Landslide surface and internal collaborative deformation reconstruction and evaluation method based on airborne LiDAR and DFOS registration

Abstract

The application discloses a kind of based on airborne LiDAR and DFOS registration's landslide surface and inside collaborative deformation reconstruction and evaluation method, belong to geological disaster monitoring technical field.Firstly, through three-dimensional coordinate conversion model, the deep strain monitoring path of DFOS is accurately mapped to the high-precision surface point cloud model obtained by airborne LiDAR;Construct deformation transfer function based on geomechanical model, use DFOS continuous strain data to invert surface deformation field initial value, and use the absolute surface deformation observation value of LiDAR discrete iteration correction, optimize inversion parameter;Finally, using the high time resolution data of DFOS, the time and space interpolation is completed to the observation gap of LiDAR, generates landslide from surface to inside, continuous dynamic collaborative deformation field.The application realizes the "transparent" three-dimensional monitoring of landslide body, significantly improves the accuracy and early warning reliability of landslide deformation process analysis.

Description

Technical Field

[0001] This invention belongs to the field of geological disaster monitoring technology, and in particular relates to a method for collaborative three-dimensional dynamic reconstruction and quantitative stability evaluation of landslide internal strain and surface deformation by fusing two heterogeneous monitoring data, namely airborne lidar (LiDAR) and distributed fiber optic sensing (DFOS). Background Technology

[0002] Accurate monitoring and early warning of landslide disasters are crucial for disaster prevention and mitigation. Currently, high-precision surface deformation monitoring mainly relies on technologies such as airborne LiDAR and UAV photogrammetry, which can quickly acquire large-scale, high-precision three-dimensional displacement fields of the surface. However, the observation objects are limited to the surface and are subject to limitations imposed by flight operation cycles and weather conditions, resulting in discontinuous monitoring data and gaps in data collection. Distributed fiber optic sensing (DFOS, such as Brillouin Optical Time Domain Analysis (BOTDA) and Distributed Acoustic Sensing (DAS)) technology can continuously, over long distances, and in a distributed manner monitor strain, temperature, and other information within soil and rock masses, directly sensing deep deformation initiation signals. It offers good spatial continuity and high temporal resolution, but its measurement results are essentially one-dimensional and linear, with limited absolute spatial positioning accuracy, making it difficult to intuitively reflect the three-dimensional deformation field. Existing technologies typically analyze LiDAR and DFOS data independently, lacking effective synergy. Because of fundamental differences in spatial reference (absolute surface coordinate system and relative path of underground optical fiber), temporal scale (discrete instantaneous sampling and continuous sequence acquisition), and observed physical quantities (three-dimensional displacement and one-dimensional strain), it is impossible to establish a quantitative causal relationship and overall mechanical mechanism analysis between deep landslide strain evolution and surface deformation response within a unified spatiotemporal framework. This fragmented "surface-interior" data severely restricts the overall understanding of the "inside-out" failure mechanism of landslides, affecting the accuracy and timeliness of early warning judgments.

[0003] Therefore, there is an urgent need for an innovative method that can deeply integrate LiDAR and DFOS heterogeneous monitoring data to achieve integrated perception, three-dimensional dynamic collaborative reconstruction, and accurate analysis and evaluation of landslide "surface deformation-internal strain". Summary of the Invention

[0004] Purpose of the invention: The purpose of this invention is to provide a method for landslide surface and internal strain co-reconstruction and evaluation based on airborne LiDAR and DFOS registration. This method can deeply integrate LiDAR and DFOS heterogeneous monitoring data to achieve integrated perception, three-dimensional dynamic co-reconstruction, and accurate analysis and evaluation of landslide surface deformation and internal strain.

[0005] Technical solution: The present invention provides a method for landslide surface-interior co-deformation reconstruction and evaluation based on airborne LiDAR and DFOS registration, comprising the following steps:

[0006] S1. Data Synchronization Acquisition and Preprocessing: Simultaneously acquire airborne lidar point cloud data and distributed fiber optic sensing deep strain monitoring data of the target landslide area, and perform coordinate calculation, denoising and classification on the airborne lidar point cloud data to generate a three-dimensional surface model, and perform temperature compensation and denoising processing on the deep strain monitoring data to construct a strain-space-time dataset.

[0007] S2. Spatial registration of heterogeneous monitoring data: Establish a unified spatial benchmark for the landslide body, and accurately map the distributed fiber optic sensing DFOS monitoring path to the surface three-dimensional model generated by the airborne lidar LiDAR through a three-dimensional coordinate transformation model to obtain the mapped DFOS strain data and realize the spatial location association between underground strain points and surface three-dimensional points.

[0008] S3. Coordinated Inversion and Correction of Surface and Interior Deformation: Based on the landslide geomechanical model, a deformation transfer function is constructed. The mapped DFOS strain data is used as input. The surface deformation field is inverted through the deformation transfer function as the first deformation estimation field. The measured surface deformation field extracted by LiDAR at the same time is used. The parameters of the deformation transfer function are corrected through optimization algorithm to obtain the corrected coordinated deformation field.

[0009] S4. Spatiotemporal fusion and dynamic reconstruction of deformation field: Using the corrected collaborative deformation field as the control node, the high temporal resolution strain data of DFOS is used for temporal interpolation to reconstruct the dynamic deformation field of the landslide from deep to the surface during the LiDAR data gap period.

[0010] S5. Quantitative evaluation and early warning of landslide stability based on collaborative deformation field: Based on the reconstructed full-time dynamic deformation field, the deformation rate time series of key parts of the landslide is extracted; according to the deformation rate time series, the real-time stability state and early warning level of the landslide are calculated using a preset early warning model, and the evaluation and early warning results are output.

[0011] Furthermore, in step S1, the process of calculating coordinates, denoising, and classifying the airborne lidar point cloud data to generate a 3D surface model specifically involves: Coordinate calculation based on the sensor's geometric calibration model, by fusing global navigation satellite system, inertial measurement unit trajectory data, and raw laser ranging data, to calculate the absolute coordinates of each laser footpoint. The calculation model is as follows:

[0012]

[0013] in, This represents the three-dimensional coordinate column vector of the laser footprint in the ground coordinate system. The column vector representing the phase center coordinates of the GNSS antenna at time t; Represents the rotation matrix from the carrier coordinate system to the ground coordinate system, which varies with time t; This represents the column vector of arm offset from the GNSS antenna phase center to the origin of the laser scanner coordinate system. This represents the fixed-mount rotation matrix from the laser scanner coordinate system to the carrier coordinate system.

[0014] This indicates that in the laser scanner coordinate system, the distance measurement value... , scanning pitch angle and azimuth A defined column vector of laser footpoint coordinates;

[0015] Point cloud filtering employs a statistical outlier removal algorithm. For any point in the point cloud... Calculate the average distance from its k nearest neighbors. and remove those that meet the requirements. The point, among which and These are the mean and standard deviation of the average distance across the entire point cloud, respectively. This is the threshold coefficient.

[0016] Furthermore, in step S1, the temperature compensation and noise reduction processing of the deep strain monitoring data specifically involves: temperature compensation for Brillouin scattering-based sensing data, using the following formula to separate the strain change. :

[0017]

[0018] in, The change in Brillouin frequency shift of the sensing optical fiber coupled to the soil mass. The Brillouin frequency shift variation is the value of a reference fiber arranged along the same path for temperature monitoring. The strain coefficient of the optical fiber;

[0019] Strain data noise reduction employs a sliding window averaging filter method, which provides smooth strain values ​​at position s on the optical fiber over time t. Calculated by the following formula:

[0020]

[0021] in, Let be the original strain value at position s at time t, and N be a positive integer representing the half-width of the preset filter window. is the spatial sampling interval of the DFOS data, and n is the index variable that takes all integer values ​​from -N to N in sequence during the summation process.

[0022] Furthermore, step S2 specifically involves: the three-dimensional coordinate transformation model is a three-dimensional similarity transformation model, which achieves coordinate transformation by solving the following equations:

[0023]

[0024] in, This is the column vector of coordinates of the DFOS monitoring points in the LiDAR coordinate system. This is the original coordinate column vector of DFOS. For rotation matrix, It is a translation vector. The scaling factor is used; the model parameters are obtained by solving the least squares method using two sets of coordinate values ​​from three ground control points.

[0025] Furthermore, in step S3, the deformation transfer function describes the physical model of the mapping relationship between deep distributed strain and surface displacement field, and its expression is:

[0026]

[0027] in, Representing surface points The deformation vector at time t, For transfer function operators, Indicates underground point The strain vector, Geological model parameters to characterize the soil and rock mechanical properties of landslides. This is the vector of function parameters to be optimized.

[0028] Furthermore, in step S3, the optimization algorithm uses minimizing the difference between the LiDAR measured deformation field and the deformation field retrieved from the deformation transfer function as the criterion for a monitoring period. Its solution is the corrected optimal parameter vector. The process can be expressed as the following least squares problem:

[0029]

[0030] in, This represents the number of surface deformation points included in the comparison. Let i be the deformation vector calculated from LiDAR data for the i-th point. The deformation vector at the corresponding point is calculated based on the DFOS strain change and the deformation transfer function. This is the vector of function parameters to be optimized.

[0031] Furthermore, in step S4, the time-series interpolation is performed at the previous LiDAR observation time. Based on the deformation field, the deformation transfer function corrected by the optimization algorithm is used. Using DFOS continuous strain data, calculate any subsequent time step. The process of deformation field is calculated using the following formula:

[0032]

[0033] in, This is the corrected optimal parameter vector. From arrive The cumulative strain change over time is measured by DFOS.

[0034] Furthermore, step S5 specifically involves: first, extracting the deformation rate time series of key landslide monitoring points from the dynamic deformation field throughout the entire time period as an evaluation index; then, comparing and analyzing the deformation rate time series with the preset warning threshold or warning model to determine the stability state of the landslide and the warning level; finally, generating and outputting the evaluation results containing the warning level and key deformation parameters.

[0035] Furthermore, the comparison and analysis of the deformation rate time series with the preset early warning threshold or early warning model specifically involves: using the tangent angle method for landslide early warning, and calculating the key points of the landslide in time based on the displacement-time curve corresponding to the deformation rate time series. displacement tangent angle Displacement tangent angle The calculation formula is:

[0036]

[0037] in, Let be the cumulative displacement over time t. For the selected reference time The displacement; setting the tangent angle threshold for near-slip warning. When the conditions are met When the time is right, it is determined that the skid is about to be cleared and the corresponding level of warning is triggered.

[0038] The present invention also discloses a computer device, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method of the present invention.

[0039] Beneficial effects: Compared with the prior art, the present invention has the following significant advantages:

[0040] 1. This invention realizes a fundamental shift in landslide monitoring from "separation" to "collaboration": through an innovative spatial mapping and model correction process, it achieves precise correlation and physical fusion of surface LiDAR point cloud and underground DFOS strain data in a unified three-dimensional spatiotemporal grid for the first time, overcoming the technical bottleneck of heterogeneous data collaborative analysis and enabling "transparent" monitoring of landslide bodies.

[0041] 2. Significantly improves the accuracy and reliability of deformation field reconstruction: The "deformation transfer function" model proposed in this invention establishes the physical relationship between deep strain and surface displacement, and uses LiDAR absolute observations to constrain and correct it, effectively overcoming the uncertainty of simply relying on model inversion or a single data source, and obtaining a cooperative deformation field that is closer to the real physical process.

[0042] 3. Overcame the challenge of continuous acquisition of high spatiotemporal resolution deformation fields: Creatively utilized the high temporal resolution characteristics of DFOS to compensate for the shortcomings of low-frequency sampling of airborne LiDAR, generating a landslide full-time deformation field animation sequence with both high spatial resolution (provided by LiDAR) and high temporal resolution (provided by DFOS), realizing "blind-spot-free" dynamic perception of the entire landslide evolution process.

[0043] 4. A complete technical closed loop from monitoring to early warning has been formed: It not only provides data fusion and reconstruction methods, but also constructs a quantitative stability evaluation and early warning process based on the fusion results, so that the cutting-edge monitoring technology results can directly and efficiently serve engineering safety judgment and disaster prevention and emergency decision-making, which is highly practical. Attached Figure Description

[0044] Figure 1 This is an overall flowchart of the method of the present invention.

[0045] Figure 2 This is a schematic diagram of heterogeneous data spatial registration and mapping in this invention.

[0046] Figure 3 This is a schematic diagram illustrating the principle of deformation transfer function inversion and LiDAR data correction in this invention.

[0047] Figure 4 This is a schematic diagram illustrating the spatiotemporal fusion completion of the invention based on DFOS high-frequency data.

[0048] Figure 5 The diagram shows a comparison of the accuracy of the surface deformation field reconstructed by different methods in this invention; where (a) is a diagram of the single LIDAR method (linear interpolation), (b) is a diagram of the single DFOS method (empirical formula inversion), and (c) is a diagram of the collaborative fusion method of this invention. Detailed Implementation

[0049] The technical solution of the present invention will be further described below with reference to the accompanying drawings.

[0050] like Figure 1As shown, the landslide “surface-interior” collaborative deformation reconstruction and evaluation method based on spatiotemporal registration of airborne LiDAR and DFOS proposed in this invention includes data synchronous acquisition and preprocessing, spatial registration of heterogeneous monitoring data, “surface-interior” collaborative deformation inversion and correction, spatiotemporal fusion and dynamic reconstruction of deformation field, and quantitative evaluation and early warning of landslide stability based on collaborative deformation field.

[0051] Data synchronous acquisition and preprocessing: Preprocessing of airborne LiDAR point cloud data includes at least coordinate calculation and point cloud filtering.

[0052] The coordinate calculation is based on the sensor geometric calibration model. By fusing trajectory data from the Global Navigation Satellite System / Inertial Measurement Unit with the original laser ranging data, the absolute coordinates of each laser footpoint are calculated. The calculation model is as follows:

[0053]

[0054] in, This represents the three-dimensional coordinate column vector of the laser footprint in the ground coordinate system. The column vector representing the phase center coordinates of the GNSS antenna at time t; Represents the rotation matrix from the carrier coordinate system to the ground coordinate system, which varies with time t; This represents the column vector of arm offset from the GNSS antenna phase center to the origin of the laser scanner coordinate system. This represents the fixed-mount rotation matrix from the laser scanner coordinate system to the carrier coordinate system. This indicates that in the laser scanner coordinate system, the distance measurement value... , scanning pitch angle and azimuth A defined column vector of laser footpoint coordinates;

[0055] Furthermore, the point cloud filtering employs a statistical outlier removal algorithm, which is used for any point in the point cloud. Calculate the average distance from its k nearest neighbors. and remove those that meet the requirements. The point, among which and These are the mean and standard deviation of the average distance across the entire point cloud, respectively. This is the threshold coefficient.

[0056] Furthermore, the preprocessing of distributed fiber optic inductive strain data includes at least temperature compensation and strain data noise reduction;

[0057] Temperature compensation, specifically for Brillouin scattering-based sensing data, separates strain changes using the following formula.

[0058]

[0059] in, The change in Brillouin frequency shift of the sensing optical fiber coupled to the soil mass. The Brillouin frequency shift variation is the value of a reference fiber arranged along the same path for temperature monitoring. The strain coefficient of the optical fiber;

[0060] Strain data noise reduction employs a sliding window averaging filter method, which provides smooth strain values ​​at position s on the optical fiber over time t. Calculated by the following formula:

[0061]

[0062] in, Let be the original strain value at position s at time t, and N be a positive integer representing the half-width of the preset filter window. is the spatial sampling interval of the DFOS data, and n is the index variable that takes all integer values ​​from -N to N in sequence during the summation process.

[0063] Figure 2 A schematic diagram illustrating the spatial registration and mapping of heterogeneous data shows the actual deployment path of the DFOS sensing fiber within the landslide body (represented by a thick solid line with uniformly distributed small dots representing sensing points) and the surface point cloud model acquired by LiDAR (represented by a dense point cloud and surface outline). To achieve spatial benchmark unification, a precise coordinate transformation relationship needs to be established between the fiber path and the surface point cloud. The DFOS "line" is precisely embedded under the LiDAR "surface". At least three (preferably six in this embodiment) ground control points are selected at locations along the fiber path that have obvious features and can be clearly identified in the LiDAR point cloud, such as fiber optic inlets / outlets and maintenance wells. Figure 2 As shown, the corresponding underground locations of the control points are marked with pentagrams, denoted as CP1, CP2, and CP3; their corresponding locations on the surface are also marked with pentagrams, denoted as CP1′, CP2′, and CP3′. The precise geodetic coordinates of the control points are obtained using GNSS static surveying methods. And extract the corresponding coordinates manually or automatically from the LiDAR point cloud. .

[0064] Furthermore, the 3D coordinate transformation model is transformed into a 3D similarity transformation model, which achieves coordinate transformation by solving the following equations:

[0065]

[0066] in, This is the column vector of coordinates of the DFOS monitoring points in the LiDAR coordinate system. This is the original coordinate column vector of DFOS. For rotation matrix, It is a translation vector. The scaling factor is used; the model parameters are obtained by solving using the least squares method with two sets of coordinate values ​​from at least three ground control points.

[0067] The “surface-interior” deformation co-inversion and correction (such as) Figure 3 As shown), the deformation transfer function is a physical model describing the mapping relationship between deep distributed strain and the surface displacement field, and its expression is:

[0068]

[0069] in, Representing surface points The deformation vector at time t, For transfer function operators, Indicates underground point The strain vector, Geological model parameters to characterize the soil and rock mechanical properties of landslides. This is a parameter vector specific to the function to be optimized.

[0070] Specifically, the optimization algorithm uses minimizing the difference between the LiDAR measured deformation field and the deformation field inverted by the deformation transfer function as the criterion for a monitoring period. Its solution for optimal parameters The process can be expressed as the following least squares problem:

[0071]

[0072] in, This represents the number of surface deformation points included in the comparison. Let i be the deformation vector calculated from LiDAR data for the i-th point. The deformation vector at the corresponding point is calculated based on the DFOS strain change and the deformation transfer function.

[0073] The spatiotemporal fusion and dynamic reconstruction of the deformation field (such as...) Figure 4 As shown), the time series interpolation is based on the previous LiDAR observation time. Based on the deformation field, the deformation transfer function corrected by the optimization algorithm is used. Using DFOS continuous strain data, calculate any subsequent time step. ( The process of deformation field, and its calculation formula is:

[0074]

[0075] in, This is the corrected optimal parameter vector. From arrive The cumulative strain change measured by DFOS at any given time.

[0076] Quantitative evaluation and early warning of landslide stability based on collaborative deformation field: The deformation rate time series of key monitoring points of landslide is extracted from the dynamic deformation field of the whole time period as the evaluation index; further, the deformation rate time series is compared and analyzed with the preset early warning threshold or early warning model to determine the stability state of the landslide and determine the early warning level; finally, the evaluation results containing the early warning level and key deformation parameters are generated and output.

[0077] Furthermore, the deformation rate time series is compared and analyzed with a preset early warning model, specifically using the tangent angle method for landslide early warning; the method includes: calculating the displacement tangent angle of the key landslide point at time t based on the displacement-time curve corresponding to the deformation rate time series. The tangent angle The calculation formula is:

[0078]

[0079] in, Let be the cumulative displacement over time t. For the selected reference time The displacement; setting the tangent angle threshold for near-slip warning. When the conditions are met When the time is right, it is determined that the skid is about to be cleared and the corresponding level of warning is triggered.

[0080] Example

[0081] 1. Experimental Site and Geological Overview

[0082] To verify the feasibility and superiority of the method of this invention, a typical landslide in a county in Sichuan Province was selected as the experimental site. The landslide is tongue-shaped, approximately 320m long and 180m wide, with an average thickness of about 15m and a total volume of about 860,000 cubic meters. The strata in the landslide area are mainly interbedded sandstone and mudstone of the Upper Jurassic Penglaizhen Formation, overlain by Quaternary residual colluvial silty clay. Three obvious surface fissures have developed on the landslide body, with a tensile fracture zone at the rear edge approximately 8-12m wide and a displacement height of about 1.5m. It is a typical traction-type soil landslide in the creep deformation stage.

[0083] 2. Monitoring System Deployment and Data Acquisition

[0084] A DJI Matrice 350RTK drone equipped with a Zenmuse L2 LiDAR system was used for three aerial surveys on March 15, 2024, June 20, 2024, and September 25, 2024. Flight parameters were: relative altitude 120m, flight speed 8m / s, lateral overlap of the flight path 60%, laser pulse frequency 240kHz, and scan angle ±30°. The acquired point cloud density averaged 45 points per square meter, with a planar accuracy better than 0.05m and an elevation accuracy better than 0.03m. The point cloud data was preprocessed to generate a 0.2m resolution digital elevation model (DEM).

[0085] One monitoring borehole, 32m deep, was drilled along the main sliding direction in the middle of the landslide, penetrating the potential sliding surface (buried at a depth of approximately 18-22m). An 8mm armored sensing optical cable was laid inside the borehole, and the entire borehole was backfilled with cement mortar to ensure good coupling between the fiber and the strata. Three surface sensing optical cables, approximately 560m long, were laid longitudinally along the landslide on the ground, using a shallow trench burial method at a depth of 0.3m. The DFOS demodulator used a Brillouin optical time domain analyzer (BOTDA), with a spatial sampling interval of 0.2m, a temporal sampling interval of 1 hour, a strain measurement accuracy of ±5με, and a temperature measurement accuracy of ±0.5℃. The monitoring period was from March 15, 2024 to December 31, 2024, with continuous acquisition of strain data.

[0086] Six ground control points were established in and around the landslide area and its surrounding stable zone. Three of these points were located at the outcrop along the fiber optic path, and three were located on stable bedrock. Static measurements were performed using a Huace X16 GNSS receiver, with each control point observed for at least 90 minutes. The calculated plane coordinate accuracy was better than ±2 cm, and the elevation accuracy was better than ±3 cm. The control point coordinates served as the reference for spatial registration.

[0087] 3. Spatial registration results

[0088] Substituting the original DFOS coordinates and LiDAR coordinates of the six control points into the 3D similarity transformation model, the transformation parameters were solved using the least squares method. The results are: scale factor λ = 1.00012, rotation angles Ω = 0.023°, Φ = -0.017°, K = 0.031°, and translation vector T = The residual statistics of the control points after registration are as follows: the root mean square error (RMSE) is 0.034m and the maximum residual is 0.048m, which meets the accuracy requirements for landslide deformation monitoring.

[0089] 4. Construction and correction of deformation transfer function

[0090] Based on the landslide geological profile, the landslide body is simplified into a three-layer medium model (overburden, slip zone, and slip bed). The deformation transfer function adopts a shear transfer model based on elasticity theory, specifically expressed as follows:

[0091]

[0092] in, This represents horizontal displacement of the Earth's surface. =22° is the inclination angle of the slip surface (determined from borehole data). =0.08 is the attenuation coefficient (obtained through inversion optimization), and H=32m is the drilling depth.

[0093] The correction period is from March 15, 2024 to June 20, 2024 (ΔT ≈ 97 days). The strain change Δε measured by DFOS during this period is substituted into the transfer function to calculate the surface deformation field D. model Compared with the LiDAR measured deformation field D during the same period lidar A comparison was performed. The least squares optimization algorithm was used, with the attenuation coefficient γ as the optimization parameter, and the objective function was:

[0094]

[0095] After eight iterations, the calculations converged, with the optimal attenuation coefficient γ*=0.073. The root mean square error between the deformation field inverted by the corrected model and the LiDAR measured deformation field decreased from 0.032 m before correction to 0.015 m, representing a 53% improvement in accuracy.

[0096] 5. Spatiotemporal fusion reconstruction results

[0097] Using the deformation field observed by LiDAR on June 20, 2024 at time t1 as a baseline, the daily deformation field from June 20, 2024 to September 25, 2024 (t2) was reconstructed using the corrected transfer function F* and DFOS continuous strain data. A total of 97 daily deformation fields (with a time interval of 1 day) were generated, forming a continuous deformation evolution sequence of the landslide from June to September 2024.

[0098] Verification: The LiDAR measured deformation field on September 25, 2024, was taken as the true value and compared with the deformation field reconstructed from DFOS data on the same day. The correlation coefficient between the two in the main sliding direction (longitudinal) of the landslide reached 0.92, and the mean absolute error was 0.009m, proving that the reconstruction result has high reliability.

[0099] 6. Landslide stability assessment and early warning effect

[0100] The cumulative displacement-time curves of key points at the rear edge of the landslide (coordinates: X=512034m, Y=3452678m) were extracted from the reconstructed deformation field sequence. The deformation rate was then calculated. The results showed that the average rate was 1.2 mm / d from June to July 2024; from August to early September, the rate gradually increased to 2.8 mm / d; after September 10, the rate accelerated significantly to 5.6 mm / d, showing accelerated deformation characteristics.

[0101] Take t r The acceleration starting point is August 1, 2024. The tangent angle α(t) is calculated. When α(t) = 45°, the actual time is September 18, 2024. According to the warning rules, an orange warning is triggered when α(t) ≥ 45°. The warning information was issued on September 18, 2024.

[0102] On September 22, 2024, a new tensile crack appeared at the rear edge of the landslide, with a width of approximately 8 cm and a downward displacement of about 12 cm, indicating that the landslide had entered an accelerated deformation stage. The warning time was 4 days earlier than the appearance of macroscopic deformation, verifying the effectiveness of the method of the present invention and the timeliness of the warning.

[0103] To verify the superiority of the method of the present invention, two sets of comparative experiments were set up, as shown in Table 1:

[0104] Table 1

[0105]

[0106] Experimental results show that the method of the present invention is significantly superior to single monitoring methods (such as deformation field reconstruction accuracy, spatiotemporal continuity, and timely early warning) in terms of deformation field reconstruction accuracy, spatiotemporal continuity, and early warning timeliness. Figure 5 (As shown).

[0107] Figure 5 The visualization comparison results of the surface deformation field reconstructed by the above three methods are presented. Among them:

[0108] Figure 5 (a) Single LiDAR method (linear interpolation): It can only obtain deformation cloud maps at three discrete time points (March, June, and September). The cloud maps are separated by monitoring data gaps (marked by dashed lines in the figure). Although the accuracy of LiDAR measured points is high (RMSE=0.028m), the time resolution is low (quarterly level), and it cannot capture the continuous evolution process of landslide deformation.

[0109] Figure 5 (b) Single DFOS method (empirical formula inversion): Although it can achieve continuous monitoring (time resolution of 1 hour), due to the lack of accurate spatial registration and physical model correction, the inverted deformation contour map has obvious spatial position offset (the position of the red high deformation area is different from the original). Figure 5 The (a) inconsistency and noise spots in the data result in an RMSE as high as 0.041m, making it difficult to use directly for quantitative evaluation.

[0110] Figure 5(c) The collaborative fusion method of the present invention: After spatial registration and parameter correction, the reconstructed deformation cloud map is spatially distributed similarly to the LiDAR measured results. Figure 5 The results in (a) show a high degree of agreement, with accurate positioning of the red high-deformation area, smooth and noise-free cloud map, and continuous dynamic monitoring (time resolution of 1 day). The RMSE was reduced to 0.015m, representing a 46.4% improvement in accuracy compared to the single LiDAR method and a 63.4% improvement compared to the single DFOS method. The white dashed line in the figure illustrates the DFOS fiber optic cable laying path.

[0111] 7. Conclusion of the Example

[0112] This embodiment fully verifies the technical feasibility and superiority of the method described in this invention using real landslide monitoring data. Spatial registration accuracy meets engineering requirements (RMSE ≤ 0.034m), accuracy is improved by 53% after deformation transfer function correction, the deformation field reconstructed by spatiotemporal fusion is highly correlated with LiDAR measured values ​​(r = 0.92), and the tangent angle method successfully provides an early warning of accelerated landslide deformation 4 days in advance. This method effectively solves the problem of collaborative analysis of LiDAR and DFOS data, realizing integrated "surface-to-interior" monitoring and dynamic early warning of landslides.

[0113] The above embodiments are merely preferred embodiments of the present invention. It should be noted that those skilled in the art can make several improvements and equivalent substitutions without departing from the principle of the present invention. All such improvements and equivalent substitutions to the claims of the present invention fall within the protection scope of the present invention.

Claims

1. A method for landslide surface-interior co-deformation reconstruction and evaluation based on airborne LiDAR and DFOS registration, characterized in that, Includes the following steps: S1. Data Synchronization Acquisition and Preprocessing: Simultaneously acquire airborne lidar point cloud data and distributed fiber optic sensing deep strain monitoring data of the target landslide area, and perform coordinate calculation, denoising and classification on the airborne lidar point cloud data to generate a three-dimensional surface model, and perform temperature compensation and denoising processing on the deep strain monitoring data to construct a strain-space-time dataset. S2. Spatial registration of heterogeneous monitoring data: Establish a unified spatial benchmark for the landslide body, and accurately map the distributed fiber optic sensing DFOS monitoring path to the surface three-dimensional model generated by the airborne lidar LiDAR through a three-dimensional coordinate transformation model to obtain the mapped DFOS strain data and realize the spatial location association between underground strain points and surface three-dimensional points. S3. Coordinated Inversion and Correction of Surface and Interior Deformation: Based on the landslide geomechanical model, a deformation transfer function is constructed. The mapped DFOS strain data is used as input. The surface deformation field is inverted through the deformation transfer function as the first deformation estimation field. The measured surface deformation field extracted by LiDAR at the same time is used. The parameters of the deformation transfer function are corrected through optimization algorithm to obtain the corrected coordinated deformation field. S4. Spatiotemporal fusion and dynamic reconstruction of deformation field: Using the corrected collaborative deformation field as the control node, the high temporal resolution strain data of DFOS is used for temporal interpolation to reconstruct the dynamic deformation field of the landslide from deep to the surface during the LiDAR data gap period. S5. Quantitative evaluation and early warning of landslide stability based on collaborative deformation field: Based on the reconstructed full-time dynamic deformation field, the deformation rate time series of key parts of the landslide is extracted; according to the deformation rate time series, the real-time stability state and early warning level of the landslide are calculated using a preset early warning model, and the evaluation and early warning results are output.

2. The landslide surface-interior co-deformation reconstruction and evaluation method based on airborne LiDAR and DFOS registration as described in claim 1, characterized in that, In step S1, the specific steps of performing coordinate calculation, denoising, and classification on the airborne lidar point cloud data to generate a 3D surface model are as follows: Coordinate calculation is based on the sensor's geometric calibration model. By fusing global navigation satellite system, inertial measurement unit trajectory data, and raw laser ranging data, the absolute coordinates of each laser footpoint are calculated. The calculation model is as follows: ； in, This represents the three-dimensional coordinate column vector of the laser footprint in the ground coordinate system. The column vector representing the phase center coordinates of the GNSS antenna at time t; Represents the rotation matrix from the carrier coordinate system to the ground coordinate system, which varies with time t; This represents the column vector of arm offset from the GNSS antenna phase center to the origin of the laser scanner coordinate system. This represents the fixed-mount rotation matrix from the laser scanner coordinate system to the carrier coordinate system. This indicates that in the laser scanner coordinate system, the distance measurement value... , scanning pitch angle and azimuth A defined column vector of laser footpoint coordinates; Point cloud filtering employs a statistical outlier removal algorithm. For any point in the point cloud... Calculate the average distance from its k nearest neighbors. and remove those that meet the requirements. The point, among which and These are the mean and standard deviation of the average distance across the entire point cloud, respectively. This is the threshold coefficient.

3. The method for landslide surface-interior co-deformation reconstruction and evaluation based on airborne LiDAR and DFOS registration as described in claim 1, characterized in that, In step S1, the temperature compensation and noise reduction processing of the deep strain monitoring data specifically involves: temperature compensation for Brillouin scattering-based sensing data, using the following formula to separate the strain change. : ； in, The change in Brillouin frequency shift of the sensing optical fiber coupled to the soil mass. The Brillouin frequency shift variation is the value of a reference fiber arranged along the same path for temperature monitoring. The strain coefficient of the optical fiber; Strain data noise reduction employs a sliding window averaging filter method, which provides smooth strain values ​​at position s on the optical fiber over time t. Calculated by the following formula: ； in, Here, represents the original strain value at position s at time t, and N is a positive integer representing the half-width of the preset filter window. is the spatial sampling interval of the DFOS data, and n is the index variable that takes all integer values ​​from -N to N in sequence during the summation process.

4. The method for landslide surface-interior co-deformation reconstruction and evaluation based on airborne LiDAR and DFOS registration as described in claim 1, characterized in that, Step S2 specifically involves: the three-dimensional coordinate transformation model is a three-dimensional similarity transformation model, which achieves coordinate transformation by solving the following equations: ； in, This is the column vector of coordinates of the DFOS monitoring points in the LiDAR coordinate system. This is the original coordinate column vector of DFOS. For rotation matrix, It is a translation vector. is the scale factor.

5. The method for landslide surface-interior co-deformation reconstruction and evaluation based on airborne LiDAR and DFOS registration as described in claim 1, characterized in that, In step S3, the deformation transfer function describes the physical model of the mapping relationship between deep distributed strain and surface displacement field, and its expression is: ； in, Representing surface points The deformation vector at time t, For transfer function operators, Indicates underground point strain vector, Geological model parameters to characterize the soil and rock mechanical properties of landslides. This is the vector of function parameters to be optimized.

6. The method for landslide surface-interior co-deformation reconstruction and evaluation based on airborne LiDAR and DFOS registration as described in claim 1, characterized in that, In step S3, the optimization algorithm uses minimizing the difference between the LiDAR measured deformation field and the deformation field retrieved from the deformation transfer function as the criterion for a monitoring period. Its solution is the corrected optimal parameter vector. The process can be expressed as the following least squares problem: ； in, This represents the number of surface deformation points included in the comparison. Let i be the deformation vector calculated from LiDAR data for the i-th point. The corresponding point deformation vector is calculated based on the DFOS strain change and the deformation transfer function. This is the vector of function parameters to be optimized.

7. The method for landslide surface-interior co-deformation reconstruction and evaluation based on airborne LiDAR and DFOS registration as described in claim 6, characterized in that, In step S4, the time-series interpolation is performed at the previous LiDAR observation time. Based on the deformation field, the deformation transfer function corrected by the optimization algorithm is used. Using DFOS continuous strain data, calculate any subsequent time step. The process of deformation field is calculated using the following formula: ； in, This is the corrected optimal parameter vector. From arrive The cumulative strain change measured by DFOS at any given time.

8. The method for landslide surface-interior co-deformation reconstruction and evaluation based on airborne LiDAR and DFOS registration as described in claim 1, characterized in that, Step S5 specifically involves: First, extracting the deformation rate time series of key landslide monitoring points from the dynamic deformation field throughout the entire time period as an evaluation index; then, comparing and analyzing the deformation rate time series with the preset warning threshold or warning model to determine the stability state of the landslide and the warning level; finally, generating and outputting the evaluation results containing the warning level and key deformation parameters.

9. The method for landslide surface-interior co-deformation reconstruction and evaluation based on airborne LiDAR and DFOS registration as described in claim 8, characterized in that, The comparison and analysis of the deformation rate time series with the preset warning threshold or warning model specifically involves: using the tangent angle method for landslide warning, and calculating the displacement tangent angle of the key landslide point at time t based on the displacement-time curve corresponding to the deformation rate time series. Displacement tangent angle The calculation formula is: ； in, Let be the cumulative displacement over time t. For the selected reference time The displacement; setting the tangent angle threshold for near-slip warning. When the conditions are met When the time is right, it is determined that the skid is about to be cleared and the corresponding level of warning is triggered.

10. A computer device comprising a memory, a processor, and a computer program stored in the memory, characterized in that, The processor executes the computer program to implement the steps of the method of claim 1.

Images