming that this percentage remains at a low level and does not show a clear, continuous upward trend (e.g., the mutation component energy percentage remains below 5%), if both conditions are met, the server classifies the current deformation stage of the monitored point as the isokinetic creep stage.

[0110] Step S1073: When the energy percentage of the slow trend term remains at its highest and the energy percentage of the mutation component shows a continuous increasing trend, the current stage is determined to be the accelerated creep stage.

[0111] In step S1073, the accelerated creep stage is a key early warning stage for the slope to transform from stable deformation to unstable failure. It is characterized by the expansion and connection of internal microcracks, which leads to a significant acceleration of the deformation rate.

[0112] Specifically, during monitoring, the server first confirms that the energy proportion of the slow-trend item remains the highest, for example, if its value exceeds 80%, indicating that long-term creep is the dominant factor in deformation. Based on this, the server focuses on analyzing the time series of the energy proportion of the mutation component. By calculating its rate of change within the most recent preset time window, such as one week, or observing whether it monotonically increases across multiple consecutive time points, the server determines whether a continuous growth trend exists. When the server confirms that the energy proportion of the mutation component is not only increasing, but this increase is also continuous, for example, rising from below 5% to above 15% within one week, the server determines that the monitoring point has entered the accelerated creep phase.

[0113] Step S1074: When the energy of the mutation component is the energy component with the highest proportion or when an intrinsic mode function component with a center frequency higher than a preset frequency threshold is identified, the current stage is determined to be the slippery stage.

[0114] In step S1074, the pre-slip stage is the final stage where the slope is about to or is undergoing overall sliding, characterized by extremely violent deformation. This is the last opportunity to issue an emergency evacuation order. The preset frequency threshold is a frequency value pre-set based on geological conditions and historical data, used to identify signals representing newly emerging high-frequency rupture events.

[0115] Specifically, when determining the current state, the server checks two independent triggering conditions. The first condition is the reversal of energy proportions: the server compares the values ​​of the three energy proportions. If it finds that the energy proportion of the abrupt change component has exceeded that of the slow trend component and the periodic fluctuation component, becoming the highest-proportioning energy component (e.g., exceeding 50%), it directly determines that the system has entered the pre-slip phase. The second condition is the emergence of a new mode: during mode decomposition, if the server identifies one or more new intrinsic mode function components, and the center frequencies of these new components are higher than a preset frequency threshold, this indicates that a new, rapid local destructive event is occurring. If either of these two conditions is met, the server will immediately determine that the current deformation stage of the monitoring point is the pre-slip phase.

[0116] If the calculated energy percentages and their trends at the current moment fail to trigger any of the stage determination or switching conditions from S1071 to S1074, the server will determine that the deformation stage of the monitoring point remains unchanged.

[0117] Specifically, the server makes a determination in each monitoring cycle. For example, a monitoring point may have been determined to be in the constant-rate creep stage. In the next monitoring cycle, after recalculation, the server finds that the energy proportion of the slow trend term is still the highest, but the energy proportion of the mutation component does not show a continuous increasing trend, nor is it the highest, and no high-frequency intrinsic mode function components are identified. At this time, since the conditions for entering the accelerated creep stage (S1073) or the near-slippery stage (S1074) are not met, the monitoring point will continue to be maintained in the constant-rate creep stage. Similarly, monitoring points in the initial creep stage or accelerated creep stage will also maintain their current stage determination if their energy proportion characteristics do not trigger new stage switching conditions.

[0118] Step S108: Based on the deformation stage of each monitoring point, generate and output the slope risk level and early warning information.

[0119] In step S108, the slope risk level refers to a comprehensive assessment of the overall hazard level of the slope, usually represented by different colors. Early warning information refers to the specific alerts and handling recommendations issued to management personnel.

[0120] Specifically, the server assigns different risk statuses to each monitoring point based on its determined deformation stage. Then, the server integrates the status of all monitoring points across the entire target slope area to generate an overall slope risk level. If most monitoring points are in the initial or isochronous creep stage, the server defines the slope risk level as green or normal. If a monitoring point enters the early stage of accelerated creep, the server defines the level as yellow or caution and generates a warning message recommending increased monitoring frequency. If it enters the middle stage of accelerated creep, the server defines the level as orange or warning and generates a warning message recommending on-site inspection. If it enters the late stage of accelerated creep or the imminent landslide stage, the server defines the level as red or danger and generates a warning message requiring immediate evacuation. Finally, the server uses different colors to render each monitoring point on a 3D visualization platform to display its deformation stage and clearly presents the overall slope risk level and recommended treatment measures.

[0121] To illustrate steps S101-S108, for a highway slope equipped with optical cameras and MIMO radar, the server instructs the cameras to capture images every 10 minutes and the radar to scan every 30 minutes for 30 consecutive days, acquiring optical image sequences and radar data covering 156 monitoring points on the slope. The server first extracts the two-dimensional planar displacement and line-of-sight deformation in three directions for each monitoring point from this raw data, and forms 156 five-dimensional joint deformation time series through spatiotemporal registration. Subsequently, the server performs multivariate variational mode decomposition on each time series, obtaining five intrinsic mode function components, and automatically classifies these components according to their center frequencies as a slow trend term representing long-term creep, a periodic fluctuation term representing daily temperature variations, and abrupt components representing local rupture. On the 15th day of monitoring, the server calculates and finds that the energy proportion of the abrupt component at a certain monitoring point has increased from 4% to 12%, meeting the criteria for entering the initial stage of accelerated creep. Therefore, the server immediately marks this monitoring point as yellow and issues a yellow warning to the system, suggesting that management personnel increase monitoring frequency. By day 18, the server calculated that the energy percentage of the abrupt change component at that point had further increased to 25%, while the energy percentage of the slow trend component reached as high as 85%, indicating that the overall sliding trend was intensifying. The system automatically upgraded the warning to orange and prompted on-site inspection. Inspectors, based on the warning, found that the cracks on the slope were expanding significantly faster. By day 20, the server detected a surge in the energy percentage of the abrupt change component to 45%, determining that the slope had entered the pre-slip stage and immediately triggering a red danger alarm. In the early morning of day 21, a partial collapse occurred on the slope. Thanks to the timely warning, all personnel and equipment had been safely evacuated beforehand.

[0122] Please refer to Figure 2In one possible implementation, based on the deformation stage of each monitoring point, slope risk level and early warning information are generated and output, specifically including steps S201-S204, as follows: Step S201: Based on the deformation stage corresponding to each monitoring point, generate a spatial distribution map of the slope deformation stage.

[0123] In step S201, the spatial distribution map of slope deformation stages refers to a diagram that visualizes the deformation stage information of all monitoring points on the slope surface. This map is usually overlaid on a three-dimensional model or two-dimensional image of the slope, and by assigning different colors to different deformation stages, it intuitively presents the overall stability state of the slope and local areas.

[0124] Specifically, the server first obtains the deformation stage of each monitoring point determined in the aforementioned steps, namely the initial creep stage, the isochronous creep stage, the accelerated creep stage, or the pre-slip stage. Simultaneously, the server retrieves the stored spatial coordinate data of each monitoring point on the slope. Then, the server matches the deformation stage information with the spatial coordinate information and renders the location of each monitoring point on a two-dimensional or three-dimensional slope geographic information model. During rendering, the server uses a preset color mapping rule, for example, using green to represent the initial or isochronous creep stage, yellow to represent the early stage of accelerated creep, orange to represent the middle stage of accelerated creep, and red to represent the late stage of accelerated creep or the pre-slip stage, ultimately generating a color cloud map covering the entire monitoring area, i.e., a spatial distribution map of the slope deformation stages.

[0125] Step S202: On the spatial distribution map, based on the preset clustering algorithm, adjacent monitoring points that are in the preset high-level deformation stage are identified and aggregated into deformation region clusters. The preset high-level deformation stage includes the accelerated creep stage and the slip stage.

[0126] In step S202, the preset clustering algorithm refers to a pre-selected mathematical algorithm that can group data points based on their spatial proximity, such as the DBSCAN density clustering algorithm. A deformation region cluster refers to a set of monitoring points that are spatially adjacent and exhibit similar deformation behavior; this set represents a coherent region with a potential overall instability risk. The preset high-level deformation stage is a clearly defined deformation stage representing the slope entering an unstable state, specifically referring to the accelerated creep stage and the near-slip stage.

[0127] Specifically, the server processes the spatial distribution map of slope deformation stages generated in the previous step. First, the server selects all monitoring points in a preset high-level deformation stage, namely the accelerated creep stage or the near-slip stage, and uses these points as input data for the clustering algorithm. Next, the server applies the preset DBSCAN clustering algorithm, setting appropriate parameters such as neighborhood radius and minimum neighbor number for core points, to calculate the spatial coordinates of the selected high-level deformation stage monitoring points. The algorithm automatically aggregates sufficiently dense monitoring points in space, forming one or more independent groups. Each such group is identified as a deformation region cluster.

[0128] Step S203: Calculate the risk index for each deformable region cluster. The risk index is calculated by weighted summation based on the number of monitoring points within the deformable region cluster, the geometric area of ​​the deformable region cluster, and the average energy of the mutation components at each monitoring point within the deformable region cluster.

[0129] In step S203, the risk index is a comprehensive quantitative indicator used to assess the degree of danger of a single deformation zone cluster; a higher value indicates a greater risk. The number of monitoring points within a deformation zone cluster refers to the total number of monitoring points constituting that cluster. The geometric area of ​​the deformation zone cluster refers to the actual physical area covered by the cluster on the slope surface. The average energy of abrupt changes at each monitoring point within the deformation zone cluster is the arithmetic mean of the proportions of abrupt changes in energy at all monitoring points within the cluster, reflecting the suddenness and severity of deformation in that area.

[0130] Specifically, the server calculates the risk index for each deformable region cluster identified in the previous step. For a given deformable region cluster, the server first counts the total number of monitoring points contained in the cluster. Then, based on the spatial coordinates of these monitoring points, the server calculates the geometric area covered by the cluster, for example, by calculating the convex hull area of ​​these point sets. Subsequently, the server extracts the energy percentage of each mutation component for each monitoring point within the cluster and calculates the average of these values ​​to obtain the average energy of the mutation component. Finally, the server applies a preset weighted summation formula, such as: Risk Index = Weight Coefficient A × Number of Monitoring Points + Weight Coefficient B × Geometric Area + Weight Coefficient C × Average Energy of Mutation Components, to calculate the final risk index of the deformable region cluster.

[0131] Step S204: Determine the slope risk level of the target area based on the highest risk index of each deformation zone cluster, and determine the corresponding early warning information based on the slope risk level.

[0132] In step S204, the slope risk level of the target area refers to the final assessment of the overall hazard status of the entire monitored slope, which is usually divided into several levels such as "normal," "caution," "warning," and "danger." Early warning information consists of specific handling suggestions and notifications corresponding to each risk level, used to guide subsequent disaster prevention and mitigation actions.

[0133] Specifically, the server first collects the risk index of all calculated deformation zone clusters. If multiple deformation zone clusters are identified across the entire slope, the server compares the risk indices of all these clusters and identifies the maximum value. This highest risk index is used as the decisive basis for assessing the risk level of the entire target area of ​​the slope. The server compares this highest risk index with a set of pre-defined risk level classification thresholds. For example, if the highest risk index is less than 50, the slope risk level is "normal"; if it is between 50 and 100, it is "caution"; and if it is higher than 100, it is "warning". Based on the comparison results, the server determines a unique slope risk level. Finally, the server matches and extracts the corresponding warning information from the warning information database according to the determined slope risk level. For example, for the "caution" level, it matches "recommend increasing the monitoring frequency," and for the "warning" level, it matches "recommend on-site inspection and preparation of emergency measures." The final slope risk level and warning information are then output to a 3D visualization warning terminal.

[0134] The following describes a slope accelerated deformation stage early warning device from the perspective of hardware processing in an embodiment of this invention. Please refer to [link / reference]. Figure 3 This is a schematic diagram of the structure of an early warning device for the accelerated deformation stage of a slope in an embodiment of this application.

[0135] It should be noted that, Figure 3 The structure of the early warning device for the accelerated deformation stage of a slope shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of the present invention.

[0136] like Figure 3As shown, a slope accelerated deformation stage early warning device includes a central processing unit (CPU) 301, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 302 or a program loaded from a storage section 308 into a random access memory (RAM) 303, such as executing the methods described in the above embodiments. The RAM 303 also stores various programs and data required for device operation. The CPU 301, ROM 302, and RAM 303 are interconnected via a bus 304. An input / output (I / O) interface 305 is also connected to the bus 304.

[0137] The following components are connected to I / O interface 305: input section 306 including audio input devices, push-button switches, etc.; output section 307 including a liquid crystal display (LCD) and audio output devices, indicator lights, etc.; storage section 308 including a hard disk, etc.; and communication section 309 including a network interface card such as a LAN (Local Area Network) card, modem, etc. Communication section 309 performs communication processing via a network such as the Internet. Drive 310 is also connected to I / O interface 305 as needed. Removable media 311, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., are installed on drive 310 as needed so that computer programs read from them can be installed into storage section 308 as needed.

[0138] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing computer programs for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 309, and / or installed from removable medium 311. When the computer program is executed by central processing unit (CPU) 301, it performs the various functions defined in the present invention.

[0139] It should be noted that specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fiber, portable compact disc read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this invention, a computer-readable storage medium can 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.

[0140] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of apparatus, methods, and computer program products according to various embodiments of the present invention. Each block in a flowchart or block diagram may represent a module, segment, or portion of code, which contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those shown in the drawings.

[0141] Specifically, the slope accelerated deformation stage early warning device of this embodiment includes a processor and a memory. The memory stores a computer program. When the computer program is executed by the processor, it implements the slope accelerated deformation stage early warning method provided in the above embodiment.

[0142] In another aspect, the present invention also provides a computer-readable storage medium, which may be included in the slope accelerated deformation stage early warning device described in the above embodiments; or it may exist independently and not assembled into the slope accelerated deformation stage early warning device. The storage medium carries one or more computer programs, which, when executed by a processor of the slope accelerated deformation stage early warning device, cause the slope accelerated deformation stage early warning device to implement the slope accelerated deformation stage early warning method provided in the above embodiments.

Claims

1. A method for early warning of accelerated deformation stage of slope, characterized in that, The method includes: Acquire optical image sequences and radar data of the target area on the slope; First deformation data is extracted based on the optical image sequence, and second deformation data is extracted based on the radar data. The first deformation data is two-dimensional planar displacement field data of multiple monitoring points in the target area of ​​the slope, and the second deformation data is line-of-sight deformation field data of multiple monitoring points. The first deformation data and the second deformation data are spatiotemporally registered to form a corresponding joint deformation time series at each of the monitoring points; Spatiotemporal mode decomposition is performed on the joint deformation time series of each monitoring point to obtain multiple eigenmode function components at different frequency scales; Based on the center frequency of the intrinsic mode function components, the intrinsic mode function components are classified into slow trend terms, periodic fluctuation terms, and abrupt change components. Calculate the modal energy and energy percentage of the slow trend term, the periodic fluctuation term, and the abrupt change component; Based on the modal energy and the energy ratio, the deformation stage of each monitoring point is determined. The deformation stage includes the initial creep stage, the constant creep stage, the accelerated creep stage, and the slippage stage. Based on the deformation stage of each monitoring point, slope risk level and early warning information are generated and output.

2. The method according to claim 1, characterized in that, The step of performing spatiotemporal registration of the first deformation data and the second deformation data to form a corresponding joint deformation time series at each of the monitoring points specifically includes: Based on the digital elevation model of the target area of ​​the slope, a three-dimensional spatial grid coordinate system is constructed, which includes the three-dimensional coordinates of each monitoring point. The two-dimensional pixel coordinates of each monitoring point in the first deformation data are mapped to the three-dimensional spatial grid coordinate system through camera model parameters to obtain the first registration data; The radar coordinates of each monitoring point in the second deformation data are registered to the three-dimensional spatial grid coordinate system to obtain the second registration data; The first registration data and the second registration data are interpolated and aligned in the time dimension to form the joint deformation time series at each monitoring point.

3. The method according to claim 1, characterized in that, The spatiotemporal modal decomposition of the joint deformation time series at each monitoring point yields multiple eigenmode function components at different frequency scales, specifically including: The joint deformation time series is used as a multivariate input signal, and the number of intrinsic mode function components to be decomposed is set. To minimize the sum of the bandwidths of the intrinsic mode function components, a constrained variational optimization model is constructed, wherein the constraint is that the sum of the intrinsic mode function components can reconstruct the joint deformation time series. The variational optimization model is iteratively solved by introducing the alternating direction multiplier method with the center frequency and Lagrange multipliers. In each iteration, the intrinsic mode function components and the center frequency of the intrinsic mode function components are alternately updated until the convergence condition is met. Output the converged intrinsic mode function components and the center frequencies corresponding to the converged intrinsic mode function components, wherein each of the intrinsic mode function components is a multidimensional vector with the same dimension as the joint deformation time series.

4. The method according to claim 1, characterized in that, The classification of intrinsic mode function components into slow trend terms, periodic fluctuation terms, and abrupt change components based on their center frequencies specifically includes: The joint deformation time series is decomposed into a predetermined number of intrinsic mode function components using a multivariate variational mode decomposition algorithm. Calculate the center frequency of each intrinsic mode function component and match the center frequency with a preset frequency feature library; The intrinsic mode function components with a center frequency lower than a first preset frequency threshold are classified as the slow trend term; The intrinsic mode function components whose center frequency matches the preset environmental period are classified as the periodic fluctuation term; The remaining intrinsic mode function components are classified as the mutation components.

5. The method according to claim 1, characterized in that, The calculation of the modal energy and energy percentage of the slow trend term, the periodic fluctuation term, and the abrupt change component specifically includes: For each intrinsic mode function component, the modal energy of the intrinsic mode function component is obtained by accumulating the square of the vector 2 norm of the intrinsic mode function component at each sampling time of the joint deformation time series. The total modal energy of the monitoring point is obtained by summing the modal energies of all intrinsic modal function components. The modal energies of each intrinsic mode function component classified as the slow trend term, the periodic fluctuation term, and the abrupt change component are summed to obtain the trend term energy, the periodic fluctuation term energy, and the abrupt change component energy. Divide the energy of the trend term, the energy of the periodic fluctuation term, and the energy of the mutation component by the total modal energy to obtain the corresponding proportions of the trend term energy, the energy of the periodic fluctuation term energy, and the energy of the mutation component energy.

6. The method according to claim 5, characterized in that, The determination of the deformation stage of each monitoring point based on the modal energy and the energy ratio specifically includes: When the energy proportion of the periodic fluctuation term is the highest, the current stage is determined to be the initial creep stage; When the energy proportion of the slow trend term replaces the periodic fluctuation term as the energy component with the highest proportion, and the energy proportion of the abrupt change component does not show a continuous increasing trend, the current stage is determined to be the constant-rate creep stage. When the energy percentage of the slow trend term remains at its highest and the energy percentage of the mutation component shows the continuous growth trend, the current stage is determined to be the accelerated creep stage; When the energy of the mutation component is the highest proportion of the energy component or when an intrinsic mode function component with a center frequency higher than a preset frequency threshold is identified, the current stage is determined to be the slippery stage.

7. The method according to claim 1, characterized in that, The process of generating and outputting slope risk level and early warning information based on the deformation stage of each monitoring point specifically includes: Based on the deformation stage corresponding to each monitoring point, a spatial distribution map of the slope deformation stage is generated. On the spatial distribution map, based on a preset clustering algorithm, adjacent monitoring points that are in a preset high-level deformation stage are identified and aggregated into deformation region clusters. The preset high-level deformation stage includes an accelerated creep stage and a slip-prone stage. Calculate the risk index for each of the deformable region clusters, the risk index being calculated by weighted summation based on the number of monitoring points within the deformable region cluster, the geometric area of ​​the deformable region cluster, and the average energy of the mutation components at each of the monitoring points within the deformable region cluster; The slope risk level of the target area is determined based on the highest risk index of each of the deformation area clusters, and the corresponding early warning information is determined based on the slope risk level.

8. A slope accelerated deformation stage early warning device, characterized in that, The early warning device for accelerated slope deformation stage includes: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code includes computer instructions, and the one or more processors call the computer instructions to cause the early warning device for accelerated slope deformation stage to perform the method as described in any one of claims 1-7.

9. A computer-readable storage medium comprising instructions, characterized in that, When the instruction is executed on the slope accelerated deformation stage early warning device, the slope accelerated deformation stage early warning device performs the method as described in any one of claims 1-7.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is run on the slope accelerated deformation stage early warning device, the slope accelerated deformation stage early warning device performs the method as described in any one of claims 1-7.

Images