Space-air-ground collaborative slope monitoring and intelligent early warning system

Abstract

The application discloses a space-air-ground cooperative slope monitoring intelligent early warning system, and relates to the technical field of monitoring and early warning.Through space-based InSAR deformation field spatial gradient anomaly detection, a global deformation anomaly area is identified, and a key detailed investigation area list is generated, according to which, an air-based unmanned aerial vehicle is triggered to perform fine inspection, and a risk urgency score is calculated based on multi-period digital elevation model difference and crack semantic segmentation, and then, a ground-based synthetic aperture radar is triggered to perform millimeter-level beam locking and real-time tracking collection on a high-risk point, and a multi-stage adaptive early warning mechanism based on a safety coefficient physical constraint is combined.The application has the advantages that: space coordinates are used for step-by-step constraint to realize the automatic progressive scheduling and resource closed-loop adjustment from surface area census to regional detailed investigation to point location fine investigation, the problems of spatial fragmentation and hierarchical triggering missing of existing monitoring means are solved, and the cooperative early warning capability from wide-area anomaly discovery to key point location accurate positioning is improved.

Description

Technical Field

[0001] This invention relates to the field of monitoring and early warning technology, specifically to an intelligent early warning system for slope monitoring that integrates air, space, and ground. Background Technology

[0002] Traditional slope monitoring methods mainly include manual inspection, GNSS single-point displacement monitoring, and deep inclination measurement, which generally have shortcomings such as limited monitoring coverage, insufficient spatial representativeness, and poor early warning timeliness.

[0003] The applicant discovered through a search that a Chinese patent, "A Three-in-One Slope Monitoring and Grading Early Warning Method for Open-Pit Mines," with publication number "CN116978187A," discloses a method that independently calculates early warning coefficients from satellite InSAR data, 3D laser scanning data, and soil monitoring data. These coefficients are then weighted and fused into a comprehensive early warning coefficient using fixed weights. This coefficient is then compared with a fixed threshold to determine the early warning level. This method only involves a numerical weighted fusion relationship between the three data sources and lacks a progressive triggering mechanism based on spatial coordinate constraints, such as "general survey → detailed survey → precise survey." After the discovery of space-based wide-area deformation anomalies, it cannot automatically trigger a space-based targeted detailed survey of the anomaly area. Furthermore, the results of the space-based detailed survey cannot automatically trigger the precise locking of high-risk points by the ground-based system. Consequently, monitoring resources cannot be dynamically configured hierarchically according to the actual risk status of the slope. In addition, another Chinese patent, "A Holographic Three-Dimensional Networked Smart Early Warning Method for Landslides Based on Multiple Monitoring Scenarios," with publication number "CN115691058A," mainly integrates three machine learning algorithms—support vector machine, random forest, and C5.0 decision tree—to obtain graded early warning results through weighted voting with fixed weight ratios. It also establishes a four-level early warning standard using the deformation tangent angle as an indicator. This method adopts an ensemble learning strategy at the data fusion level, but its weights remain unchanged during deployment after being determined during model training. This is a static weight at the statistical level, failing to use the physical and mechanical state of the slope (such as the safety factor) as a constraint for weight adjustment. Consequently, the early warning model lacks generalization ability and physical interpretability across engineering scenarios when facing the nonlinear evolution process of slopes from stability to creep to accelerated deformation to instability. Therefore, we propose a space-air-ground collaborative intelligent early warning system for slope monitoring. Summary of the Invention

[0004] The purpose of this invention is to provide an intelligent early warning system for slope monitoring that integrates air, space, and ground technologies.

[0005] To achieve the above objectives, the present invention provides the following technical solution: an intelligent early warning system for slope monitoring that integrates air, space, and ground systems, the early warning system comprising: Space-based global survey module: The data output end is connected to the data input end of the progressive trigger control module. It is used to acquire spaceborne synthetic aperture radar remote sensing data of the entire slope area and process it to obtain the global deformation rate field. The global deformation rate field is then sent to the progressive trigger control module. The progressive trigger control module has its control output connected to the control inputs of the airborne key survey module, the ground-based key detailed survey module, and the space-based global survey module, respectively. It is used to identify deformation anomaly clusters based on the received global deformation rate field, trigger the airborne key survey module to start, and constrain its inspection range by the spatial boundary of the anomaly area; it triggers the ground-based key detailed survey module to start based on the risk results returned by the airborne key survey module, and constrains its monitoring targets by the coordinates of the risk points; it is also used to adjust the monitoring frequency of the space-based global survey module and the resource allocation status of the ground-based key detailed survey module in reverse based on the displacement data returned by the ground-based key detailed survey module. The data output terminal of the airborne key detailed investigation module is connected to the data input terminals of the progressive trigger control module and the multi-stage adaptive early warning module respectively; it is used to perform refined inspection of the designated area by UAV, obtain regional elevation change and crack development characteristic data, calculate and generate risk urgency results and send them back to the progressive trigger control module, and at the same time send the original inspection data to the multi-stage adaptive early warning module. The data output terminal of the foundation key detailed investigation module is communicatively connected to the data input terminals of the progressive trigger control module, the multi-stage adaptive early warning module, and the parameter inversion calibration module, respectively; it is used to perform high-precision continuous deformation monitoring on specified risk points, obtain millimeter-level displacement time series data, and send them to the corresponding modules respectively. The data input end of the multi-stage adaptive early warning module is connected to the data output ends of the space-based comprehensive survey module, the air-based key detailed survey module, and the ground-based key fine survey module, respectively. The data output end is connected to the visualization early warning module. It is used to divide the evolution stages based on the physical and mechanical state of the slope, dynamically adjust the fusion weight and early warning threshold of the three types of monitoring data, and output graded early warning results. The data input end of the parameter inversion calibration module is connected to the data output ends of the space-based global survey module and the foundation key detailed survey module, respectively. The data output end is connected to the data input end of the multi-stage adaptive early warning module. It is used to combine space-based surface deformation data and foundation point displacement data to invert and calibrate slope geotechnical parameters, and feed the calibration results back to the multi-stage adaptive early warning module. The data input of the remaining safety time prediction module is connected to the data output of the parameter inversion calibration module and the external rainfall forecast data interface. The data output is connected to the data input of the visualization early warning module. It is used to deduce the temporal change of the slope safety factor based on the calibrated mechanical parameters and rainfall forecast data, and output the remaining safety time and the corresponding confidence interval. The data input end of the visualization early warning module is connected to the data output ends of the multi-stage adaptive early warning module and the remaining safety time prediction module, respectively. It is used to integrate multi-source monitoring data and early warning results for three-dimensional visualization and to match and push corresponding early warning response plans.

[0006] As a further aspect of the present invention: the space-based global survey module is specifically used to: acquire the original image data of the spaceborne synthetic aperture radar, and sequentially perform image registration, interferometric phase generation, phase unwrapping, and geocoding processing to obtain the global deformation rate field in the geodetic coordinate system, wherein each spatial pixel of the deformation rate field corresponds to the annual average deformation rate of the corresponding location of the slope.

[0007] As a further aspect of the present invention: the airborne key investigation module is specifically used for: controlling a drone equipped with a lidar and a high-resolution optical camera to perform autonomous flight data collection within a designated inspection range, obtaining high-density lidar point cloud data and high-resolution optical images; constructing a digital elevation model for the lidar point cloud data, and obtaining the spatial distribution of regional elevation changes through multi-period elevation model difference; performing pixel-level crack semantic segmentation on the optical images to obtain the crack's location, aperture, and propagation rate; and calculating a risk urgency score by comprehensively considering elevation changes, crack propagation rate, and anomaly area area, using the following formula: ; in, The risk urgency level is scored, dimensionless. The maximum elevation change within the monitoring area, in meters. This is a critical reference value for elevation change, in meters, pre-determined according to the safety level of the slope engineering. To monitor the maximum crack propagation rate within the area, the unit is meters per day. This is a critical reference value for crack propagation rate, expressed in meters per day, pre-calibrated according to the slope's soil and rock type. This represents the total area of ​​the deformation anomaly region, in square meters. This is a critical reference value for the area of ​​the abnormal region, in square meters, and is pre-determined based on the scale of the slope engineering project. These are the normalized weighting coefficients corresponding to the three indicators, and the sum of the three is 1.

[0008] As a further aspect of the present invention: the foundation key fine investigation module is specifically used to: control the foundation synthetic aperture radar equipment, automatically adjust the azimuth and elevation angles of the antenna according to the coordinates of the target point, so that the radar beam completely covers the target risk area, and calculate the continuous deformation displacement time series of the target point through synthetic aperture imaging and differential interferometry, with a sampling frequency of not less than once every 5 minutes and deformation measurement accuracy reaching the sub-millimeter level.

[0009] As a further aspect of the present invention: the progressive triggering control module specifically performs forward progressive triggering control and reverse closed-loop feedback control, specifically including: The deformation rate field output by the space-based global survey module is received, and the spatial gradient anomaly detection algorithm is used to identify deformation anomaly clusters: the spatial gradient amplitude of the deformation rate of each pixel is calculated, and pixels with gradient amplitudes exceeding the adaptive threshold are used as seed points. A continuous deformation anomaly cluster is generated through a density clustering algorithm, and a list of key areas for detailed investigation is generated according to risk priority. The list is sent to the space-based key area for detailed investigation module, and the spatial boundary of the anomaly area is used as the spatial constraint range for UAV inspection. The formula for calculating the spatial gradient magnitude of the deformation rate is: ; in, For the deformation rate along gradient components in the direction, For the deformation rate along gradient components in the direction, This represents the gradient magnitude corresponding to a single pixel. The formula for calculating the adaptive threshold is: ; in, The mean of the global deformation rate gradient magnitude. The standard deviation of the global deformation rate gradient magnitude; The system receives the risk urgency score output by the airborne key investigation module. When the score exceeds the preset trigger threshold, it sends the coordinates of the high-risk points to the ground-based key investigation module, using the point coordinates as the target constraint for radar beam locking. Reverse closed-loop feedback control: The spatial boundary coordinates of the anomaly area and the coordinates of the high-risk points respectively constitute a direct constraint of the spatial search range of the next monitoring scale on the output of the previous monitoring scale, realizing multi-scale progressive coordination with progressively narrowing coverage area and progressively improving spatial resolution.

[0010] As a further aspect of the present invention: the progressive triggering control module performs reverse closed-loop feedback control, specifically including: The system receives the displacement time series output by the ground-based key detailed survey module. When the displacement rate is lower than the safety threshold for no less than 3 consecutive observation periods, it determines that the risk point is stabilizing, controls the space-based full-domain survey module to reduce to the frequency of routine survey monitoring, and releases ground-based radar resources. When the displacement rate shows a continuous accelerating trend, the risk is judged to be aggravated. The space-based global survey module is controlled to increase the monitoring frequency, increase the satellite data ordering cycle, and maintain the high-frequency acquisition state of the ground-based radar. The adjustment targets of the reverse closed-loop feedback are the operating resources and acquisition rhythm of the monitoring system itself, including shortening or extending the ordering cycle of space-based satellite data, allocating or releasing ground-based radar resources, and increasing or decreasing the frequency of UAV inspections. The system achieves fully automatic dynamic optimal configuration of monitoring resources without human intervention.

[0011] As a further aspect of the present invention: the multi-stage adaptive early warning module specifically performs multi-stage adaptive early warning based on physical mechanism constraints, including: (1) Calculation of real-time safety factor of slope based on simplified limit equilibrium theory By combining deformation acceleration to jointly determine the evolution stage of the slope, the slope evolution process is divided into four stages: stable period, creep period, accelerated deformation period, and instability period. (2) According to different evolution stages, the weight ratio of space-based monitoring data, air-based monitoring data and ground-based monitoring data in the early warning fusion model is dynamically adjusted. The stage switching is triggered by the joint event of the safety factor crossing the physical threshold and the change of deformation acceleration. The timing of the fusion weight switching of each stage is rooted in the change of the physical and mechanical state of the slope. (3) A dynamic early warning threshold mechanism is adopted, in which the early warning threshold is dynamically adjusted according to the real-time safety factor. The calculation formula is as follows: ; in, The dynamic early warning threshold is dimensionless. The baseline warning threshold is dimensionless and is preset according to the project's safety level. The slope safety factor is a dimensionless real-time factor. To adjust the parameters, the value range is 0.5-1.5, and it should be adjusted according to the type of slope rock and soil. The lower the safety factor, the smaller the dynamic warning threshold and the higher the warning sensitivity.

[0012] As a further aspect of the present invention: the parameter inversion calibration module is specifically used to: construct an inversion objective function using space-based surface deformation observation data and ground-based point displacement observation data as joint constraints, and solve for the cohesion of the slope soil and rock mass. and internal friction angle The optimal value is obtained; the time-varying characteristics of the parameters are continuously tracked using a rolling time window, and the calibrated parameters are updated in real time to the safety factor calculation model; the expression of the inversion objective function is: ; in, Let be the objective function value, and the objective is to minimize it. The values ​​represent the area deformation obtained from space-based INSAR observations, in meters. These are the displacement values ​​of points obtained from ground-based radar observations, in meters. Based on cohesion and internal friction angle The deformation prediction values, calculated using the forward model, are in meters. The weighting coefficients for the space-based observation term are dimensionless. The weighting coefficients for the ground-based observation terms are dimensionless. is the spatial smoothing regularization constraint coefficient, dimensionless, used to suppress ill-conditioned oscillations in the inversion results. This is the parameter space gradient regularization term, used to ensure the smoothness of the spatial distribution of geotechnical parameters; The spatial correspondence strategy between space-based surface data and ground-based point data adopts buffer sampling and shape function interpolation: a circular buffer is established with the ground-based observation point as the center and a radius of 1.5 times the INSAR pixel resolution. The median of all INSAR deformation values ​​in the buffer is taken as the representative value of the surface observation at that point. The predicted values ​​of the forward model grid nodes are mapped to the coordinates of the ground-based observation point through shape function interpolation to achieve spatial matching of multi-scale data.

[0013] As a further aspect of the present invention: the remaining safety time prediction module is specifically used for: based on calibrated geotechnical parameters and combined with future rainfall forecast data, to extrapolate the temporal changes in groundwater level and slope safety factor caused by rainfall infiltration day by day, until the critical date when the safety factor first drops below 1.0, and to calculate the time difference between the current date and the critical date as the remaining safety time; simultaneously, the Monte Carlo simulation method is used to quantify uncertainty and output the 95% confidence interval of the remaining safety time: using the rainfall forecast probability distribution, the permeability coefficient probability distribution, and the posterior covariance of the inversion parameters as inputs, at least 1000 random forward extrapolations are performed to obtain the empirical probability distribution of the remaining safety time, and the 2.5% quantile and 97.5% quantile are taken as the upper and lower limits of the 95% confidence interval, and the output format is as follows: , This is the lower limit of the 95% confidence interval for the remaining safe time. This is the upper limit of the 95% confidence interval for the remaining safe time.

[0014] As a further aspect of the present invention: the visualization early warning module is specifically used to: uniformly map the space-based deformation field, the air-based real-scene 3D model, the ground displacement curve, and the rainfall and groundwater level data to the real-scene 3D model base under the 2000 National Geodetic Coordinate System, and spatially visualize the four warning levels using four colors: green, yellow, orange, and red, respectively. Based on the warning level and risk type, it automatically matches the disposal measures in the contingency plan database and pushes the warning information and disposal plan to the designated terminal device.

[0015] Compared with the prior art, the beneficial effects of the present invention by adopting the above technical solution are as follows: 1. This invention constructs a multi-scale collaborative monitoring architecture that progresses step-by-step from area screening to detailed regional investigation and precise point-based investigation. It establishes a spatial constraint and two-way linkage mechanism among space-based, airborne, and ground-based monitoring methods, solving the core problems of spatial information loss and diluted advantages of different scale monitoring methods caused by parallel multi-source data acquisition and flattened fusion in existing technologies. In the forward progressive triggering link, the output results of the previous level monitoring directly serve as the spatial constraint conditions for the next level monitoring, achieving orderly collaboration with progressively narrowing coverage and progressively improving monitoring accuracy. In the reverse closed-loop feedback link, the system autonomously adjusts the intensity and allocation strategy of monitoring resources based on the real-time risk status of the slope. Ultimately, this achieves an architectural upgrade of the monitoring system from open-loop parallel operation to closed-loop autonomous scheduling, improving the utilization efficiency of high-cost monitoring equipment while fully releasing the differentiated technological advantages of monitoring methods at various spatial scales. 2. This invention constructs a multi-stage adaptive early warning architecture constrained by physical mechanisms, embedding the slope's mechanical safety state as the core anchor point into a data-driven early warning fusion model. This solves the core problem of existing early warning models, which either rely on fixed weight thresholds or are purely data-driven and lack physical basis, thus failing to adapt to the nonlinear evolution of slopes. Based on the actual mechanical evolution stage of the slope, the fusion weights of monitoring data from different sources are dynamically adjusted. Simultaneously, the sensitivity of early warning triggering is dynamically adjusted based on the mechanical state, and physical mechanism verification further reduces the risk of false alarms caused by data noise. Ultimately, this upgrades the early warning model from static parameter configuration to physical-driven dynamic adaptation, significantly improving the model's generalization ability and judgment accuracy under different geological conditions and different evolution stages. 3. This invention establishes a two-way value closed-loop mechanism for joint inversion of multi-source monitoring data and prediction of remaining safety time. This upgrades monitoring data from a single input for current state assessment to a core basis for continuous calibration of model parameters, solving the problems in existing technologies where monitoring data is only used for state assessment, cannot feed back into model iteration, and the output results are insufficient to support proactive emergency decision-making. It jointly calibrates geotechnical parameters with space-based surface deformation data and high-precision ground-based point data, and uses meteorological forecast data to extrapolate the temporal changes in slope stability, while quantifying the confidence interval of the prediction results. Ultimately, it achieves a leap in monitoring value from post-event state evaluation to pre-event instability prediction, and the output quantifiable time window information can directly provide clear decision-making references for engineering emergency response. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the system flow in an embodiment of the present invention; Figure 2 This is a schematic diagram of the system architecture in an embodiment of the present invention; Figure 3 This is a schematic diagram of the multi-stage adaptive early warning workflow in an embodiment of the present invention. Detailed Implementation

[0017] The specific embodiments of the present invention will be further described below with reference to the accompanying drawings. It should be noted that the description of these embodiments is for the purpose of helping to understand the present invention, but does not constitute a limitation of the present invention.

[0018] Furthermore, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0019] Please see the appendix Figure 1 -Appendix Figure 3 This invention discloses an intelligent early warning system for slope monitoring that integrates air, space, and ground systems. The early warning system includes: Space-based global survey module: The data output end is connected to the data input end of the progressive trigger control module. It is used to acquire spaceborne synthetic aperture radar remote sensing data of the entire slope area and process it to obtain the global deformation rate field, and send the global deformation rate field to the progressive trigger control module. The progressive trigger control module has its control output connected to the control inputs of the airborne key survey module, the ground-based key detailed survey module, and the space-based global survey module, respectively. It is used to identify deformation anomaly clusters based on the received global deformation rate field, trigger the airborne key survey module to start, and constrain its inspection range by the spatial boundary of the anomaly area; it triggers the ground-based key detailed survey module to start based on the risk results returned by the airborne key survey module, and constrains its monitoring targets by the coordinates of the risk points; it is also used to adjust the monitoring frequency of the space-based global survey module and the resource allocation status of the ground-based key detailed survey module in reverse based on the displacement data returned by the ground-based key detailed survey module. The data output end of the airborne key detailed investigation module is connected to the data input ends of the progressive trigger control module and the multi-stage adaptive early warning module respectively; it is used to perform refined inspection of the designated area by UAV, obtain regional elevation change and crack development characteristic data, calculate and generate risk urgency results and send them back to the progressive trigger control module, and at the same time send the original inspection data to the multi-stage adaptive early warning module. The data output terminal of the foundation key detailed investigation module is communicatively connected to the data input terminals of the progressive trigger control module, the multi-stage adaptive early warning module, and the parameter inversion calibration module, respectively; it is used to perform high-precision continuous deformation monitoring on specified risk points, obtain millimeter-level displacement time series data, and send them to the corresponding modules respectively. The data input end of the multi-stage adaptive early warning module is connected to the data output ends of the space-based comprehensive survey module, the air-based key detailed survey module, and the ground-based key fine survey module, respectively. The data output end is connected to the visualization early warning module. It is used to divide the evolution stages based on the physical and mechanical state of the slope, dynamically adjust the fusion weight and early warning threshold of the three types of monitoring data, and output graded early warning results. The data input end of the parameter inversion calibration module is connected to the data output ends of the space-based global survey module and the foundation key detailed survey module, respectively. The data output end is connected to the data input end of the multi-stage adaptive early warning module. It is used to combine space-based surface deformation data and foundation point displacement data to invert and calibrate slope geotechnical parameters, and feed the calibration results back to the multi-stage adaptive early warning module. The data input of the remaining safety time prediction module is connected to the data output of the parameter inversion calibration module and the external rainfall forecast data interface. The data output is connected to the data input of the visualization early warning module. It is used to deduce the temporal change of the slope safety factor based on the calibrated mechanical parameters and rainfall forecast data, and output the remaining safety time and the corresponding confidence interval. The data input end of the visualization early warning module is connected to the data output ends of the multi-stage adaptive early warning module and the remaining safety time prediction module, respectively. It is used to integrate multi-source monitoring data and early warning results for three-dimensional visualization and to match and push corresponding early warning response plans. Example

[0020] This example was applied to a rock slope section from K120+300 to K122+100 of a highway in a mountainous area of ​​Southwest China. The slope is 82m high with a slope angle of 38°, and the rock type is moderately weathered sandstone. The total monitoring area is approximately 1.8km². 2 The slope has complex terrain and poor accessibility, making it suitable for large-scale routine monitoring using a combined air-space-ground system.

[0021] 1.1 Operation process of the space-based global survey module The space-based global survey module acquires C-band synthetic aperture radar images from the Sentinel-1 satellite with a spatial resolution of 20m×22m and a revisit period of 12 days.

[0022] Image registration, interferometric phase generation, phase unwrapping, and geocoding are sequentially performed on the original image to obtain the global deformation rate field in the WGS-84 geodetic coordinate system, with the annual average deformation rate corresponding to the location of the slope for each pixel. In this embodiment, the average global deformation rate is 1.8 mm / year, indicating that the overall process is stable.

[0023] 1.2 Forward triggering process of the progressive triggering control module After receiving the deformation rate field, the progressive trigger control module uses a spatial gradient anomaly detection algorithm to identify deformation anomaly clusters. The specific process is as follows: The deformation rate of each cell is calculated using the Sobel operator. Directional gradient components and Directional gradient components Synthesized gradient magnitude: ; The mean of the global deformation rate gradient magnitude Standard deviation Calculate the adaptive threshold: ; With gradient magnitude exceeding Using the pixels as seed points, the DBSCAN density clustering algorithm is used to generate continuous deformation anomaly clusters, with a neighborhood radius of 3 pixels (approximately 60m) and a minimum number of points of 5.

[0024] In this embodiment, two deformation anomaly clusters were identified, designated A1 and A2. Area A1 has an area of ​​1260 m², an average deformation rate of 8.7 mm / year, and an average gradient amplitude of 4.2 mm / (year·pixel); area A2 has an area of ​​890 m², with an average deformation rate of 6.3 mm / year. A list of key areas for detailed investigation was generated after prioritizing by risk, and this list was sent to the airborne key area investigation module. The spatial boundary coordinates of the anomaly areas were directly used as the spatial constraint range for UAV inspection.

[0025] 1.3 Operation process of the air-based key investigation module The airborne key inspection module controls a multi-rotor UAV equipped with lidar and a 20-megapixel optical camera to perform autonomous flight inspections of areas A1 and A2, with a relative flight altitude of 150m and a lidar point cloud density of 52 points / m². 2 The optical imagery has a ground resolution of 4.8 cm. The acquired laser point cloud data was denoised, classified, and interpolated to construct a regional digital elevation model (DEM) with a resolution of 0.1 m. The current DEM was registered with the baseline DEM using the ICP algorithm, achieving a registration accuracy of 0.042 m RMS. The spatial distribution of elevation changes was obtained through pixel-by-pixel differencing. Cracks were identified using the U-Net semantic segmentation network on the optical imagery, yielding the crack location, aperture, and propagation rate. The risk urgency score for area A1 was calculated by combining the elevation changes, crack propagation rate, and anomaly area. ; The parameter values ​​are: , ; sky, sky; , Weighting coefficient .

[0026] Substituting into the calculation, we get: ; Preset trigger threshold If the score of area A1 exceeds the threshold, the system will send the coordinates of high-risk points in area A1 to the foundation key investigation module, triggering beam locking.

[0027] 1.4 Operation process of the key foundation investigation module The key ground-based detailed investigation module controls the ground-based synthetic aperture radar equipment, automatically adjusting the antenna azimuth and elevation angles based on the coordinates of high-risk points to ensure complete radar beam coverage of the A1 risk area. The radar operates in the Ku-band, with a range resolution of 0.75m, an azimuth resolution of 1.2m, a deformation measurement accuracy of 0.1mm, and a sampling frequency set to once every 5 minutes, continuously acquiring the deformation displacement time series of target points.

[0028] 1.5 Progressive Trigger Control Module Reverse Feedback Process After 15 consecutive days of ground-based monitoring, the displacement rate at point A1 stabilized at 0.008 mm / day, and remained below the safety threshold of 0.01 mm / day for 10 consecutive observation periods. The progressive trigger control module determined that the risk point was stabilizing and executed a reverse closed-loop adjustment: controlling the space-based global survey module to return to the regular survey frequency of once every 12 days, eliminating the need to subscribe to additional commercial satellite data; and releasing ground-based radar resources, which could be reassigned to other monitoring points. In this embodiment, through the progressive triggering and closed-loop feedback mechanism, the ground-based radar is only deployed to risk points. Compared to full-time, full-area monitoring, the effective utilization rate of the equipment is increased by approximately 65%, significantly reducing monitoring and operating costs. Example

[0029] This example was applied to the north slope of an open-pit coal mine in Inner Mongolia. The slope is a silty clay slope with a height of 118m, a slope angle of 32°, and a monitoring range of approximately 2.6km. 2 The slope is significantly affected by mining disturbances and rainfall, exhibiting distinct stages of evolution and requiring high accuracy and timeliness in early warning systems.

[0030] 2.1 Operation process of parameter inversion calibration module The parameter inversion calibration module uses space-based INSAR surface deformation data and ground-based radar point displacement data as joint constraints to construct an inversion objective function and solve for the cohesion of the slope soil and rock mass. and internal friction angle The optimal value. The objective function expression is: ; In this embodiment, the weighting coefficient =0.3, =0.7, spatial smoothness regularization coefficient =0.05; The forward model adopts a simplified Bishop limit equilibrium model, the soil and rock constitutive model is based on the Mohr-Coulomb criterion, and the calculation assumptions are a circular arc slip surface and horizontal forces between blocks. The space-based surface data and the ground-based point data adopt a spatial correspondence strategy of "buffer sampling + shape function interpolation": a circular buffer is established with the ground-based radar observation point as the center and a radius of 1.5 times (30m) of the INSAR pixel resolution as the radius. The median of all INSAR deformation values ​​in the buffer is taken as the representative value of the surface observation at that point. The predicted values ​​of the forward model grid nodes are mapped to the coordinates of the ground-based observation points through linear shape function interpolation to achieve spatial matching of multi-scale data.

[0031] The Levenberg-Marquardt nonlinear optimization algorithm is used to solve for the minimum value of the objective function. After iterative convergence, the inversion result is obtained: cohesion. =21.8 kPa, internal friction angle The time-varying characteristics of the parameters are continuously tracked using a 30-day rolling time window, and the calibrated parameters are updated in real time to the safety factor calculation model.

[0032] 2.2 Operation process of multi-stage adaptive early warning module The multi-stage adaptive early warning module calculates the real-time slope safety factor Fs based on the simplified Bishop method, combines it with deformation acceleration to jointly determine the slope evolution stage, and dynamically adjusts the fusion weights and early warning thresholds of the three types of monitoring data. The specific process is as follows: Stability determination: Initial safety factor of the slope before mining disturbance. =1.42, deformation acceleration ≈0, indicating a stable period. The weighting ratio of space-based, air-based, and ground-based data fusion is 7:2:1, with space-based INSAR data dominating, maintaining a large-scale survey.

[0033] At this time, the baseline warning threshold =0.6, adjust parameter =1.0, calculate the dynamic early warning threshold: ; Creep period determination: After entering the rainy season, the groundwater level rises by 2.1m due to rainfall infiltration, and the safety factor decreases to =1.12, the displacement rate continues to increase slowly and the acceleration is positive, indicating the creep stage. The fusion weight ratio automatically switches to 3:5:2, and the frequency of airborne UAV inspections increases from once a month to once every two weeks. At this time, the dynamic warning threshold is: ; The early warning sensitivity is about 17% higher than that during the stable period, enabling the detection of deformation anomalies earlier.

[0034] Accelerated deformation period assessment: Due to the combined effects of continuous heavy rainfall and mining disturbances, the safety factor has decreased to [a certain value]. =0.93, displacement acceleration remains positive, indicating an accelerated deformation period. The fusion weighting ratio automatically switches to 1:3:6, with ground-based radar data dominating, and the sampling frequency increased to once every 2 minutes. At this time, the dynamic warning threshold is: ; The system's early warning sensitivity has been further improved, and it automatically triggers an orange alert, matching the corresponding emergency response plan.

[0035] In this embodiment, the stage switching is driven by a joint event of the safety factor crossing the physical threshold and the deformation acceleration changing. The fusion weight and the early warning threshold are dynamically adjusted according to the slope mechanical state. Compared with the traditional scheme with fixed weights and fixed thresholds, the false alarm rate of the early warning is reduced by about 42%, and the missed alarm rate is reduced by about 35%. Example

[0036] This embodiment is applied to the bank slope of a large reservoir in southern China. The slope is a soil-rock mixed slope with a height of 67m and a slope angle of 35°. It is significantly affected by seasonal rainfall and reservoir water level fluctuations. The risk of slope instability directly threatens the safety of the reservoir waterway and the residential areas along the bank, and there is a strong need for advance prediction and emergency decision support.

[0037] 3.1 Remaining Safety Time Prediction Process The remaining safety time prediction module is based on the real-time geotechnical parameters (cohesion) output by the parameter inversion calibration module. internal friction angle Permeability coefficient Based on the meteorological and rainfall forecast data for the next 7 days (cumulative rainfall of 125 mm and maximum daily rainfall of 32 mm), the changes in groundwater level and slope safety factor caused by rainfall infiltration are extrapolated day by day.

[0038] The deduction process is as follows: calculate the infiltration rate based on the daily rainfall, and calculate the daily groundwater level rise based on the slope hydrogeological parameters. The updated groundwater level is then substituted into the simplified Bishop method to calculate the daily safety factor. until The critical date when the value first drops below 1.0. In this embodiment, the initial safety factor... The groundwater level rose by 0.7m after the first day of rainfall. The groundwater level rose by a cumulative 1.4 meters after the second day of rainfall. After the third day of rainfall, the groundwater level rose by a cumulative 1.9 meters. After four days of continuous rainfall, the groundwater level rose by a cumulative 2.3 meters. It dropped below 1.0 for the first time. Therefore, the critical date is day 4, and the remaining safe time... sky.

[0039] 3.2 Uncertainty Quantification and Confidence Interval Calculation Monte Carlo simulation was used to quantify uncertainty. The simulation was run 1200 times. Input random variables included: the normal distribution error of rainfall forecast (mean 0, standard deviation 15%), the log-normal distribution of permeability coefficient (coefficient of variation 0.3), and the posterior covariance matrix of the inversion parameters. For each simulation, a set of parameters was randomly sampled to perform a forward extrapolation, resulting in 1200 samples of remaining safe time. An empirical probability distribution was constructed. The 2.5% quantile and 97.5% quantile were used as the upper and lower limits of the 95% confidence interval. The final output 95% confidence interval for the remaining safe time was... .

[0040] This confidence interval provides the most optimistic and most pessimistic time boundaries for engineering decisions, significantly improving reliability and engineering reference value compared to single-point prediction results.

[0041] 3.3 Operation process of the visual early warning module The visualization and early warning module maps space-based INSAR deformation fields, airborne real-scene 3D models, ground displacement time-series curves, and rainfall and groundwater level data to a real-scene 3D model base under the 2000 National Geodetic Coordinate System. Four colors—green, yellow, orange, and red—correspond to four warning levels, respectively, and are spatially visualized in a 3D scene, clearly marking the location, level, and evolution trend of risk areas. Based on the current orange warning level and the risk type of rainfall-induced landslide, the system automatically matches corresponding response measures from the contingency plan database, including increased monitoring frequency, slope drainage and reinforcement, and personnel evacuation route planning. The system then pushes the early warning information and response plan to the terminal devices of the reservoir management department and the village committees along the river. In this embodiment, the system's output of the remaining safety time and response plan provides ample preparation window for emergency response, effectively reducing the risk of disaster losses.

[0042] In summary, this system is stably adaptable to three typical slope scenarios: high mountain and canyon highways, open-pit mines, and water conservancy reservoirs, meeting the monitoring and early warning needs of different geological conditions and risk levels. By improving the efficiency of monitoring resource utilization through a multi-scale progressive and collaborative scheduling mechanism, ensuring the accuracy of early warnings through a dynamic model constrained by physical mechanisms, and supporting proactive emergency decision-making through parameter inversion and time prediction, it can provide a complete technical solution for the safety control of various slope engineering projects.

[0043] While the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the invention. Any variations and modifications can be made by those skilled in the art without departing from the spirit and scope of the invention. Therefore, any modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention, without departing from the scope of the invention, fall within the protection scope defined by the claims of the present invention.

Claims

1. A space-air-ground integrated intelligent early warning system for slope monitoring, characterized in that, The early warning system includes: Space-based global survey module: used to acquire spaceborne synthetic aperture radar remote sensing data of the entire slope area and process it to obtain the global deformation rate field; The progressive trigger control module is used to identify deformation anomaly clusters based on the received global deformation rate field, trigger the launch of the space-based key detailed survey module, and constrain its inspection range with the spatial boundary of the anomaly area; it also triggers the launch of the ground-based key detailed survey module based on the risk results returned by the space-based key detailed survey module and constrains its monitoring targets with the coordinates of the risk points; and it is also used to adjust the monitoring frequency of the space-based global survey module and the resource allocation status of the ground-based key detailed survey module in reverse based on the displacement data returned by the ground-based key detailed survey module. The airborne key detailed investigation module is used to perform refined inspections of designated areas by UAVs, obtain data on regional elevation changes and crack development characteristics, calculate and generate risk urgency results, and send them back to the progressive trigger control module. The foundation key investigation module is used to perform high-precision continuous deformation monitoring on designated risk points to obtain millimeter-level displacement time series data. Multi-stage adaptive early warning module: used to divide the evolution stages based on the physical and mechanical state of the slope, dynamically adjust the fusion weight and early warning threshold of the three types of monitoring data, and output graded early warning results; Parameter inversion and calibration module: used to combine space-based surface deformation data and ground-based point displacement data to invert and calibrate slope geotechnical parameters; The remaining safety time prediction module is used to extrapolate the temporal changes of the slope safety factor based on calibrated mechanical parameters and rainfall forecast data, and output the remaining safety time and the corresponding confidence interval. Visualized early warning module: This module integrates multi-source monitoring data and early warning results for three-dimensional visualization and pushes corresponding early warning response plans.

2. The air-space-ground integrated intelligent early warning system for slope monitoring according to claim 1, characterized in that, The space-based global survey module is specifically used to: acquire the original image data of the spaceborne synthetic aperture radar, and sequentially perform image registration, interferometric phase generation, phase unwrapping, and geocoding to obtain the global deformation rate field in the geodetic coordinate system. Each spatial pixel of the deformation rate field corresponds to the annual average deformation rate of the corresponding location of the slope.

3. The air-space-ground integrated intelligent early warning system for slope monitoring according to claim 2, characterized in that, The airborne key investigation module is specifically used for: controlling a drone equipped with a lidar and a high-resolution optical camera to perform autonomous flight data collection within a designated inspection area, obtaining high-density lidar point cloud data and high-resolution optical images; constructing a digital elevation model from the lidar point cloud data, and obtaining the spatial distribution of regional elevation changes through multi-period elevation model difference; performing pixel-level crack semantic segmentation on the optical images to obtain the crack location, aperture, and propagation rate; and calculating a risk urgency score by combining elevation changes, crack propagation rate, and anomaly area area, using the following formula: ; in, Rate the urgency of the risk. The maximum elevation change within the monitoring area The critical reference value for elevation change is pre-calibrated based on the safety level of the slope engineering. To monitor the maximum crack propagation rate within the area, The critical reference value for crack propagation rate is pre-calibrated based on the slope soil and rock type. This represents the total area of ​​the deformation anomaly region. This serves as a critical reference value for the area of ​​the abnormal region, pre-calibrated based on the scale of the slope engineering project. These are the normalized weighting coefficients corresponding to the three indicators, and the sum of the three is 1.

4. The air-space-ground integrated intelligent early warning system for slope monitoring according to claim 3, characterized in that, The specific function of the ground-based key detailed investigation module is to: control the ground-based synthetic aperture radar equipment, automatically adjust the azimuth and elevation angles of the antenna according to the coordinates of the target point, so that the radar beam completely covers the target risk area, and calculate the continuous deformation displacement time series of the target point through synthetic aperture imaging and differential interferometry, with a sampling frequency of not less than once every 5 minutes.

5. The air-space-ground integrated intelligent early warning system for slope monitoring according to claim 4, characterized in that, The progressive trigger control module specifically executes forward progressive trigger control and reverse closed-loop feedback control, including: The deformation rate field output by the space-based global survey module is received, and the spatial gradient anomaly detection algorithm is used to identify deformation anomaly clusters: the spatial gradient amplitude of the deformation rate of each pixel is calculated, and pixels with gradient amplitudes exceeding the adaptive threshold are used as seed points. A continuous deformation anomaly cluster is generated through a density clustering algorithm, and a list of key areas for detailed investigation is generated according to risk priority. The list is sent to the space-based key area for detailed investigation module, and the spatial boundary of the anomaly area is used as the spatial constraint range for UAV inspection. The formula for calculating the spatial gradient magnitude of the deformation rate is: ； in, For the deformation rate along gradient components in the direction, For the deformation rate along gradient components in the direction, This represents the gradient magnitude corresponding to a single pixel. The formula for calculating the adaptive threshold is: ； in, The mean of the global deformation rate gradient magnitude. The standard deviation of the global deformation rate gradient magnitude; The system receives the risk urgency score output by the airborne key investigation module. When the score exceeds the preset trigger threshold, the coordinates of the high-risk points are sent to the ground-based key investigation module, and the point coordinates are used as the target constraint for radar beam locking. Reverse closed-loop feedback control: The spatial boundary coordinates of the anomaly area and the coordinates of the high-risk points respectively constitute a direct constraint of the spatial search range of the next monitoring scale on the output of the previous monitoring scale, realizing multi-scale progressive coordination with progressively narrowing coverage area and progressively improving spatial resolution.

6. The intelligent early warning system for slope monitoring based on air-space-ground coordination according to claim 5, characterized in that, The progressive trigger control module performs reverse closed-loop feedback control, specifically including: The system receives the displacement time series output by the ground-based key detailed survey module. When the displacement rate is lower than the safety threshold for no less than 3 consecutive observation periods, it determines that the risk point is stabilizing, controls the space-based full-domain survey module to reduce to the frequency of routine survey monitoring, and releases ground-based radar resources. When the displacement rate shows a continuous accelerating trend, the risk is judged to be aggravated. The space-based global survey module is controlled to increase the monitoring frequency, increase the satellite data ordering cycle, and maintain the high-frequency acquisition state of the ground-based radar. The adjustment targets of the reverse closed-loop feedback are the operating resources and acquisition rhythm of the monitoring system itself, including shortening or extending the ordering cycle of space-based satellite data, allocating or releasing ground-based radar resources, and increasing or decreasing the frequency of UAV inspections.

7. The air-space-ground integrated intelligent early warning system for slope monitoring according to claim 6, characterized in that, The multi-stage adaptive early warning module specifically performs multi-stage adaptive early warning based on physical mechanism constraints, including: (1) Calculation of real-time safety factor of slope based on simplified limit equilibrium theory By combining deformation acceleration to jointly determine the evolution stage of the slope, the slope evolution process is divided into four stages: stable period, creep period, accelerated deformation period, and instability period. (2) According to different evolution stages, the weight ratio of space-based monitoring data, air-based monitoring data and ground-based monitoring data in the early warning fusion model is dynamically adjusted. The stage switching is triggered by the joint event of the safety factor crossing the physical threshold and the change of deformation acceleration. The timing of the fusion weight switching of each stage is rooted in the change of the physical and mechanical state of the slope. (3) A dynamic early warning threshold mechanism is adopted, in which the early warning threshold is dynamically adjusted according to the real-time safety factor. The calculation formula is as follows: ； in, For dynamic early warning thresholds, The baseline warning threshold is preset according to the project's safety level. The real-time safety factor for the slope. To adjust the parameters, the value range is 0.5-1.5, and it should be adjusted according to the type of slope rock and soil. The lower the safety factor, the smaller the dynamic warning threshold and the higher the warning sensitivity.

8. The air-space-ground integrated intelligent early warning system for slope monitoring according to claim 7, characterized in that, The parameter inversion calibration module is specifically used to: construct an inversion objective function using space-based surface deformation observation data and ground-based point displacement observation data as joint constraints, and solve for the cohesion of the slope soil and rock mass. and internal friction angle The optimal value is obtained; the time-varying characteristics of the parameters are continuously tracked using a rolling time window, and the calibrated parameters are updated in real time to the safety factor calculation model; the expression of the inversion objective function is: ； in, Let be the objective function value, and the objective is to minimize it. These are the area deformation values ​​obtained from space-based INSAR observations. These are the displacement values ​​of points obtained from ground-based radar observations. Based on cohesion and internal friction angle Deformation prediction values ​​calculated using a forward model These are the weighting coefficients for the space-based observation term. The weighting coefficients for the foundation observation terms. These are the spatial smoothing regularization constraint coefficients, used to suppress ill-conditioned oscillations in the inversion results. This is the parameter space gradient regularization term, used to ensure the smoothness of the spatial distribution of geotechnical parameters; The spatial correspondence strategy between space-based surface data and ground-based point data adopts buffer sampling and shape function interpolation: a circular buffer is established with the ground-based observation point as the center and a radius of 1.5 times the INSAR pixel resolution. The median of all INSAR deformation values ​​in the buffer is taken as the representative value of the surface observation at that point. The predicted values ​​of the forward model grid nodes are mapped to the coordinates of the ground-based observation point through shape function interpolation.

9. A space-air-ground integrated intelligent early warning system for slope monitoring according to claim 8, characterized in that, The remaining safety time prediction module is specifically used for: based on calibrated geotechnical parameters and combined with future rainfall forecast data, to extrapolate the temporal changes in groundwater level and slope safety factor caused by rainfall infiltration on a daily basis, until the critical date when the safety factor first drops below 1.0, and to calculate the time difference between the current date and the critical date as the remaining safety time; simultaneously, it uses the Monte Carlo simulation method to quantify uncertainty and outputs the 95% confidence interval of the remaining safety time: using the rainfall forecast probability distribution, the permeability coefficient probability distribution, and the posterior covariance of the inversion parameters as inputs, it performs no less than 1000 random forward extrapolations to obtain the empirical probability distribution of the remaining safety time, and takes the 2.5% quantile and the 97.5% quantile as the upper and lower limits of the 95% confidence interval, and outputs the result in the following format: , This is the lower limit of the 95% confidence interval for the remaining safe time. This is the upper limit of the 95% confidence interval for the remaining safe time.

10. The air-space-ground integrated intelligent early warning system for slope monitoring according to claim 9, characterized in that, The visualization early warning module is specifically used to: map the space-based deformation field, the air-based real-scene 3D model, the ground displacement curve, and the rainfall and groundwater level data to the real-scene 3D model base under the 2000 National Geodetic Coordinate System, and display the spatial visualization using four colors—green, yellow, orange, and red—corresponding to the four levels of early warning. It automatically matches the response measures in the contingency plan database according to the early warning level and risk type, and pushes the early warning information and response plan to the designated terminal device.

Images