Multi-source data fusion-based urban land subsidence analysis method and system

Abstract

The invention discloses an urban land subsidence analysis method and system based on multi-source data fusion, relates to the technical field of subsidence analysis, and constructs a multi-source three-dimensional monitoring network, performs air-space-ground multi-source data fusion analysis and atmospheric delay error correction, constructs a land subsidence prediction model, and provides an urban land subsidence analysis method and system based on multi-source data fusion. According to the method, a ground surface and underground three-dimensional monitoring network is constructed, the level, layered settlement and underground water data are integrated, minute-level global monitoring is achieved in combination with the Internet of Things, and the method is high in practicability and high in practicability. Through air-sky-ground data fusion and atmospheric correction, monitoring errors are reduced to a millimeter level, an evaluation system combining subjective and objective weights generates a five-level risk map, a visual platform supports dynamic simulation and factor traceability, a whole-process technical closed loop is formed, and the intelligent level of urban settlement prevention and control is remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of settlement analysis, and particularly to a method and system for urban ground settlement analysis based on multi-source data fusion. Background Art

[0002] With the acceleration of urbanization and the over-exploitation of groundwater, ground settlement has become a key geological disaster threatening urban safety. Traditional monitoring means have problems such as low data resolution, insufficient spatio-temporal coverage, and difficulty in multi-source data fusion, making it difficult to meet the requirements of high-precision and real-time dynamic monitoring.

[0003] Currently, the Chinese invention patent with the application number CN202410116472.7 discloses a method for analyzing the driving force of ground settlement by fusing multi-source observation data, including: dividing the target area into settlement areas and non-settlement areas, and further dividing the settlement areas into regional settlement and local settlement; judging the influencing components causing the settlement in the settlement areas and judging the deformation patterns of the settlement areas; classifying the types of ground objects in the settlement areas; for local settlement, according to the overall shape of the time-series curve, deformation pattern, and ground object type of the settlement area; for regional settlement, according to the groundwater level information and the data of the layered marker group, respectively complete the analysis of the driving force of ground settlement.

[0004] The above technology is difficult to solve the defects of low monitoring accuracy, one-sided prediction, strong subjectivity in evaluation, and insufficient decision support, and it is difficult to achieve all-dimensional precise analysis and intelligent control of urban ground settlement. Summary of the Invention

[0005] The technical problem solved by the present invention is that the existing technology is difficult to solve the defects of low monitoring accuracy, one-sided prediction, strong subjectivity in evaluation, and insufficient decision support, and it is difficult to achieve all-dimensional precise analysis and intelligent control of urban ground settlement.

[0006] To solve the above technical problem, the present invention provides the following technical solutions: A method for urban ground settlement analysis based on multi-source data fusion, including the following steps: Step S1, constructing a multi-source three-dimensional monitoring network to conduct synchronous monitoring of the ground settlement of the target city at multiple levels on the surface and underground; Step S2, conducting multi-source data fusion analysis of space-air-ground to unify the spatio-temporal reference of each monitoring data; Step S3, correcting the atmospheric delay error of the InSAR monitoring data; Step S4, constructing a ground settlement prediction model to model and predict the development trend of ground settlement; Step S5, carrying out an evaluation of the susceptibility of ground settlement, determining the weights of influencing factors and dividing the settlement susceptibility levels; Step S6: Develop a visualization interface for land subsidence analysis to integrally display and interact with the monitoring data, prediction results, and evaluation information.

[0007] Preferably, the step S1 includes the following sub-steps: Step S101: Layout a second-order leveling control network on the surface layer as the reference framework for land subsidence monitoring; Step S102: Install hierarchical settlement markers in the deep underground and carry GNSS monitoring equipment to establish a hierarchical settlement dynamic monitoring system to obtain real-time data feedback on land subsidence at different depths; Step S103: Deploy high-precision water level monitoring sensors in the groundwater monitoring layer and construct a pressure-water level joint monitoring system in typical monitoring wells to collect the buried depth of the groundwater level and pore water pressure parameters at a minute-level frequency; Step S104: Integrate the leveling control network, hierarchical settlement monitoring system, and groundwater monitoring system through Internet of Things technology to form a real-time data transmission network.

[0008] Preferably, the step S2 includes the following sub-steps: Step S201: Take the elevation change obtained from the second-order leveling control network as the absolute reference, and use the GNSS monitoring data to establish a dynamic conversion model between the GNSS phase elevation and the leveling elevation to achieve the unification of the absolute reference for multi-source data. The dynamic conversion model satisfies: ; Wherein, is the phase elevation, is the leveling elevation, is the conversion coefficient between the phase elevation and the leveling elevation, is the elevation change amount, is the leveling elevation correction factor, is the temperature under a standard atmospheric pressure, is the atmospheric delay compensation term; Step S202: Use the persistent scatterer InSAR technology to obtain the relative deformation data of high-density surface deformation points, establish the spatial constraint relationship between the InSAR deformation points and the leveling control points by constructing a Delaunay triangulation network, and register the relative deformation field to the absolute reference using the weighted least squares algorithm.

[0009] Preferably, the step S3 includes the following sub-steps: Step S301: Based on the observation data of the ground GNSS continuous stations, use the Saastamoinen model to calculate the zenith total delay ZTD of each station; Step S302: Perform spatial interpolation on the ZTD calculated by the GNSS stations to construct grid data of atmospheric water vapor content with a resolution of 1 km × 1 km, improving the accuracy of the description of the spatial distribution of tropospheric water vapor; Step S303: Separate the residual atmospheric phase delay error observed by InSAR into stratospheric delay component and turbulent delay component; Step S304: Eliminate the stratospheric delay component and the turbulent delay component. Among them, the stratospheric delay component is eliminated by establishing an elevation-related model using MERRA-2 reanalysis data and polynomial fitting, and the turbulent delay component is corrected by constructing a spatial filtering operator based on GNSS water vapor grid data.

[0010] Preferably, step S4 includes the following sub-steps: Step S401: Construct a physical mechanism model and conduct numerical simulation analysis on the land subsidence process, including: Establish a fully coupled seepage-stress finite element model based on Biot consolidation theory, divide the study area into multiple soil layer partitions according to the lithological characteristics of the soil layers, and assign mechanical and seepage parameters respectively. At the same time, introduce the hydraulic hysteresis effect correction algorithm of unsaturated soil to improve the simulation accuracy of the settlement caused by the change of the groundwater level. Use the InSAR deformation time series data, groundwater level dynamic monitoring data, and layered settlement monitoring data as the input boundary conditions of the model, adopt an implicit iterative solution strategy, and combine the parallel computing architecture based on CUDA to accelerate the stiffness matrix solution process.

[0011] Preferably, step S4 further includes the following sub-steps: Step S402: Construct a data-driven machine learning model for intelligent prediction of land subsidence, including: Design a neural network architecture based on a meta-heuristic optimization algorithm, use the buried depth of the groundwater level, drawdown of shallow groundwater, drawdown of deep groundwater, thickness of clay layer, thickness of sand layer, annual rainfall, and permeability coefficient as the input features of the model, and use a spatio-temporal attention mechanism to extract key influencing features; Adopt a BP neural network as the settlement prediction model, and optimize the model parameters in real time through a meta-heuristic algorithm. Further embed a transfer learning module to update the model periodically online. Import the latest monitoring data every 24 hours and fine-tune the network weights to maintain the adaptability of the model to environmental changes, and output the land subsidence prediction amount and its spatial distribution within the next 72 hours.

[0012] Preferably, step S5 includes the following sub-steps: Step S501: Screen the buried depth of the groundwater level, drawdown of shallow groundwater, drawdown of deep groundwater, thickness of clay layer, thickness of sand layer, annual rainfall, and permeability coefficient from geological and environmental factors as evaluation factors for land subsidence susceptibility; Step S502: Calculate the subjective weights of each evaluation factor using the analytic hierarchy process. Construct a judgment matrix for comparing evaluation indicators based on expert judgment, and determine the subjective weight values of each factor through consistency testing; Step S503: Calculate the objective weights of each evaluation factor based on historical monitoring data using the CRITIC method, and obtain the objective weight values by comprehensively considering the numerical variability and mutual correlation of each indicator; Combine the subjective weight values and objective weight values, and calculate the final combined weights of each factor using the weighted average method; Step S504: Calculate the comprehensive evaluation value of land subsidence susceptibility based on the combined weights of each factor and relevant industry specifications, and divide the land subsidence susceptibility into extremely low level, low level, medium level, high level, and extremely high level.

[0013] Preferably, the step S6 includes the following sub-steps: Step S601: Develop a visualization interface, and the visualization interface obtains the buried depth of the groundwater level, drawdown of shallow groundwater, drawdown of deep groundwater, thickness of clay layer, thickness of sand layer, annual rainfall, and permeability coefficient input by the user; Step S602: Based on the land subsidence susceptibility evaluation model and the settlement prediction model, calculate the grade prediction result of land subsidence susceptibility in real time in the visualization interface; Step S603: Visualize and display the predicted settlement susceptibility grade results through the graphical user interface.

[0014] Preferably, the visualization interface further provides an index risk level analysis function, and based on the evaluation results of land subsidence susceptibility, conducts a disaster-causing risk assessment for each influencing factor respectively, and divides the disaster-causing risks of each influencing factor into levels.

[0015] A city land subsidence analysis system based on multi-source data fusion, which is applied to the above-mentioned city land subsidence analysis method based on multi-source data fusion, includes a network monitoring module, a fusion analysis module, an error correction module, a settlement prediction module, a settlement division module, and a visual interaction module; The network monitoring module is used to construct a multi-source three-dimensional monitoring network to synchronously monitor the land subsidence of the target city at multiple levels on the surface and underground; The fusion analysis module is used to conduct multi-source data fusion analysis of space-air-ground, and unify the spatio-temporal benchmarks of each monitoring data; The error correction module is used to correct the atmospheric delay error of InSAR monitoring data; The settlement prediction module is used to construct a land subsidence prediction model to model and predict the development trend of land subsidence; The settlement division module is used to carry out the susceptibility evaluation of land subsidence, determine the weights of influencing factors, and divide the susceptible levels of subsidence. The visual interaction module is used to develop a visualization interface for land subsidence analysis, and integrate and display the monitoring data, prediction results, and evaluation information for interaction.

[0016] Advantages of the present invention: The present invention constructs a three-dimensional monitoring network on the surface and underground, integrates leveling, layer settlement, and groundwater data, realizes minute-level full-domain monitoring through the Internet of Things, reduces the monitoring error to the millimeter level through air-space-ground data fusion and atmospheric correction, generates a five-level risk map with an evaluation system combining subjective and objective weights, and the visualization platform supports dynamic simulation and factor tracing, forming a full-process technical closed-loop, significantly improving the intelligent level of urban subsidence prevention and control. Description of the Drawings

[0017] Figure 1 It is a step flow chart of a method for analyzing urban land subsidence based on multi-source data fusion provided by an embodiment of the present invention; Figure 2 It is a basic process schematic diagram of a system for analyzing urban land subsidence based on multi-source data fusion provided by an embodiment of the present invention. Detailed Embodiments

[0018] To make the above objects, features, and advantages of the present invention more obvious and understandable, the following detailed description of the specific embodiments of the present invention is made in conjunction with the accompanying drawings of the specification. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention.

[0019] Embodiment 1, referring to Figure 1 , provides a method for analyzing urban land subsidence based on multi-source data fusion, including the following steps: Step S1, construct a multi-source three-dimensional monitoring network to synchronously monitor the land subsidence on the surface and underground levels of the target city.

[0020] Step S2, perform air-space-ground multi-source data fusion analysis to unify the spatio-temporal benchmarks of each monitoring data.

[0021] Step S3, perform atmospheric delay error correction on the InSAR monitoring data.

[0022] Step S4, construct a land subsidence prediction model to model and predict the development trend of land subsidence.

[0023] Step S5, carry out the susceptibility evaluation of land subsidence, determine the weights of influencing factors, and divide the susceptible levels of subsidence.

[0024] Step S6, develop a visualization interface for land subsidence analysis, and integrate and display the monitoring data, prediction results, and evaluation information for interaction.

[0025] Step S1 includes the following sub-steps: Step S101, lay out a second-order leveling control network on the surface layer as the reference framework for land subsidence monitoring.

[0026] Step S101 establishes a high-precision vertical displacement monitoring reference through the second-order leveling control network, ensuring the unity of the absolute reference framework for land subsidence monitoring, with the error controlled at the millimeter level, and providing a reference anchor point for subsequent data registration.

[0027] Step S102, install layer settlement markers in the deep underground and carry GNSS monitoring equipment to establish a layer settlement dynamic monitoring system, and obtain real-time data feedback on land subsidence at different depths.

[0028] Step S102 uses the layer settlement markers and GNSS for collaborative monitoring to capture the independent settlement amounts of soil layers at different depths underground, with a resolution of up to centimeter level, revealing the non-uniform settlement characteristics in the vertical direction and breaking through the limitations of traditional single-surface monitoring.

[0029] Step S103, deploy high-precision water level monitoring sensors in the groundwater monitoring layer and build a pressure-water level joint monitoring system in typical monitoring wells to collect parameters such as the buried depth of the groundwater level and pore water pressure at a minute-level frequency.

[0030] The pressure-water level joint monitoring system in Step S103 collects data at a minute-level frequency, accurately depicts the dynamic changes of the groundwater level and the fluctuations of pore water pressure, and provides high-frequency and multi-parameter observational evidence for analyzing the water-soil coupling mechanism.

[0031] Step S104, integrate the leveling control network, layer settlement monitoring system, and groundwater monitoring system through Internet of Things technology to form a real-time data transmission network.

[0032] Step S104 breaks data silos through Internet of Things integration, realizes the real-time transmission and cloud convergence of multi-source monitoring data, with a data delay of less than 1 second, supports dynamic monitoring and instant analysis, and significantly improves the timeliness and coordination of the monitoring system.

[0033] Step S1 constructs a three-dimensional monitoring network with surface-underground linkage, realizes full coverage from the surface deformation reference to the settlement of deep soil layers and the change of groundwater level, and provides high-precision and high spatio-temporal resolution basic data support for multi-source data fusion analysis.

[0034] Step S2 includes the following sub-steps: Step S201: Using the elevation change obtained from the second-order leveling control network as the absolute reference, a dynamic conversion model between GNSS phase elevation and leveling elevation is established using GNSS monitoring data to achieve the unification of the absolute reference for multi-source data. The dynamic conversion model satisfies: ; where is the phase elevation, is the leveling elevation, is the conversion coefficient between the phase elevation and the leveling elevation, is the elevation change, is the leveling elevation correction factor, is the temperature under standard atmospheric pressure, is the atmospheric delay compensation term.

[0035] In step S201, a dynamic conversion model between GNSS phase elevation and leveling elevation is established to achieve the absolute unification of the vertical reference, eliminate the systematic error between GNSS and leveling measurements, ensure the comparability of multi-source data in vertical deformation analysis, and break through the problem of inconsistent reference in traditional data fusion.

[0036] Step S202: Using the permanent scatterer InSAR technology to obtain the relative deformation data of high-density surface deformation points, establishing the spatial constraint relationship between InSAR deformation points and leveling control points by constructing a Delaunay triangulation network, and registering the relative deformation field to the absolute reference using the weighted least squares algorithm.

[0037] In step S2, the spatial topological relationship between InSAR deformation points and leveling control points is constructed through Delaunay triangulation, and the InSAR relative deformation field is accurately registered to the absolute reference using the weighted least squares algorithm. The registration accuracy of the deformation field is improved to the sub-centimeter level, and at the same time, the spatial detail features of high-density InSAR deformation points are retained, solving the problem of insufficient absolute positioning accuracy of traditional InSAR data.

[0038] The effect of step S2 is to eliminate the systematic deviation of GNSS, InSAR, and leveling monitoring data through spatio-temporal reference unification and multi-source data fusion, construct a high-precision and highly reliable surface deformation field, and provide a consistent spatial reference framework for land subsidence analysis.

[0039] Step S3 includes the following sub-steps: Step S301: Based on the observation data of ground GNSS continuous stations, the zenith total delay ZTD of each station is calculated using the Saastamoinen model.

[0040] Step S301 calculates the zenith total delay of GNSS stations based on the Saastamoinen model, quantifies the influence of atmospheric water vapor on the electromagnetic wave propagation path, provides a high-precision initial field for subsequent spatial interpolation, and breaks through the bottleneck of the lack of water vapor parameters in traditional InSAR atmospheric correction.

[0041] In step S302, the ZTD calculated by GNSS stations is spatially interpolated to construct grid data of atmospheric water vapor content with a resolution of 1 km × 1 km, improving the accuracy of the description of the spatial distribution of tropospheric water vapor.

[0042] In step S302, grid data of atmospheric water vapor with a resolution of 1 km × 1 km is generated through spatial interpolation, finely depicting the spatial heterogeneity of tropospheric water vapor, and providing high-resolution input for turbulent delay correction compared with traditional low-resolution models.

[0043] In step S303, the residual atmospheric phase delay error observed by InSAR is separated into stratospheric delay component and turbulent delay component.

[0044] In step S303, the InSAR residual atmospheric phase is decomposed into stratospheric and turbulent components, realizing the physical separation of error components, providing a theoretical basis for targeted correction, and solving the problem of incomplete correction of traditional single models.

[0045] In step S304, the stratospheric delay component and the turbulent delay component are eliminated. Among them, the stratospheric delay component is eliminated by establishing an elevation-related model using MERRA-2 reanalysis data and polynomial fitting, and the turbulent delay component is corrected based on the GNSS water vapor grid data to construct a spatial filtering operator.

[0046] Stratospheric delay correction in step S304: Using MERRA-2 reanalysis data to construct an elevation-related model, eliminating the systematic deviation related to terrain through polynomial fitting, and breaking through the limitation of poor adaptability of traditional models to complex terrain.

[0047] Construct a spatial filtering operator based on GNSS water vapor grid data, effectively suppressing the random phase fluctuations caused by turbulence, and making the InSAR deformation monitoring results reach sub-centimeter accuracy.

[0048] The effect of step S3 is to reduce the InSAR deformation monitoring error to the millimeter level through multi-scale atmospheric delay correction, significantly improving the compensation accuracy of atmospheric phase delay under complex terrain conditions, and ensuring the reliability of the surface deformation inversion results.

[0049] Step S4 includes the following sub-steps: In step S401, a physical mechanism model is constructed to conduct numerical simulation analysis on the land subsidence process, including: Based on Biot consolidation theory, a fully coupled seepage-stress finite element model is established. According to the lithological characteristics of the soil layer, the study area is divided into multiple soil layer partitions, and mechanical and seepage parameters are assigned respectively. At the same time, a modified algorithm for the hydraulic hysteresis effect of unsaturated soil is introduced to improve the simulation accuracy of the model for the settlement caused by the change of the groundwater level. The InSAR deformation time series data, groundwater level dynamic monitoring data, and layered settlement monitoring data are used as the input boundary conditions of the model. An implicit iterative solution strategy is adopted and combined with a CUDA-based parallel computing architecture to accelerate the stiffness matrix solution process.

[0050] Step S401 constructs a fully coupled seepage-stress model based on Biot consolidation theory, quantifies the dynamic interaction between soil layer compression and groundwater flow, introduces a hydraulic hysteresis effect algorithm, accurately captures the hysteretic deformation caused by groundwater level fluctuations, uses InSAR deformation time series, water level dynamic monitoring, and layered settlement data as boundary inputs, realizes the real-time calibration of model parameters, and avoids the deviation caused by the simplification of assumptions in traditional models. The CUDA parallel architecture improves the solution efficiency of the million-degree-of-freedom model.

[0051] Step S402 constructs a data-driven machine learning model for intelligent prediction of land subsidence, including: Design a neural network architecture based on a metaheuristic optimization algorithm, use the buried depth of the groundwater level, drawdown of shallow groundwater, drawdown of deep groundwater, thickness of clay layer, thickness of sand layer, annual rainfall, and permeability coefficient as the input features of the model, and use a spatio-temporal attention mechanism to extract key influencing features.

[0052] Adopt a BP neural network as the settlement prediction model, and optimize the model parameters in real time through a metaheuristic algorithm. Further embed a transfer learning module to update the model periodically online. Import the latest monitoring data every 24 hours and fine-tune the network weights to maintain the adaptability of the model to environmental changes, and output the predicted land subsidence amount and its spatial distribution within the next 72 hours.

[0053] Step S402 automatically screens key influencing factors through a spatio-temporal attention mechanism, suppresses the interference of redundant information, and the metaheuristic algorithm adjusts the neural network weights in real time, shortening the prediction response time of the model to extreme settlement events. Through 24-hour periodic model fine-tuning, it adapts to the seasonal changes of soil layer parameters.

[0054] Step S4 realizes the accurate prediction of the long-term evolution law and short-term dynamic changes of land subsidence through the collaboration of physical mechanisms and data-driven models, supporting urban-level refined settlement risk early warning.

[0055] Step S5 includes the following sub-steps: Step S501: Screen the groundwater table depth, shallow groundwater drawdown, deep groundwater drawdown, clay layer thickness, sand layer thickness, annual rainfall, and permeability coefficient from geological and environmental factors as evaluation factors for land subsidence susceptibility.

[0056] In step S501, core factors are screened from two dimensions of geological structure and hydrological process, covering the main control factors of subsidence driving force, eliminating redundant indicators, and ensuring that the physical meaning of the evaluation system is clear and quantifiable.

[0057] Step S502: Calculate the subjective weights of each evaluation factor using the analytic hierarchy process (AHP). Construct a judgment matrix for comparing evaluation indicators based on expert judgment, and determine the subjective weight values of each factor through consistency testing.

[0058] In step S502, an expert judgment matrix is constructed through the analytic hierarchy process (AHP) to quantify the relative importance among factors, reflecting the empirical cognition in the field of geological environment and making up for the interpretability defect of data-driven methods.

[0059] Step S503: Calculate the objective weights of each evaluation factor using the CRITIC method based on historical monitoring data, and obtain the objective weight values by comprehensively considering the numerical variability and mutual correlation of each indicator.

[0060] Combine the subjective weight values and objective weight values, and use the weighted average method to calculate the final combined weights of each factor.

[0061] In step S503, the standard deviation and correlation coefficient of factors are calculated based on historical subsidence data through the CRITIC method. The objective weight is positively correlated with the factor standard deviation and negatively correlated with the absolute value of the correlation coefficient, eliminating the interference of information overlap between indicators. Compared with the single entropy weight method, the weight stability is improved.

[0062] Step S504: Calculate the comprehensive evaluation value of land subsidence susceptibility based on the combined weights of each factor and relevant industry specifications, and divide the land subsidence susceptibility into extremely low level, low level, medium level, high level, and extremely high level.

[0063] In step S504, the weighted average combined weights are used to balance experience and data rules. Five-level thresholds are divided according to the "Specifications for the Preparation of Land Subsidence Prevention and Control Plans" to generate a susceptibility map, which directly supports the delineation of limited groundwater extraction areas and the control of building loads.

[0064] The effect of step S5 is an evaluation method for susceptibility that integrates subjective and objective weights, scientifically quantifies the coupled action of multi-factors of geology-environment, generates a spatially continuous subsidence risk classification map, and provides a hierarchical decision-making basis for urban planning, groundwater control, and engineering protection.

[0065] Step S6 includes the following sub-steps: Step S601: Develop a visualization interface. The visualization interface obtains the depth of the groundwater table, drawdown of shallow groundwater, drawdown of deep groundwater, thickness of clay layer, thickness of sand layer, annual rainfall, and permeability coefficient input by the user.

[0066] In step S601, a B / S architecture 3D visualization engine is constructed, which is compatible with access from both PC and mobile devices, supports real-time input of parameters such as groundwater level and soil layer thickness and historical data retrieval, reduces the user operation response time, and lowers the threshold for using professional software.

[0067] Step S602: Based on the land subsidence susceptibility evaluation model and the subsidence prediction model, calculate the grade prediction result of land subsidence susceptibility in real time in the visualization interface.

[0068] In step S602, a lightweight susceptibility evaluation model and prediction model are embedded, and CUDA accelerated calculation is automatically triggered after the user inputs parameters.

[0069] Step S603: Visually display the predicted subsidence susceptibility grade result through the graphical user interface.

[0070] The visualization interface further provides an index risk grade analysis function. According to the evaluation result of land subsidence susceptibility, the disaster-causing risks of the influencing factors are evaluated respectively, and the disaster-causing risks of each influencing factor are classified.

[0071] In step S603, a heat map - contour line - three-dimensional profile multi-modal display technology is adopted. Based on the susceptibility evaluation model, the contribution degree of each factor is output to support the formulation of differentiated prevention and control strategies.

[0072] Step S6 realizes the dynamic display and intelligent decision-making support of the whole process of land subsidence monitoring - prediction - evaluation through a visualization interaction platform. After the user inputs parameters, a subsidence risk map is generated in seconds, which supports multi-dimensional risk tracing and emergency plan simulation, and significantly improves the management efficiency of geological disasters and the public's risk awareness.

[0073] Example 2, referring to Figure 2 , provides an urban land subsidence analysis system based on multi-source data fusion, including a network monitoring module, a fusion analysis module, an error correction module, a subsidence prediction module, a subsidence division module, and a visual interaction module.

[0074] The network monitoring module is used to build a multi-source three-dimensional monitoring network to synchronously monitor the land subsidence of the target city at multiple levels on the surface and underground.

[0075] The fusion analysis module is used to perform air - space - ground multi-source data fusion analysis to unify the spatio-temporal benchmarks of each monitoring data.

[0076] The error correction module is used to correct the atmospheric delay error of InSAR monitoring data.

[0077] The settlement prediction module is used to construct a ground settlement prediction model for modeling and predicting the development trend of ground settlement.

[0078] The settlement division module is used to conduct an evaluation of ground settlement susceptibility, determine the weights of influencing factors, and divide the settlement susceptibility levels.

[0079] The visual interaction module is used to develop a visualization interface for ground settlement analysis, and integrate and display monitoring data, prediction results, and evaluation information for interaction.

[0080] Through the coordinated deployment of the surface leveling network, underground layered settlement marks, and groundwater monitoring system, all-dimensional monitoring from the surface to the deep soil layer is achieved, with a data resolution of the minute level, a spatial coverage extended to the entire urban area, a dynamic conversion model eliminating the benchmark difference between GNSS and leveling elevations, Delaunay triangulation network constraints improving the registration accuracy of the InSAR deformation field, the overall monitoring error reduced to the millimeter level, atmospheric water vapor raster data combined with MERRA-2 reanalysis data significantly improving the InSAR atmospheric delay correction effect, a physical model introducing a non-saturated soil hydraulic hysteresis correction algorithm to accurately simulate the settlement hysteresis effect caused by water level fluctuations, a machine learning model capturing key influencing factors through a spatio-temporal attention mechanism and realizing 24-hour periodic model updates through transfer learning, and the combination of the analytic hierarchy process and the CRITIC method, which not only retains expert experience but also quantifies the objective laws of data.

[0081] The 3D visualization engine integrates monitoring data, prediction results, and risk assessment, supports spatio-temporal dynamic playback and multi-scenario simulation, the index risk level analysis function can trace the disaster-causing contribution of a single factor, assists in formulating targeted prevention and control measures, proposes a full-process technical closed-loop of "monitoring - fusion - correction - prediction - evaluation - display", fills the methodological gap in the field of comprehensive analysis of urban ground settlement, and meets the application requirements of super-large cities.

[0082] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system or a computer program product. Therefore, the present invention can take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program code. Among them, the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk. These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured article including an instruction device, and the instruction device implements the process Figure 1 in one process or multiple processes and / or boxes Figure 1 the functions specified in one box or multiple boxes.

[0083] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered by the scope of the claims of the present invention.

Claims

1. A method for analyzing urban land subsidence based on multi-source data fusion, characterized in that It includes the following steps: Step S1, construct a multi-source three-dimensional monitoring network to conduct synchronous monitoring of the ground settlement of the target city at multiple levels on the surface and underground; Step S2, conduct multi-source data fusion analysis of air-space-ground, and unify the spatio-temporal benchmarks of each monitoring data; Step S3, correct the atmospheric delay error of InSAR monitoring data; Step S4, construct a ground settlement prediction model to model and predict the development trend of ground settlement; Step S5, carry out an evaluation of the susceptibility of ground settlement, determine the weights of influencing factors and divide the settlement susceptibility levels; Step S6, develop a visualization interface for ground settlement analysis, and integratively display and interact the monitoring data, prediction results and evaluation information.

2. The urban land subsidence analysis method based on multi-source data fusion according to claim 1, wherein The said Step S1 includes the following sub-steps: Step S101, layout a second-class leveling control network on the surface layer as the reference framework for ground settlement monitoring; Step S102, install layered settlement markers in the deep underground and carry GNSS monitoring equipment to establish a layered settlement dynamic monitoring system to obtain real-time data feedback on ground settlement at different depths; Step S103, deploy high-precision water level monitoring sensors in the groundwater monitoring layer, and construct a pressure-water level joint monitoring system in typical monitoring holes to collect the buried depth of the groundwater level and pore water pressure parameters at a minute-level frequency; Step S104, integrate the leveling control network, layered settlement monitoring system and groundwater monitoring system through Internet of Things technology to form a real-time data transmission network.

3. The urban land subsidence analysis method based on multi-source data fusion according to claim 2, characterized in that The said Step S2 includes the following sub-steps: Step S201, take the elevation change obtained by the second-class leveling control network as the absolute reference, and use GNSS monitoring data to establish a dynamic conversion model between GNSS phase elevation and leveling elevation to achieve the unification of the absolute reference of multi-source data. The dynamic conversion model satisfies: ； Among them, is the phase elevation, is the leveling elevation, is the conversion coefficient between the phase elevation and the leveling elevation, is the elevation change amount, is the leveling elevation correction factor, is the temperature under a standard atmospheric pressure, is the atmospheric delay compensation term; Step S202, use the persistent scatterer InSAR technology to obtain the relative deformation data of high-density surface deformation points, establish the spatial constraint relationship between InSAR deformation points and leveling control points by constructing a Delaunay triangulation network, and use the weighted least squares algorithm to register the relative deformation field to the absolute reference.

4. The urban land subsidence analysis method based on multi-source data fusion according to claim 3, wherein The said Step S3 includes the following sub-steps: Step S301, based on the observation data of ground GNSS continuous stations, use the Saastamoinen model to calculate the zenith total delay ZTD of each station; Step S302, conduct spatial interpolation on the ZTD calculated by GNSS stations to construct a grid data of atmospheric water vapor content with a resolution of 1km×1km to improve the accuracy of the description of the spatial distribution of tropospheric water vapor; Step S303, separate the residual atmospheric phase delay error observed by InSAR into stratospheric delay component and turbulent delay component; Step S304, eliminate the stratospheric delay component and turbulent delay component. Among them, the stratospheric delay component is eliminated by establishing an elevation-related model using MERRA-2 reanalysis data and polynomial fitting, and the turbulent delay component is corrected based on the GNSS water vapor grid data to construct a spatial filtering operator.

5. The urban land subsidence analysis method based on multi-source data fusion according to claim 4, characterized in that The said Step S4 includes the following sub-steps: Step S401, construct a physical mechanism model to conduct numerical simulation analysis on the ground settlement process, including: Based on Biot consolidation theory, a fully coupled seepage-stress finite element model is established. According to the lithological characteristics of the soil layer, the study area is divided into multiple soil layer partitions, and mechanical and seepage parameters are assigned respectively. At the same time, a correction algorithm for the hydraulic hysteresis effect of unsaturated soil is introduced to improve the simulation accuracy of the model for the settlement caused by the change of the groundwater level. The InSAR deformation time series data, the dynamic monitoring data of the groundwater level, and the layered settlement monitoring data are used as the input boundary conditions of the model. An implicit iterative solution strategy is adopted and combined with a parallel computing architecture based on CUDA to accelerate the stiffness matrix solution process.

6. The urban land subsidence analysis method based on multi-source data fusion according to claim 5, characterized in that The step S4 further includes the following sub-steps: Step S402, constructing a data-driven machine learning model for intelligent prediction of land subsidence, including: Designing a neural network architecture based on a metaheuristic optimization algorithm, taking the groundwater level depth, shallow groundwater drawdown, deep groundwater drawdown, clay layer thickness, sand layer thickness, annual rainfall, and permeability coefficient as the input features of the model, and using a spatio-temporal attention mechanism to extract key influencing features; Using a BP neural network as the settlement prediction model, and real-time optimizing the model parameters through a metaheuristic algorithm. Further, a transfer learning module is embedded to periodically update the model online. Import the latest monitoring data every 24 hours and fine-tune the network weights to maintain the adaptability of the model to environmental changes, and output the land subsidence prediction amount and its spatial distribution within the next 72 hours.

7. The urban land subsidence analysis method based on multi-source data fusion according to claim 6, wherein The step S5 includes the following sub-steps: Step S501, screening the groundwater level depth, shallow groundwater drawdown, deep groundwater drawdown, clay layer thickness, sand layer thickness, annual rainfall, and permeability coefficient from geological and environmental factors as the evaluation factors for land subsidence susceptibility; Step S502, using the analytic hierarchy process to calculate the subjective weights of each evaluation factor, constructing a judgment matrix for comparing evaluation indicators according to expert judgment, and determining the subjective weight values of each factor through consistency test; Step S503, using the CRITIC method to calculate the objective weights of each evaluation factor based on historical monitoring data, and obtaining the objective weight values by comprehensively considering the numerical variability and mutual correlation of each index; Combining the subjective weight values and objective weight values, using the weighted average method to calculate the final combined weights of each factor; Step S504, based on the combined weights of each factor and relevant industry specifications, calculating the comprehensive evaluation value of land subsidence susceptibility, and dividing the land subsidence susceptibility into extremely low level, low level, medium level, high level, and extremely high level.

8. The method for analyzing urban land subsidence based on multi-source data fusion according to claim 7, wherein, The step S6 includes the following sub-steps: Step S601, developing a visualization interface, which obtains the groundwater level depth, shallow groundwater drawdown, deep groundwater drawdown, clay layer thickness, sand layer thickness, annual rainfall, and permeability coefficient input by the user; Step S602, based on the land subsidence susceptibility evaluation model and the settlement prediction model, calculating the grade prediction result of land subsidence susceptibility in real time in the visualization interface; Step S603, visually displaying the predicted settlement susceptibility grade result through a graphical user interface.

9. The urban land subsidence analysis method based on multi-source data fusion according to claim 8, characterized in that, The visualization interface further provides an index risk level analysis function, which conducts a disaster-causing risk assessment for each of the influencing factors according to the evaluation results of the ground settlement susceptibility, and classifies the disaster-causing risks of each influencing factor.

10. A urban land subsidence analysis system based on multi-source data fusion, which is applied to a urban land subsidence analysis method based on multi-source data fusion as described in any one of claims 1-9, and is characterized in that, It includes a network monitoring module, a fusion analysis module, an error correction module, a settlement prediction module, a settlement classification module, and a visual interaction module; The network monitoring module is used to construct a multi-source three-dimensional monitoring network to synchronously monitor the ground settlement of the target city at multiple levels on the surface and underground; The fusion analysis module is used to conduct multi-source data fusion analysis of space-air-ground to unify the spatio-temporal benchmarks of each monitoring data; The error correction module is used to correct the atmospheric delay error of InSAR monitoring data; The settlement prediction module is used to construct a ground settlement prediction model to model and predict the development trend of ground settlement; The settlement classification module is used to conduct an evaluation of ground settlement susceptibility, determine the weights of influencing factors, and classify the settlement susceptibility levels; The visual interaction module is used to develop a visualization interface for ground settlement analysis to integratively display and interact with monitoring data, prediction results, and evaluation information.

Images