Cobweb diffusion mine roadway modeling method based on point cloud data

Abstract

The invention discloses a cobweb diffusion mine roadway modeling method based on point cloud data, and the method comprises the steps: point cloud data acquisition and geometric feature preprocessing, multi-core-point-driven annular fan-shaped adaptive diffusion modeling, local structure fusion and topological connection path optimization, and flexible triangular surface splicing and global model closed construction. And finally generating a continuous and smooth global surface model. The method disclosed by the invention can efficiently adapt to complex and irregular environments such as mine roadways and the like, remarkably improves the modeling precision and stability, and is suitable for scenes such as three-dimensional modeling, structure reconstruction and safety monitoring.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of three-dimensional laser point cloud data processing, and particularly relates to a cobweb type diffusion mine roadway modeling method based on point cloud data. BACKGROUND

[0002] With the continuous increase of mining depth, the underground space environment is increasingly complex, and the three-dimensional modeling and structure monitoring of mine roadway, as a key structure for transportation and ventilation, are of great significance to mine safety operation and disaster warning. However, the traditional roadway modeling and measurement method is limited by factors such as narrow space, insufficient light and irregular wall surface, and has long measurement period, difficult to guarantee precision and relies on a large number of human resources, which is difficult to meet the needs of current intelligent mine construction for efficient and refined modeling.

[0003] In recent years, three-dimensional laser scanning technology has been widely used in underground space, tunnel deformation monitoring and other fields due to its high precision, high density and non-contact acquisition capability, and has gradually become an important means to replace the traditional roadway surveying and mapping method. Using three-dimensional laser point cloud to model the mine roadway can not only record the geometric shape completely, but also support subsequent digital analysis and structure risk assessment.

[0004] However, most of the current point cloud modeling methods still rely on regular profile reconstruction, global surface fitting or projection modeling, and when facing complex scenes such as non-uniform point cloud density, local occlusion and curvature mutation, the model may be discontinuous, the boundary may be incomplete, and the topological structure may be unreasonable. Therefore, it is urgent to propose a three-dimensional point cloud modeling method for complex structures of mine roadway, with structure control ability and local adaptive ability, to complete the roadway surface reconstruction task with higher efficiency and stronger adaptability. SUMMARY

[0005] To solve the above technical problems, the purpose of the present application is to provide a cobweb type diffusion mine roadway modeling method based on point cloud data. For the point cloud data after preprocessing, the multi-core point is used to realize the structured three-dimensional topological modeling with adaptive angle and step length, and the flexible fusion and closure completion between different regions are completed by combining the minimum spanning tree and graph optimization algorithm, and the flexible triangular surface is constructed for connection.

[0006] The purpose of the present application is achieved by the following steps: S100, point cloud data acquisition and geometric feature preprocessing: acquiring three-dimensional point cloud data of the mine roadway to be modeled, preprocessing the point cloud data to denoise and reduce the data amount, estimating the normal and curvature of the point cloud data after preprocessing, and selecting a plurality of core points in each local region according to the normal and curvature estimation results for subsequent topological growth of ring sector diffusion; S200, ring-fan adaptive diffusion modeling driven by multiple core points: respectively starting from different core points, a ring-fan diffusion method is used for topological growth to form multiple local spiderweb-like grid structures; the diffusion method specifically divides multiple diffusion directions by taking the core point as the center and using an adaptive angle control mechanism, determines the step length in each diffusion direction and expands outward by using an adaptive step length control mechanism, and similarity calculation is performed based on the curvature difference and spatial distance between points during the expansion process, and points that meet the conditions are selected to construct a connection relationship; S300, local structure fusion and topological connection path optimization: detecting the proximity state of the boundary of each diffusion region; specifically, different scenarios are combined for selection and use, for regions with sparse structure distribution, balanced point density or regular boundary, a minimum spanning tree algorithm is used for local fusion, for regions with inconsistent boundary direction, large normal difference or curvature mutation, a graph optimization algorithm is used for local fusion, for regions with large-scale and complex morphology, first use the minimum spanning tree algorithm to construct the initial connection, and then use the graph optimization algorithm to optimize the connection quality, realize continuous coverage modeling; then optimize the connection path between the points, and output the optimal connection path point set; S400, flexible triangular face splicing and global model closure construction: a flexible triangular face connection area is constructed at the fusion boundary to form a transition grid between irregular sectors, ensuring smooth connection and local consistency of the topological structure, and completing the generation of a continuous and smooth global three-dimensional surface model.

[0007] Preferably, the preprocessing of step S100 includes: De-noising processing: using conventional statistical filtering methods in the art to remove obvious outliers, and smoothing the point cloud data to remove high-curvature points with large curvature, and retaining important concave and convex features and key areas in the mine tunnel; Adaptive downsampling: using the existing voxel grid filtering (Voxel Grid Filter) method in the art to downsample the point cloud according to the set voxel size, reducing the data amount. Voxel grid filtering effectively reduces the sampling by segmenting the three-dimensional space, while retaining key geometric features.

[0008] Preferably, the normal estimation of step S100 uses principal component analysis (PCA) to estimate the normal by fitting a plane in the neighborhood of each point, and the curvature estimation is calculated by fitting the neighborhood points and calculating the fitting error. The shape change of the corner and crack in the mine tunnel is a high-curvature area.

[0009] Preferably, the core point of step S100 is selected using the following formula: Assuming that the curvature of a point is , the normal vector is , and the determination condition of whether the point is a core store is: wherein, is the curvature value of point i, is the curvature threshold, is the normal variation threshold, is the normal variation rate of point i; wherein, is the neighborhood point set of point i, is the Euclidean norm.

[0010] Preferably, the adaptive angle control mechanism of S200 step is specifically: when the ring fan is diffused, the angle determines the division of each diffusion direction, and the dynamic adjustment depends on the local geometric features of the point cloud; when the local area curvature is large and the normal variation is severe, the angle is reduced, and the complex surface is finely fitted; when the area is flat, the angle is enlarged, and the efficiency is improved; the specific calculation formula is: Assuming that the diffusion angle is , the adaptive adjustment angle is: wherein, : initial standard angle; : average curvature of the neighborhood of the core point; : normal fluctuation index; , : adjustment weight parameter.

[0011] The adaptive angle control mechanism dynamically adjusts the angle based on the local geometric complexity of the point cloud, uses the average curvature value and the normal direction variation rate in the local neighborhood of each core point diffusion to construct a complexity index, determines the angle interval size of the current ring fan direction, makes the diffusion more dense in the high complexity area, and makes the diffusion more sparse in the flat area.

[0012] Preferably, the adaptive step control mechanism of S200 step is specifically: the step controls the diffusion distance of the point cloud in each diffusion direction; in the complex structure, that is, in the concave-convex severe or dense area, the dynamic adjustment of the step can improve the modeling precision, avoid 'jump connection', reduce redundant sampling in the flat area, and improve the diffusion efficiency, which can guarantee the structural coherence and topological stability in each expansion, and the specific calculation formula is as follows: wherein, is the default step; is the curvature variance, indicating the curvature change on the diffusion; is the distance variance between points, indicating the point density in the diffusion direction; , Adjustment coefficient to control the flexibility of step length adjustment.

[0013] The adaptive step length control mechanism adjusts the step length according to the variation of the geometric characteristics on the diffusion path, ensuring that a smaller step length is used in complex surface areas to improve modeling accuracy, and a larger step length is used in flat areas to improve diffusion efficiency.

[0014] Preferably, during the multiple core point diffusion process of S200, points that meet the conditions can be screened and connection relationships can be constructed based on similarity, and the specific judgment formula is: wherein, is the point normal; is the point curvature; is the point similarity between and is the normal direction similarity; is the curvature difference similarity; is the spatial distance similarity.

[0015] Preferably, the proximity state of the boundary of each diffusion area detected in S300 is specifically: S301, proximity determination: after each local diffusion area completes topological growth, a boundary point set is formed , which contains all points located on the edge of the diffusion area, and the specific formula of the boundary point set is as follows: wherein, p is a point in the local grid; is the number of adjacent points connected to point p; is the set maximum topological connection threshold; S302, proximity calculation: the boundary point set of multiple local areas is set as the shortest distance between the boundary point sets of multiple local diffusion areas , to determine the areas that need to be fused, and the calculation formula is: wherein, is the Euclidean distance between point p and point q; when , the fusion process is triggered, is the preset proximity threshold, which is the maximum allowed distance between boundary points, and is an empirical value or dynamically adjusted.

[0016] Preferably, the minimum spanning tree algorithm is used for local fusion in S300, specifically: for the boundary point sets of the two local structures to be fused, the boundary point sets of the two local structures to be fused are taken as the node set of the graph​ Each edge in the figure represents the connection between two points, and the edge weight represents the Euclidean distance between the points The specific formula is: Wherein, is the distance between points and ; is the Euclidean distance calculation formula; The algorithm aims to connect all nodes by selecting the smallest edge weight set , so as to ensure that the total length of the connection path is the smallest, and the specific formula is: Wherein, is the edge set of the minimum spanning tree; is the weight of each edge; The minimum spanning tree fusion algorithm is suitable for the case where the point cloud boundary distribution is regular and the direction is clear; The basic idea is to establish a spatial distance graph between all possible boundary point pairs, and select a set of edges with the smallest connection cost from them to form a connected structure; The S300 step adopts a graph optimization algorithm for local fusion, specifically: set the boundary point set of the local structure and as the node set of the graph , and the weight of each edge contains the Euclidean distance, the normal angle, the curvature difference and the spatial distance, and the calculation formula is as follows: Wherein, is the normal direction similarity; is the curvature difference normalized similarity; is the spatial distance normalized similarity; are the normal direction weight, the curvature difference weight and the spatial distance weight, respectively; is the upper limit of normalization; The graph optimization fusion algorithm is suitable for point cloud regions with complex boundary morphology, drastic structural changes or local distortion. By constructing a weighted graph model containing boundary point pairs, the normal angle, the curvature difference and the spatial distance are used as similarity indexes for multi-feature fusion to optimize the connection relationship of the boundary, and the most optimal connection path of geometric continuity is extracted from it, so as to realize high-quality local structure fusion and transition splicing; The formula for optimizing the connection path between points in the S300 step is: Wherein, is the optimal connection path set; For the connected point set.

[0017] Preferably, the S400 step specifically comprises the following steps: S401, splice region determination: for the connection path obtained in the S300 step, the joint region is demarcated, the boundary point set of each local region is taken as the target region for triangular face splicing, and the shortest distance between the boundary point set and is checked If it is less than a preset distance threshold, subsequent splicing steps are performed. S402, triangular face splicing and filling: triangular patches are constructed, gaps between local regions are filled, model connectivity is ensured, three-point combinations are selected, that is, non-collinearity, normal consistency and topological consistency are met, specifically: Non-collinearity: ensure that the selected three points are not on the same straight line, and triangular shapes with an area of zero are discarded, and the formula is: Normal consistency: ensure that the normal of the constructed triangular face is consistent with the normal direction of the adjacent region, and is compared with the normal of the surrounding structure, if the included angle θ is less than a set threshold , it is considered that the triangular face is legal: Topological consistency: ensure that the newly generated triangular face does not intersect with the existing structure, and avoid topological conflicts, after generating a triangular shape each time, it is necessary to check whether its edge is repeated or intersected with the existing grid structure; S403, triangular face construction and weighted interpolation: after the three-point combination is completed, the triangular patch is constructed , and is added to the local grid structure, and the normal interpolation of the newly generated triangular face is performed, the normals of the three vertices are respectively , weighted interpolation is performed, and the specific formula is: Wherein, is the weight parameter of the normal of each point; S404, loop face construction until the boundary is closed: after constructing a triangular face each time, the boundary point set and the ring structure are updated, and the next triangular point group that can be constructed is selected; this process is repeatedly executed until all boundary regions are closed, or the remaining point number is less than three points, and the face construction condition is no longer met.

[0018] S400 step adopts flexible triangular face splicing at boundary fusion, realizes smooth transition at boundary fusion, and combines normal direction, curvature change and topological constraint conditions to control triangular surface quality, ensures that the finally generated global surface model has structural continuity, geometric smoothness and splicing consistency.

[0019] Compared with the prior art, the present application has the following technical effects: 1、The ring fan diffusion mechanism of the method of the present application, combined with adaptive angle and step control, can accurately adapt to different surface characteristics in the modeling process, significantly improving the modeling efficiency and accuracy in complex environments such as mine tunnels, and is suitable for mine three-dimensional modeling, structure reconstruction and safety monitoring scenes; 2、The present application effectively solves the problems of irregular boundary shape, large normal difference and obvious curvature change of point cloud by combining the minimum spanning tree algorithm with the graph optimization algorithm, realizes the smooth fusion and geometric continuity of local structure; 3、The present application adopts flexible triangular face splicing strategy to fill the boundary gap and ensure the continuity and smooth transition of the model surface, and is suitable for processing gaps caused by construction shielding, local missing, etc.; 4、Compared with the traditional method based on global fitting or region growing, the present application can maintain high precision and adaptability under the condition of non-uniform point cloud density, complex geometric shape and local shielding, and ensure efficient modeling in various environments; 5、The present application provides a clear topological control to ensure that each step of the modeling process is interpretable and controllable, suitable for subsequent fine analysis and engineering application, and avoids the "black box" modeling problem in traditional methods. BRIEF DESCRIPTION OF DRAWINGS

[0020] Figure 1 The present application is a flowchart; Figure 2 The present application is a ring fan diffusion structure diagram of multi-core point driving; Figure 3 The present application is a ring fan diffusion modeling model diagram; Figure 4 The present application is a flexible triangular splicing structure diagram; Figure 5 The present application is a modeling effect diagram of flexible triangular splicing; Figure 6 The present application is a final modeling effect diagram. DETAILED DESCRIPTION

[0021] The present application will be further described below in conjunction with the embodiments and drawings, but in any way limit the present application, any transformation or replacement based on the teaching of the present application belongs to the protection scope of the present application.

[0022] Embodiment 1 The embodiment is implemented for a certain lead-silver mine underground tunnel in Yunnan. The mine is underground mining, and the mined minerals are lead and silver. The production scale is 100,000 tons / year, the mining area is 0.3850 km², and the mine is located in a county. The mine was built in 2006, and the current mining focus is shifting to the 2050m elevation.

[0023] As shown in the accompanying Figure 1 The spider web diffusion mine tunnel modeling method based on point cloud data in the embodiment includes the following steps: S100, point cloud data acquisition and geometric feature preprocessing: acquire the three-dimensional point cloud data of the mine tunnel to be modeled, and preprocess the point cloud data to denoise and reduce the data volume; estimate the normal and curvature of the preprocessed point cloud data; and select a plurality of core points in each local area according to the normal and curvature estimation results for subsequent topological growth of ring-fan diffusion; S200, multi-core point driven ring-fan adaptive diffusion modeling: respectively starting from different core points, topological growth is performed in a ring-fan diffusion manner to form a plurality of local spider web-like grid structures; the diffusion manner is to divide a plurality of diffusion directions by taking the core point as the center and using an adaptive angle control mechanism, and in each diffusion direction, the step is determined by an adaptive step control mechanism and is extended outward, and in the extension process, similarity calculation is performed based on the curvature difference and spatial distance between points to select points that meet the conditions to construct a connection relationship; S300, local structure fusion and topological connection path optimization: detect the proximity state of the boundary of each diffusion region; for regions with sparse structure distribution, balanced point density, or regular boundary, use the minimum spanning tree algorithm for local fusion; for regions with inconsistent boundary directions, large normal differences, or curvature mutations, use a graph optimization algorithm for local fusion; for regions with large-scale and complex point clouds, first use the minimum spanning tree algorithm to construct an initial connection, and then use the graph optimization algorithm to optimize the connection quality; and then optimize the connection path between points, and output the optimal connection path point set; S400, flexible triangular face splicing and global model closure construction: construct a flexible triangular face connection area at the fusion boundary to form a transition grid between irregular sectors, and complete the generation of a continuous and smooth global three-dimensional surface model.

[0024] Embodiment 2 The spider web diffusion mine tunnel modeling method based on point cloud data in the embodiment is based on Embodiment 1, wherein the point cloud data obtained in step S100 includes the three-dimensional coordinates (x, y, z) of each point, as well as additional information such as reflectivity and color. The preprocessing operation on the point cloud data includes: De-noising: using statistical filtering to remove obvious noise points, judging whether it is an outlier by calculating the local neighborhood standard deviation of point cloud data, and deleting these points, and smoothing the point cloud data, removing high curvature points with large curvature, and retaining important concave-convex features and key areas in the mine tunnel; Adaptive down-sampling: using the Voxel Grid Filter method, the point cloud is down-sampled according to the set voxel size, thereby reducing the data amount and improving the subsequent processing efficiency; Normal estimation: the embodiment uses Principal Component Analysis (PCA) to estimate the normal by fitting a plane in the neighborhood of each point, so as to facilitate subsequent diffusion modeling; Curvature estimation: fitting the neighborhood points by least squares method and calculating the fitting error, the corner, crack and other morphological changes in the mine tunnel are high curvature areas; Core point selection: selecting multiple core points according to the curvature extreme points and the difference in normal direction, usually selecting points with large curvature and large normal change direction in places with significant local geometric features, to ensure that complex geometric areas can be covered, and the core point selection condition is as follows: Assuming that the curvature of a point is , the normal vector is , and the judgment condition of whether the point is a core store is: Wherein, is the curvature value of point i, is the curvature threshold, is the normal change rate threshold, is the normal change rate of point i; Wherein, is the neighborhood point set of point i, is the Euclidean norm.

[0025] Embodiment 3 The spider web diffusion mine tunnel modeling method based on point cloud data in this embodiment is based on embodiment 2, wherein the S200 step mainly expands layer by layer outward from the core point as the center, as shown in the accompanying Figure 2 Fig. 2 shows the ring-fan diffusion topology growth process with multiple core points as the center. Compared with the traditional global triangulation, the angle control and step control dynamically adjust the grid structure density and accuracy in the diffusion process, and the dynamic diffusion has local controllability, clear direction, and stable growth of structure layer by layer, which is suitable for mine tunnel scenes with uneven and irregular shapes. At the same time, each core point independently controls its own diffusion area, achieving multi-core collaborative fusion. The adaptive angle control mechanism is specifically as follows: during the annular fan-shaped diffusion, the angle... The division of each diffusion direction is dynamically adjusted based on the local geometric features of the point cloud; when the curvature of a local region is large and the normal changes drastically, the angle is reduced to more accurately fit complex surfaces; when the region is flat, the angle is increased to improve efficiency; the specific calculation formula is as follows: Assuming the diffusion angle is Then the angle is adjusted adaptively: in, Initial standard angle; : Average curvature of the neighborhood of the core point; : Normal volatility indicator; , Adjusting the weighting parameters; The adaptive step size control mechanism specifically refers to: step size The diffusion distance of the point cloud in each diffusion direction is controlled, and the specific calculation formula is as follows: in, This is the default step size; Let be the curvature variance, representing the curvature change on the diffusion; Here, represents the variance of the distance between points, and represents the point density along the diffusion direction; , To adjust the coefficient and control the flexibility of step size adjustment; When multiple core points diffuse, a local spiderweb-like grid forms around each core point. During the diffusion process, points that meet certain conditions are selected based on similarity, and connection relationships are established. The specific judgment formula is as follows: in, For point normals; Point curvature; For point and The similarity between them; Similarity in normal directions; Similarity of curvature differences; Spatial distance similarity; As attached Figure 3As shown, a 10m long section of the underground mine tunnel is extracted, presenting the multi-core point driven annular fan-shaped adaptive diffusion modeling of step S200 in Example 3. The grid structure formed by the annular fan-shaped diffusion closely fits the shape of the tunnel wall, and the grid density is significantly improved. In the smooth wall area, the step size is increased to 0.3m~0.5m, which effectively reduces redundant data and verifies the effectiveness of the adaptive angle and step size control mechanism. It achieves the expected goal of "high-precision modeling of complex areas and efficient processing of flat areas", and uses this as the basis for the next step of flexible triangular face splicing.

[0026] Example 4 This embodiment of the spiderweb-like diffusion mine roadway modeling method based on point cloud data is based on embodiment 3. After the above-mentioned multi-core point diffusion, the proximity of the boundary needs to be detected first. Specifically: S301. Proximity Detection: After completing topological growth, each local diffusion region forms a set of boundary points. This point set contains all points located at the edge of the diffusion region. The specific formula for the boundary point set is as follows: Where p is a point in the local mesh; Let p be the number of adjacent points connected to it. This is the maximum topology connection threshold set. S302. Proximity Calculation: The boundary point set of multiple local regions is set as the boundary point set of multiple local diffusion regions. and shortest distance between Calculations are performed to determine the areas that need to be merged. The calculation formula is as follows: in, Let p be the Euclidean distance between points p and q; when Triggering the fusion process The preset proximity threshold is the maximum allowable distance between boundary points, which can be either an empirically set value or dynamically adjusted. During the structural fusion process, for scenarios with sparse structural distribution and regular boundaries, the minimum spanning tree algorithm is introduced to ensure the shortest overall connection path. For regions with complex boundary shapes or requiring strong geometric continuity, a graph optimization strategy can be further introduced. An optimized graph is constructed using multi-dimensional weights such as shortest path, normal angle constraint, and curvature similarity to achieve higher quality structural matching. For regions with sparse structural distribution, balanced point density, or regular boundaries, the minimum spanning tree algorithm is used for local fusion. Specifically, for the boundary point sets of two local structures to be fused, these boundary point sets are used as the node sets of the graph. , each edge in the graph represents a connection between two points, and the edge weight represents the Euclidean distance between the points , the specific formula is: wherein, is the distance between points and ; is the formula for calculating the Euclidean distance; The algorithm aims to connect all nodes by selecting the smallest edge weight set to minimize the total length of the connection path, and the specific formula is: wherein, is the edge set of the minimum spanning tree; is the weight of each edge; For areas with inconsistent boundary directions, large normal differences, or curvature mutations, local fusion is performed using a graph optimization algorithm, which is: set the boundary point set of the local structure and as the node set of the graph , and the weight of each edge contains the Euclidean distance, the normal angle, the curvature difference, and the spatial distance, and the calculation formula is as follows: wherein, is the normal direction similarity; is the curvature difference normalized similarity; is the spatial distance normalized similarity; are the normal direction weight, curvature difference weight, and spatial distance weight, respectively; is the upper limit of normalization; For large-scale and complex point cloud regions, the above two fusion methods are combined, as follows: in the boundary rule and point density balanced region, the minimum spanning tree is used to construct the fusion path, and the global topology is quickly connected; in the area with inconsistent boundary directions, large normal differences, or curvature mutations, the graph optimization fusion method is used to ensure geometric continuity; for large-scale and complex regions, the minimum spanning tree can also be used to construct the initial connection, and then the local graph optimization is used to reinforce the connection quality; The next step is to optimize the connection path between the points, the goal of which is to maximize geometric consistency, and on the basis of the above maximum similarity, to achieve smooth structure fusion, the formula is: wherein, is the optimal connection path set; is the connected point set; The optimal connection path point set output by this step can be used to drive the subsequent flexible triangular face splicing, realize the continuous transition between multiple local structures and global model construction.

[0027] Embodiment 5 The cobweb diffusion mine tunnel modeling method based on point cloud data in this embodiment is based on embodiment 4, wherein the S400 step performs flexible triangle splicing, introduces flexible triangular face splicing, as shown in the theoretical model diagram of flexible triangular face splicing, presents the spatial distribution difference of the triangular face adapting to the boundary points, and is used to cope with the point cloud density difference, normal difference and local geometric mutation after splicing of the above algorithm. Figure 4 S401, splice region determination: the connection path obtained in the S300 step is labeled to determine the joint region, and the boundary point set of each local region is taken as the target region for triangular face splicing, and the shortest distance between the boundary point set and is checked , and if it is less than a preset distance threshold, it can be spliced; S402, triangular face splicing and filling: construct a triangular patch, fill the gap between local regions, ensure model connectivity, select three-point combination, that is, satisfy non-collinearity, normal consistency and topological consistency, specifically: Non-collinearity: ensure that the selected three points are not on the same straight line, and the triangular area is zero, and the formula is: Normal consistency: ensure that the normal of the constructed triangular face is consistent with the normal direction of the adjacent region, and is compared with the normal of the surrounding structure, and if the included angle θ is less than a set threshold , it is considered that the triangular face is legal: Topological consistency: ensure that the newly generated triangular face does not intersect with the existing structure to avoid topological conflict, and after generating a triangular face each time, it needs to be checked whether its edge is repeated or intersected with the existing grid structure.

[0028] S403, triangular face construction and weighted interpolation: after the three-point combination is completed, the triangular patch is constructed , and is added to the local grid structure, and the normal of the newly generated triangular face is interpolated, the normals of the three vertices are respectively , and the weighted interpolation is performed, and the specific formula is: wherein, is the weight parameter of the normal of each point.​

[0029] S404, loop until the boundary closed: after each construction of triangular surface, update the boundary point set and ring structure, reselect the next available triangular point group. Loop this process until all the boundary area is closed, or the remaining number of points is less than three points, no longer meet the conditions of the surface; As shown in the accompanying Figure 5 The actual effect diagram of flexible triangular surface splicing is based on the point cloud data of a mine in Yunnan at an elevation of 2050m, and the continuous surface formed by the whole process of flexible triangular surface refinement and filling. The spider web skeleton constructs the preliminary topological connection through the multi-core point ring fan-shaped diffusion, and the final flexible triangular surface fills the gap through weighted interpolation. The triangular surface density is adaptively adjusted according to the curvature, which verifies the adaptive ability of the present application in processing non-uniform point cloud.

[0030] As shown in the accompanying Figure 6 The global modeling effect diagram of the mine tunnel generated by the method of the present application is shown in the accompanying drawings. The diagram is a global three-dimensional surface model of a part of the mine tunnel. The model is continuous and smooth as a whole, and there is no obvious joint gap. The smoothness of the tunnel wall surface is reflected. Compared with the same modeling method, the model distortion at the corner of the tunnel is reduced, and the modeling efficiency and accuracy are improved. At the same time, the result verifies the effectiveness of the present application in achieving the closed construction of the global model of the complex tunnel through multi-core diffusion and flexible splicing. The expected purpose of "efficient adaptation to complex environment, improving modeling accuracy and stability" can be achieved.

Claims

1. A method for modeling spiderweb-like diffusion mine roadways based on point cloud data, characterized in that... Includes the following steps: S100. Point cloud data acquisition and geometric feature preprocessing: Acquire the 3D point cloud data of the mine roadway to be modeled, and preprocess the point cloud data to remove noise and reduce the amount of data. Normal and curvature estimation are performed on the preprocessed point cloud data. Based on the results of normal and curvature estimation, several core points are selected in each local region for subsequent annular fan-shaped diffusion topology growth. S200, Multi-core-point driven annular fan-shaped adaptive diffusion modeling: Starting from different core points, topological growth is performed using an annular fan-shaped diffusion method to form multiple local spider web-like mesh structures; The diffusion method is specifically based on the core point as the center and uses an adaptive angle control mechanism to divide multiple diffusion directions. In each diffusion direction, the step size is determined by an adaptive step size control mechanism and the diffusion is expanded outward. During the expansion process, similarity is calculated based on the curvature difference between points and the spatial distance, and points that meet the conditions are selected to build connection relationships; S300, Local Structure Fusion and Topological Connection Path Optimization: Detect the proximity status of the boundaries of each diffusion region; for regions with sparse structure distribution, balanced point density, or regular boundaries, the minimum spanning tree algorithm is used for local fusion; for regions with inconsistent boundary directions, large differences in normals, or abrupt curvature changes, a graph optimization algorithm is used for local fusion; for regions with large-scale, complex point cloud morphology, the minimum spanning tree algorithm is first used to construct initial connections, and then the graph optimization algorithm is used to optimize and strengthen the connection quality; finally, the connection paths between point pairs are optimized, and the optimal connection path point set is output. S400, Flexible Triangular Facet Assembly and Global Model Closure Construction: A flexible triangular facet connection region is constructed at the fusion boundary to form a transition mesh between irregular sectors, thus completing the generation of a continuous and smooth global 3D surface model.

2. The spiderweb-like diffusion mine roadway modeling method based on point cloud data according to claim 1, characterized in that... The preprocessing described in step S100 includes: Noise reduction: Statistical filtering is used to remove obvious outliers and smooth the point cloud data. High curvature points with large verification curvature are removed without hands-free processing, while important concave and convex features and key areas in the mine roadway are preserved. Adaptive downsampling: A voxel grid filtering method is used to downsample the point cloud according to the set voxel size, thereby reducing the amount of data.

3. The spiderweb-like diffusion mine roadway modeling method based on point cloud data according to claim 1, characterized in that... The S100 step normal estimation uses principal component analysis to estimate the normal by fitting a plane in the neighborhood of each point. Curvature estimation is performed by fitting neighborhood points using the least squares method and calculating the fitting error.

4. The spiderweb-like diffusion mine roadway modeling method based on point cloud data according to claim 1, characterized in that... The core points of the S100 step are selected using the following formula: Assume the curvature at a certain point is The normal vector is The criteria for determining whether a point qualifies as a core store are: in, Let i be the curvature value at point i. For curvature threshold, This is the threshold for the rate of change of the normal. Let be the rate of change of the normal to point i; in, Let i be the set of neighborhood points. It is the Euclidean norm.

5. The spiderweb-like diffusion mine roadway modeling method based on point cloud data according to claim 1, characterized in that... The S200 adaptive angle control mechanism specifically works as follows: during annular fan-shaped diffusion, the angle... The division of each diffusion direction is dynamically adjusted based on the local geometric features of the point cloud; when the curvature of a local region is large and the normal changes drastically, the angle is reduced to more accurately fit complex surfaces; when the region is flat, the angle is increased to improve efficiency; the specific calculation formula is as follows: Assuming the diffusion angle is Then the angle is adjusted adaptively: in, Initial standard angle; : Average curvature of the neighborhood of the core point; : Normal fluctuation indicator; , Adjust the weight parameters.

6. The spiderweb-like diffusion mine roadway modeling method based on point cloud data according to claim 1, characterized in that... The S200 step-adaptive step-size control mechanism specifically involves: step size... The diffusion distance of the point cloud in each diffusion direction is controlled, and the specific calculation formula is as follows: in, This is the default step size; Let be the curvature variance, representing the curvature change on the diffusion; Here, represents the variance of the distance between points, and represents the point density along the diffusion direction; , This is an adjustment coefficient to control the flexibility of step size adjustment.

7. The spiderweb-like diffusion mine roadway modeling method based on point cloud data according to claim 1, characterized in that... In the S200 step, during the diffusion of multiple core points, points that meet certain conditions are selected based on similarity, and connection relationships are constructed. The specific judgment formula is as follows: in, For point normals; Point curvature; For point and The similarity between them; Similarity in normal direction; Similarity of curvature differences; This represents spatial distance similarity.

8. The spiderweb-like diffusion mine roadway modeling method based on point cloud data according to claim 1, characterized in that... The S300 step for detecting the proximity status of each diffusion region boundary is as follows: S301. Proximity Detection: After completing topological growth, each local diffusion region forms a set of boundary points. This point set contains all points located at the edge of the diffusion region. The specific formula for the boundary point set is as follows: Where p is a point in the local mesh; Let p be the number of adjacent points connected to it. This is the maximum topology connection threshold set. S302. Proximity Calculation: The boundary point set of multiple local regions is set as the boundary point set of multiple local diffusion regions. and shortest distance between Calculations are performed to determine the areas that need to be merged. The calculation formula is as follows: in, Let p be the Euclidean distance between points p and q; when Triggering the fusion process The preset proximity threshold is the maximum allowable distance between boundary points, which can be either an empirically set value or dynamically adjusted.

9. The spiderweb-like diffusion mine roadway modeling method based on point cloud data according to claim 1, characterized in that... Step S300 employs the minimum spanning tree algorithm for local fusion. Specifically, for the boundary point sets of two local structures to be merged, these boundary point sets are used as the node sets of the graph. In the graph, each edge represents a connection between two points, and the edge weight represents the Euclidean distance between the points. The specific formula is as follows: in, For point and The distance between them; The Euclidean distance calculation formula; The algorithm aims to select the minimum set of edge weights. Connect all nodes to minimize the total length of the connected paths. The specific formula is: in, Let be the edge set of the minimum spanning tree; The weight of each edge; Step S300 employs a graph optimization algorithm for local fusion, specifically: setting a set of boundary points for the local structure. and As a set of nodes in a graph Meanwhile, the weight of each edge includes the Euclidean distance, the angle between the normals, the difference in curvature, and the spatial distance, calculated using the following formula: in, Similarity in normal direction; Normalized similarity based on curvature difference; Similarity is normalized to spatial distance; These are the weights for the normal direction, curvature difference, and spatial distance, respectively. This is the upper limit of normalization; The formula for optimizing the connection path between point pairs in step S300 is: in, This is the set of optimal connection paths; For the set of connected points.

10. The spiderweb-like diffusion mine roadway modeling method based on point cloud data according to claim 1, characterized in that... The S400 process specifically includes the following steps: S401. Determining the splicing area: Based on the connection path obtained in step S300, mark the seam area. The boundary point set of each local area is used as the target area for triangular face splicing. Check the boundary point set. and shortest distance between If the distance is less than the preset distance threshold, then proceed with the subsequent splicing steps; S402, Triangular Facet Assembly and Filling: Construct triangular facets to fill gaps between local areas, ensuring model connectivity. Select a combination of three points that satisfy non-collinearity, normal consistency, and topological consistency. Specifically: Noncollinearity: Ensure the selection of three points If the triangles are not on the same straight line, discard the triangle with a zero area. The formula is: Normal consistency: Ensure the normals of the constructed triangles are consistent. The direction of the normal to the adjacent region is consistent with the normal to the surrounding structure. If the included angle θ is less than a set threshold, the result is considered valid. If the triangle is valid, then the triangle is considered legal. Topological consistency: Ensure that newly generated triangle faces do not intersect with existing structures to avoid topological conflicts. After each triangle is generated, check whether its edges are repeated or intersecting with existing mesh structures. S403. Triangle Face Construction and Weighted Interpolation: After combining the three points, construct the triangle facet. This is then added to the local mesh structure, and normal interpolation is performed on the newly generated triangles, with the three vertices... The normals are respectively We then perform weighted interpolation, using the following formula: in, The weight parameters for the normal line at each point; S404, Cyclic face construction until boundary closure: After each construction of a triangular face, update the boundary point set and loop structure, and reselect the next set of triangular points for face construction; repeat this process until all boundary regions are closed, or the number of remaining points is less than three, and the face construction conditions are no longer met.

Images