Three-dimensional network model construction method and device, electronic equipment and storage medium

Abstract

The invention relates to a three-dimensional network model construction method and device, electronic equipment and a storage medium, and the method comprises the steps: carrying out the preprocessing, extraction and fitting of a roadway three-dimensional point cloud, obtaining an edge parameter equation of two road surface edge lines, and determining a center line parameter equation of a roadway center line; gradually advancing along the direction of the center line, determining a space region containing a current section based on a current roadway depth parameter and an edge parameter equation, and generating a closed curve parameter equation by utilizing point cloud fitting in the region; the current roadway depth parameter, the closed curve parameter equation and the curve length parameter form section elements, and the section elements are added into the vectorization section set; and constructing a three-dimensional grid model of the roadway based on the vectorized profile set. According to the method, the irregular change of the section of the roadway caused by undercut, overexcavation, chambers and deformation can be accurately described, the generated model is highly matched with a real roadway form, and a completely parameterized and vectorized roadway digital model is established.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to a method, apparatus, electronic device and storage medium for constructing a three-dimensional network model. Background Technology

[0002] With the development of the internet, smart mines are an inevitable trend in industries such as coal mining. Existing 3D modeling methods for tunnels mainly rely on the regularization of tunnel geometry for modeling.

[0003] However, existing 3D modeling methods for tunnels cannot accurately represent the irregular deformation and local complex structure of tunnels. The position of any point inside the constructed model cannot be described and queried through precise mathematical parameters. There is a lack of direct and computable mapping relationship between spatial coordinates and the logical position of the tunnel, which limits the efficiency and accuracy of the model in advanced digital applications such as path planning, precise measurement, and spatial analysis. Summary of the Invention

[0004] This application provides a method, apparatus, electronic device, and storage medium for constructing a three-dimensional network model, in order to solve the technical problem that existing three-dimensional modeling methods cannot construct vectorized network models, have low modeling efficiency and low intelligence level, and thus lead to low model utilization efficiency.

[0005] Firstly, this application provides a method for constructing a three-dimensional network model, the method comprising: The acquired three-dimensional point cloud data of the tunnel is preprocessed to obtain the point cloud set of the tunnel surface. The point cloud sets of the two edge lines of the tunnel surface are extracted from the point cloud set of the tunnel surface. The point cloud sets of the two edge lines are fitted to obtain the edge parameter equations corresponding to the two edge lines. Based on the edge parameter equations, the center line parameter equation of the tunnel center line is determined. Along the centerline of the tunnel, based on the tunnel depth parameters corresponding to the current advancing position and the edge parameter equation, a spatial region containing the current tunnel profile is determined. Using the point cloud data extracted from the spatial region, a closed curve parameter equation of the spatial region is fitted and generated. The current tunnel depth parameter corresponding to the spatial region, the closed curve parametric equation, and the curve length parameter are combined to form a profile element, which is then added to the vectorized profile set of the tunnel. Based on the vectorized profile set, a three-dimensional mesh model of the tunnel is constructed.

[0006] In one possible implementation, the preprocessing of the acquired three-dimensional point cloud data of the tunnel to obtain a point cloud set on the tunnel surface includes: A spatial index structure for the 3D point cloud data is constructed, and density statistical analysis is performed on the point cloud based on the spatial index structure to remove outlier noise points. Calculate the normal vector and curvature of each point in the point cloud after removing outlier noise points, and filter based on the curvature threshold to remove points that are not on the roadway surface, thus obtaining the point cloud set of the roadway surface.

[0007] In one possible implementation, the step of extracting point clouds of two edge lines of the roadway surface from the point cloud set of the roadway surface, and fitting the point clouds of the two edge lines respectively to obtain the edge parameter equations corresponding to the two edge lines includes: Based on the curvature values ​​of each point in the point cloud set on the surface of the alley, a subset of candidate edge point clouds is selected by using a curvature threshold. Distance clustering is performed on the candidate edge point cloud subset to obtain multiple initial clusters; from the multiple initial clusters, the initial clusters that have a linear spatial distribution and contain more than a preset threshold of points are selected as the target candidate edge point cloud subset. The target candidate edge point cloud subset is further clustered by distance to obtain point cloud sets of two edge lines that run through the length of the tunnel and are linearly spatially distributed. If the number of point cloud sets is greater than two, then the average spatial coordinates of each point cloud set are determined, and the two point cloud sets with the smallest vertical height are selected as the final point cloud sets for the left and right road edge lines. Piecewise continuously differentiable spatial curve fitting is performed on the point clouds of the left and right road edge lines respectively to obtain the piecewise parametric equations of the two edge lines as edge parametric equations.

[0008] In one possible implementation, determining the centerline parameter equation of the roadway centerline based on the edge parameter equation includes: Based on the piecewise parameter equations of the two edge lines in the edge parameter equation, for the same target parameter value, calculate the coordinates of the corresponding three-dimensional spatial points on the left and right edge lines respectively. Based on the target parameter value, determine the coordinates of the midpoint between the corresponding point on the left side and the corresponding point on the right side; Using the target parameter value as the independent variable and the midpoint coordinates as the dependent variable, the centerline parametric equation of the roadway centerline is formed.

[0009] In one possible implementation, determining the spatial region encompassing the current tunnel profile along the centerline direction of the tunnel, based on the tunnel depth parameter corresponding to the current advancing position and the edge parameter equation, includes: Based on the current tunnel depth parameter corresponding to the current advance position, and based on the piecewise parameter equations of the two edge lines in the edge parameter equation, the first edge point on the left edge line and the second edge point on the right edge line are calculated respectively. Connect the first edge point and the second edge point to obtain the current road surface cross section; Based on the current road cross section and combined with the preset height range of the tunnel, a hexahedral spatial region containing the current tunnel profile is determined in three-dimensional space.

[0010] In one possible implementation, the step of fitting and generating a closed curve parametric equation for the spatial region using point cloud data extracted from the spatial region includes: If there is no historical closed curve parametric equation corresponding to the current spatial region, then a new closed curve parametric equation is directly fitted using the currently extracted point cloud data. If a historical closed curve parametric equation exists corresponding to the current spatial region, the fitting error between the currently extracted point cloud data and the historical closed curve parametric equation is calculated. Based on the fitting error and a preset error threshold, the closed curve parametric equation for the spatial region is determined. If the fitting error is less than or equal to the preset error threshold, the historical closed curve parametric equation is used or finely adjusted as the current closed curve parametric equation. If the fitting error is greater than the preset error threshold, a new closed curve parametric equation is generated by refitting the currently extracted point cloud data.

[0011] In one possible implementation, the curve length parameter is determined in the following way: Obtain the parametric equation of the closed curve, and based on the parametric equation of the closed curve, calculate the arc length function and the total arc length of the closed curve; The arc length function is normalized relative to the total arc length to obtain the normalized arc length function. The range of the normalized arc length function is determined as the curve length parameter.

[0012] Secondly, this application provides a three-dimensional network model construction device, the device comprising: The parameter equation determination module is used to preprocess the acquired three-dimensional point cloud data of the tunnel to obtain the point cloud set of the tunnel surface, extract the point cloud sets of the two edge lines of the tunnel surface from the point cloud set of the tunnel surface, fit the point cloud sets of the two edge lines respectively to obtain the edge parameter equations corresponding to the two edge lines, and determine the center line parameter equation of the tunnel center line based on the edge parameter equations. The closed curve parametric equation determination module is used to determine the spatial region containing the current roadway profile along the centerline direction of the roadway, based on the roadway depth parameter corresponding to the current advancing position and the edge parametric equation, and to fit and generate the closed curve parametric equation of the spatial region using point cloud data extracted from the spatial region. The vectorized profile set construction module is used to construct a profile element by taking the current tunnel depth parameter corresponding to the spatial region, the closed curve parameter equation and the curve length parameter, and adding it to the vectorized profile set of the tunnel. A 3D network model construction module is used to construct a 3D mesh model of the tunnel based on the vectorized profile set.

[0013] In one possible implementation, the parametric equation determination module is specifically used for: A spatial index structure for the 3D point cloud data is constructed, and density statistical analysis is performed on the point cloud based on the spatial index structure to remove outlier noise points. Calculate the normal vector and curvature of each point in the point cloud after removing outlier noise points, and filter based on the curvature threshold to remove points that are not on the roadway surface, thus obtaining the point cloud set of the roadway surface.

[0014] In one possible implementation, the parametric equation determination module is specifically used for: Based on the curvature values ​​of each point in the point cloud set on the surface of the alley, a subset of candidate edge point clouds is selected by using a curvature threshold. Distance clustering is performed on the candidate edge point cloud subset to obtain multiple initial clusters; from the multiple initial clusters, the initial clusters that have a linear spatial distribution and contain more than a preset threshold of points are selected as the target candidate edge point cloud subset. The target candidate edge point cloud subset is further clustered by distance to obtain point cloud sets of two edge lines that run through the length of the tunnel and are linearly spatially distributed. If the number of point cloud sets is greater than two, then the average spatial coordinates of each point cloud set are determined, and the two point cloud sets with the smallest vertical height are selected as the final point cloud sets for the left and right road edge lines. Piecewise continuously differentiable spatial curve fitting is performed on the point clouds of the left and right road edge lines respectively to obtain the piecewise parametric equations of the two edge lines as edge parametric equations.

[0015] In one possible implementation, the parametric equation determination module is specifically used for: Based on the piecewise parameter equations of the two edge lines in the edge parameter equation, for the same target parameter value, calculate the coordinates of the corresponding three-dimensional spatial points on the left and right edge lines respectively. Based on the target parameter value, determine the coordinates of the midpoint between the corresponding point on the left side and the corresponding point on the right side; Using the target parameter value as the independent variable and the midpoint coordinates as the dependent variable, the centerline parametric equation of the roadway centerline is formed.

[0016] In one possible implementation, the closed curve parametric equation determination module is specifically used for: Based on the current tunnel depth parameter corresponding to the current advance position, and based on the piecewise parameter equations of the two edge lines in the edge parameter equation, the first edge point on the left edge line and the second edge point on the right edge line are calculated respectively. Connect the first edge point and the second edge point to obtain the current road surface cross section; Based on the current road cross section and combined with the preset height range of the tunnel, a hexahedral spatial region containing the current tunnel profile is determined in three-dimensional space.

[0017] In one possible implementation, the closed curve parametric equation determination module is specifically used for: If there is no historical closed curve parametric equation corresponding to the current spatial region, then a new closed curve parametric equation is directly fitted using the currently extracted point cloud data. If a historical closed curve parametric equation exists corresponding to the current spatial region, the fitting error between the currently extracted point cloud data and the historical closed curve parametric equation is calculated. Based on the fitting error and a preset error threshold, the closed curve parametric equation for the spatial region is determined. If the fitting error is less than or equal to the preset error threshold, the historical closed curve parametric equation is used or finely adjusted as the current closed curve parametric equation. If the fitting error is greater than the preset error threshold, a new closed curve parametric equation is generated by refitting the currently extracted point cloud data.

[0018] In one possible implementation, the curve length parameter is determined in the following way: Obtain the parametric equation of the closed curve, and based on the parametric equation of the closed curve, calculate the arc length function and the total arc length of the closed curve; The arc length function is normalized relative to the total arc length to obtain the normalized arc length function. The range of the normalized arc length function is determined as the curve length parameter.

[0019] Thirdly, this application provides an electronic device, including: a processor and a memory, wherein the processor is configured to execute a three-dimensional network model building program stored in the memory to implement the three-dimensional network model building method described in any one of the first aspects.

[0020] Fourthly, this application provides a storage medium storing one or more programs that can be executed by one or more processors to implement the three-dimensional network model construction method described in any one aspect.

[0021] Compared with the prior art, the technical solution provided in this application has the following advantages: The method provided in this application preprocesses the three-dimensional point cloud of the tunnel and extracts and fits the edge parameter equations of two road surface edge lines to determine the centerline parameter equation of the tunnel centerline; it advances step by step along the centerline direction, determines the spatial region containing the current profile based on the current tunnel depth parameter and the edge parameter equation, and uses the point cloud within the region to fit and generate the closed curve parameter equation; it uses the current tunnel depth parameter, the closed curve parameter equation, and the curve length parameter to form the profile element and adds it to the vectorized profile set; it constructs a three-dimensional mesh model of the tunnel based on the vectorized profile set. By dynamically constructing an adaptive spatial region along the tunnel centerline and using the local point cloud within it to directly fit a closed curve of arbitrary shape to express the tunnel cross section, it can accurately depict the irregular changes in the tunnel cross section caused by under-excavation, over-excavation, chambers, and deformation. The generated model is highly consistent with the real tunnel morphology, and a fully parameterized and vectorized tunnel digital model is established. Attached Figure Description

[0022] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0023] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] One or more embodiments are illustrated by way of example with reference numerals in the accompanying drawings. These illustrations do not constitute a limitation on the embodiments. Elements with the same reference numerals in the drawings are denoted as similar elements. Unless otherwise stated, the figures in the drawings are not to be limited by scale.

[0025] Figure 1 A flowchart illustrating an embodiment of a three-dimensional network model construction method provided in this application; Figure 2 A hexahedral schematic diagram of the current spatial region of a roadway is provided for an embodiment of this application; Figure 3 A flowchart illustrating an embodiment of another three-dimensional network model construction method provided in this application; Figure 4 A flowchart illustrating another embodiment of a three-dimensional network model construction method provided in this application; Figure 5 A structural block diagram of a three-dimensional network model construction device provided in this application embodiment; Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0027] The following disclosure provides numerous different embodiments or examples for implementing various structures of this application. To simplify the disclosure, specific examples of components and arrangements are described below. These are merely examples and are not intended to limit the scope of this application. Furthermore, reference numerals and / or letters may be repeated in different examples. Such repetition is for simplification and clarity and does not in itself indicate a relationship between the various embodiments and / or arrangements discussed.

[0028] To address the technical problems of existing 3D modeling methods being unable to construct vectorized network models, resulting in low modeling efficiency and low intelligence levels, and consequently low model utilization efficiency, this application provides a 3D network model construction method, device, electronic device, and storage medium. It can realize the expression of the roadway cross section by dynamically constructing an adaptive spatial region along the roadway centerline and directly fitting closed curves of arbitrary shapes using local point clouds within it. It can accurately depict the irregular changes in the roadway cross section caused by under-excavation, over-excavation, chambers, and deformation. The generated model highly matches the actual roadway morphology, establishing a fully parameterized and vectorized roadway digital model.

[0029] Figure 1 A flowchart illustrating an embodiment of a three-dimensional network model construction method provided in this application is shown below. Figure 1 As shown, it includes the following steps: Step 101: Preprocess the acquired 3D point cloud data of the tunnel to obtain the point cloud set of the tunnel surface. Extract the point cloud sets of the two edge lines of the tunnel surface from the point cloud set of the tunnel surface. Fit the point cloud sets of the two edge lines to obtain the edge parameter equations corresponding to the two edge lines. Based on the edge parameter equations, determine the centerline parameter equation of the tunnel centerline.

[0030] Three-dimensional point cloud data refers to a collection of massive three-dimensional coordinate points (x, y, z) on the surface of a tunnel collected by devices such as lidar and RGB-D cameras. Some points contain information such as reflection intensity and color, and are an original digital snapshot of the tunnel morphology.

[0031] Point cloud set on the surface of the tunnel can refer to the point cloud that retains only the tunnel body (ground, left and right walls, and arch) after preprocessing, and removes irrelevant point cloud data such as pipes, equipment, and noise points.

[0032] The two edge lines of the alleyway surface can refer to the boundary lines between the road surface and the left and right walls (like the two lines close to the wall when a person walks in the alleyway), which are the key baselines for locating the road surface range and subsequently constructing the profile.

[0033] Edge parametric equations refer to the use of mathematical functions (such as piecewise polynomials) to express the orientation of an edge line in three-dimensional space. Typically, the arc length is used as the parameter; given a parameter value, the three-dimensional coordinates of a point on the line can be calculated. Essentially, it can be understood as an equation describing the spatial shape of an edge line using mathematical formulas (such as piecewise quadratic polynomials); inputting the length parameter along the edge directly calculates the corresponding three-dimensional coordinates.

[0034] The centerline parametric equation can be defined as an equation based on the locus of the midpoints of the two edge lines. It describes the core axis of the tunnel's extension direction and serves as the benchmark for subsequent tunnel depth determination. Essentially, it can be understood as the mathematical expression of a spatial curve located at the center of the tunnel's cross-section and parallel to the two edge lines.

[0035] In one embodiment, a spatial index (e.g., a KD-tree) is established for the 3D point cloud of the tunnel to accelerate neighbor point search. For each point, the number of points within a certain range is counted, isolated noise points are removed, and density filtering is performed to obtain a point cloud set on the tunnel surface. Principal component analysis is performed on the local neighborhood of each point in the point cloud set on the tunnel surface. The eigenvector corresponding to the minimum eigenvalue is the normal direction, and the curvature is calculated based on the ratio of eigenvalues. A high curvature threshold is set, and points with curvature greater than the threshold are retained to obtain a candidate point cloud set. Spatial clustering is performed on the candidate point cloud set to obtain multiple clusters. The linearity of each cluster is calculated, and the two clusters that most resemble a straight line and have the most points are retained as the point clouds along the left and right sides.

[0036] Furthermore, for the two point clusters along the left and right edges, a three-dimensional spatial curve fitting is performed using piecewise polynomial curves to obtain the edge parameter equations corresponding to the left edge and the right edge. Substituting the same parameter into the two edge parameter equations, two edge points are obtained. The center point of the roadway is determined based on the edge points, and then the centerline parameter equation corresponding to the center point is obtained.

[0037] For example, suppose we scan a 100-meter-long tunnel using a laser scanner, obtaining 10 million points. After preprocessing, dust points floating in the air and cable points hanging on the walls are removed, leaving 8 million points representing the tunnel surface. The algorithm calculates the curvature of each point and finds that a row of points at the junction of the ground and the wall has particularly high curvature. These high points are clustered together, and two long, winding point bands are found through clustering. Curve fitting is performed on these two point bands to obtain two mathematical curves that accurately describe the direction of the left and right boundaries of the tunnel. On these two boundary curves, a point is taken every 1% of the length, and the corresponding points on the left and right are connected to obtain the midpoint of this line segment. Connecting all these midpoints forms a centerline running through the center of the tunnel.

[0038] Step 102: Along the centerline of the tunnel, based on the tunnel depth parameters and edge parameter equations corresponding to the current advancing position, determine the spatial region containing the current tunnel profile, and use the point cloud data extracted from the spatial region to fit and generate the closed curve parameter equation of the spatial region.

[0039] The tunnel depth parameter can refer to the position coordinates along the centerline direction, which is usually the normalized arc length from the starting point (e.g., 0.0 to 1.0). t=0.3 indicates the position at approximately 30% of the total tunnel length.

[0040] A spatial region can refer to a localized area in three-dimensional space surrounding the current profile location, used to capture the point cloud at that point. It is typically a hexahedron, such as... Figure 2 The diagram shown is a hexahedral schematic of the current spatial region of a tunnel according to an embodiment of this application.

[0041] A closed curve parametric equation can describe a closed, two-dimensional mathematical curve equation that represents the cross-sectional shape of a tunnel at a specific depth.

[0042] In one embodiment, based on the current tunnel depth parameter t, the center point P_center(t) is obtained by substituting it into the centerline parameter equation, and the left edge point P_left(t) and right edge point P_right(t) are obtained by substituting them into the edge parameter equation. Taking the left edge point P_left(t) (see...) Figure 2 The points p_pl and right edge P_right(t) shown in the diagram (see also p_pl) and right edge P_right(t) (see also p_right(t)) Figure 2 The line connecting pr (as shown) is taken as the current road cross-section. Using this cross-section as a reference, extend it upwards and downwards by a certain distance (e.g., 50% and 105% of the estimated tunnel height) to form the top and bottom faces of a hexahedron. Take a small step before and after t to obtain two more cross-sections. The vertical faces containing these two cross-sections constitute the front and back faces of the hexahedron. This defines a spatial region encompassing the current tunnel cross-section, such as... Figure 2Given a hexahedral spatial region, quickly query all point clouds located within this region. Project these points onto a two-dimensional profile plane perpendicular to the centerline, and use a closed curve fitting algorithm to fit these projected points into a smooth, closed two-dimensional curve, i.e., the parametric equation of the closed curve.

[0043] For example, suppose we need to process a profile of a tunnel 50 meters deep (t=0.5). First, calculate the precise locations of the center point, left edge point, and right edge point at 50 meters. Then, based on this, draw a rectangular box in 3D space that perfectly fits this tunnel section. The algorithm quickly filters out all points from 8 million points, identifying approximately 5,000 points that fall within this box. Next, flatten these 5,000 3D points onto a 2D plane perpendicular to the tunnel's orientation. Finally, like using an elastic, closed rope to lasso these scattered points, fit a closed curve that best represents the distribution shape of these points. This curve may not be a standard semicircular arch but an irregular shape, realistically reflecting over-excavation or deformation at this location.

[0044] Step 103: Combine the current tunnel depth parameters, closed curve parameter equations, and curve length parameters corresponding to the spatial region into a profile element and add it to the vectorized profile set of the tunnel.

[0045] A profile element can refer to a structured data unit that contains all the information describing a specific cross-section of a roadway.

[0046] A vectorized profile set can refer to a collection of all profile elements arranged in depth order, which is a complete, digital skeleton model of the tunnel.

[0047] The curve length parameter can refer to the parameter used to locate any point on a closed curve. The perimeter of the entire closed curve is normalized to 1, and the position of any point on the curve is represented by the relative distance (a decimal between 0 and 1) traveled along the curve from the starting point.

[0048] In one embodiment, the curve length parameter is determined by: obtaining the parametric equation of the closed curve; calculating the arc length function and the total arc length of the closed curve based on the parametric equation of the closed curve; normalizing the arc length function relative to the total arc length to obtain the normalized arc length function; and determining the range of the normalized arc length function as the curve length parameter.

[0049] For example, for the closed curve parametric equation fitted in the above steps, the cumulative arc length from the starting point to any point on the curve is calculated through numerical integration. The total arc length (circumference) of the entire closed curve is calculated. The cumulative arc length is divided by the total arc length to obtain the normalized arc length value s. The value range of s is [0,1), and s=0.5 indicates that the point is in the middle of the curve. The current tunnel depth parameter t, the closed curve parametric equation, and the total arc length of the curve are packaged into a data structure, which is a profile element. This profile element is added to a global list or array in order of the current tunnel depth parameter t value to form a vectorized profile set.

[0050] For example, for an irregular profile curve at a depth of 50 meters (t=0.5), the algorithm calculates its total perimeter to be 15.8 meters. There is a point on the curve that has traveled approximately 4.74 meters counterclockwise along the curve from its defined starting point (e.g., the intersection point along the right side of the curve). Therefore, the curve length parameter s at this point is approximately 4.74 / 15.8 ≈ 0.3. The algorithm then creates a record: {tunnel depth parameter: 0.5, curve equation: [...], curve length parameter: 0.3, ...} and stores this record in a table. The first column of the table is the tunnel depth parameter t, ranging from 0.0, 0.001, 0.002, ... up to 1.0, with each row corresponding to a profile element. This table can be understood as a vectorized profile set, precisely describing the entire tunnel in tabular form.

[0051] Step 104: Construct a three-dimensional mesh model of the tunnel based on the vectorized profile set.

[0052] In one embodiment, for each profile element in the vectorized profile set, its closed curve is decomposed into multiple geometric segments connected in sequence; each geometric segment is discretized into an ordered sequence of points; the two ordered sequence of points corresponding to the same geometric segment in two adjacent profile elements are connected to generate a set of triangular facets; and the sets of triangular facets corresponding to all geometric segments are combined to form a three-dimensional mesh model of the tunnel.

[0053] A geometric section can refer to several segments into which a closed tunnel profile curve is decomposed according to its geometric meaning. These are typically divided into: ground section, left wall section, right wall section, and arch section.

[0054] A triangular facet can refer to a triangle formed by three spatial points, and it is the basic unit of a three-dimensional mesh model (usually referring to a triangular mesh).

[0055] A three-dimensional mesh model can refer to a mesh model composed of thousands of connected triangular facets that simulates the inner and outer surfaces of a tunnel. It can be used for visualization, simulation analysis, etc.

[0056] For example, for each profile in the vectorized profile set, the intersection points of the curve and the left and right edges are identified, dividing the closed curve into four logical segments. For each segment (e.g., the arch segment), a series of two-dimensional coordinates of points are calculated according to a certain sampling density (e.g., one sample per 0.1-meter arc length), forming an ordered list. Two adjacent profiles are taken (e.g., the i-th and i+1-th profiles). For the same segment (e.g., both are arch segments), the point sequences of that segment on the two profiles are mapped one-to-one. The point A_j of the i-th profile is connected with the corresponding point B_j of the i+1-th profile and the next point B_{j+1}, forming two triangles (A_j, B_j, B_{j+1}) and (A_j, A_{j+1}, B_{j+1}). Iterate through all j, where j is an integer starting from 0 and incrementing (usually j = 0, 1, 2, ..., N-2), representing the position of the point in the two ordered point sequences currently being processed. This creates a ring of triangular facets between the two profiles. Repeat this process for all segments and all adjacent profiles.

[0057] For example, let's say we want to transform that data table (vectorized profile set) into a 3D tunnel that can be displayed on a computer. We take two adjacent profiles with tunnel depth parameters t=0.5 and t=0.501. We divide the closed curves of these two profiles into four segments: the ground surface, the left wall, the right wall, and the crown. Taking the crown segment as an example, the crown segment of the t=0.5 profile is discretized into 20 points, and the one of t=0.501 is also discretized into 20 points. We list these two points together like a zipper: using the first point of the first profile, the first point of the second profile, and the second point of the second profile, we form the first triangle; then using the first point of the first profile, the second point of the second profile, and the second point of the second profile, we form the second triangle... and so on. Between these two very close profiles, we use many small triangles to create a small section of the crown surface.

[0058] Furthermore, after performing this operation on all sections and all segments, countless tiny triangles are pieced together to form a complete, smooth, hollow three-dimensional tunnel model, which can be rotated and viewed in 3D software, and even used for ventilation simulation, safety distance analysis, etc.

[0059] The method provided in this application preprocesses the three-dimensional point cloud of the tunnel and extracts and fits the edge parameter equations of two road surface edges to determine the centerline parameter equation of the tunnel centerline. It then proceeds step by step along the centerline, determining a spatial region containing the current profile based on the current tunnel depth parameter and the edge parameter equation, and using the point cloud within the region to generate a closed curve parameter equation. The current tunnel depth parameter, the closed curve parameter equation, and the curve length parameter are used to construct profile elements and added to a vectorized profile set. Based on this vectorized profile set, a three-dimensional mesh model of the tunnel is constructed. By dynamically constructing an adaptive spatial region along the tunnel centerline and directly fitting closed curves of arbitrary shapes using the local point cloud within it to represent the tunnel cross-section, this method can accurately depict irregular changes in the tunnel cross-section caused by under-excavation, over-excavation, chambers, and deformation. The generated model highly matches the actual tunnel morphology, establishing a fully parameterized and vectorized digital tunnel model.

[0060] Figure 3 A flowchart illustrating another embodiment of a three-dimensional network model construction method provided in this application is shown below. Figure 1 Based on the illustrated process, this section mainly describes how to determine the edge parameter equations corresponding to the two edge lines of the tunnel, as well as the centerline parameter equation of the tunnel, based on the collected 3D point cloud data; see [link to documentation]. Figure 3 As shown, it includes the following steps: Step 301: Construct a spatial index structure for the 3D point cloud data, and perform density statistical analysis on the point cloud based on the spatial index structure to remove outlier noise points.

[0061] Spatial index structures refer to data structures that efficiently organize three-dimensional spatial data, enabling quick answers to queries about nearby points. Common examples include KD-trees and octrees.

[0062] Density statistical analysis can refer to a filtering method based on local point density. Specifically, it can count the number of points in the neighborhood of each point based on a spatial index structure to determine the distribution density of the points. Outlier noise points have a much lower number of neighborhood points than the main point cloud, and this can be used as a basis for filtering out outlier noise points. Furthermore, it can be understood that in three-dimensional space, points on the surface of an object should be dense and continuous, while noise points are often isolated and sparse.

[0063] Outlier noise points refer to isolated points in a 3D point cloud that are far from the main point cloud (such as points on the surface of a tunnel). Causes include multipath reflection from lidar, scattering from environmental dust / raindrops, and sensor errors. They manifest as abnormal coordinates and a lack of spatial continuity. Essentially, they can be understood as isolated points that do not accurately represent the surface of the tunnel, resulting from scanning errors, airborne dust, and reflection interference.

[0064] In one embodiment, three-dimensional point cloud data is used as input, and the space is recursively divided using coordinate axes (alternating X, Y, and Z) to form a binary tree structure. Each node represents a spatial region and contains points falling within that region. For each point in the point cloud data, the binary tree structure is used to quickly find its nearest neighbor points with a preset value (e.g., K), and the average distance and standard deviation from each point to these K neighbors are calculated. Points with an average distance greater than a preset distance threshold are identified as outlier noise points and are removed.

[0065] For example, suppose the lane point cloud has 5 million points. The program first constructs a binary tree structure (KD-tree), which can be understood as creating a 3D map index for all points. Then, it examines each point: for example, point A, and finds that there are an average of 40 points within a 0.1-meter radius around it, evenly distributed. Point B has only 3 points within the same radius, and they are far apart. Point B is therefore identified as an isolated outlier and removed from the data. This deletion step removes all outliers.

[0066] Step 302: Calculate the normal vector and curvature of each point in the point cloud after removing outlier noise points, and filter based on the curvature threshold to remove points that are not on the roadway surface, thus obtaining the point cloud set of the roadway surface.

[0067] A normal vector refers to the vector perpendicular to the local surface of a point in a point cloud. For smooth walls in a tunnel, the normal directions are roughly the same; for edge points, the normal directions change abruptly.

[0068] Curvature can be a mathematical quantity that describes the degree of curvature of a local surface. The curvature of a flat area is close to 0, while the curvature values ​​of edges, corners, and cylindrical surfaces (such as pipes) are large.

[0069] Points not on the roadway surface can refer to points in the mining roadway environment that, in addition to the top, bottom, and walls, may include pipes, cables, equipment, debris, etc., which are not part of the roadway structure itself.

[0070] In one embodiment, for each point after removing outlier noise points, a KD-tree is used to find all neighboring points within its radius R, where the radius R can be preset (e.g., R = 0.2 meters). This group of neighboring points is analyzed to calculate the covariance matrix. The eigenvector corresponding to the smallest eigenvalue of the covariance matrix is ​​the normal direction of that point (because the data changes least in the normal direction). The curvature c can be approximately calculated as: c = λ1 ÷ (λ1 + λ2 + λ3), where λ1, λ2, and λ3 are the eigenvalues ​​of the covariance matrix, and λ1 ≤ λ2 ≤ λ3. A larger c value indicates a larger curvature. A low first curvature threshold (e.g., 0.01) is set, and all points with curvature greater than the first curvature threshold are considered non-channel surface points and discarded.

[0071] For example, calculating the curvature of each point reveals that the curvature of points on the tunnel wall is less than 0.005, while the curvature of points on the surface of a pipe crossing the tunnel is greater than 0.1. Therefore, a threshold of 0.01 can be set, and all points in the pipe with curvature exceeding the threshold of 0.1 are marked as non-tunnel surface points and removed. What remains is a relatively flat point cloud of the tunnel top, bottom, and walls.

[0072] Step 303: Based on the curvature values ​​of each point in the point cloud set on the roadway surface, filter the candidate edge point cloud subset by using the curvature threshold.

[0073] Candidate edge point cloud subsets can refer to the set of high curvature points selected from the point cloud on the road surface. These points are potential components of the edge line (because the edge is where the shape of the road surface and the wall changes abruptly, and the local curvature is much higher than that of a flat ground / wall).

[0074] In one embodiment, based on the point cloud of the tunnel surface obtained by the above steps, a higher second curvature threshold (e.g., 0.02-0.05) is reset, all points are traversed, and points with curvature greater than the second curvature threshold are retained.

[0075] For example, in a point cloud of a clean alleyway surface, the program searches for points that are not so flat, i.e., points with higher curvature. For instance, the curvature of a point in the middle of a wall is 0.001, the curvature of a point in the middle of the ground is 0.002, while the curvature of a point at the boundary between the ground and the wall can reach 0.03. Therefore, the program can set a threshold of 0.02 to filter all candidate points with curvature greater than the threshold of 0.2, because the higher curvature of these points indicates that they are located on the boundary edge, and these candidate points are used to construct a subset of candidate edge point clouds.

[0076] Step 304: Perform distance clustering on the candidate edge point cloud subsets to obtain multiple initial clusters; select the initial clusters that have a linear spatial distribution and contain more than a preset threshold as the target candidate edge point cloud subsets from the multiple initial clusters.

[0077] Distance clustering can refer to clustering algorithms based on spatial distance (such as Euclidean distance clustering), which group points whose distance between two points is less than or equal to the cluster radius into the same class, achieving grouping of similar points densely and dissimilar points sparsely.

[0078] Linear spatial distribution can refer to the continuous linear arrangement of points within a cluster along a certain direction (such as the left and right road edges extending along the length of the alley). Essentially, it can refer to the distribution of point sets in space as slender, strip-shaped rather than clumped or spherical.

[0079] In one embodiment, a spatial clustering algorithm is run on a subset of candidate edge point clouds. The algorithm groups density-connected points that are mutually reachable into the same initial cluster. For each initial cluster, the covariance matrix of all its points is calculated and analyzed. If the largest eigenvalue is much larger than the other two eigenvalues ​​(e.g., by an order of magnitude), it indicates that the point set extends along a main direction and is linear. Simultaneously, the number of points within each cluster is checked, and initial clusters with too few points (e.g., <50) are discarded. These initial clusters may be noisy or have short edges and therefore need to be removed.

[0080] For example, suppose there are 10,000 points in the candidate edge point cloud subset. After spatial clustering, five initial clusters are obtained. Initial cluster 1 has 3,000 points, which are elongated and thin (high linearity); initial cluster 2 has 2,500 points, which are also elongated and thin; initial cluster 3 has 200 points, which are small clusters (possibly in a device corner); initial clusters 4 and 5 each have fewer than 50 points. The program selects initial cluster 1 and initial cluster 2 as the target candidate edge point cloud subset.

[0081] Step 305: Perform distance clustering again on the target candidate edge point cloud subset to obtain point cloud sets of two edge lines that run through the length of the tunnel and are linearly spatially distributed.

[0082] Re-distance clustering refers to performing finer distance clustering on the already selected linear clusters. By adjusting the clustering parameters, it addresses potential issues from the initial clustering, such as over-merging of similar points or misclassification of dissimilar points. The goal is to ensure that each cluster strictly corresponds to a continuous edge line. Due to the long tunnels, scans may be discontinuous, or the edge line may be partially obscured by obstacles, causing the same edge line to be spatially segmented into several segments. Therefore, it is necessary to use a larger connection distance to glue these segments together, i.e., cluster them together.

[0083] In one embodiment, each subset of target candidate edge point clouds selected in the above steps (which may correspond to a segment of the left or right edge) is subjected to Euclidean space clustering. The distance threshold used this time is much larger than that used in the first distance clustering (e.g., 0.5 meters or even 1 meter), with the aim of merging spatially adjacent segments that belong to the same physical edge line but are broken for some reason into a complete edge line point cloud set.

[0084] For example, the target candidate initial cluster 1 (left edge) may consist of 3 segments, each spaced 0.3 meters apart (not detected due to scanning blind spots). Using a distance threshold of 0.5 meters for Euclidean clustering, these 3 segments, because their distance from each other is less than 0.5 meters, will be merged into a large initial cluster, representing the complete point cloud set of the left edge. The same applies to the right edge, ultimately resulting in point clouds set along both edges.

[0085] Step 306: If the number of point cloud sets is greater than two, determine the average spatial coordinates of each point cloud set, and select the two point cloud sets with the smallest vertical height as the final point cloud sets for the left and right road edge lines.

[0086] The average spatial coordinates can refer to the average of the x, y, and z coordinates of all points in each point cluster, which is used to reflect the overall spatial location of the cluster.

[0087] Vertical height can refer to the average z-coordinate (vertical height) of the cluster. For example, the road edge is located at the junction of the road surface and the wall, with a z-coordinate close to the ground level (e.g., 1.0-1.2m), while the top z-coordinate is close to the vault (e.g., 3.5-4.0m), and the bottom z-coordinate is close to below the ground level (e.g., 0.5m).

[0088] In one embodiment, when more than two point sets with linear spatial distribution are obtained, the average spatial coordinates of each point set are calculated, and the two point sets with the smallest average spatial coordinates are selected as the final left and right edge point sets.

[0089] For example, suppose clustering results in three linear point clouds: point cloud A with an average height of -50 meters, point cloud B with an average height of -50.1 meters, and point cloud C with an average height of -48 meters (possibly a tall pipeline rack). In this case, point cloud A and point cloud B with the lowest heights are selected and identified as the left and right edges of the line.

[0090] Step 307: Perform piecewise continuous differentiable spatial curve fitting on the point cloud sets of the left and right road edge lines respectively to obtain the piecewise parametric equations of the two edge lines as edge parametric equations.

[0091] Piecewise continuous differentiable spatial curve fitting refers to dividing the edge line into several segments according to the curvature change, and fitting each segment with a quadratic / cubic polynomial. Adjacent segments satisfy coordinate continuity (x, y, z are equal) and derivative continuity (tangent direction is consistent) at the segmentation point, ensuring that the entire curve is smooth without discontinuities or sharp corners.

[0092] Piecewise parametric equations can refer to equations that use arc length as a parameter, with each segment having its own function expression. For a specific arc length parameter, the corresponding coordinates can be calculated using the corresponding function expression.

[0093] In one embodiment, for each edge line point cloud, an approximate arc length parameter value is assigned to each point by accumulating the actual distances between adjacent points, using the starting point as the corresponding arc length parameter. The entire line is divided into several segments according to arc length (e.g., every 10 meters). For points within each segment, parametric equations consisting of three functions x(t), y(t), and z(t) are fitted with t as the independent variable, constraining that the three function values ​​and the first derivatives are equal at the connection points of adjacent segments.

[0094] For example, there are 1000 points along the left side of the line, with a total length of 150 meters. This line is divided into 15 segments, each 10 meters long. For the points in the first segment (0-10 meters), the parametric equations of x_l1(t), y_l1(t), and z_l1(t) along the left side of the line are fitted. For the points in the second segment (10-20 meters), the fitting requires that the function value of the second segment when the arc length parameter is 10 is equal to the function value of the first segment when the arc length parameter is 10, and that the derivatives are also the same. This results in a smooth mathematical curve describing the spatial orientation of the left side of the line.

[0095] Step 308: Based on the piecewise parametric equations of the two edge lines in the edge parameter equation, calculate the coordinates of the corresponding three-dimensional points on the left and right edge lines respectively for the same target parameter value.

[0096] Step 309: Based on the target parameter value, determine the midpoint coordinates of the corresponding points along the left and right sides of the line.

[0097] Step 310: Using the target parameter value as the independent variable and the midpoint coordinate as the dependent variable, form the centerline parametric equation of the roadway centerline.

[0098] The following is a unified explanation of steps 308-310 above: The target parameter value can refer to a parameter t, which can be any value. This parameter can be substituted into the corresponding piecewise parametric equation to calculate the edge points on the corresponding edge line.

[0099] In one embodiment, for any given parameter t (e.g., t=0.25), substitute it into the piecewise parametric equations along the left and right sides of the tunnel to obtain two three-dimensional point coordinates, P_left(t) and P_right(t). Calculate the midpoint P_mid(t) = (P_left(t) + P_right(t)) ÷ 2. Iterate through the parameter t from 0 to 1 to calculate a series of midpoints P_mid(t). Perform piecewise curve fitting on these midpoints to obtain the parametric equations for the centerline of the tunnel.

[0100] For example, at 25% of the total tunnel length (t=0.25), the equation along the left side gives the point (10, 5, -50), and the equation along the right side gives the point (30, 5, -50). The midpoint is (20, 5, -50). Fitting all the midpoints calculated from the starting point (t=0) to the ending point (t=1) yields a parametric equation for a centerline that runs through the center of the tunnel, parallel to both sides, and equidistant from them. This centerline is not directly extracted from the point cloud, but rather a perfect spine mathematically derived from the two edge lines, which will serve as a unified reference benchmark for all subsequent modeling steps.

[0101] Step 311: Along the centerline of the tunnel, based on the tunnel depth parameters and edge parameter equations corresponding to the current advancing position, determine the spatial region containing the current tunnel profile, and use the point cloud data extracted from the spatial region to fit and generate the closed curve parameter equation of the spatial region.

[0102] Step 312: Combine the current tunnel depth parameters, closed curve parameter equations, and curve length parameters corresponding to the spatial region into a profile element and add it to the vectorized profile set of the tunnel.

[0103] Step 313: Based on the vectorized profile set, construct a three-dimensional mesh model of the tunnel.

[0104] For steps 311-313 above, please refer to the above. Figure 1 Detailed description of the relevant embodiments.

[0105] pass Figure 3 The illustrated embodiment describes how spatial indexing and density filtering effectively remove scanning noise, and how normal and curvature features are used to accurately separate non-tunnel structures such as pipes. Then, a curvature threshold and two-level distance clustering strategy is employed to stably identify linearly distributed left and right roadway edges from complex point clouds, and height filtering is used to eliminate interference. Finally, piecewise, continuously differentiable spatial curves are fitted to the edges, and the parametric equations for the tunnel centerline are automatically generated through mathematical derivation. This process overcomes the limitations of traditional methods in complex, irregular mine tunnel environments, such as sensitivity to noise, reliance on manual intervention, and difficulty in handling deformation and occlusion. It lays a high-precision, highly reliable data foundation for subsequent vectorized modeling and is a key step in realizing fully automated digital modeling of smart mine tunnels.

[0106] Figure 4 A flowchart illustrating another embodiment of the three-dimensional network model construction method provided in this application is shown below. Figure 1 Based on the illustrated process, this section mainly describes how to generate the parametric equations of the closed curve and how to construct the three-dimensional network model of the tunnel. (See [link to relevant documentation]). Figure 4 As shown, it includes the following steps: Step 401: Preprocess the acquired 3D point cloud data of the tunnel to obtain the point cloud set of the tunnel surface. Extract the point cloud sets of the two edge lines of the tunnel surface from the point cloud set of the tunnel surface. Fit the point cloud sets of the two edge lines to obtain the edge parameter equations corresponding to the two edge lines. Based on the edge parameter equations, determine the centerline parameter equation of the tunnel centerline.

[0107] For step 401 above, please refer to the above. Figure 1 Detailed description of the relevant embodiments.

[0108] Step 402: Based on the current tunnel depth parameter corresponding to the current advance position, and based on the piecewise parameter equations of the two edge lines in the edge parameter equation, calculate the first edge point on the left edge line and the second edge point on the right edge line respectively.

[0109] The current advancing position can refer to a specific location along the centerline of the roadway that the algorithm is currently processing.

[0110] The current tunnel depth parameter can be a normalized value representing the relative distance between the current processing position and the tunnel start point. Essentially, it's a kilometer that advances progressively along the centerline; for example, t=0.35 could represent the current processing position being at 35% of the total tunnel length.

[0111] The first edge point can refer to the precise three-dimensional spatial point at the junction of the road surface and the wall on the left side of the tunnel on the tunnel cross-section corresponding to the depth parameter.

[0112] The second edge point can refer to the precise three-dimensional spatial point at the junction of the road surface and the wall on the right side of the tunnel on the tunnel cross-section corresponding to the depth parameter.

[0113] In one embodiment, the same tunnel depth parameter value is substituted into both the left and right edge parameter equations to obtain the coordinates of the first edge point and the second edge point.

[0114] For example, suppose the algorithm processes 35% of the total tunnel length (t=0.35), and inputs this "mileage" of 0.35 into the piecewise parametric equations of the left and right edges. The formula immediately outputs two points: the coordinates of the first edge point on the left may be (105.2, 45.7, -120.5), and the coordinates of the second edge point on the right may be (125.8, 45.5, -120.3).

[0115] Step 403: Connect the first edge point and the second edge point to obtain the current road surface cross section.

[0116] The current road cross section can refer to the points connecting the left and right edges of the current profile in three-dimensional space (e.g., ...). Figure 2 A straight line segment (pl and pr) shown represents the visual representation of the road width in three-dimensional space at the current location of the alley.

[0117] In one embodiment, in a three-dimensional coordinate system, points pl and pr define a line segment. This line segment is the road surface cross section.

[0118] For example, in three-dimensional space, a line segment is drawn from the first edge point on the left (105.2, 45.7, -120.5) to the second edge point on the right (125.8, 45.5, -120.3). This line segment, approximately 20.6 meters long, accurately marks the actual width and spatial orientation of the road surface at 35% depth.

[0119] Step 404: Based on the current road cross section and combined with the preset height range of the tunnel, determine a hexahedral spatial region in three-dimensional space that contains the current tunnel profile.

[0120] A hexahedral spatial region can refer to a cubic region in three-dimensional space, resembling a flattened box. Its core design goal is to exactly enclose the cross-section at the current tunnel depth, so as to extract a local point cloud from the entire point cloud to fit the shape of that section.

[0121] The preset height range can refer to the estimated height value of the tunnel based on the tunnel design or scanned point cloud. For example, the height from the floor to the vault can be preset to 4 meters.

[0122] In one embodiment, the vertical plane containing the current road surface cross section is used as a symmetry reference plane. This plane is shifted downwards by 5% of the preset tunnel height (e.g., 5% of 4 meters is 0.2 meters) to determine the bottom surface of the spatial region, ensuring it includes any possible road surface undulations or debris. This plane is then shifted upwards by 105% of the preset tunnel height (e.g., 105% of 4 meters is 4.2 meters) to determine the top surface of the spatial region, ensuring it includes the complete vault. A dynamic step size is taken before and after the current tunnel depth parameter containing the current road surface cross section. Steps 402-403 are repeated based on this dynamic compensation to obtain the previous and next cross sections. The vertical planes containing these two cross sections constitute the front and back of the hexahedron. The four edges of the front and back (line segments connecting the left and right edge points) naturally enclose four sides in space. The spatial curved surface (approximate plane) formed by connecting the left edge points is the left side, and the same applies to the right side. Ultimately, the three-dimensional region enclosed by these six faces is a slice of pipe that closely follows the surface of the tunnel and bends with the curves of the tunnel, perfectly encompassing all surface point clouds near the current profile.

[0123] For example, see Figure 2 As shown, A' and B' are two adjacent points on the left edge, calculated based on the first edge point pl on the left edge and the step length of the left edge. C' and D' are two adjacent points on the left edge, calculated based on the first edge point pr on the right edge and the step length of the right edge. Figure 2The front and back faces of the hexahedral spatial region shown are formed by the vertical planes containing the extended transverse sections AC (obtained by extending A'C') and BD (obtained by extending B'D'). Similarly, the left and right faces are formed by the vertical planes (or approximately vertical curved surfaces) containing the edge connecting lines AB and CD. The top and bottom faces are formed by two horizontal planes (or inclined planes) determined according to the height range.

[0124] Step 405: Using the point cloud data extracted from the spatial region, fit and generate the parametric equation of the closed curve of the spatial region.

[0125] In one embodiment, if no historical closed curve parametric equation exists corresponding to the current spatial region, a new closed curve parametric equation is directly generated by fitting the currently extracted point cloud data; if a historical closed curve parametric equation exists corresponding to the current spatial region, the fitting error between the currently extracted point cloud data and the historical closed curve parametric equation is calculated; if the fitting error is less than or equal to a preset error threshold, the historical closed curve parametric equation is used or finely adjusted as the current closed curve parametric equation; if the fitting error is greater than the preset error threshold, a new closed curve parametric equation is refitted using the currently extracted point cloud data.

[0126] The parametric equation of a historical closed curve can refer to the curve equation that fits the profile at the previous advance position (e.g., t=0.349).

[0127] Fitting error can be used to measure how well newly extracted point cloud data matches an existing curve model. It is commonly calculated using the average distance from a point to the curve.

[0128] The preset error threshold can refer to a pre-defined tolerance value, such as 5 centimeters. If the error is less than this value, the change in cross-sectional shape is considered negligible.

[0129] For example, at t=0.35, 5000 points are extracted, and the average distance from each point to the historical curve at t=0.349 is calculated to be 2 cm. Since 2 cm is less than the preset threshold of 5 cm, the shape is considered to have not changed significantly, and the previous historical closed curve parametric equation can be used, saving computation. At t=0.65, the extracted point cloud shows that the right wall suddenly recesses into a large hole (tunnel). The average distance of these points to the historical curve reaches 30 cm, far exceeding the threshold, indicating a drastic change in shape! The old curve is no longer suitable. At this point, the 5000 new points are used to refit the curve, resulting in a completely new closed curve parametric equation with a concave right side, thus accurately capturing the opening shape of the tunnel.

[0130] Step 406: Combine the current tunnel depth parameters, closed curve parameter equations, and curve length parameters corresponding to the spatial region into a profile element and add it to the vectorized profile set of the tunnel.

[0131] Step 407: Based on the vectorized profile set, construct a three-dimensional mesh model of the tunnel.

[0132] For steps 406-407 above, please refer to the above. Figure 1 Detailed description of the relevant embodiments.

[0133] Through the above Figure 4 The illustrated embodiment describes how, by dynamically constructing a hexahedral spatial region that adapts to the tunnel morphology, accurate extraction and high-fidelity fitting of local point clouds at any location within the tunnel are achieved. Its core lies in introducing an adaptive fitting mechanism based on error judgment, which intelligently uses, fine-tunes, or reconstructs the closed-section curves of the profile. This ensures modeling efficiency while accurately depicting abrupt changes in the tunnel's cross-section caused by factors such as chambers and deformation. This process ultimately generates a structured vectorized profile set, providing a reliable data foundation for constructing a high-precision, computable, and analyzable 3D mesh model, significantly improving the automation and geometric realism of digital modeling of complex mine tunnels.

[0134] Figure 5 This application provides a structural block diagram of a three-dimensional network model construction device, the device comprising: The parameter equation determination module 51 is used to preprocess the acquired three-dimensional point cloud data of the tunnel to obtain the point cloud set of the tunnel surface, extract the point cloud sets of the two edge lines of the tunnel surface from the point cloud set of the tunnel surface, fit the point cloud sets of the two edge lines respectively to obtain the edge parameter equations corresponding to the two edge lines, and determine the center line parameter equation of the tunnel center line based on the edge parameter equations. The closed curve parametric equation determination module 52 is used to determine the spatial region containing the current roadway profile along the centerline direction of the roadway, based on the roadway depth parameter corresponding to the current advancing position and the edge parametric equation, and to fit and generate the closed curve parametric equation of the spatial region using the point cloud data extracted from the spatial region. The vectorized profile set construction module 53 is used to construct a profile element by taking the current tunnel depth parameter corresponding to the spatial region, the closed curve parameter equation and the curve length parameter, and adding it to the vectorized profile set of the tunnel. The three-dimensional network model construction module 54 is used to construct a three-dimensional mesh model of the tunnel based on the vectorized profile set.

[0135] In one possible implementation, the parametric equation determination module 51 is specifically used for: A spatial index structure for the 3D point cloud data is constructed, and density statistical analysis is performed on the point cloud based on the spatial index structure to remove outlier noise points. Calculate the normal vector and curvature of each point in the point cloud after removing outlier noise points, and filter based on the curvature threshold to remove points that are not on the roadway surface, thus obtaining the point cloud set of the roadway surface.

[0136] In one possible implementation, the parametric equation determination module 51 is specifically used for: Based on the curvature values ​​of each point in the point cloud set on the surface of the alley, a subset of candidate edge point clouds is selected by using a curvature threshold. Distance clustering is performed on the candidate edge point cloud subset to obtain multiple initial clusters; from the multiple initial clusters, the initial clusters that have a linear spatial distribution and contain more than a preset threshold of points are selected as the target candidate edge point cloud subset. The target candidate edge point cloud subset is further clustered by distance to obtain point cloud sets of two edge lines that run through the length of the tunnel and are linearly spatially distributed. If the number of point cloud sets is greater than two, then the average spatial coordinates of each point cloud set are determined, and the two point cloud sets with the smallest vertical height are selected as the final point cloud sets for the left and right road edge lines. Piecewise continuously differentiable spatial curve fitting is performed on the point clouds of the left and right road edge lines respectively to obtain the piecewise parametric equations of the two edge lines as edge parametric equations.

[0137] In one possible implementation, the parametric equation determination module 51 is specifically used for: Based on the piecewise parameter equations of the two edge lines in the edge parameter equation, for the same target parameter value, calculate the coordinates of the corresponding three-dimensional spatial points on the left and right edge lines respectively. Based on the target parameter value, determine the coordinates of the midpoint between the corresponding point on the left side and the corresponding point on the right side; Using the target parameter value as the independent variable and the midpoint coordinates as the dependent variable, the centerline parametric equation of the roadway centerline is formed.

[0138] In one possible implementation, the closed curve parametric equation determination module 52 is specifically used for: Based on the current tunnel depth parameter corresponding to the current advance position, and based on the piecewise parameter equations of the two edge lines in the edge parameter equation, the first edge point on the left edge line and the second edge point on the right edge line are calculated respectively. Connect the first edge point and the second edge point to obtain the current road surface cross section; Based on the current road cross section and combined with the preset height range of the tunnel, a hexahedral spatial region containing the current tunnel profile is determined in three-dimensional space.

[0139] In one possible implementation, the closed curve parametric equation determination module 52 is specifically used for: If there is no historical closed curve parametric equation corresponding to the current spatial region, then a new closed curve parametric equation is directly fitted using the currently extracted point cloud data. If a historical closed curve parametric equation exists corresponding to the current spatial region, the fitting error between the currently extracted point cloud data and the historical closed curve parametric equation is calculated. Based on the fitting error and a preset error threshold, the closed curve parametric equation for the spatial region is determined. If the fitting error is less than or equal to the preset error threshold, the historical closed curve parametric equation is used or finely adjusted as the current closed curve parametric equation. If the fitting error is greater than the preset error threshold, a new closed curve parametric equation is generated by refitting the currently extracted point cloud data.

[0140] In one possible implementation, the curve length parameter is determined in the following way: Obtain the parametric equation of the closed curve, and based on the parametric equation of the closed curve, calculate the arc length function and the total arc length of the closed curve; The arc length function is normalized relative to the total arc length to obtain the normalized arc length function. The range of the normalized arc length function is determined as the curve length parameter.

[0141] like Figure 6 As shown in the figure, this application provides an electronic device, including a processor 111, a communication interface 112, a memory 113, and a communication bus 114, wherein the processor 111, the communication interface 112, and the memory 113 communicate with each other through the communication bus 114. Memory 113 is used to store computer programs; In one embodiment of this application, when the processor 111 executes the program stored in the memory 113, it implements the three-dimensional network model construction method provided in any of the foregoing method embodiments, including: The acquired three-dimensional point cloud data of the tunnel is preprocessed to obtain the point cloud set of the tunnel surface. The point cloud sets of the two edge lines of the tunnel surface are extracted from the point cloud set. The point cloud sets of the two edge lines are fitted to obtain the edge parameter equations corresponding to the two edge lines. Based on the edge parameter equations, the center line parameter equation of the tunnel center line is determined. Along the centerline of the tunnel, based on the tunnel depth parameters corresponding to the current advancing position and the edge parameter equation, a spatial region containing the current tunnel profile is determined. Using the point cloud data extracted from the spatial region, a closed curve parameter equation of the spatial region is fitted and generated. The current tunnel depth parameter corresponding to the spatial region, the closed curve parametric equation, and the curve length parameter are combined to form a profile element, which is then added to the vectorized profile set of the tunnel. Based on the vectorized profile set, a three-dimensional mesh model of the tunnel is constructed.

[0142] This application also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the three-dimensional network model construction method provided in any of the foregoing method embodiments.

[0143] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0144] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented using software plus a general-purpose hardware platform, or of course, using hardware. Based on this understanding, the above technical solutions, in essence or the parts that contribute to the related technology, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0145] It should be understood that the terminology used herein is for the purpose of describing particular exemplary embodiments only and is not intended to be limiting. Unless the context clearly indicates otherwise, the singular forms “a,” “an,” and “described” as used herein may also mean including the plural forms. The terms “comprising,” “including,” “containing,” and “having” are inclusive and therefore indicate the presence of the stated features, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, elements, components, and / or combinations thereof. The method steps, processes, and operations described herein are not construed as requiring them to be performed in a particular order described or illustrated unless the order of performance is explicitly indicated. It should also be understood that additional or alternative steps may be used.

[0146] The above description is merely a specific embodiment of this application, enabling those skilled in the art to understand or implement this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features claimed herein.

Claims

1. A method for constructing a three-dimensional network model, characterized in that, The method includes: The acquired three-dimensional point cloud data of the tunnel is preprocessed to obtain the point cloud set of the tunnel surface. The point cloud sets of the two edge lines of the tunnel surface are extracted from the point cloud set of the tunnel surface. The point cloud sets of the two edge lines are fitted to obtain the edge parameter equations corresponding to the two edge lines. Based on the edge parameter equations, the center line parameter equation of the tunnel center line is determined. Along the centerline of the tunnel, based on the tunnel depth parameters corresponding to the current advancing position and the edge parameter equation, a spatial region containing the current tunnel profile is determined. Using the point cloud data extracted from the spatial region, a closed curve parameter equation of the spatial region is fitted and generated. The current tunnel depth parameter corresponding to the spatial region, the closed curve parametric equation, and the curve length parameter are combined to form a profile element, which is then added to the vectorized profile set of the tunnel. Based on the vectorized profile set, a three-dimensional mesh model of the tunnel is constructed.

2. The method according to claim 1, characterized in that, The preprocessing of the acquired 3D point cloud data of the tunnel to obtain a point cloud set on the tunnel surface includes: A spatial index structure for the 3D point cloud data is constructed, and density statistical analysis is performed on the point cloud based on the spatial index structure to remove outlier noise points. Calculate the normal vector and curvature of each point in the point cloud after removing outlier noise points, and filter based on the curvature threshold to remove points that are not on the roadway surface, thus obtaining the point cloud set of the roadway surface.

3. The method according to claim 1, characterized in that, The step of extracting point cloud sets of two edge lines of the roadway surface from the point cloud set of the roadway surface, and fitting the point cloud sets of the two edge lines to obtain the edge parameter equations corresponding to the two edge lines, includes: Based on the curvature values ​​of each point in the point cloud set on the surface of the alley, a subset of candidate edge point clouds is selected by using a curvature threshold. Distance clustering is performed on the candidate edge point cloud subset to obtain multiple initial clusters; from the multiple initial clusters, the initial clusters that have a linear spatial distribution and contain more than a preset threshold of points are selected as the target candidate edge point cloud subset. The target candidate edge point cloud subset is further clustered by distance to obtain point cloud sets of two edge lines that run through the length of the tunnel and are linearly spatially distributed. If the number of point cloud sets is greater than two, then the average spatial coordinates of each point cloud set are determined, and the two point cloud sets with the smallest vertical height are selected as the final point cloud sets for the left and right road edge lines. Piecewise continuously differentiable spatial curve fitting is performed on the point clouds of the left and right road edge lines respectively to obtain the piecewise parametric equations of the two edge lines as edge parametric equations.

4. The method according to claim 1, characterized in that, The determination of the centerline parameter equation of the roadway centerline based on the edge parameter equation includes: Based on the piecewise parameter equations of the two edge lines in the edge parameter equation, for the same target parameter value, calculate the coordinates of the corresponding three-dimensional spatial points on the left and right edge lines respectively. Based on the target parameter value, determine the coordinates of the midpoint between the corresponding point on the left side and the corresponding point on the right side; Using the target parameter value as the independent variable and the midpoint coordinates as the dependent variable, the centerline parametric equation of the roadway centerline is formed.

5. The method according to claim 1, characterized in that, Along the centerline direction of the tunnel, based on the tunnel depth parameters corresponding to the current advancing position and the edge parameter equation, a spatial region containing the current tunnel profile is determined, including: Based on the current tunnel depth parameter corresponding to the current advance position, and based on the piecewise parameter equations of the two edge lines in the edge parameter equation, the first edge point on the left edge line and the second edge point on the right edge line are calculated respectively. Connect the first edge point and the second edge point to obtain the current road surface cross section; Based on the current road cross section and combined with the preset height range of the tunnel, a hexahedral spatial region containing the current tunnel profile is determined in three-dimensional space.

6. The method according to claim 1, characterized in that, The step of using point cloud data extracted from the spatial region to fit and generate the parametric equation of the closed curve of the spatial region includes: If there is no historical closed curve parametric equation corresponding to the current spatial region, then a new closed curve parametric equation is directly fitted using the currently extracted point cloud data. If a historical closed curve parametric equation exists corresponding to the current spatial region, the fitting error between the currently extracted point cloud data and the historical closed curve parametric equation is calculated. Based on the fitting error and a preset error threshold, the closed curve parametric equation for the spatial region is determined. If the fitting error is less than or equal to the preset error threshold, the historical closed curve parametric equation is used or finely adjusted as the current closed curve parametric equation. If the fitting error is greater than the preset error threshold, a new closed curve parametric equation is generated by refitting the currently extracted point cloud data.

7. The method according to claim 1, characterized in that, The curve length parameter is determined in the following way: Obtain the parametric equation of the closed curve, and based on the parametric equation of the closed curve, calculate the arc length function and the total arc length of the closed curve; The arc length function is normalized relative to the total arc length to obtain the normalized arc length function. The range of the normalized arc length function is determined as the curve length parameter.

8. A three-dimensional network model construction device, characterized in that, The device includes: The parameter equation determination module is used to preprocess the acquired three-dimensional point cloud data of the tunnel to obtain the point cloud set of the tunnel surface, extract the point cloud sets of the two edge lines of the tunnel surface from the point cloud set of the tunnel surface, fit the point cloud sets of the two edge lines respectively to obtain the edge parameter equations corresponding to the two edge lines, and determine the center line parameter equation of the tunnel center line based on the edge parameter equations. The closed curve parametric equation determination module is used to determine the spatial region containing the current roadway profile along the centerline direction of the roadway, based on the roadway depth parameter corresponding to the current advancing position and the edge parametric equation, and to fit and generate the closed curve parametric equation of the spatial region using point cloud data extracted from the spatial region. The vectorized profile set construction module is used to construct a profile element by taking the current tunnel depth parameter corresponding to the spatial region, the closed curve parameter equation and the curve length parameter, and adding it to the vectorized profile set of the tunnel. A 3D network model construction module is used to construct a 3D mesh model of the tunnel based on the vectorized profile set.

9. An electronic device, characterized in that, include: A processor and a memory, the processor being configured to execute a three-dimensional network model building program stored in the memory to implement the three-dimensional network model building method according to any one of claims 1-7.

10. A storage medium, characterized in that, The storage medium stores one or more programs, which can be executed by one or more processors to implement the three-dimensional network model construction method according to any one of claims 1-7.

Images