A method for collecting data of red sandstone roadbed slope surface and stratum

Abstract

The present application belongs to the technical field of geotechnical engineering investigation, especially to a red sandstone roadbed slope surface and stratum data acquisition method, comprising the following steps: step one: stratum data acquisition and preprocessing before construction: drilling points are arranged in the area of the to-be-excavated cutting slope, survey drilling data is obtained, the drilling data is input into the Kriging interpolation R language program, the zoning fitting strategy of the variogram is adopted, different ranges and base values are set for different weathering zones of the red sandstone, and an initial stratum information model is generated. The red sandstone roadbed slope surface and stratum data acquisition method sets the red sandstone roadbed slope surface and stratum data acquisition method, and integrates the data acquisition of the three key stages of the pre-construction drilling exploration, the excavation process onboard radar scanning and the backfilling unmanned aerial vehicle radar scanning into a unified technical framework through the integrated acquisition strategy of "three stages-double control surface-multi source fusion".

Description

Technical Field

[0001] This invention relates to the field of geotechnical engineering investigation technology, and in particular to a method for collecting data on red sandstone roadbed slopes and strata. Background Technology

[0002] Red sandstone is widely distributed in southern my country, and many highway projects use it as roadbed fill. However, red sandstone contains a large amount of hydrophilic clay minerals (such as montmorillonite), exhibiting undesirable engineering characteristics such as low strength, high weathering degree, and softening and disintegration upon contact with water. Studies have shown that red sandstone is easily disintegrated and broken by external factors such as wet-dry cycles, pH value, temperature, and disturbance. Directly using it as roadbed fill can lead to uneven roadbed settlement, slope collapse, and other problems. Therefore, accurately obtaining information on the slope morphology and stratigraphic distribution of red sandstone roadbeds is of significant engineering importance for roadbed stability assessment and deformation prediction.

[0003] Traditional methods for collecting and processing roadbed slope data have the following main technical drawbacks: 1. The data source is singular and fragmented, making it difficult to reflect the evolution of red sandstone roadbed throughout its entire life cycle. In existing technologies, geological exploration typically uses borehole sampling to obtain stratigraphic information, while slope morphology measurement employs total stations, GPS-RTK, or conventional photogrammetry. These two data acquisition methods are independent, with borehole data (point data) and surface mapping data (area data) stored and processed separately. There is a lack of technical means to organically integrate internal stratigraphic information with slope surface morphology. Red sandstone exhibits strong spatial heterogeneity, with significant differences in engineering characteristics across different weathering zones (strongly weathered, moderately weathered, and slightly weathered). A single data source cannot comprehensively reflect the true three-dimensional structure of the roadbed.

[0004] 2. Poor timing matching, making it difficult to coordinate data throughout the entire construction process. Red sandstone subgrade engineering includes multiple stages such as pre-construction exploration, excavation, and backfilling / compaction. In existing technologies, data acquisition at each stage often faces challenges in comparative analysis and fusion modeling on the same 3D platform due to inconsistent coordinate systems and mismatched acquisition precision. For example, pre-construction borehole data uses a geodetic coordinate system, airborne radar data during excavation uses a construction coordinate system, and UAV data after backfilling uses a GPS coordinate system, making it difficult to use multi-source data collaboratively.

[0005] 3. The interpolation method has limitations and cannot accurately reconstruct abrupt transition interfaces in red sandstone strata. Conventional spatial interpolation methods (such as inverse distance weighting and ordinary kriging) typically employ globally uniform variability function parameters when dealing with spatially variable soil and rock masses like red sandstone, without setting differentiated range and sill values ​​for different weathering zones. More importantly, conventional interpolation methods often produce smooth transitions at critical engineering interfaces such as excavation and backfill surfaces, failing to accurately reproduce the abrupt changes at these interfaces, resulting in significant deviations between the 3D model and the actual engineering conditions.

[0006] 4. Insufficient precision in point cloud data processing and lack of systematic error correction mechanisms. In existing technologies, although point cloud data acquired by airborne radar and UAV radar can reflect the surface morphology of slopes, the original point cloud contains systematic errors due to equipment vibration, environmental factors, and other influences. Conventional processing methods only perform operations such as noise reduction and registration, without introducing on-site detection control points to correct for systematic errors, making it difficult to guarantee the absolute accuracy of the point cloud data.

[0007] 5. Difficulty in fusing multi-source heterogeneous data, lack of a unified interpolation framework. The current technical challenge lies in fusing and modeling three types of heterogeneous data—borehole data, excavated slope point cloud data, and backfilled slope point cloud data—on the same platform. Existing methods typically process each type of data separately before stitching them together, without establishing a unified collaborative interpolation framework. This results in discontinuities or distortions at data boundaries in the fused model. Summary of the Invention

[0008] To address the technical problems of existing methods for acquiring data on red sandstone roadbed slopes and strata, such as fragmented data acquisition and processing, insufficient interpolation accuracy, and difficulty in accurately reconstructing the three-dimensional structure of red sandstone roadbeds, this invention proposes a new method for acquiring data on red sandstone roadbed slopes and strata.

[0009] The present invention proposes a method for acquiring data on the slope surface and strata of red sandstone roadbed, including step one: strata data acquisition and preprocessing before construction: drilling points are set up in the slope area of ​​the road cut to be excavated to obtain survey drilling data, the drilling data is input into the Kriging interpolation R language program, and a variation function partition fitting strategy is adopted to set different ranges and sill values ​​for different weathering zones of red sandstone to generate an initial strata information model; Step 2: Identification and Marking of Special Interpolation Control Surfaces: The structural surfaces of the cut slope after cold excavation and the surface structural surfaces of the subgrade slope after backfilling are taken as special interpolation control surfaces and marked as constrained interpolation boundaries in three-dimensional space, so that they have a higher weight coefficient than ordinary data points in subsequent interpolation calculations. Step 3: Slope Data Acquisition and Preprocessing during Excavation: During the cold excavation of the red sandstone subgrade, the three-dimensional point cloud data of the excavated slope is acquired using the cold excavation machine-mounted scanning radar. The structural surface model of the road cut slope before backfilling is formed through noise reduction and registration processing. On-site detection control points are set up for error correction. The elevation and planar coordinate information of the dense point cloud in the point cloud model are processed based on Surfer software to restore the actual shape of the slope surface and extract the corresponding coordinate data. Step 4: Data Acquisition and 3D Reconstruction of Backfilled Slope: For the backfilled red sandstone roadbed slope, a 3D laser scanning radar on a UAV component is used for scanning. Combined with GPS positioning data and ground control point data, the SFM motion recovery structure algorithm and MFS multi-view stereo matching algorithm are used for joint calculation. GPS data and control point data are used as global optimization constraints to eliminate cumulative errors and construct a point cloud model of the surface slope of the red sandstone roadbed. After processing by Surfer, the coordinate data of the backfilled slope surface is extracted. Step 5: Multi-source heterogeneous data collaborative fusion and 3D modeling steps: Import the backfilled slope surface coordinate data, the R language preprocessed stratum coordinate data, and the cold excavation cut slope coordinate data into the 3D modeling platform. Using the two special interpolation control surfaces marked in Step 2 as constraint boundaries, the three types of data are fused and interpolated using the co-kriging interpolation method that considers abrupt interfaces to generate a spatial interpolation control surface that integrates stratum information and slope surface information. Step Six: Model Accuracy Assessment and Output Step: Based on the deviation data between the field detection control points and the 3D model, the accuracy of this modeling is assessed using the root mean square error calculation formula, and the 3D geological model data after error correction is automatically output.

[0010] Preferably, the borehole data processing in step one employs a variogram partitioning fitting strategy, generating an initial formation information model using the variogram calculation formula and the partitioning range setting formula. The formula for calculating the function of variation is: ,in, The value of the variation function represents the spatial distance. The degree of variation between sample points The spatial distance between sample points; The distance is The number of sample point logs; For position The borehole data values ​​at the location include formation depth and thickness parameters; For position Drilling data values ​​at the location; The formulas for setting the partition range and base value are as follows: ,in, The values ​​of the variation function after partitioning; Let be the nugget constant, representing the random variation caused by changes in microstructure and measurement errors, expressed as an experimental variogram function. The intercept is determined at that time; The sill value represents the overall variability of the regionalized variable and is determined by the stable value of the experimental variogram. For variable range, representing the maximum distance of spatial correlation, when > The time variable no longer has spatial correlation; For the spherical model function, the expression is: , ; For different weathering zones (strongly weathered layer, moderately weathered layer, slightly weathered layer) and weak interlayers of red sandstone, range, sill value and nugget constant are set to match their spatial variation characteristics.

[0011] Preferably, the special interpolation control surface marking in step two employs a constrained interpolation boundary marking method, using a constrained boundary weight coefficient formula to enhance the weight of the special interpolation control surface. The formula for the constraint boundary weight coefficient is: ,in, These are the weighting coefficients for the special interpolation control surface, and are dimensionless. These are the weighting coefficients for ordinary data points, and are dimensionless. As the interface enhancement factor, the optimal value was determined through cross-validation experiments based on the stability requirements of the red sandstone slope structure. The marking method for special interpolation control surfaces in three-dimensional space is as follows: In the interpolation mesh system, the mesh nodes belonging to the special interpolation control surfaces are marked as constraint nodes. During the co-Kriging equations solution process, the weight coefficients of the constraint nodes are forced to execute the above formula to ensure that the key engineering interfaces are accurately restored in the interpolation results.

[0012] Preferably, the point cloud data processing of the excavated slope surface in step three is performed using a point cloud error correction formula and a Surfer meshing formula: The point cloud error correction formula is: ,in, These are the point cloud coordinates after error correction, including three-dimensional coordinate components; These are the original point cloud coordinates, containing three-dimensional coordinate components; To ensure the accuracy of calibration, control points are evenly distributed to determine the number of control points for on-site testing. For the first The measured coordinates of each on-site control point were obtained by total station or GPS-RTK measurement. For the point cloud model corresponding to the first The coordinates of the control points; The Surfer meshing formula is: ,in, For grid points Elevation value at the location; The number of point clouds involved in the interpolation is determined based on the search radius; For the first The weight coefficients of each point cloud satisfy the following conditions: The weighting coefficients are determined by the Kriging interpolation method and are related to the spatial distance between point clouds and the variogram. For the first Elevation values ​​of point clouds; The Kriging gridding method was adopted, and the grid spacing, search radius and anisotropy ratio were reasonably set according to the accuracy requirements of the surface layer of the red sandstone slope to ensure that the gridded digital elevation model can accurately reproduce the actual shape of the slope surface.

[0013] Preferably, the three-dimensional reconstruction of the UAV point cloud data in step four is processed using the SFM-MFS joint solution formula and the global optimization constraint formula: The formula for SFM motion recovery structure is: ,in, For the first The rotation matrix of the image has dimensions of... , indicating the camera orientation; For the first The translation vector of the image has dimension . , indicating the camera position; For the first The coordinates of a three-dimensional point, with dimensions of . ; For the first The three-dimensional point at the th t The coordinates of the image points on the image are in dimension [dimensional value]. ; This is a camera projection function that projects 3D points onto a 2D image plane. For the number of images, This represents the number of three-dimensional points. Using Euclidean distance, minimize reprojection error to optimize camera parameters and 3D point coordinates; The MFS multi-view stereo matching formula is: ,in, For pixels disparity value; For pixels The set of neighboring pixels; This serves as an image index identifier, used to distinguish different neighboring images. This is a robust cost function used to suppress the influence of outliers; For the first The grayscale value of the image; The grayscale value is the reference image. For parallax The corresponding mapping function maps pixels Mapped to the corresponding location in the neighboring image; The global optimization constraint formula is: ,in, The coordinates of the three-dimensional point are optimized using GPS and ground control point constraints, with dimensions of [dimension number missing]. ; The original 3D point coordinates calculated by SFM, with dimensions of ; The Kalman gain matrix has a dimension of . The covariance matrix of the noise is calculated by measuring the noise using GPS and estimating the noise using SFM. This is a GPS coordinate transformation function that converts the coordinates of a 3D point to coordinates in the GPS coordinate system. These are the measured GPS coordinates.

[0014] Preferably, the multi-source heterogeneous data fusion interpolation in step five uses a co-kriging interpolation formula that considers abrupt interfaces to process the three types of data: The co-kriging interpolation formula is: ,in, Unknown point The interpolated estimates include information on stratum depth and slope elevation; For the coordinate data of the slope surface after backfilling, The number of surface data points participating in the interpolation. The weight coefficients for the surface data points are dimensionless and satisfy the following conditions: ; For preprocessed stratigraphic coordinate data in R language, The number of stratigraphic data points used in the interpolation. The weighting coefficients for the stratigraphic data points are dimensionless and satisfy the following conditions: ; This is the coordinate data of the road cut slope after cold excavation. The number of data points from the excavation face participating in the interpolation. The weighting coefficients for the data points on the excavation face are dimensionless and satisfy the following conditions: ; The weighting coefficients are determined by a solution logic that satisfies the unbiased estimation condition. They are combined with the variation functions of surface data, stratum data, and excavation face data, as well as the cross variation functions between them, and solved using Lagrange multipliers to ensure the accuracy and rationality of the interpolation results without the need for specific equation sets to limit them. The weighting coefficients are determined by a solution logic that satisfies the unbiased estimation condition. They are combined with the variation functions of surface data, stratum data, and excavation face data, as well as the cross variation functions between them, and solved using Lagrange multipliers to ensure the accuracy and rationality of the interpolation results without the need for specific equation sets to limit them. The variogram and cross-variogram are represented by simplified characters: The variation function characterizing the surface data itself, The variation function characterizing the stratigraphic data itself, The variability function characterizing the excavation face data itself, (and (Equal) represents the cross-variance function of surface data and stratigraphic data. (and (Equal) represents the cross-variance function of surface data and excavation face data. (and (Equal) is a cross-variance function characterizing the stratigraphic data and the excavation face data. The cross-variance function characterizing the surface data and the estimated points. The cross-variance function characterizes the stratigraphic data and the estimated points. The cross-variance function characterizing the excavation face data and the estimated points; the Lagrange multipliers are represented by simplified characters. , , Characterization, used to satisfy the unbiased estimation condition; The formula for the interface constraint in case of a sudden change is: ,in, To consider the variation function of the abrupt interface; For the variation function of a continuous region, a spherical model or an exponential model is used; The variation function is for the discontinuous region, and the nugget model is used; The influence distance of the special interpolation control surface is determined by analyzing the variation characteristics of the data on both sides of the special interpolation control surface.

[0015] Preferably, the root mean square error (RMSE) calculation formula in step six is ​​as follows: ,in, To assess the number of control points, , , For the first Measured three-dimensional coordinates of each control point , , For the first The model's three-dimensional coordinates for each control point.

[0016] Preferably, the drone component in step four includes a drone, and the lower surface of the drone is provided with a mounting device, the mounting device including a fixing frame, the fixing frame fixing the three-dimensional laser scanning radar under the drone; The mounting device further includes a mounting bracket, which is fixedly mounted on the lower surface of the UAV fuselage. A transmission gear ring is rotatably connected to the outer surface of the mounting bracket via a bearing. Two transmission gear rings are connected by a synchronous belt and a synchronous pulley. A drive motor is fixedly mounted on the outer surface of the mounting bracket via a support frame. One end of the output shaft of the drive motor drives one of the transmission gear rings to rotate via a gear. A lifting screw is threadedly connected to the inner wall of the transmission gear ring. A push plate is fixedly mounted on the lower end of the lifting screw. A push frame is slidably inserted into the outer surface of one end of the push plate. A return spring is fixedly mounted on the outer surface of the push frame. One end of the return spring is fixedly mounted to the lower surface of the push plate.

[0017] Preferably, a lifting frame is slidably inserted into the outer surface of the mounting bracket via a limiting rod. A drive gear is rotatably connected to the upper surface of the lifting frame via a bearing. A rotating sleeve with an arc groove is fixedly installed on the upper surface of the drive gear. A plug-in rod with a column is fixedly installed on the lower surface of the push frame. One end of the plug-in rod is slidably inserted into the inner wall of the rotating sleeve, and the outer surface of the column is slidably inserted into the inner wall of the arc groove. A movable rack is slidably inserted into the upper surface of the lifting frame, and the movable rack meshes with the drive gear. A connecting spring is fixedly installed on the outer surface of the limiting rod of the lifting frame. One end of the connecting spring is fixedly installed on the outer surface of the mounting bracket. Both ends of the push plate are slidably inserted into the inner wall of the mounting bracket and slidably sleeved with the outer surface of the limiting rod of the lifting frame.

[0018] Preferably, the upper surface of the lifting frame is slidably connected to a slider via a slide rail, the outer surface of the slider is fixedly installed to the lower surface of the fixed frame, one end of the moving rack is fixedly installed with a linkage rod, the outer surface of the linkage rod is slidably connected to the outer surface of the fixed frame via a sleeve, the lower surface of the three-dimensional laser scanning radar is fixedly installed with a mounting base, and the mounting base is clamped after the multiple fixed frames move together.

[0019] The beneficial effects of this invention are as follows: 1. By setting up data acquisition methods for red sandstone subgrade slopes and strata, and employing an integrated acquisition strategy of "three stages - dual control surfaces - multi-source fusion," data acquisition from three key stages—pre-construction borehole exploration, airborne radar scanning during excavation, and UAV radar scanning after backfilling—is integrated into a unified technical framework. The structural surfaces of the cut slope after cold excavation and the surface structural surfaces of the subgrade slope after backfilling are used as special interpolation control surfaces for weight enhancement. A partitioned adaptive variogram and a co-kriging interpolation algorithm considering abrupt interfaces are employed, along with a fully automated processing chain including point cloud error correction, SFM-MFS joint solution, and root mean square error assessment. The modeling error at the key engineering interface is reduced by 35%-40%, the model matches the field measured data by more than 97%, the interpolation accuracy of the stratum interface is improved by about 30%, and the 3D reconstruction accuracy is improved by about 25% compared with conventional methods. It fundamentally solves the technical problems of difficult multi-source data fusion and inaccurate reconstruction of key interfaces in red sandstone, a rock and soil body with strong spatial heterogeneity. It provides reliable data support for stability assessment and deformation prediction of red sandstone subgrade, and solves the technical problems of fragmented data acquisition and processing, insufficient interpolation accuracy, and difficulty in accurately reconstructing the 3D structure of red sandstone subgrade in existing red sandstone subgrade slope and stratum data acquisition methods.

[0020] 2. By designing the UAV components and adopting a quick-release structure of "dual-motor synchronous drive - lead screw lifting - composite transmission - multi-directional clamping," the drive motor transmits the synchronous rotation of the gear ring via a synchronous belt, driving the lifting lead screw to rise and fall. The arc groove sliding mechanism of the rotating sleeve and plug-in rod converts the lifting motion into rotational motion, driving the moving rack to synchronously converge multiple fixed frames for multi-directional adaptive clamping of the 3D laser scanning radar. Installation and disassembly of the radar can be completed within 30 seconds without any tools, improving field relocation efficiency by over 90%. Simultaneously, the return spring and connecting spring provide continuous preload, ensuring that the radar mounting seat displacement is less than 0.05mm in a 20Hz-200Hz vibration environment, improving stability by 45% compared to conventional bolt fixing methods, and flight attitude stability by 30%. Point cloud plane error ≤ ±1.5cm and elevation error ≤ ±2cm fully meet the engineering requirements for high-precision mapping of red sandstone roadbed slopes. It can also drive the 3D laser scanning radar to rise and fall, preventing scanning limitations caused by the UAV's legs. Attached Figure Description

[0021] Figure 1 This is a schematic diagram of a method for acquiring data on red sandstone roadbed slopes and strata proposed in this invention; Figure 2 This is a three-dimensional view of the installation support structure for a method of acquiring data on red sandstone roadbed slopes and strata proposed in this invention; Figure 3This is a three-dimensional view of the lifting screw structure of a method for acquiring data on red sandstone roadbed slope and strata proposed in this invention. Figure 4 This is a three-dimensional view of the push plate structure of a method for acquiring data on red sandstone roadbed slopes and strata proposed in this invention. Figure 5 This is a three-dimensional view of the fixing frame structure for a method of acquiring data on red sandstone roadbed slopes and strata proposed in this invention; Figure 6 This is a flowchart of a method for acquiring data on red sandstone roadbed slopes and strata, as proposed in this invention.

[0022] In the diagram: 1. UAV; 11. Mounting base; 12. 3D laser scanning radar; 2. Mounting bracket; 21. Transmission gear ring; 22. Drive motor; 23. Lifting screw; 24. Push plate; 25. Push frame; 26. Return spring; 3. Lifting frame; 31. Drive gear; 32. Rotating sleeve; 33. Connecting rod; 34. Moving rack; 35. Connecting spring; 4. Slider; 41. Linkage rod; 42. Fixing frame. Detailed Implementation

[0023] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0024] Reference Figures 1-6 A method for acquiring data on red sandstone roadbed slopes and strata includes the following steps: Step 1: Pre-construction geological data acquisition and preprocessing steps: Drilling Point Layout: In the area of ​​the road cut slope to be excavated, drilling points are strictly laid out according to a density of "one per 100-200㎡". Each drilling point is at least 2m away from the slope edge. The drilling density remains consistent across different weathering zones, including strongly weathered, moderately weathered, slightly weathered, and weak interlayers, to ensure that the drilling data comprehensively covers the entire slope area and reflects the characteristics of different weathering layers. Based on the actual slope area (e.g., approximately 1500㎡), a total of 9 drilling points are laid out, evenly distributed throughout the slope area. The drilling depth is determined based on the stratum thickness, ranging from 10-20m, ensuring penetration of all weathered layers and weak interlayers. The drilling locations avoid areas disturbed by slope construction to guarantee the accuracy of the drilling data.

[0025] Data acquisition from exploration boreholes: Drilling operations are carried out using geological drilling equipment (such as the XY-1 geological drilling rig). The stratigraphic type, burial depth, thickness, and lithological parameters at different depths of each borehole are recorded simultaneously to form a complete data ledger for exploration boreholes. The data accuracy is controlled within ±0.05m, consistent with the accuracy requirements implied in this invention. After the data acquisition is completed, the borehole data is initially screened to remove abnormal data such as sudden changes in stratigraphic parameters caused by drilling errors or equipment failures. Valid data is retained for subsequent preprocessing to ensure the accuracy and representativeness of the data.

[0026] Data preprocessing and initial stratigraphic information model generation: The filtered borehole data is imported into a Kriging interpolation R language program (R4.2.0 version can be used, with the gstat package to implement Kriging interpolation function). For different weathering zones of red sandstone, the range, sill value and nugget constant are set within a clear parameter range. The specific values ​​are determined by fine-tuning within the range defined by this invention, combined with the actual weathering degree of the slope. The specific values ​​are as follows (all within the parameter range of this invention): Strongly weathered layer: range 5m (3-8m range), sill value 1.6m (0.8-2.5m range), nugget constant 0.2m (0.1-0.3m range); Moderately weathered layer: range 12m (8-15m range), sill value 3.8m (2.5-5.0m range), nugget constant 0.12m (0.05-0.2m range); Slightly weathered layer: range 20m (15-25m), slab value 6.5m (5.0-8.0m), nugget constant 0.09m (0.03-0.15m).

[0027] Weak interlayer: range 3.5m (within the range of 2-5m), sill value 1.2m (within the range of 0.5-1.8m), nugget constant 0.28m (within the range of 0.15-0.4m); the values ​​can be slightly adjusted within the above range according to the actual weathering degree of the red sandstone.

[0028] Specifically, in step one, the variogram fitting is performed as follows: the variogram partitioning is fitted using the following formula to generate the initial stratigraphic information model: The formula for calculating the function of variation is: ,in, The value of the variation function represents the spatial distance. The degree of variation between sample points The spatial distance between sample points; The distance is The number of sample point logs; For position The borehole data values ​​at the location include formation depth and thickness parameters; For position Drilling data values ​​at the location.

[0029] The formulas for setting the partition range and base value are as follows: ,in, The values ​​of the variation function after partitioning; Let be the nugget constant, representing the random variation caused by changes in microstructure and measurement errors, expressed as an experimental variogram function. The intercept is determined at that time; The sill value represents the overall variability of the regionalized variable and is determined by the stable value of the experimental variogram. For variable range, representing the maximum distance of spatial correlation, when > The time variable no longer has spatial correlation; For the spherical model function, the expression is: , In this embodiment, all weathering zones are fitted using the spherical model, with a fitting accuracy of R²≥0.92, ensuring reliable fitting results.

[0030] Model Generation and Validation: An initial stratigraphic model was generated by performing variogram partitioning and kriging interpolation on the borehole data using the R language program. The model had a resolution of 0.5m × 0.5m and clearly presented the distribution range, burial depth, and thickness of each weathering layer. After generation, three borehole data points were selected as validation points. The model predictions were compared with the measured values, and the error was ≤0.08m, ensuring that the initial stratigraphic model met the technical requirements of this invention and provided reliable basic data for subsequent steps.

[0031] Step Two: Identification and Marking of Special Interpolation Control Surfaces Identification of Special Interpolation Control Surfaces: Combining engineering design drawings, field survey results, and the definitions of this invention, the specific ranges of two special interpolation control surfaces are clarified: First, the structural surface of the cut slope after cold excavation (i.e., the exposed red sandstone slope surface after excavation, serving as the interface between the strata and air), corresponding to the structural surface formed by the excavated slope surface data collected in subsequent step three; second, the surface structural surface of the roadbed slope after backfilling (i.e., the slope surface formed after backfilling, serving as the interface between the backfill layer and air), corresponding to the structural surface formed by the backfilled slope surface data collected in subsequent step four. Both interfaces are key constraint boundaries for subsequent co-kriging interpolation calculations, directly affecting the accuracy of the 3D model. The identification accuracy is controlled within ±0.05m.

[0032] Three-dimensional spatial marking: Using three-dimensional modeling software (AutoCAD Civil 3D 2024 version can be used), the two special interpolation control surfaces mentioned above are marked as constrained interpolation boundaries in three-dimensional space. The specific method is as follows: import the plane coordinates and elevation data of the slope area (data accuracy ≤ ±0.05m), draw the three-dimensional contour lines of the two interfaces, and mark the grid nodes within the contour line range as constrained nodes. The coordinate accuracy of the constrained nodes is controlled within ±0.03m (higher than the basic accuracy of this invention to ensure the reliability of the marking), ensuring that the marking position is consistent with the actual engineering interface and fully meets the marking requirements of "constrained interpolation boundaries" in this invention.

[0033] Weight enhancement processing: A constrained interpolation boundary marking method is used to enhance the weights of special interpolation control surfaces through the constrained boundary weight coefficient formula, and an interface enhancement factor is applied. The value ranges from 0.3 to 1.2, where the structural surface of the cut slope after cold excavation corresponds to... The value is 0.8-1.2, corresponding to the surface structure of the roadbed slope after backfilling. The value is 0.3-0.8, and the optimal value can be determined iteratively by using 30-50 sets of verification samples and taking the minimum interpolation error as the verification index.

[0034] Specifically, in step two, the special interpolation control surface marking adopts a constrained interpolation boundary marking method, and the special interpolation control surface is weighted by the constrained boundary weight coefficient formula: The formula for the constraint boundary weight coefficient is: ,in, These are the weighting coefficients for the special interpolation control surface, and are dimensionless. These are the weighting coefficients for ordinary data points, and are dimensionless. As an interface enhancement factor, based on the stability requirements of the red sandstone slope structure, the optimal value is determined iteratively using 30-50 sets of verification samples and with the minimum interpolation error as the verification index. The marking method for special interpolation control surfaces in three-dimensional space is as follows: In the interpolation mesh system, the mesh nodes belonging to the special interpolation control surfaces are marked as constraint nodes. During the co-Kriging equations solution process, the weight coefficients of the constraint nodes are forced to execute the above formula to ensure that the key engineering interfaces are accurately restored in the interpolation results.

[0035] Step 3: Slope Data Acquisition and Preprocessing During Excavation: Point cloud data acquisition: During the cold excavation of the red sandstone roadbed, the cold excavation airborne scanning radar (in this embodiment, the RIEGL VZ-4000 airborne radar with a scanning accuracy of ±0.01m) is used to scan synchronously with the excavation progress. The scanning frequency is 10Hz, and the scanning range covers the entire excavated slope to acquire the three-dimensional point cloud data information of the excavated slope. The point cloud density is controlled at 100-150 points per square meter to ensure the density and integrity of the data and meet the needs of subsequent preprocessing and modeling.

[0036] Point cloud preprocessing (denoising and registration): The collected raw point cloud data is imported into point cloud processing software (CloudCompare version 2.12 is used in this embodiment). Strictly following the limitations of this invention, Gaussian filtering denoising algorithm (3×3 filtering window) and ICP registration algorithm (registration accuracy threshold ≤0.05m) are used for denoising and registration processing. Noise reduction: A Gaussian filtering noise reduction algorithm (with a strict 3×3 filtering window) is used to remove noise points (such as dust reflection points in the air and equipment interference points) from the point cloud data, while retaining the effective slope point cloud data. The noise error of the point cloud data after noise reduction is ≤0.02m, ensuring the accuracy of the point cloud data.

[0037] Registration processing: The ICP registration algorithm is adopted, and the registration accuracy threshold is strictly controlled within ≤0.05m. The point cloud data collected at different excavation stages are registered and fused to form a complete structural surface model of the road cut slope before backfilling, ensuring that the point cloud data of different areas are in the same coordinate system. The overall deviation of the model after registration is ≤0.08m.

[0038] Point cloud error correction: On-site detection control points are evenly distributed on the excavated slope surface. The actual coordinates of the control points are measured using a total station (accuracy ±0.005m). The distribution method follows the principle of "even distribution and coverage of the entire slope surface". Error correction is used to ensure the accuracy of the point cloud data.

[0039] Surfer meshing and coordinate data extraction: Import the error-corrected point cloud data into Surfer software (Surfer version 16.0 can be used) and perform meshing processing strictly according to the specified parameters.

[0040] The grid spacing for Surfer meshing is 0.1-0.5m (0.3-0.5m for slopes ≤30° and 0.1-0.3m for slopes >30°), the search radius is 1.5-5.0m, the ratio of the search radius to the grid spacing is controlled between 3:1 and 10:1, the anisotropy ratio is 1.0-2.5, 1.0-1.5 along the slope direction and 1.5-2.5 perpendicular to the slope direction.

[0041] Specifically, the point cloud error correction formula in step three is as follows: ,in, These are the point cloud coordinates after error correction, including three-dimensional coordinate components; These are the original point cloud coordinates, containing three-dimensional coordinate components; To ensure the accuracy of calibration, control points are evenly distributed to determine the number of control points for on-site testing. For the first The measured coordinates of each on-site control point were obtained by total station or GPS-RTK measurement. For the point cloud model corresponding to the first The coordinates of the control points; The Surfer meshing formula is: ,in, For grid points Elevation value at the location; The number of point clouds involved in the interpolation is determined based on the search radius; For the first The weight coefficients of each point cloud satisfy the following conditions: The weighting coefficients are determined by the Kriging interpolation method and are related to the spatial distance between point clouds and the variogram. For the first Elevation values ​​of point clouds; The Kriging gridding method was adopted, and the grid spacing, search radius and anisotropy ratio were reasonably set according to the accuracy requirements of the surface layer of the red sandstone slope to ensure that the gridded digital elevation model can accurately reproduce the actual shape of the slope surface.

[0042] Step 4: Slope Data Acquisition and 3D Reconstruction After Backfilling: Slope data acquisition: For the backfilled red sandstone roadbed slope, a 3D laser scanning radar 12 (which can be a DJI Phantom 4 RTK with a laser radar module, scanning accuracy ±0.02m) of the drone component is used for scanning. The 3D laser scanning radar 12 can rotate, fly at an altitude of 20m, and fly at a speed of 5m / s. The scanning range covers the entire backfilled slope. GPS positioning data (positioning accuracy ±0.01m) and ground control point data (8 ground control points are set up, and GPS-RTK measurement is used, with an accuracy of ±0.008m) are collected simultaneously to ensure that the collected data meets the requirement of "combining GPS positioning data and ground control point data" in this invention.

[0043] SFM-MFS Joint Solving: Import the acquired image data, GPS data, and control point data into the 3D reconstruction software (Agisoft Metashape Professional version 2.0 can be used), and perform joint solving through the SFM (Structure of Motion) algorithm and the MFS (Multi-View Stereo Matching) algorithm.

[0044] Global optimization constraints: GPS data and ground control point data are used as global optimization constraints to eliminate the cumulative error in the SFM-MFS joint solution process and ensure the accuracy of the 3D point cloud data, with a cumulative error ≤0.06m.

[0045] Point cloud model processing and coordinate extraction: After optimization, a point cloud model of the surface slope of the red sandstone roadbed is constructed. The model is imported into Surfer software and processed using the same gridding parameters as in step three (fully in accordance with the limitations of this invention). The coordinate data of the backfilled slope surface (including X, Y, and Z three-dimensional coordinates) is extracted to form a coordinate data ledger. The coordinate accuracy is controlled within ±0.05m. Together with the excavation face coordinate data extracted in step three and the stratum coordinate data in step one, it serves as the data basis for multi-source fusion.

[0046] Specifically, the SFM-MFS joint solution and global optimization constraints in step four are achieved through the following formula: The formula for SFM motion recovery structure is: ,in, For the first The rotation matrix of the image has dimensions of... , indicating the camera orientation; For the first The translation vector of the image has dimension . , indicating the camera position; For the first The coordinates of a three-dimensional point, with dimensions of . ; For the first The three-dimensional point at the th t The coordinates of the image points on the image are in dimension [dimensional value]. ; This is a camera projection function that projects 3D points onto a 2D image plane. For the number of images, This represents the number of three-dimensional points. Using Euclidean distance, minimize reprojection error to optimize camera parameters and 3D point coordinates; The MFS multi-view stereo matching formula is: ,in, For pixels disparity value; For pixels The set of neighboring pixels; This serves as an image index identifier, used to distinguish different neighboring images. This is a robust cost function used to suppress the influence of outliers; For the first The grayscale value of the image; The grayscale value is the reference image. For parallax The corresponding mapping function maps pixels Mapped to the corresponding location in the neighboring image; The global optimization constraint formula is: ,in, The coordinates of the three-dimensional point are optimized using GPS and ground control point constraints, with dimensions of [dimension number missing]. ; The original 3D point coordinates calculated by SFM, with dimensions of ; The Kalman gain matrix has a dimension of . The covariance matrix of the noise is calculated by measuring the noise using GPS and estimating the noise using SFM. This is a GPS coordinate transformation function that converts the coordinates of a 3D point to coordinates in the GPS coordinate system. These are the measured GPS coordinates.

[0047] Step 5: Collaborative Fusion and 3D Modeling of Multi-Source Heterogeneous Data: Multi-source data import and preprocessing: The stratigraphic coordinate data obtained in step one (exported from the initial stratigraphic information model), the cold excavation face coordinate data extracted in step three, and the backfill surface coordinate data extracted in step four are all converted into the same coordinate system (geographic coordinate system can be used). Abnormal data (such as data with excessive coordinate deviation) are removed, and missing data are supplemented by linear interpolation to ensure the consistency and integrity of multi-source data. The coordinate accuracy of the three types of data is controlled within ±0.05m.

[0048] Interpolation constraint boundary settings: Use the two special interpolation control surfaces marked in step two as the constraint boundaries for co-Kriging interpolation, import them into the 3D modeling platform (AutoCAD Civil 3D 2024 version), and set the weight coefficients of the constraint boundaries (as determined in step two). (Value selection) to ensure that the constraint boundary plays a dominant role in the interpolation process and avoids interpolation distortion.

[0049] Mutation interface constraint settings: The mutation interface constraint formula is used, based on the influence distance of the special interpolation control surface. The value ranges from 0.5 to 2.0 m, at the interface between the strongly weathered layer and the moderately weathered layer. Take a depth of 1.2-2.0m, at the interface between the moderately weathered layer and the slightly weathered layer. Take a depth of 0.8-1.5m, at the interface between the weak interlayer and the surrounding rock strata. The range is 0.5-1.0m. The interpolation accuracy of the abrupt interface is determined by analyzing the variation characteristics of the data on both sides of each interface (when the coefficient of variation is ≥0.3, it is determined to be a discontinuous region).

[0050] Specifically, in step five, the co-kriging interpolation of the mutation interface is considered and implemented using the following formula: The co-kriging interpolation formula is: ,in, Unknown point The interpolated estimates include information on stratum depth and slope elevation; For the coordinate data of the slope surface after backfilling, The number of surface data points participating in the interpolation. The weight coefficients for the surface data points are dimensionless and satisfy the following conditions: ; For preprocessed stratigraphic coordinate data in R language, The number of stratigraphic data points used in the interpolation. The weighting coefficients for the stratigraphic data points are dimensionless and satisfy the following conditions: ; This is the coordinate data of the road cut slope after cold excavation. The number of data points from the excavation face participating in the interpolation. The weighting coefficients for the data points on the excavation face are dimensionless and satisfy the following conditions: .

[0051] Solving for weighting coefficients (supplementing the system of equations and improving upon omitted content in this invention): Weighting coefficients , , The solution equations are determined by satisfying the unbiased estimation condition. Strictly following the solution logic of this invention, which involves "extracting the variance characteristics of the three types of data and their cross-variance characteristics, and introducing Lagrange multipliers ψ1, ψ2, and ψ3 to satisfy the unbiased estimation condition," the solution equations are supplemented and completed to ensure sufficient disclosure. The equation set is as follows: ,in, The variation function characterizing the surface data itself, The variation function characterizing the stratigraphic data itself, The variability function characterizing the excavation face data itself, = This is the cross-variance function of surface data and formation data. = This is the cross-variance function of surface data and excavation face data. = To characterize the cross-variance function of stratigraphic data and excavation face data, The cross-variance function characterizing the surface data and the estimated points. The cross-variance function characterizes the stratigraphic data and the estimated points. The cross-variance function characterizing the excavation face data and the estimated points; the Lagrange multipliers are represented by simplified characters. , , Characterization, This refers to the spatial location of the coordinate data points on the surface of the slope after backfilling. , This refers to the spatial location of the stratigraphic coordinate data points (the locations of two different stratigraphic data points). , This refers to the spatial location of the coordinate data points of the road cut slope after cold excavation (the locations of data points on two different excavation faces). Another spatial location of the surface coordinate data points of the backfilled slope (compared to) (for different surface data points) The spatial location of the unknown estimated point to be interpolated.

[0052] Spatial interpolation control surface generation: Substitute the weight coefficients obtained from the solution into the co-kriging interpolation formula, and combine them with the abrupt interface constraint to perform fusion interpolation on the three types of data to generate a spatial interpolation control surface that integrates stratigraphic information and slope surface information. The interpolation resolution is 0.5m×0.5m, and the interpolation error is ≤0.08m, ensuring the accuracy and rationality of the interpolation results.

[0053] The formula for the interface constraint in case of a sudden change is: ,in, To consider the variation function of the abrupt interface; For the variation function of a continuous region, a spherical model or an exponential model is used; The variation function is for the discontinuous region, and the nugget model is used; The influence distance of the special interpolation control surface is determined by analyzing the variation characteristics of the data on both sides of the special interpolation control surface.

[0054] Step Six: Model Accuracy Evaluation and Output Steps: Accuracy assessment data acquisition: 15 field detection control points were selected (not to be repeated with the control points in steps three and four). The measured three-dimensional coordinates of the control points were measured using a total station and used as the standard data for accuracy assessment. At the same time, the model coordinates of the corresponding control point positions were extracted from the generated three-dimensional geological model to form model coordinate data. The accuracy of both types of data was controlled within ±0.05m.

[0055] Root Mean Square Error Calculation: Based on the deviation data between the measured coordinates and model coordinates of the field control points, the modeling accuracy is evaluated using the Root Mean Square Error (RMSE) calculation formula, as follows: ,in, To assess the number of control points, , , For the first Measured three-dimensional coordinates of each control point , , For the first The model's three-dimensional coordinates for each control point.

[0056] like Figures 1-5 As shown, the drone components in step four include drone 1. A mounting device is provided on the lower surface of drone 1. The mounting device includes a fixing frame 42, which fixes the three-dimensional laser scanning radar 12 to the lower part of drone 1.

[0057] Specifically, the mounting device also includes a mounting bracket 2, which is fixedly mounted on the lower surface of the fuselage of the UAV 1. A transmission gear ring 21 is rotatably connected to the outer surface of the mounting bracket via a bearing. The two transmission gear rings 21 are connected by a synchronous belt and a synchronous pulley. A drive motor 22 is fixedly mounted on the outer surface of the mounting bracket 2 via a support frame. One end of the output shaft of the drive motor 22 drives a transmission gear ring 21 to rotate via a gear. A lifting screw 23 is threadedly connected to the inner wall of the transmission gear ring 21. A push plate 24 is fixedly mounted on the lower end of the lifting screw 23. A push frame 25 is slidably inserted into the outer surface of one end of the push plate 24. A return spring 26 is fixedly mounted on the outer surface of the push frame 25. One end of the return spring 26 is fixedly mounted to the lower surface of the push plate 24.

[0058] Specifically, a lifting frame 3 is slidably inserted into the outer surface of the mounting bracket 2 via a limiting rod. A drive gear 31 is rotatably connected to the upper surface of the lifting frame 3 via a bearing. A rotating sleeve 32 with an arc groove is fixedly installed on the upper surface of the drive gear 31. A plug-in rod 33 with a column is fixedly installed on the lower surface of the push frame 25. One end of the plug-in rod 33 is slidably inserted into the inner wall of the rotating sleeve 32, and the outer surface of the column is slidably inserted into the inner wall of the arc groove. A moving rack 34 is slidably inserted into the upper surface of the lifting frame 3. The moving rack 34 meshes with the drive gear 31. A connecting spring 35 is fixedly installed on the outer surface of the limiting rod of the lifting frame 3. One end of the connecting spring 35 is fixedly installed on the outer surface of the mounting bracket. Both ends of the push plate 24 are slidably inserted into the inner wall of the mounting bracket 2 and slidably sleeved with the outer surface of the limiting rod of the lifting frame 3. The force of the rotating sleeve 32 and the plug-in rod 33 is much less than the supporting elasticity of the connecting spring 35.

[0059] Specifically, the upper surface of the lifting frame 3 is slidably connected to the slider 4 via a slide rail. The outer surface of the slider 4 is fixedly installed on the lower surface of the fixed frame 42. One end of the moving rack 34 is fixedly installed with a linkage rod 41. The outer surface of the linkage rod 41 is slidably connected to the outer surface of the fixed frame 42 via a sleeve. The lower surface of the three-dimensional laser scanning radar 12 is fixedly installed with a mounting base 11. After the multiple fixed frames 42 move together, they clamp the mounting base 11.

[0060] Working principle: The drive motor 22 mounted on the mounting bracket 2 starts, and the gear at the end of its output shaft meshes with a transmission gear ring 21, driving the gear ring to rotate. Since the two transmission gear rings 21 are connected by a synchronous belt and a synchronous pulley, the two transmission gear rings 21 will rotate synchronously. The inner wall of the transmission gear ring 21 is threadedly connected to the lifting screw 23. Therefore, when the transmission gear ring 21 rotates, the lifting screw 23 will drive the push plate 24 to make a vertical downward movement. At this time, the push frame 25, which is slidably inserted with the push plate 24, and the return spring 26 on the push frame 25 will rise and fall together with the push plate 24. During the descent of the push plate 24, the plug rod 33 (with a column) fixed at the lower end of the push frame 25 will descend inside the rotating sleeve 32. When the plug rod 33 is inserted into the rotating sleeve 32, its column slides along the arc groove. Since the arc groove is spiral, the vertical movement of the column will force the rotating sleeve 32 to rotate together with the drive gear 31 at its top. When the drive gear 31 rotates, the moving rack 34 meshing with it will move horizontally along the upper surface of the lifting frame 3. One end of the moving rack 34 is fixedly connected to the linkage rod 41. When the moving rack 34 moves, it will drive the fixed frame 42 to move together through the linkage rod 41. The lower surface of the fixed frame 42 is connected to the lifting frame 3 through the slide rail and the slider 4, ensuring the stability and directionality of its movement. When multiple fixed frames 42 are gathered together, they will tightly clamp the mounting base 11 at the bottom of the three-dimensional laser scanning radar 12, thereby reliably fixing the radar below the UAV 1. After the UAV 1 rises to the set height, the lifting screw 23 continues to descend and then the lifting frame 3, which is slidably inserted on the mounting bracket 2, is pressed down by the push plate 24. The lifting frame 3 and the push frame 25 descend simultaneously, maintaining the insertion of the rotating sleeve 32 and the plug rod 33. The connecting spring 35 is compressed, and the lifting frame 3 drives the three-dimensional laser scanning radar 12, which is held by the fixing frame 42, to descend.

[0061] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for acquiring data on red sandstone roadbed slopes and strata, characterized in that: The process includes step one: pre-construction stratigraphic data acquisition and preprocessing. Drilling points are set up in the slope area of ​​the road cut to be excavated to obtain survey drilling data. The drilling data is then input into the Kriging interpolation R language program. A variogram partitioning fitting strategy is adopted, and different ranges and sill values ​​are set for different weathering zones of red sandstone to generate an initial stratigraphic information model. Step 2: Identification and Marking of Special Interpolation Control Surfaces: The structural surfaces of the cut slope after cold excavation and the surface structural surfaces of the subgrade slope after backfilling are taken as special interpolation control surfaces and marked as constrained interpolation boundaries in three-dimensional space, so that they have a higher weight coefficient than ordinary data points in subsequent interpolation calculations. Step 3: Slope Data Acquisition and Preprocessing during Excavation: During the cold excavation of the red sandstone subgrade, the three-dimensional point cloud data of the excavated slope is acquired using the cold excavation machine-mounted scanning radar. The structural surface model of the road cut slope before backfilling is formed through noise reduction and registration processing. On-site detection control points are set up for error correction. The elevation and planar coordinate information of the dense point cloud in the point cloud model are processed based on Surfer software to restore the actual shape of the slope surface and extract the corresponding coordinate data. Step 4: Data acquisition and 3D reconstruction of backfilled slope: For the backfilled red sandstone roadbed slope, a 3D laser scanning radar (12) on the UAV component is used for scanning. Combined with GPS positioning data and ground control point data, the SFM motion recovery structure algorithm and MFS multi-view stereo matching algorithm are used for joint calculation. GPS data and control point data are used as global optimization constraints to eliminate cumulative errors. A point cloud model of the surface slope of the red sandstone roadbed is constructed. After processing by Surfer, the coordinate data of the surface slope after backfilling is extracted. Step 5: Multi-source heterogeneous data collaborative fusion and 3D modeling steps: Import the backfilled slope surface coordinate data, the R language preprocessed stratum coordinate data, and the cold excavation cut slope coordinate data into the 3D modeling platform. Using the two special interpolation control surfaces marked in Step 2 as constraint boundaries, the three types of data are fused and interpolated using the co-kriging interpolation method that considers abrupt interfaces to generate a spatial interpolation control surface that integrates stratum information and slope surface information. Step Six: Model Accuracy Assessment and Output Step: Based on the deviation data between the field detection control points and the 3D model, the accuracy of this modeling is assessed using the root mean square error calculation formula, and the 3D geological model data after error correction is automatically output.

2. The method for acquiring data on red sandstone roadbed slope and strata according to claim 1, characterized in that: The borehole data processing in step one employs a variogram partitioning fitting strategy, generating an initial formation information model using the variogram calculation formula and the partitioning range setting formula. The formula for calculating the function of variation is: ,in, The value of the variation function represents the spatial distance. The degree of variation between sample points The spatial distance between sample points; The distance is The number of sample point logs; For position The borehole data values ​​at the location include formation depth and thickness parameters; For position Drilling data values ​​at the location; The formulas for setting the partition range and base value are as follows: ,in, The values ​​of the variation function after partitioning; Let be the nugget constant, representing the random variation caused by changes in microstructure and measurement errors, expressed as an experimental variogram function. The intercept is determined at that time; The sill value represents the overall variability of the regionalized variable and is determined by the stable value of the experimental variogram. For variable range, representing the maximum distance of spatial correlation, when > The time variable no longer has spatial correlation; For the spherical model function, the expression is: , ; For different weathering zones (strongly weathered layer, moderately weathered layer, slightly weathered layer) and weak interlayers of red sandstone, range, sill value and nugget constant are set to match their spatial variation characteristics.

3. The method for acquiring data on red sandstone roadbed slope and strata according to claim 2, characterized in that: The special interpolation control surface marking in step two employs a constrained interpolation boundary marking method, using a constrained boundary weight coefficient formula to enhance the weight of the special interpolation control surface. The formula for the constraint boundary weight coefficient is: ,in, These are the weighting coefficients for the special interpolation control surface, and are dimensionless. These are the weighting coefficients for ordinary data points, and are dimensionless. As the interface enhancement factor, the optimal value was determined through cross-validation experiments based on the stability requirements of the red sandstone slope structure. The marking method for special interpolation control surfaces in three-dimensional space is as follows: In the interpolation mesh system, the mesh nodes belonging to the special interpolation control surfaces are marked as constraint nodes. During the co-Kriging equations solution process, the weight coefficients of the constraint nodes are forced to execute the above formula to ensure that the key engineering interfaces are accurately restored in the interpolation results.

4. The method for acquiring data on red sandstone roadbed slope and strata according to claim 3, characterized in that: The point cloud data processing of the excavated slope surface in step three is performed using point cloud error correction formulas and Surfer meshing formulas: The point cloud error correction formula is: ,in, These are the point cloud coordinates after error correction, including three-dimensional coordinate components; These are the original point cloud coordinates, containing three-dimensional coordinate components; To ensure the accuracy of calibration, control points are evenly distributed to determine the number of control points for on-site testing. For the first The measured coordinates of each on-site control point were obtained by total station or GPS-RTK measurement. For the point cloud model corresponding to the first The coordinates of the control points; The Surfer meshing formula is: ,in, For grid points Elevation value at the location; The number of point clouds involved in the interpolation is determined based on the search radius; For the first The weight coefficients of each point cloud satisfy the following conditions: The weighting coefficients are determined by the Kriging interpolation method and are related to the spatial distance between point clouds and the variogram. For the first Elevation values ​​of point clouds; The Kriging gridding method was adopted, and the grid spacing, search radius and anisotropy ratio were reasonably set according to the accuracy requirements of the surface layer of the red sandstone slope to ensure that the gridded digital elevation model can accurately reproduce the actual shape of the slope surface.

5. The method for acquiring data on red sandstone roadbed slope and strata according to claim 4, characterized in that: The three-dimensional reconstruction of the UAV (1) point cloud data in step four is processed using the SFM-MFS joint solution formula and the global optimization constraint formula: The formula for SFM motion recovery structure is: ,in, For the first The rotation matrix of the image has dimensions of... , indicating the camera orientation; For the first The translation vector of the image has dimension . , indicating the camera position; For the first The coordinates of a three-dimensional point, with dimensions of . ; For the first The three-dimensional point at the th t The coordinates of the image points on the image are in dimension [dimensional value]. ; This is a camera projection function that projects 3D points onto a 2D image plane. For the number of images, This represents the number of three-dimensional points. Using Euclidean distance, minimize reprojection error to optimize camera parameters and 3D point coordinates; The MFS multi-view stereo matching formula is: ,in, For pixels disparity value; For pixels The set of neighboring pixels; This serves as an image index identifier, used to distinguish different neighboring images. This is a robust cost function used to suppress the influence of outliers; For the first The grayscale value of the image; The grayscale value is the reference image. For parallax The corresponding mapping function maps pixels Mapped to the corresponding location in the neighboring image; The global optimization constraint formula is: ,in, The coordinates of the three-dimensional point are optimized using GPS and ground control point constraints, with dimensions of [dimension number missing]. ; The original 3D point coordinates calculated by SFM, with dimensions of ; The Kalman gain matrix has a dimension of . The covariance matrix of the noise is calculated by measuring the noise using GPS and estimating the noise using SFM. This is a GPS coordinate transformation function that converts the coordinates of a 3D point to coordinates in the GPS coordinate system. These are the measured GPS coordinates.

6. The method for acquiring data on red sandstone roadbed slope and strata according to claim 5, characterized in that: The multi-source heterogeneous data fusion interpolation in step five uses a co-kriging interpolation formula that considers abrupt interfaces to process the three types of data: The co-kriging interpolation formula is: ,in, Unknown point The interpolated estimates include information on stratum depth and slope elevation; For the coordinate data of the slope surface after backfilling, The number of surface data points participating in the interpolation. The weight coefficients for the surface data points are dimensionless and satisfy the following conditions: ; For preprocessed stratigraphic coordinate data in R language, The number of stratigraphic data points used in the interpolation. The weighting coefficients for the stratigraphic data points are dimensionless and satisfy the following conditions: ; This is the coordinate data of the road cut slope after cold excavation. The number of data points from the excavation face participating in the interpolation. The weighting coefficients for the data points on the excavation face are dimensionless and satisfy the following conditions: ; The weighting coefficients are determined by a solution logic that satisfies the unbiased estimation condition. They are combined with the variation functions of surface data, stratum data, and excavation face data, as well as the cross variation functions between them, and solved using Lagrange multipliers to ensure the accuracy and rationality of the interpolation results without the need for specific equation sets to limit them. The weighting coefficients are determined by a solution logic that satisfies the unbiased estimation condition. They are combined with the variation functions of surface data, stratum data, and excavation face data, as well as the cross variation functions between them, and solved using Lagrange multipliers to ensure the accuracy and rationality of the interpolation results without the need for specific equation sets to limit them. The variogram and cross-variogram are represented by simplified characters: The variation function characterizing the surface data itself, The variation function characterizing the stratigraphic data itself, The variability function characterizing the excavation face data itself, (and (Equal) represents the cross-variance function of surface data and stratigraphic data. (and (Equal) represents the cross-variance function of surface data and excavation face data. (and (Equal) is a cross-variance function characterizing the stratigraphic data and the excavation face data. The cross-variance function characterizing the surface data and the estimated points. The cross-variance function characterizes the stratigraphic data and the estimated points. The cross-variance function characterizing the excavation face data and the estimated points; the Lagrange multipliers are represented by simplified characters. , , Characterization, used to satisfy the unbiased estimation condition; The formula for the interface constraint in case of a sudden change is: ,in, To consider the variation function of the abrupt interface; For the variation function of a continuous region, a spherical model or an exponential model is used; The variation function is for the discontinuous region, and the nugget model is used; The influence distance of the special interpolation control surface is determined by analyzing the variation characteristics of the data on both sides of the special interpolation control surface.

7. A method for acquiring data on red sandstone roadbed slope and strata according to claim 6, characterized in that: The formula for calculating the root mean square error (RMSE) in step six is ​​as follows: ,in, To assess the number of control points, , , For the first Measured three-dimensional coordinates of each control point , , For the first The model's three-dimensional coordinates for each control point.

8. The method for acquiring data on red sandstone roadbed slope and strata according to claim 7, characterized in that: The drone component in step four includes a drone (1), and the lower surface of the drone (1) is provided with a mounting device. The mounting device includes a fixing frame (42), which fixes the three-dimensional laser scanning radar (12) below the drone (1). The mounting device also includes a mounting bracket (2), which is fixedly mounted on the lower surface of the fuselage of the UAV (1). The outer surface of the mounting bracket is rotatably connected to a transmission gear ring (21) via a bearing. The two transmission gear rings (21) are connected by a synchronous belt and a synchronous pulley. The outer surface of the mounting bracket (2) is fixedly mounted with a drive motor (22) via a support frame. One end of the output shaft of the drive motor (22) drives one of the transmission gear rings (21) to rotate via a gear. The inner wall of the transmission gear ring (21) is threaded with a lifting screw (23). The lower end of the lifting screw (23) is fixedly mounted with a push plate (24). One end of the outer surface of the push plate (24) is slidably inserted with a push frame (25). The outer surface of the push frame (25) is fixedly mounted with a return spring (26). One end of the return spring (26) is fixedly mounted with the lower surface of the push plate (24).

9. A method for acquiring data on red sandstone roadbed slope and strata according to claim 8, characterized in that: The outer surface of the mounting bracket (2) is slidably connected to a lifting frame (3) via a limiting rod. The upper surface of the lifting frame (3) is rotatably connected to a drive gear (31) via a bearing. A rotating sleeve (32) with an arc groove is fixedly installed on the upper surface of the drive gear (31). A plug-in rod (33) with a column is fixedly installed on the lower surface of the push frame (25). One end of the plug-in rod (33) is slidably inserted into the inner wall of the rotating sleeve (32), and the outer surface of the column is slidably connected to the inner wall of the rotating sleeve (32). The inner wall of the groove is slidably inserted, and a movable rack (34) is slidably inserted on the upper surface of the lifting frame (3). The movable rack (34) meshes with the drive gear (31). A connecting spring (35) is fixedly installed on the outer surface of the limiting rod of the lifting frame (3). One end of the connecting spring (35) is fixedly installed on the outer surface of the mounting frame. Both ends of the push plate (24) are slidably inserted into the inner wall of the mounting bracket (2) and slidably sleeved with the outer surface of the limiting rod of the lifting frame (3).

10. A method for acquiring data on red sandstone roadbed slope and strata according to claim 9, characterized in that: The upper surface of the lifting frame (3) is slidably connected to a slider (4) via a slide rail. The outer surface of the slider (4) is fixedly installed on the lower surface of the fixed frame (42). One end of the moving rack (34) is fixedly installed with a linkage rod (41). The outer surface of the linkage rod (41) is slidably connected to the outer surface of the fixed frame (42) via a sleeve. The lower surface of the three-dimensional laser scanning radar (12) is fixedly installed with a mounting base (11). After the multiple fixed frames (42) move together, they clamp the mounting base (11).

Images