Visualization method and system for oversized in-situ filling space under composite roof

Abstract

The invention discloses a visualization method and system for an oversized in-situ filling space under a composite roof, and belongs to the technical field of information processing. The method comprises the following steps: performing multi-angle data acquisition on a filling space under a composite roof structure to obtain original point cloud data; obtaining complete three-dimensional point cloud data based on the original point cloud data; preprocessing the three-dimensional point cloud data to obtain standard point cloud data; performing boundary contour recognition based on the standard point cloud data to obtain space boundary contour data; obtaining a spatial three-dimensional geometric model based on the spatial boundary contour data; performing section slicing based on the space coordinate range of the space three-dimensional geometric model, and obtaining size distribution data based on the obtained contour size parameters of the section; obtaining a total volume and a segmented volume of the filling space based on the size distribution data; and obtaining a three-dimensional visual model of the filling space based on the total volume and the segmented volume. The method provides reliable technical support for mine filling design and resource evaluation.

Description

Technical Field

[0001] This invention belongs to the field of information processing technology, specifically relating to a visualization method and system for a composite top-and-bottom ultra-large in-situ filling space. Background Technology

[0002] In the fields of mining engineering and geological exploration, accurately understanding the geometric characteristics of underground spaces is crucial for safe construction and resource utilization. This is especially true for ultra-large in-situ filling spaces under composite roof structures, where the size and shape directly affect the design and implementation of filling schemes. However, existing measurement and visualization methods often fall short of the requirements for high precision and comprehensiveness in complex geological environments. Current technologies rely heavily on traditional surveying tools or single two-dimensional imaging methods, which are ill-suited to the diversity and dynamic changes in spatial morphology under composite roofs, particularly in complex boundary identification and three-dimensional geometric reconstruction. These methods often fail to accurately capture changes in the boundary contours within the space and have low measurement accuracy for irregular cross-sectional dimensions, resulting in a lack of reliable spatial parameters for filling scheme design.

[0003] The core challenge lies in effectively identifying and extracting the complex boundary contours of ultra-large filling spaces, and on this basis, accurately measuring the internal dimensions of the space. Due to the irregular spatial morphology under the composite roof structure, automatic boundary contour identification becomes the primary technical challenge. For example, in actual mine operations, the boundary of the filling space may exhibit discontinuous curved surface features due to rock fractures or ore body mining. Traditional scanning techniques struggle to accurately distinguish the boundary from noise interference, thus affecting the reliability of subsequent dimensional measurements. The inadequacy of boundary identification further leads to distortion of spatial geometric parameters during 3D reconstruction, particularly in generating cross-sectional dimension distribution maps, making it difficult to accurately reflect the true shape of the space. For instance, in the filling design of a certain mining area, boundary identification errors caused cross-sectional dimension deviations, which in turn affected the estimation of filling material usage, resulting in resource waste and construction risks.

[0004] Therefore, the key issue is how to automatically identify the boundary contour of the ultra-large filling space and accurately measure the inner diameter of the composite top plate structure, and generate accurate three-dimensional spatial geometric parameters. Summary of the Invention

[0005] This invention aims to address the shortcomings of existing technologies by proposing a visualization method for a composite top-and-bottom ultra-large in-situ filling space. The method obtains the total volume and segmented volume of the filling space through acquired three-dimensional point cloud data, and then obtains a three-dimensional visualization model of the filling space based on the total volume and segmented volume.

[0006] To achieve the above objectives, the present invention provides the following solution: a visualization method for a composite subsurface ultra-large in-situ filling space, comprising the following steps:

[0007] S1. Use laser scanning equipment to collect data from multiple angles of the filling space under the composite roof structure to obtain raw point cloud data; and obtain complete three-dimensional point cloud data based on the raw point cloud data.

[0008] S2. Preprocess the three-dimensional point cloud data to obtain standard point cloud data;

[0009] S3. Based on the standard point cloud data, perform boundary contour recognition to obtain complete spatial boundary contour data;

[0010] S4. Obtain a three-dimensional spatial geometric model based on the spatial boundary contour data;

[0011] S5. Based on the spatial coordinate range of the three-dimensional geometric model, perform cross-sectional slicing, and based on the obtained profile dimension parameters of the cross-section, obtain dimension distribution data.

[0012] S6. Based on the size distribution data, obtain the total volume and segmented volume of the filling space;

[0013] S7. Based on the total volume and segmented volumes, obtain a three-dimensional visualization model of the filled space.

[0014] More preferably, S1 includes the following steps:

[0015] S11. Use laser scanning equipment to collect data from multiple angles of the filling space under the composite roof structure to obtain raw point cloud data;

[0016] S12. Perform point cloud registration on the original point cloud data to obtain registered point cloud data with unified coordinates;

[0017] S13. Spatial segmentation is performed on the registered point cloud data to obtain voxelized point cloud data;

[0018] S14. Surface reconstruction is performed based on the voxelized point cloud data to obtain three-dimensional surface data;

[0019] S17. Repair the voids in the three-dimensional surface data to obtain the three-dimensional point cloud data.

[0020] More preferably, S2 includes the following steps:

[0021] S21. Obtain the spatial coordinates and reflection intensity of the three-dimensional point cloud data, calculate the statistical characteristics of each three-dimensional point cloud based on the spatial coordinates and reflection intensity, and obtain an initial feature point cloud set.

[0022] S22. Calculate the neighborhood statistics for each initial feature point cloud to identify noise points;

[0023] S23. When the neighborhood statistics of a noise point are greater than a preset neighborhood statistics threshold, the noise point is removed to obtain a preliminary denoised point cloud.

[0024] S24. Calculate the point cloud density of the preliminary denoised point cloud. When the point cloud density is lower than the preset point cloud density threshold, use the difference method to supplement the point cloud to obtain the supplemented point cloud.

[0025] S25. Recalculate the spatial coordinates and reflection intensity of the supplemented point cloud, and repeat steps S21-S24 until all supplemented point clouds meet the neighborhood statistics and point cloud density verification to obtain the standard point cloud data.

[0026] More preferably, S3 includes the following steps:

[0027] S31. Calculate the local neighborhood vector of each point cloud based on the standard point cloud data to obtain the point cloud normal vector;

[0028] S32. Calculate the curvature of each point cloud based on the point cloud normal vector to obtain curvature distribution data;

[0029] S33. When the curvature distribution data is greater than a preset curvature threshold, the curvature distribution data is marked as a boundary point;

[0030] S34. Based on the boundary points, determine the preliminary spatial boundary outline;

[0031] S35. The preliminary spatial boundary contour is smoothed and reconstructed in three dimensions to obtain the spatial boundary contour data.

[0032] More preferably, S4 includes the following steps:

[0033] S41. Generate a structured point set based on the spatial boundary contour data to obtain an initial point cloud model;

[0034] S42. Generate a set of triangular facets based on the initial point cloud model to obtain an initial closed surface;

[0035] S43. Perform a closure test on the initial closed surface. If there are voids, fill the voids to obtain a complete closed surface.

[0036] S44. Adjust the distribution of face patches based on the complete closed surface to obtain optimized triangular face patches;

[0037] S45. Based on the optimized triangular facets, the spatial three-dimensional geometric model is obtained.

[0038] More preferably, S5 includes the following steps:

[0039] S51. Based on the spatial coordinate range of the aforementioned three-dimensional geometric model, obtain the bounding box coordinates;

[0040] S52. Based on the bounding box coordinates, generate equally spaced cross sections in the horizontal and vertical directions using a uniform sampling method to obtain a set of cross sections;

[0041] S53. Obtain the contour dimension parameters based on the set of cross sections;

[0042] S54. Based on the contour size parameters, obtain the morphological feature vector;

[0043] S55. When the Euclidean distance of the morphological feature vector is greater than a preset distance threshold, the cross section is marked as an abnormal cross section, thereby obtaining a set of abnormal cross sections.

[0044] S56. The abnormal cross-section set is removed, and the removed contour data is supplemented by interpolation method to obtain the complete size distribution data.

[0045] More preferably, S6 includes the following steps:

[0046] S61. Obtain standard cross-sectional data based on the size distribution data;

[0047] S62. Based on the standard cross-section data, obtain the local volume corresponding to each cross-section, and then obtain the segmented volume;

[0048] S63. Integrate the segmented volumes to obtain the total volume of the filling space.

[0049] This invention also provides a visualization system for a composite under-top ultra-large in-situ filling space, comprising:

[0050] The 3D point cloud data acquisition module is used to collect data from the filling space under the composite roof structure from multiple angles using a laser scanning device to obtain raw point cloud data; and to obtain complete 3D point cloud data based on the raw point cloud data.

[0051] The point cloud preprocessing module is used to preprocess the three-dimensional point cloud data to obtain standard point cloud data.

[0052] The contour recognition module is used to perform boundary contour recognition based on the standard point cloud data to obtain complete spatial boundary contour data.

[0053] A spatial three-dimensional geometric model construction module is used to obtain a spatial three-dimensional geometric model based on the spatial boundary contour data;

[0054] The size distribution data acquisition module is used to slice the cross section based on the spatial coordinate range of the three-dimensional geometric model, and obtain size distribution data based on the obtained profile size parameters of the cross section.

[0055] The volume calculation module is used to obtain the total volume and segmented volume of the filling space based on the size distribution data;

[0056] A visualization model building module is used to obtain a three-dimensional visualization model of the filled space based on the total volume and segmented volumes.

[0057] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0058] 1. High-precision 3D reconstruction capability: Through multi-angle laser scanning and point cloud registration technology, high-precision, full-coverage data acquisition and 3D reconstruction of complex filling space under composite roof is achieved. The spatial reconstruction accuracy can reach the millimeter level, which is significantly better than traditional 2D surveying methods.

[0059] 2. Intelligent boundary recognition and noise processing: The boundary recognition algorithm based on curvature analysis and neighborhood statistics can effectively distinguish between real boundaries and noise interference. It is especially suitable for filling spaces with complex geological structures and irregular surfaces, improving the accuracy and robustness of boundary extraction.

[0060] 3. Complete geometric modeling and volume calculation: Through cross-sectional slicing and contour dimension analysis, the accurate calculation of segmented volume and total volume of the filling space is realized, providing reliable data support for the estimation of filling material usage and resource assessment, and avoiding resource waste or construction risks caused by dimensional deviations.

[0061] 4. Strong visualization and decision support capabilities: The generated 3D visualization model not only intuitively displays the spatial form, but also supports dynamic simulation and multi-dimensional analysis, which helps in mine engineering design, safety assessment and construction plan optimization, and improves the scientific nature and efficiency of engineering management.

[0062] 5. High system integration and strong automation: From data acquisition, preprocessing, boundary identification, modeling to volume calculation and visualization, the entire process achieves a high degree of automation and modular integration, reducing the need for manual intervention and making it suitable for large-scale, multi-scenario engineering applications.

[0063] 6. Strong adaptability and good scalability: This method and system can be widely used in mining, tunnels, underground engineering and other fields, and is especially suitable for complex geological environments such as composite roofs and irregular spaces. It has good engineering applicability and technical promotion value. Attached Figure Description

[0064] To more clearly illustrate the technical solution of the present invention, the drawings used in the embodiments are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0065] Figure 1 This is a schematic diagram of a visualization method for a composite top-and-bottom ultra-large in-situ filling space provided in an embodiment of the present invention. Detailed Implementation

[0066] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0067] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0068] Example 1:

[0069] Combined with appendix Figure 1 This embodiment provides a visualization method for a large in-situ filling space under a composite structure, specifically including the following steps:

[0070] S1. Use laser scanning equipment to collect data from multiple angles of the filling space under the composite roof structure to obtain raw point cloud data; and obtain complete three-dimensional point cloud data based on the raw point cloud data.

[0071] A further implementation involves S1 including the following steps:

[0072] S11. Use laser scanning equipment to collect data from multiple angles of the filling space under the composite roof structure to obtain raw point cloud data.

[0073] In this embodiment, the laser scanning device used is a FaroFocus S350, with a scanning range covering 0.6 to 350 meters and an accuracy of ±1 mm. For the space under the composite roof, the scanner collects data from multiple angles at 30-degree intervals. Each point cloud contains approximately 5 million points, covering the entire filling space. The data includes XYZ coordinates and reflection intensity.

[0074] S12. Perform point cloud registration on the original point cloud data to obtain registered point cloud data with unified coordinates.

[0075] Specifically, the original point cloud data is first coarsely registered.

[0076] Step 1: Take two frames of original point cloud data as the source point cloud data and the target point cloud data respectively. Through coordinate transformation, the coordinate system of the source point cloud data is transformed to the coordinate system of the target point cloud data through scaling, rotation, projection and translation.

[0077] In this embodiment, the coordinate system transformation is achieved through a unified transformation matrix T:

[0078]

[0079] in, Represents the basic transformations of scaling, rotation, and shearing maps; [a 41 a 42 a 43 [ represents translation along the X, Y, and Z directions; a] 44 This is a global scaling transformation; This is a projection transformation.

[0080] Step 2: Construct the target transformation matrix based on the unified transformation matrix.

[0081] Specifically, m reference points are selected, and a coordinate transformation overdetermined system is constructed using (x,y,z) information collected by a laser scanning device, the transformed (x,y,z) information, and a unified transformation matrix.

[0082] AT = B;

[0083] In the formula, A is a matrix composed of (x,y,z) information collected by the laser scanning device; B is a matrix composed of (x,y,z) information after transformation.

[0084] To solve an overdetermined coordinate transformation system, and to obtain the most approximate result, a residual square function is introduced:

[0085] S(T) = ||AT-B||.

[0086] when When S(T) reaches its minimum value, then:

[0087]

[0088] in, The target transformation matrix.

[0089] When the value of S(T) is differentiated to its minimum, we have:

[0090]

[0091] The superscript T indicates transpose.

[0092] When A T When A is a non-singular matrix, the target transformation matrix has a unique solution, thus yielding the target transformation matrix:

[0093]

[0094] In the formula, the superscript -1 indicates taking the inverse.

[0095] Step 3: Transform the source point cloud data into the target point cloud data using the target transformation matrix.

[0096]

[0097] In the formula, P r Represents target point cloud data; P s This represents the source point cloud data.

[0098] Next, fine registration is performed on the original point cloud data after coarse registration.

[0099] Step 1: Key point selection; In this embodiment, key points are mainly selected from specific corner locations. Then, the key points are stored in the target point cloud data P. r Neutralize the original point cloud data P t middle.

[0100] Step 2: Search for the corresponding point of each point in the registered point cloud data in the target point cloud data using the shortest distance.

[0101] Step 3: Construct the transformation parameters: rotation matrix R* and translation vector t*.

[0102]

[0103] In the formula, P t j These represent the matrix representations of data point i in the target point cloud data and data point j in the original point cloud data, respectively.

[0104] Step 4: Obtain the registered point cloud data using the two conversion parameters mentioned above.

[0105] P m =R * P r +t * ;

[0106] In the formula, P m This represents the point cloud data after registration.

[0107] S13. Spatial segmentation is performed on the registered point cloud data to obtain voxelized point cloud data.

[0108] In this embodiment, the space is divided into regular grids using the voxel grid method, with the voxel size set to 5 cm, to generate voxelized point cloud data.

[0109] S14. Surface reconstruction is performed based on voxelized point cloud data to obtain three-dimensional surface data.

[0110] In this embodiment, a three-dimensional surface model is generated using the Poisson surface reconstruction algorithm, and the surface curvature is estimated based on the point cloud normal vectors to generate smooth three-dimensional surface data.

[0111] S17. Repair the voids in the three-dimensional surface data to obtain three-dimensional point cloud data.

[0112] The reconstructed surface model clearly reflects the uneven shape of the filling space under the top plate, with an accuracy of ±5 mm. If there are voids in the surface model, such as data loss due to occlusion, the voids can be repaired using spline interpolation algorithms.

[0113] S2. Preprocess the 3D point cloud data to obtain standard point cloud data.

[0114] A further implementation involves S2 including the following steps:

[0115] S21. Obtain the spatial coordinates and reflection intensity of the 3D point cloud data, calculate the statistical features of each 3D point cloud based on the spatial coordinates and reflection intensity, and obtain the initial feature point cloud set.

[0116] In this embodiment, the three-dimensional coordinates and reflection intensity values ​​of each point are directly extracted using a laser scanning device. In the filling space scenario under the composite roof slab, the laser scanner scans at a rate of 1 million points per second, acquiring point coordinates with millimeter-level accuracy. Reflection intensity reflects the reflective characteristics of the material surface; for example, concrete roof slabs have higher reflection intensity, while loose filling materials have lower intensity. The initial feature point cloud set includes the coordinates, intensity, and statistical information of neighboring points, such as the number and density of neighboring points.

[0117] S22. Calculate the neighborhood statistics for each initial feature point cloud to identify noise points.

[0118] Specifically, when calculating the neighborhood statistics of each initial feature point cloud using statistical filtering, the average distance and standard deviation of the local point cloud are calculated based on the k nearest neighbors of the point (k=20 in this embodiment). If the average distance from a point to its neighboring points is 5 mm and the standard deviation is 2 mm, and the preset threshold is the average distance ± 2 times the standard deviation, then points outside this range are considered noise points.

[0119] S23. When the neighborhood statistics of a noise point are greater than the preset neighborhood statistics threshold, the noise point is removed to obtain a preliminary denoised point cloud.

[0120] S24. Calculate the point cloud density of the initial denoised point cloud. When the point cloud density is lower than the preset point cloud density threshold, use the difference method to supplement the point cloud to obtain the supplemented point cloud.

[0121] When calculating point cloud density, the space can be divided into voxels of 1 cubic centimeter, and the number of points within each voxel can be counted. If the number of points within a voxel is less than 5, which is below a preset threshold of 10, it is considered a low-density region. An interpolation method, such as weighted average interpolation based on neighboring points, is used to form a supplementary point cloud, thus obtaining the supplemented point cloud.

[0122] S25. Recalculate the spatial coordinates and reflection intensity of the supplemented point cloud, and repeat steps S21-S24 until all supplemented point clouds meet the neighborhood statistics and point cloud density verification to obtain standard point cloud data.

[0123] When recalculating the spatial coordinates and reflection intensity of the supplementary point cloud, the positions of the new points are adjusted using a weighted average. For example, the coordinates of the new points are calculated based on the distance weighted average of the eight neighboring points to ensure smooth integration with the surrounding point cloud. After optimizing the point cloud data, the data quality can be verified by statistically analyzing the average point spacing (e.g., 2 mm) and performing integrity checks (coverage up to 98%), generating the final denoised point cloud.

[0124] Data standardization normalizes point cloud coordinates to a range of 0 to 1, and reflectsance intensity to a standard normal distribution. For example, the coordinate range is mapped from (-10 meters, 10 meters) to (0, 1), and the intensity values ​​are converted from (0, 255) to a distribution with a mean of 0 and a standard deviation of 1. Standardized point cloud data can be used for subsequent modeling to ensure data consistency.

[0125] S3. Based on standard point cloud data, perform boundary contour recognition to obtain complete spatial boundary contour data.

[0126] Further implementation involves S3 including the following steps:

[0127] S31. Calculate the local neighborhood vector of each point cloud based on the standard point cloud data to obtain the point cloud normal vector.

[0128] When calculating the set of point cloud normal vectors, the local planar orientation can be determined by analyzing the positional relationships of points in the surrounding neighborhood of each point. For example, for a point in the point cloud of a rock sample, a neighborhood with a radius of 0.5 mm can be selected, and the direction of the point cloud normal vector can be calculated based on the spatial distribution of points within the neighborhood. This method can effectively characterize the local geometric features of the point cloud, providing a foundation for subsequent curvature analysis.

[0129] S32. Calculate the curvature of each point cloud based on the point cloud normal vector to obtain curvature distribution data.

[0130] S33. When the curvature distribution data is greater than the preset curvature threshold, the curvature distribution data is marked as a boundary point.

[0131] For point clouds of rock samples, curvature values ​​reflect the degree of surface unevenness. If a point has a curvature value of 0.8, which is higher than a preset threshold of 0.5, it is marked as a potential boundary point. This method helps identify geometric changes on the sample surface by quantifying the curvature distribution. For example, when generating a boundary point dataset, edge points of the rock sample can be filtered out by marking high-curvature points. Assuming the curvature values ​​of the sample edge region are concentrated between 0.7 and 0.9, these points are marked to form a boundary point dataset. This method can effectively separate the boundary regions of the sample, facilitating subsequent contour extraction.

[0132] S34. Based on the boundary points, determine the preliminary spatial boundary outline.

[0133] Boundary extraction algorithms can construct preliminary spatial boundary contours by connecting adjacent boundary points. For example, based on a boundary point dataset of rock samples, the nearest neighbor algorithm can be used to connect adjacent points to form a closed contour line. This method can quickly delineate the edges of samples and is suitable for the analysis of complex geological structures.

[0134] S35. Smooth the preliminary spatial boundary contour and perform three-dimensional reconstruction to obtain spatial boundary contour data.

[0135] For points with abrupt curvature changes in the initial spatial boundary contour, a weighted average method is used for smoothing, making the contour line more consistent with the actual sample morphology. This method reduces boundary jaggedness and improves contour accuracy. Using 3D reconstruction technology, based on the smoothed spatial boundary contour, the interface between the filling space and the surrounding rock is generated. Assuming the rock sample contains cavities, the boundary between the cavities and the surrounding rock is delineated using contour lines, and the 3D interface is reconstructed. This method clearly characterizes the internal structure of the sample and is suitable for geological modeling. The reliability of the final spatial boundary contour can be ensured by verifying the continuity and integrity of the contour lines. For example, checking whether the rock sample contour is closed verifies whether the interface accurately reflects the actual boundary between the cavity and the surrounding rock. This method enhances the application value of point cloud data and is suitable for scenarios such as geological exploration.

[0136] S4. Obtain a spatial three-dimensional geometric model based on spatial boundary contour data.

[0137] Further implementation involves S4 including the following steps:

[0138] S41. Generate a structured point set based on the spatial boundary contour data to obtain the initial point cloud model.

[0139] When generating a structured point set through point cloud preprocessing, the ore body sample can be scanned using a stereomicroscope to obtain raw point cloud data containing millions of points. Preferably, downsampling is performed first, reducing the point cloud density from 1000 points per cubic centimeter to 100 points per cubic centimeter, preserving key geometric features and reducing subsequent computation. Next, the kd-tree algorithm is used for spatial segmentation to form a regular grid structure, resulting in the initial point cloud model.

[0140] S42. Generate a set of triangular facets based on the initial point cloud model to obtain the initial closed surface.

[0141] This embodiment generates a set of triangular facets using the Delaunay triangulation algorithm, and constructs a triangular mesh with point cloud vertices as nodes based on the point cloud data in the initial point cloud model.

[0142] For example, for a point cloud of a 10-centimeter square rock sample, approximately 5,000 triangular patches are generated, ensuring that the side length of each triangle does not exceed 2 millimeters. This method guarantees the geometric accuracy of the initial closed surface, providing a reliable foundation for subsequent surface optimization.

[0143] S43. Perform a closure test on the initial closed surface. If there are voids, fill the voids to obtain a complete closed surface.

[0144] S44. Adjust the distribution of face patches based on the complete closed surface to obtain optimized triangular face patches.

[0145] If 10 unconnected boundary edges are detected at the edge of the surface, it indicates the presence of holes. A local point cloud interpolation method is used to generate new points based on the geometric properties of neighboring points to fill the holes until all boundary edges are closed, forming a complete surface. This method ensures the topological integrity of the surface.

[0146] S45. A spatial three-dimensional geometric model is obtained based on optimized triangular facets.

[0147] The geometric properties of triangular facets are calculated by analyzing the area and angle distribution of each triangle. For example, a minimum angle threshold of 30 degrees is set; if a triangle's angle is less than this value, it is marked as a low-quality facet. Then, the Laplacian smoothing algorithm is used to adjust vertex positions, making the triangle angle distribution more uniform. After optimization, the minimum angle of the triangle is increased to 40 degrees. This smoothing process significantly improves the surface quality.

[0148] When extracting geometric features to determine the quality of facets, the consistency of the normal vectors of triangular facets can be statistically analyzed, with a threshold of 15 degrees for the angle between the normal vectors. If the angle between the normal vectors of adjacent facets within a certain region exceeds 15 degrees, the vertex positions are iteratively adjusted until the angle meets the requirement. This method can effectively reduce local abrupt changes on the surface and improve model stability.

[0149] After constructing the 3D geometric model, the Catmull-Clark surface subdivision algorithm is used to subdivide the triangular facets into smaller units, improving the smoothness of the surface. For example, the original surface contains 5,000 facets, which increases to 20,000 after subdivision, resulting in a more natural surface transition.

[0150] S5. Based on the spatial coordinate range of the three-dimensional geometric model, perform cross-sectional slicing, and based on the obtained profile dimension parameters of the cross-section, obtain dimension distribution data.

[0151] Further implementation involves S5 including the following steps:

[0152] S51. Based on the spatial coordinate range of the three-dimensional geometric model, obtain the bounding box coordinates.

[0153] To obtain the spatial coordinate range of a 3D geometric model, the minimum and maximum coordinate values ​​of the model in 3D space are first determined using imaging data from a stereomicroscope, combined with depth information. Assuming the model is an irregular geometric shape, the coordinate range obtained after scanning is 0 to 100 mm along the x-axis, 0 to 80 mm along the y-axis, and 0 to 50 mm along the z-axis. Based on this, a bounding box is constructed to define the external contour of the model, facilitating subsequent sampling and analysis. This method clearly defines the spatial range of the model, providing a foundation for cross-section generation.

[0154] S52. Based on the bounding box coordinates, a uniform sampling method is used to generate equally spaced cross sections in the horizontal and vertical directions to obtain a set of cross sections.

[0155] A uniform sampling method is used to generate equally spaced cross-sections. Within the bounding box, samples are generated at fixed intervals along the horizontal direction (xy plane) and the vertical direction (z-axis). For example, a cross-section is generated every 5 mm in the horizontal direction and every 2 mm in the vertical direction, resulting in a set of 20 horizontal cross-sections and 25 vertical cross-sections. This uniform sampling method comprehensively covers the geometric features of the model, ensuring the accuracy of subsequent intersection line calculations.

[0156] S53. Obtain contour dimension parameters based on cross-section set.

[0157] The intersection lines between the cross-sections and the 3D geometric model are calculated using a polygon approximation algorithm, and the intersection points of each cross-section and the model are connected to form a closed polygon. For example, a horizontal cross-section intersects the model to form a set of points, which are then fitted into a polygonal outline with a perimeter of 150 mm and an area of ​​800 mm². 2 These contour dimension parameters provide a quantitative basis for subsequent morphological analysis.

[0158] S54. Based on the contour size parameters, obtain the morphological feature vector.

[0159] The feature vectors of the cross-section shape are extracted by principal component analysis algorithm. The polygonal contour of each cross-section is decomposed to obtain the feature vectors of the principal direction.

[0160] S55. When the Euclidean distance of the morphological feature vector is greater than the preset distance threshold, the cross section is marked as an abnormal cross section, and thus a set of abnormal cross sections is obtained.

[0161] If the feature vector of a cross section indicates that its main extension direction deviates from the expected angle by 10 degrees, and the Euclidean distance is 2.5, exceeding the threshold of 1.5, then it is marked as an abnormal cross section. This method can effectively identify cross sections with abnormal shapes, facilitating subsequent processing.

[0162] S56. Remove abnormal cross-sections and supplement the removed contour data using interpolation methods to obtain complete size distribution data.

[0163] In one possible implementation, spline interpolation is used to reconstruct missing contours for the interpolation supplementation of a set of abnormal cross sections. For example, if the contour data of a missing portion of an abnormal cross section is used, continuous contour curves are generated by interpolating the contour points of adjacent cross sections, thus restoring the complete size distribution. This interpolation method can ensure the continuity and consistency of the data.

[0164] S6. Obtain the total volume and segmented volume of the filling space based on the size distribution data.

[0165] Further implementation involves S6 including the following steps:

[0166] S61. Obtain cross-sectional dimension distribution data from the dimension distribution data, and preprocess the data using a stereomicroscopic algorithm to obtain standardized cross-sectional data.

[0167] S62. Based on the standard cross-sectional data, obtain the local volume corresponding to each cross-section, and then obtain the segmented volume.

[0168] Local volumes are calculated using stereo integrals. For each cross section in the standardized cross section dataset, and considering the spacing between adjacent cross sections (e.g., 0.5 mm), the volume contribution between cross sections is estimated using stereo integrals.

[0169] S63. Integrate the segmented volumes to obtain the total volume of the filled space.

[0170] S7. A 3D visualization model of the filled space is obtained based on the total volume and segmented volumes.

[0171] Example 2:

[0172] This embodiment provides a visualization system for a composite subsurface ultra-large in-situ filling space, including:

[0173] The 3D point cloud data acquisition module is used to collect data from the filling space under the composite roof structure from multiple angles using laser scanning equipment to obtain raw point cloud data; and to obtain complete 3D point cloud data based on the raw point cloud data.

[0174] The point cloud preprocessing module is used to preprocess 3D point cloud data to obtain standard point cloud data;

[0175] The contour recognition module is used to identify boundary contours based on standard point cloud data to obtain complete spatial boundary contour data.

[0176] The spatial 3D geometric model construction module is used to obtain a spatial 3D geometric model based on spatial boundary contour data.

[0177] The dimension distribution data acquisition module is used to slice the cross section based on the spatial coordinate range of the three-dimensional geometric model, and obtain dimension distribution data based on the obtained contour dimension parameters of the cross section.

[0178] The volume calculation module is used to obtain the total volume and segmented volume of the filling space based on the size distribution data.

[0179] The visualization model building module is used to obtain a 3D visualization model of the filled space based on the total volume and segmented volumes.

[0180] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A visualization method for a composite subsurface ultra-large in-situ filling space, characterized in that, Includes the following steps: S1. Use laser scanning equipment to collect data from multiple angles of the filling space under the composite roof structure to obtain raw point cloud data; and obtain complete three-dimensional point cloud data based on the raw point cloud data. S2. Preprocess the three-dimensional point cloud data to obtain standard point cloud data; S3. Based on the standard point cloud data, perform boundary contour recognition to obtain complete spatial boundary contour data; S4. Obtain a three-dimensional spatial geometric model based on the spatial boundary contour data; S5. Based on the spatial coordinate range of the three-dimensional geometric model, perform cross-sectional slicing, and based on the obtained profile dimension parameters of the cross-section, obtain dimension distribution data. S6. Based on the size distribution data, obtain the total volume and segmented volume of the filling space; S7. Based on the total volume and segmented volumes, obtain a three-dimensional visualization model of the filled space.

2. The visualization method for a composite under-top ultra-large in-situ filling space according to claim 1, characterized in that, S1 includes the following steps: S11. Use laser scanning equipment to collect data from multiple angles of the filling space under the composite roof structure to obtain raw point cloud data; S12. Perform point cloud registration on the original point cloud data to obtain registered point cloud data with unified coordinates; S13. Spatial segmentation is performed on the registered point cloud data to obtain voxelized point cloud data; S14. Surface reconstruction is performed based on the voxelized point cloud data to obtain three-dimensional surface data; S17. Repair the voids in the three-dimensional surface data to obtain the three-dimensional point cloud data.

3. The visualization method for a composite under-top ultra-large in-situ filling space according to claim 1, characterized in that, S2 includes the following steps: S21. Obtain the spatial coordinates and reflection intensity of the three-dimensional point cloud data, calculate the statistical characteristics of each three-dimensional point cloud based on the spatial coordinates and reflection intensity, and obtain an initial feature point cloud set. S22. Calculate the neighborhood statistics for each initial feature point cloud to identify noise points; S23. When the neighborhood statistics of a noise point are greater than a preset neighborhood statistics threshold, the noise point is removed to obtain a preliminary denoised point cloud. S24. Calculate the point cloud density of the preliminary denoised point cloud. When the point cloud density is lower than the preset point cloud density threshold, use the difference method to supplement the point cloud to obtain the supplemented point cloud. S25. Recalculate the spatial coordinates and reflection intensity of the supplemented point cloud, and repeat steps S21-S24 until all supplemented point clouds meet the neighborhood statistics and point cloud density verification to obtain the standard point cloud data.

4. The visualization method for a composite under-top ultra-large in-situ filling space according to claim 1, characterized in that, S3 includes the following steps: S31. Calculate the local neighborhood vector of each point cloud based on the standard point cloud data to obtain the point cloud normal vector; S32. Calculate the curvature of each point cloud based on the point cloud normal vector to obtain curvature distribution data; S33. When the curvature distribution data is greater than the preset curvature threshold, the curvature distribution data is marked as a boundary point; S34. Based on the boundary points, determine the preliminary spatial boundary outline; S35. The preliminary spatial boundary contour is smoothed and reconstructed in three dimensions to obtain the spatial boundary contour data.

5. The visualization method for a composite under-top ultra-large in-situ filling space according to claim 1, characterized in that, S4 includes the following steps: S41. Generate a structured point set based on the spatial boundary contour data to obtain an initial point cloud model; S42. Generate a set of triangular facets based on the initial point cloud model to obtain an initial closed surface; S43. Perform a closure test on the initial closed surface. If there are voids, fill the voids to obtain a complete closed surface. S44. Adjust the distribution of face patches based on the complete closed surface to obtain optimized triangular face patches; S45. Based on the optimized triangular facets, the spatial three-dimensional geometric model is obtained.

6. The visualization method for a composite under-top ultra-large in-situ filling space according to claim 1, characterized in that, S5 includes the following steps: S51. Based on the spatial coordinate range of the aforementioned three-dimensional geometric model, obtain the bounding box coordinates; S52. Based on the bounding box coordinates, generate equally spaced cross sections in the horizontal and vertical directions using a uniform sampling method to obtain a set of cross sections; S53. Obtain the contour dimension parameters based on the set of cross sections; S54. Based on the contour size parameters, obtain the morphological feature vector; S55. When the Euclidean distance of the morphological feature vector is greater than a preset distance threshold, the cross section is marked as an abnormal cross section, thereby obtaining a set of abnormal cross sections. S56. The abnormal cross-section set is removed, and the removed contour data is supplemented by interpolation method to obtain the complete size distribution data.

7. The visualization method for a composite under-top ultra-large in-situ filling space according to claim 1, characterized in that, S6 includes the following steps: S61. Obtain standard cross-sectional data based on the size distribution data; S62. Based on the standard cross-section data, obtain the local volume corresponding to each cross-section, and then obtain the segmented volume; S63. Integrate the segmented volumes to obtain the total volume of the filling space.

8. A visualization system for a composite under-top ultra-large in-situ filling space, the visualization system being used to implement the visualization method according to any one of claims 1-7, characterized in that, include: The 3D point cloud data acquisition module is used to collect data from the filling space under the composite roof structure from multiple angles using laser scanning equipment to obtain raw point cloud data. And based on the original point cloud data, complete three-dimensional point cloud data is obtained; The point cloud preprocessing module is used to preprocess the three-dimensional point cloud data to obtain standard point cloud data. The contour recognition module is used to perform boundary contour recognition based on the standard point cloud data to obtain complete spatial boundary contour data. A spatial three-dimensional geometric model construction module is used to obtain a spatial three-dimensional geometric model based on the spatial boundary contour data; The size distribution data acquisition module is used to slice the cross section based on the spatial coordinate range of the three-dimensional geometric model, and obtain size distribution data based on the obtained profile size parameters of the cross section. The volume calculation module is used to obtain the total volume and segmented volume of the filling space based on the size distribution data; A visualization model building module is used to obtain a three-dimensional visualization model of the filled space based on the total volume and segmented volumes.

Images