Multi-shape tunnel three-dimensional modeling method and system based on laser point cloud

Abstract

The invention provides a multi-shape tunnel three-dimensional modeling method and system based on laser point clouds, and relates to the technical field of tunnel surveying and mapping data processing.The method comprises the steps that tunnel girdles and tunnel sections are extracted from point cloud data of a tunnel in a segmented mode; calculating the minimum outer surrounding circle of the tunnel section; determining a projection cylinder of the tunnel girdle; projecting the point cloud of the tunnel girdle to the surface of the projection cylinder through the minimum outer surrounding circle; generating a two-dimensional intensity image according to the projection cylinder and the tunnel girdle; obtaining a standard section through a tunnel section; correcting the two-dimensional intensity image according to the minimum outer surrounding circle and the standard section; based on an improved Poisson reconstruction algorithm, obtaining a tunnel three-dimensional grid model; and generating a tunnel three-dimensional texture model in combination with the tunnel three-dimensional grid model and the corrected two-dimensional intensity image. According to the method, rapid browsing of the three-dimensional texture model of the subway tunnel in multiple shapes is realized, and an effective way is provided for visually and stereoscopically detecting the safety state of the tunnel structure.

Description

Technical Field

[0001] This application relates to the technical field of tunnel surveying and mapping data processing, and particularly to a three-dimensional modeling method and system for multi-shaped tunnels based on laser point clouds. Background Art

[0002] High-quality textures and high-precision three-dimensional tunnel models are the prerequisites for realizing digital tunnels. Mobile laser scanning technology provides important basic data for the construction of three-dimensional models. It is not only widely used to obtain detailed features of the tunnel lining surface but also to obtain high-precision three-dimensional point clouds of spatial coordinates. However, it is still a challenge to establish a three-dimensional tunnel model with structural integrity and high-quality textures using mobile laser scanning data, especially for tunnels of different shapes.

[0003] To visually and stereoscopically detect the safety status of the tunnel structure, using three-dimensional laser point clouds as the data source, first generate a two-dimensional texture image and a three-dimensional tunnel model based on the tunnel point clouds, then establish the mapping relationship between the two-dimensional texture image and the three-dimensional tunnel model, and finally inlay the two-dimensional texture onto the three-dimensional model to achieve three-dimensional reconstruction. However, the existing methods and technologies have the following problems:

[0004] (1) Currently, the methods for generating two-dimensional texture images mainly target circular or elliptical tunnels. However, in actual engineering environments, the common shapes of tunnels also include horseshoe-shaped, arched, rectangular, etc. Most methods need to preset cross-section fitting formulas and projection parameters according to the shape of the tunnel. When facing complex tunnel shapes, there are often large deformation problems in the resulting images.

[0005] (2) Among the current methods and technologies for constructing three-dimensional tunnel models, triangular meshes are mostly used to construct models and traditional Poisson reconstruction algorithms. The former has a fast modeling speed, but is extremely sensitive to point cloud noise and is not suitable for large-scale point cloud reconstruction. The latter has a high tolerance for noise and incomplete point clouds, can process large-scale point cloud data in a parallel computing environment, and generate smooth models. However, when applied to the construction of non-closed tunnel models, it is easy to generate pseudo patches, resulting in low model accuracy and affecting the correctness of the safety monitoring and maintenance decisions of the tunnel by the staff.

[0006] (3) Currently, many methods and technologies mainly rely on three-dimensional grid tunnel models to reflect the actual situation of the tunnel, but these models lack detailed information such as textures, so they cannot accurately present the true situation inside the tunnel. Summary of the Invention

[0007] The purpose of the present invention is to provide a three-dimensional modeling method and system for multi-shaped tunnels based on laser point clouds to solve the problems of insufficient reconstruction accuracy, processing speed, and automation degree of different-shaped tunnels in the prior art when performing three-dimensional reconstruction based on large-scale tunnel point clouds.

[0008] The above object of the present application is achieved by the following technical solutions:

[0009] S1: Obtain the point cloud data of a tunnel with any shape and perform preprocessing;

[0010] S2: Segment and extract the tunnel annulus and the tunnel cross-section from the preprocessed point cloud data;

[0011] S3: Calculate the minimum circumscribed circle of the tunnel cross-section; determine the projection cylinder of the tunnel annulus; project the point cloud of the tunnel annulus onto the surface of the projection cylinder through the minimum circumscribed circle;

[0012] S4: Generate a two-dimensional intensity image of the inner surface of the tunnel lining according to the projection cylinder and the point cloud intensity values of the tunnel annulus;

[0013] S5: Standardize the initial cross-section point set of the tunnel cross-section to obtain a standard cross-section; correct the two-dimensional intensity image according to the minimum circumscribed circle and the standard cross-section;

[0014] S6: Based on an improved Poisson reconstruction algorithm with adaptive threshold segmentation, model the preprocessed point cloud data to obtain a three-dimensional grid model of the tunnel;

[0015] S7: Use the automatic texture mapping technology to generate a three-dimensional texture model of the tunnel in combination with the three-dimensional grid model of the tunnel and the corrected two-dimensional intensity image.

[0016] Optionally, step S1 includes:

[0017] Adopt a mobile laser scanner to perform on-site three-dimensional laser scanning operation in the tunnel to obtain the point cloud data of a tunnel with any shape;

[0018] Use CloudCompare point cloud processing software to delete the noise inside the tunnel and the outlier points outside the tunnel in the point cloud data.

[0019] Optionally, step S2 includes:

[0020] S21: Perform coordinate transformation on the preprocessed point cloud data, and then extract a preset mileage interval along the x-axis to obtain the tunnel annulus within the mileage interval and unify the mileage values;

[0021] S22: Select a point p(x,y) in the two-dimensional point set located on the YOZ plane of the segmented tunnel annulus;

[0022] Preset a radius R of a circle, and search for the set Q of all points within a distance of 2R from the point p(x,y) among the remaining points in the two-dimensional point set;

[0023] S23: Select the point p in the set Q1 (x 1 , y 1 ), according to the coordinate values of point p and point p 1 and radius R, combined with the resection method, calculate the coordinates of two center points p 2 (x 2 , y 2 ) and p 3 (x 3 , y 3 ); Remove point p 1 from set Q, and then calculate the distances from other points in set Q to the center points p 2 and p 3 .

[0024] S24: If the distances from other points in set Q to the two center points are all greater than R, determine that point p is a boundary point; if the distances from other points in set Q to the two center points are not all greater than R, determine that point p is not a boundary point, that is, a non-boundary point;

[0025] S25: Traverse the remaining points in set Q in sequence, take the traversed remaining points as point p(x, y), and repeat steps S23 and S24 to determine the boundary points in set Q;

[0026] S26: Determine the tunnel cross-section extracted from each segmented point cloud according to the boundary points.

[0027] Optionally, step S3 includes:

[0028] S31: Set the coordinates of the first point P 1 in the initial cross-section point set as the center of the minimum circumscribed circle, and set the radius to a minimum value. Denote this circle as C 1 ;

[0029] S32: Take the second point P 2 in the initial cross-section point set. If P 2 is not inside circle C 1 , then update circle C 1 with P 2 and P 1 as the diameter and denote it as C 2 ;

[0030] S33: Take the third point P 3 in the initial cross-section point set. If P 3 is inside circle C 2 , then do not update circle C 2 , otherwise construct a new circle C 1 with the coordinate values of P 2 , P 3 and P 3 , and P 1 and P 2, P 3 The three - point coordinates are on the circle C 3 ;

[0031] S34: Traverse the remaining points P in the initial cross - section point set in sequence i , if the remaining point P i is inside the circle C j and j ≤ i, then do not update the circle C j ; if the remaining point P i is not inside the circle C j , then take P i as a boundary point and update the circle C j to C j+1 such that C j+1 contains all the previously traversed points;

[0032] S35: After traversing all the remaining points in the initial cross - section point set, obtain the center and radius parameters of the minimum outer - enclosing circle of the initial cross - section point set of the tunnel cross - section according to the three - point coordinates and the equation of the circle;

[0033] S36: Take the point - cloud mileage as the height of the projection cylinder. The specific steps include: Set the height of the projection cylinder of the tunnel annulus as the difference between the maximum and minimum values of the x - coordinates in the point - cloud data of the tunnel annulus; The projection cylinder is the minimum outer - enclosing cylinder;

[0034] S37: Traverse all the points of the tunnel annulus based on the projection cylinder. Let the X - coordinate of any point in the tunnel mileage direction be x i , and the clockwise angle β i with the y - axis. Then the coordinates of the point after coordinate mapping are (β i , x i ), that is, convert the three - dimensional coordinates of all the points of the tunnel annulus into two - dimensional coordinates;

[0035] The calculation formula of β i is as follows:

[0036]

[0037] where O(y 0 , z 0 ) is the center of the minimum outer - enclosing circle; y i , z i represent the Y - coordinate and Z - coordinate of any point in the tunnel mileage direction.

[0038] Optionally, step S4 includes:

[0039] S41: Unroll the three - dimensional projection cylinder along the track center line to obtain a flattened diagram; The flattened diagram is a two - dimensional rectangle;

[0040] S42: Search for the point cloud corresponding to each pixel in the flattened image;

[0041] S43: When the point cloud of the tunnel annulus only contains reflection intensity information, calculate the average reflection intensity of the points within each pixel;

[0042] S44: Determine the color value C of the pixel through the average reflection intensity of the points within each pixel, that is, obtain the color values of all pixels of the flattened image;

[0043] S45: Generate a two-dimensional intensity image of the inner surface of the tunnel lining through the color values of all pixels of the flattened image.

[0044] Optionally, step S5 includes:

[0045] S51: Sort the initial cross-section point set of the tunnel section in two-dimensional coordinate form according to the magnitude of the clockwise angle with the y-axis;

[0046] S52: Loop to determine whether there are points in each preset angle interval. If there are multiple points, select the point with the angle value closest to the midpoint value of the angle interval as the standard point of the standard cross-section in this angle interval; if there are no points in this angle interval, perform smoothing processing and point filling processing to obtain the final standard cross-section;

[0047] S53: Calculate the correction ratio of each sub-image in the two-dimensional intensity image according to the straight-line length and arc length of two points of the standard cross-section and the minimum circumscribed circle in the same angle interval;

[0048] S54: Correct the two-dimensional intensity image according to the image correction ratio, combined with cubic spline interpolation.

[0049] Optionally, step S6 includes:

[0050] S61a: Determine the boundary of the tunnel and the indicator function Set the outside of the tunnel to 0 and the inside to 1, and calculate the indicator function of each point P i in the point cloud data

[0051] S61b: Calculate the normal vector of each point of the tunnel x n = -x, y n = 0, z n = -z;

[0052] S61c: Use the smoothed filtered indicator function and the smoothing function to calculate the gradient

[0053]

[0054] Among them, F p (q 0 ) represents a smoothing function, and q 0 represents an arbitrary point on the surface to be reconstructed; represents the boundary the normal vector of point p on the boundary; p represents a point on the boundary ;

[0055] S61d: Divide the surface of the tunnel into small pieces P s , replace the surface integral with the sum of integrals of the interior points in each small piece, and the calculation formula satisfies

[0056]

[0057] Among them, F s.p (q) represents the function value of point q at point p on the surface patch s; q represents a specific point at which the gradient needs to be calculated during the Poisson reconstruction process; represents the dot product of the normal vector of the surface patch s and the normal vector of point p; represents the gradient vector at point q;

[0058] S61e: Generate the Poisson model of the tunnel through the Marching Cube algorithm;

[0059] S61f: Calculate the distribution probability of the model triangular patches with the same perimeter in the Poisson model;

[0060] S61g: Use the vertices of the triangular patches with the largest distribution probability and the same perimeter as the sampling points;

[0061] Suppose the number of input point clouds in the preprocessed point cloud data is k, and the set of input point clouds P = {p 1 , p 2 , …, p k}, and the coordinates of the input point cloud are (x p , y p , z p );

[0062] The coordinates of the sampling points are (x q , y q , z q ), and the set of sampling points Q = {q 1 , q 2 , …, q m}, calculate the Euclidean distance d pq between the sampling points and the sampling points of the input point cloud;

[0063] S61h: Use the vertices of the triangular patches as rows and the input point cloud as columns, and calculate according to the Euclidean distance d from the vertices of the triangular patches to the input point cloudpq Perform ascending sorting to obtain the density of the vertices of the Poisson model as ρ′;

[0064] S61i: Round ρ′ down to obtain [ρ′], and take the [ρ′]-th row of d pq The [ρ′]-th row represents the distance from the [ρ′]-th vertex to the input point cloud

[0065] S61j: Traverse all the vertices of the Poisson model and the input point cloud and calculate the corresponding distance values. If the distance value is greater than then mark the vertex as 0. If the distance value is less than or equal to mark the vertex as 1;

[0066] S61k: Remove all triangular patches marked as 0 to obtain an unclosed tunnel three-dimensional mesh model.

[0067] Optionally, step S7 includes:

[0068] S71: Determine the pixel unit where the vertex of the tunnel three-dimensional mesh model is located according to the point cloud flattening process; normalize the position coordinates of the model vertex to obtain texture mapping coordinates;

[0069] S72: Construct a tunnel three-dimensional texture model according to the tunnel three-dimensional mesh model, the two-dimensional intensity image, and the texture mapping coordinates.

[0070] A multi-shape tunnel three-dimensional modeling system based on laser point cloud, the system includes: a data loading and display module, a tunnel point cloud denoising and thinning module, a tunnel point cloud texture flattening module, and a tunnel point cloud three-dimensional reconstruction module;

[0071] The data loading and display module, the tunnel point cloud denoising and thinning module, the tunnel point cloud texture flattening module, and the tunnel point cloud three-dimensional reconstruction module are sequentially connected;

[0072] The data loading and display module is used to load and parse point cloud data, three-dimensional models, and two-dimensional images;

[0073] The tunnel point cloud denoising and thinning module is used to preprocess the point cloud data;

[0074] The tunnel point cloud texture flattening module is used to calculate the minimum circumscribed circle of the tunnel cross-section; determine the projection cylinder of the tunnel annulus; project the point cloud of the tunnel annulus onto the surface of the projection cylinder through the minimum circumscribed circle;

[0075] The tunnel point cloud texture flattening module is also used to generate a two-dimensional intensity image of the inner surface of the tunnel lining according to the projection cylinder and the point cloud intensity value of the tunnel annulus;

[0076] The tunnel point cloud texture flattening module is also used to standardize the initial cross-section point set of the tunnel to obtain a standard cross-section; correct the two-dimensional intensity image according to the minimum circumscribed circle and the standard cross-section;

[0077] The tunnel point cloud three-dimensional reconstruction module is used to model the preprocessed point cloud data based on an improved Poisson reconstruction algorithm with adaptive threshold segmentation to obtain a tunnel three-dimensional grid model;

[0078] The tunnel point cloud three-dimensional reconstruction module is also used to generate a tunnel three-dimensional texture model by using automatic texture mapping technology in combination with the tunnel three-dimensional grid model and the corrected two-dimensional intensity image.

[0079] The beneficial effects brought by the technical solution provided by this application are:

[0080] 1. Considering the complex multi-shape situation of the tunnel, a corrected two-dimensional intensity image for multi-shape two-dimensional textures is established to reduce the degree of image deformation; the present invention can handle tunnels with not only circular or elliptical shapes, but also complex-shaped tunnels such as horseshoe-shaped, arched, and rectangular tunnels, without presetting cross-section fitting formulas and projection parameters. The technical solution of the present invention reduces the problem of result image deformation caused by shape mismatch, thereby improving the accuracy and applicability of generating two-dimensional texture images of tunnels with different shapes.

[0081] 2. Improve the traditional Poisson model to reduce the problem of easy generation of pseudo patches during the modeling of non-closed tunnel models. By improving the deficiencies in the existing three-dimensional modeling technology, such as reducing the pseudo patches generated by the Poisson reconstruction algorithm in non-closed tunnels, the present invention provides a tunnel three-dimensional model construction method with higher accuracy. This helps to improve the accuracy of staff for safety monitoring and maintenance decision-making and ensure the safety and stability of the tunnel.

[0082] 3. Compared with the method relying on the three-dimensional grid model, the present invention not only constructs the three-dimensional structure of the tunnel, but also successfully adds high-quality texture information, making the generated tunnel model more realistic and detailed. This model containing texture details can more accurately reflect the actual situation inside the tunnel and provide a more reliable basis for engineering evaluation. Description of the Drawings

[0083] The following will further illustrate the present application in conjunction with the drawings. In the drawings:

[0084] Figure 1 is the step diagram of the multi-shape tunnel three-dimensional modeling based on laser point cloud in the embodiment of the present invention;

[0085] Figure 2 is the point cloud of four-shaped tunnels after denoising in the embodiment of the present invention;

[0086] Figure 3 Schematic diagram for extracting the tunnel annular cross-section and unifying the mileage in the embodiments of the present invention;

[0087] Figure 4 Tunnel cross-sections extracted after preprocessing the four types of tunnel point clouds in the embodiments of the present invention;

[0088] Figure 5 Schematic diagram of the minimum circumcircle calculation principle in the embodiments of the present invention;

[0089] Figure 6 Minimum circumcircles of the cross-sections of four types of tunnel shapes in the embodiments of the present invention;

[0090] Figure 7 Schematic diagram of coordinate mapping in the embodiments of the present invention;

[0091] Figure 8 Schematic diagram of cylindrical projection flattening in the embodiments of the present invention;

[0092] Figure 9 Standardized cross-sections of the point sets of four types of tunnel cross-sections in the embodiments of the present invention;

[0093] Figure 10 Schematic diagram of image correction in the embodiments of the present invention;

[0094] Figure 11 Diagrams showing the effects before and after image correction of four types of tunnels in the embodiments of the present invention;

[0095] Figure 12 Three-dimensional tunnel grid model in wireframe mode in the embodiments of the present invention;

[0096] Figure 13 Three-dimensional tunnel texture models after mapping of tunnels of four shapes in the embodiments of the present invention. Detailed implementation manners

[0097] For a clearer understanding of the technical features, objectives, and effects of the present application, the detailed implementation manners of the present application will now be described with reference to the accompanying drawings.

[0098] The embodiments of the present application provide a method for three-dimensional modeling of multi-shaped tunnels based on laser point clouds.

[0099] Please refer to Figure 1 , Figure 1 which is a step diagram of a method for three-dimensional modeling of multi-shaped tunnels based on laser point clouds in the embodiments of the present application, and includes:

[0100] S1: Obtain the point cloud data of a tunnel with any shape and perform preprocessing;

[0101] S2: Segment and extract the tunnel annulus and the tunnel cross-section from the preprocessed point cloud data;

[0102] S3: Calculate the minimum circumscribed circle of the tunnel cross-section; determine the projection cylinder of the tunnel annulus; project the point cloud of the tunnel annulus onto the surface of the projection cylinder through the minimum circumscribed circle;

[0103] S4: Generate a two-dimensional intensity image of the inner surface of the tunnel lining based on the projection cylinder and the point cloud intensity values of the tunnel annulus;

[0104] S5: Standardize the initial cross-section point set of the tunnel cross-section to obtain a standard cross-section; correct the two-dimensional intensity image according to the minimum circumscribed circle and the standard cross-section;

[0105] S6: Based on an improved Poisson reconstruction algorithm with adaptive threshold segmentation, model the preprocessed point cloud data to obtain a three-dimensional mesh model of the tunnel;

[0106] S7: Use the automatic texture mapping technology to combine the three-dimensional mesh model of the tunnel and the corrected two-dimensional intensity image to generate a three-dimensional texture model of the tunnel.

[0107] As an embodiment, the present invention realizes the rapid browsing of three-dimensional texture models of subway tunnels with multiple shapes. It provides an effective way to intuitively and stereoscopically detect the safety state of the tunnel structure, and promotes the digital management of the tunnel.

[0108] Step S1 includes:

[0109] Adopt a mobile laser scanner to perform on-site three-dimensional laser scanning operations in the tunnel to obtain point cloud data of tunnels with arbitrary shapes;

[0110] Use CloudCompare point cloud processing software to delete the noise points inside the tunnel and the outlier points outside the tunnel in the point cloud data.

[0111] As an embodiment, the Uniform sampling and SOR algorithms are used for the point cloud data to remove the interference of pipelines, wires, light bands, etc. inside the tunnel. As Figure 2 shown, the point cloud data after tunnel denoising, and the tunnel shapes are rectangular, elliptical, horseshoe-shaped, and circular respectively.

[0112] Step S2 includes:

[0113] S21: Perform coordinate transformation on the preprocessed point cloud data, and then extract a preset mileage interval along the x-axis to obtain the tunnel annulus within the mileage interval and unify the mileage values;

[0114] S22: Select a point p(x, y) in the two-dimensional point set located on the YOZ plane of the segmented tunnel annulus;

[0115] Preset the radius R of a circle, and in the remaining points of the two-dimensional point set, search for the set Q of all points within a distance of 2R from the point p(x, y).

[0116] S23: Select the point p in the set Q 1 (x 1 , y 1 ). According to the coordinate values of the point p and the point p 1 and the radius R, combined with the resection method, calculate the coordinates of the two center points p 2 (x 2 , y 2 ) and p 3 (x 3 , y 3 ); Remove the point p 1 from the set Q, and then calculate the distances from the other points in the set Q to the center points p 2 , p 3 .

[0117] S24: If the distances from the other points in the set Q to the two center points are all greater than R, then determine that the point p is a boundary point; if the distances from the other points in the set Q to the two center points are not all greater than R, then determine that the point p is not a boundary point, that is, a non-boundary point.

[0118] S25: Traverse the remaining points in the set Q in sequence, take the traversed remaining points as the point p(x, y), and repeat steps S23 and S24 to determine the boundary points in the set Q.

[0119] S26: Determine the tunnel cross-section extracted from each segmented point cloud according to the boundary points.

[0120] As an embodiment, segment and extract the tunnel annulus and tunnel cross-section from the point cloud data, as Figure 3 、 Figure 4 shown.

[0121] Step S3 includes:

[0122] S31: Set the coordinate of the first point P 1 in the initial cross-section point set of the tunnel cross-section as the center of the minimum circumscribed circle, and set the radius as a minimum value, and denote this circle as C 1 .

[0123] S32: Take the second point P 2 in the initial cross-section point set. If P 2 is not inside the circle C 1 , then update the circle C 1 with P 2 and P 1 as the diameter, and denote it as C 2 .

[0124] S33: Take the third point P from the initial cross-section point set 3 , if P 3 is inside the circle C 2 , then do not update the circle C 2 , otherwise construct a new circle C 1 with the coordinates of points P 2 , P 3 , and P 3 , and the coordinates of points P 1 , P 2 , and P 3 are on the circle C 3 ;

[0125] As an example, the coordinates of points P 1 , P 2 , and P 3 are on the circle C 3 to ensure that the newly added points are definitely on the boundary of the updated circle.

[0126] S34: Traverse the remaining points P i in the initial cross-section point set in sequence. If the remaining point P i is inside the circle C j and j ≤ i, then do not update the circle C j ; if the remaining point P i is not inside the circle C j , then take P i as a boundary point and update the circle C j to C j+1 such that C j+1 contains all the previously traversed points;

[0127] S35: After traversing all the remaining points in the initial cross-section point set, obtain the center and radius parameters of the minimum outer bounding circle of the initial cross-section point set of the tunnel section according to the coordinates of three points and the equation of the circle;

[0128] S36: Take the point cloud mileage as the height of the projection cylinder. The specific steps include: Set the height of the projection cylinder of the tunnel annulus to the difference between the maximum and minimum values of the x coordinates in the point cloud data of the tunnel annulus; The projection cylinder is the minimum outer bounding cylinder;

[0129] As an example, set the height of the cylinder of the tunnel annulus to the difference between the maximum and minimum values of the x coordinates in the point cloud data of the tunnel annulus to ensure that the cylinder can completely cover all the denoised point cloud data.

[0130] S37: Traverse all the points of the tunnel annulus based on the projection cylinder. Let the X coordinate of any point in the tunnel mileage direction be x i , and the clockwise angle β i with the coordinate axis y, then the coordinates of the point after coordinate mapping are (βi , x i ), that is, converting the three-dimensional coordinates of all points on the tunnel annulus into two-dimensional coordinates;

[0131] β i The calculation formula of is as follows:

[0132]

[0133] where O(y 0 , z 0 ) is the center of the smallest enclosing circle; y i , z i represents the Y coordinate and Z coordinate of any point in the tunnel mileage direction.

[0134] As an embodiment, calculating the smallest enclosing circle of the tunnel cross-section point cloud is as Figure 5 shown. As Figure 6 shown, it shows the initial cross-section of the tunnel point cloud with different shapes and its corresponding smallest enclosing circle. It can be seen that regardless of the cross-section shape of the tunnel, the corresponding smallest enclosing circle can be found to cover all points. Especially for horseshoe-shaped and circular tunnels, the fitting effect of the smallest enclosing circle is particularly obvious.

[0135] Step S4 includes:

[0136] S41: Unfolding the three-dimensional projection cylinder along the track center line to obtain a flattened diagram; the flattened diagram is a two-dimensional rectangle;

[0137] As an embodiment, the length and width of the flattened diagram satisfy the calculation formula:

[0138]

[0139] where px, py are the grid resolutions, representing the central angle range corresponding to one pixel, L is the tunnel length, and r is the radius of the smallest enclosing circle. According to the actual tunnel point cloud data test, the tunnel image accuracy and quality are the best when the resolution is set to 0.0012. The track center line is the center position of the track, located at an equal distance between each track.

[0140] S42: Search for the point cloud corresponding to each pixel in the flattened diagram;

[0141] As an embodiment, searching for the point cloud corresponding to each pixel in the image, the points within the pixel and the pixel coordinates should satisfy the following range relationship:

[0142] {(β i , x i )|(x i - 1)×px ≤ β i ≤ x i ×px, (yi (-1)×py×r ≤ x i ≤ y i ×py×r}

[0143] Where (x i , y i ) is the coordinate of a pixel in the flattened image, and the cloud points of all points within this pixel are (β i , x i ), where i = 1, 2, 3, …, k, and k is the number of points contained in the pixel.

[0144] S43: When the point cloud of the tunnel annulus only contains reflection intensity information, calculate the average reflection intensity of the points within each pixel;

[0145] S44: Determine the color value C of the pixel through the average reflection intensity of the points within each pixel, that is, obtain the color values of all pixels of the flattened image;

[0146] S45: Generate a two-dimensional intensity image of the inner surface of the tunnel lining through the color values of all pixels of the flattened image.

[0147] As an embodiment, as Figure 7 shown, Figure 7 it represents that the tunnel point cloud is coordinate-mapped based on a projection cylinder, thereby forming a YOZ plane coordinate system. Unroll the projection cylinder with the track center line as the axis, and generate a two-dimensional intensity image of the inner surface of the tunnel lining according to the point cloud intensity value, as Figure 8 shown. The reflection intensity information is the energy intensity of the light pulse emitted by the lidar reflected back by the target surface and exists as an attribute value.

[0148] Step S5 includes:

[0149] S51: Sort the initial cross-section point set of the tunnel section in the form of two-dimensional coordinates according to the size of the clockwise angle with the y-axis;

[0150] S52: Circularly judge whether there are points in each preset angle interval. If there are multiple points, select the point whose angle value is closest to the midpoint value of the angle interval as the standard point of the standard cross-section in this angle interval; if there are no points in this angle interval, perform smoothing processing and point supplementation processing to obtain the final standard cross-section;

[0151] As an embodiment, each angle starting from zero is used as an angle interval, and circularly judge whether there are points in each angle interval. If there are no points in the angle interval, it means that there are abnormal depressions, protrusions, and missing parts in the initial cross-section.

[0152] Perform point supplementation operations on the initial cross-section, and the coordinates of the supplemented points satisfy the calculation formula:

[0153]

[0154] Among them, for the angle α between the straight line connecting two adjacent points A(y 1 , z 1 ) and B(y 2 , z 2 ) and the coordinate axis y, and the length of AB is L AB , the number of points to be supplemented is N, then the length of each evenly divided small segment should be y i , z i respectively represent the horizontal and vertical coordinate values of the i-th point to be supplemented. As Figure 9 shown, it is the standardized cross-section of the point set of four tunnel cross-sections, with one point for each angle.

[0155] S53: Calculate the correction ratio of each sub-image in the two-dimensional intensity image according to the straight-line length and arc length between two points in the same angular interval of the standard cross-section and the minimum circumscribed circle;

[0156] S54: Correct the two-dimensional intensity image according to the image correction ratio in combination with cubic spline interpolation.

[0157] As an embodiment, the correction schematic diagram, as Figure 10 shown, when the correction ratio K is less than 1, the sub-image will be compressed: when K is greater than 1, the sub-image will be stretched. The correction ratio K satisfies the calculation formula:

[0158]

[0159] where r is the radius of the minimum circumscribed circle, A 2 (y A2 , z A2 ), B 2 (y B2 , z B2 ) are two points on the standard cross-section, α is the central angle corresponding to two points in the same angular interval of the standard cross-section and the minimum circumscribed circle, the arc length of the specified angular interval of the projected cross-section is L 1 , and the straight-line length between the cross-section points in the same angular interval of the standard cross-section is L 2 . As Figure 11 shown, it is the flattening diagram of the effects before and after the correction of four tunnel images. a is the original point cloud, b is the initial image, and c is the corrected image. It can be seen that there are obvious deformations in the image before correction compared with the original point cloud, and the image after correction is basically the same as the original point cloud.

[0160] Step S6 includes:

[0161] S61a: Determine the boundary of the tunnel and the indicator function Set the outside of the tunnel to 0 and the inside to 1, and calculate each point P in the point cloud data i of the indicator function

[0162] S61b: Calculate the normal vector of each point of the tunnel x n = -x, y n = 0, z n = -z;

[0163] S61c: Use the smoothed indicator function and the smoothing function to calculate the gradient

[0164]

[0165] where F p (q 0 ) represents the smoothing function, q 0 represents any point on the surface to be reconstructed; represents the boundary the normal vector of point p on the boundary; p represents a point on the boundary ;

[0166] S61d: Divide the surface of the tunnel into small pieces P s , replace the surface integral with the sum of the integrals of the interior points in each small piece, and the calculation formula satisfies

[0167]

[0168] where F s.p (q) represents the function value of point q at point p on the surface patch s; q represents a specific point where the gradient needs to be calculated in the Poisson reconstruction process; represents the dot product of the normal vector of the surface patch s and the normal vector of point p; represents the gradient vector at point q;

[0169] S61e: Generate the Poisson model of the tunnel through the Marching Cube algorithm;

[0170] As an example, the formula of the Poisson equation is: The sum of the second partial derivatives of the function at a certain point. It is a way to measure the rate of change of the function and here represents the degree of change of a certain quantity. is the gradient operator, representing the directional derivative of the function at a certain point. Here, it represents the vector field direction. is the vector field, representing the magnitude and direction of the velocity at each point in space.

[0171] S61f: Calculate the distribution probability of model triangular patches with the same perimeter in the Poisson model;

[0172] As an embodiment, the distribution probability F(v j ) satisfies the calculation formula:

[0173]

[0174] where n is the total number of patches in the Poisson model, and the perimeter of each triangle is L i , before processing, L i can be multiplied by a value to make it an integer. Denote the minimum and maximum perimeters as L min and L max , then L min and L max should also be integers. The number of triangular faces with perimeter L j is v j .

[0175] S61g: Use the vertices of the triangular patches with the maximum distribution probability as sampling points;

[0176] Let the number of input point clouds in the preprocessed point cloud data be k, and the set of input point clouds be P = {p 1 , p 2 , …, p k}, and the coordinates of the input point cloud be (x p , y p , z p );

[0177] The coordinates of the sampling points are (x q , y q , z q ), and the set of sampling points is Q = {q 1 , q 2 , …, q m}. Calculate the Euclidean distance d pq between the sampling points and the input point cloud sampling points;

[0178] S61h: Take the vertices of the triangular patches as rows and the input point cloud as columns, and sort them in ascending order according to the Euclidean distance d pq from the vertices of the triangular patches to the input point cloud, to obtain the density ρ′ of the vertices of the Poisson model;

[0179] As an embodiment, the density ρ′ of the vertices of the Poisson model

[0180]

[0181] where ρ is the density of the input point cloud, and k′ is the number of vertices of the triangular faces of the Poisson model.

[0182] S61i: Round down ρ′ to obtain [ρ′], and take the [ρ′]-th row of d pq The [ρ′]-th row represents the distance from the [ρ′]-th vertex to the input point cloud

[0183] S61j: Traverse all vertices of the Poisson model and the input point cloud and calculate the corresponding distance values. If the distance value is greater than then mark the vertex as 0. If the distance value is less than or equal to mark the vertex as 1;

[0184] S61k: Remove all triangular patches marked as 0 to obtain an unclosed tunnel three-dimensional mesh model.

[0185] The present application provides an embodiment as follows. An improved Poisson reconstruction algorithm based on adaptive threshold segmentation is used to model the preprocessed tunnel point cloud, as Figure 12 shown.

[0186] Step S7 includes:

[0187] S71: Determine the pixel unit where the vertex of the tunnel three-dimensional mesh model is located according to the point cloud flattening process; normalize the position coordinates of the model vertex to obtain texture mapping coordinates;

[0188] As an embodiment, the model vertices are distributed around the original point cloud and the projection cylinder. Determine the pixel unit where the vertex of the tunnel three-dimensional mesh model is located according to the point cloud flattening process; normalize the position coordinates of the model vertex to obtain texture mapping coordinates;

[0189] S72: Construct a tunnel three-dimensional texture model according to the tunnel three-dimensional mesh model, the two-dimensional intensity image, and the texture mapping coordinates.

[0190] As an embodiment, the calculation formula for the uv coordinates of the vertices of the tunnel three-dimensional mesh model satisfies:

[0191]

[0192] where px and py are the image resolutions, and p i (x i , y i ) are the coordinates of the i-th vertex of the model, and height and width are the height and width of the flattened image.

[0193] As an embodiment, as Figure 13As shown, it is a three-dimensional texture model of tunnels in four shapes after mapping. It can be seen that the tunnel 3D model can comprehensively and highly accurately reproduce the tunnel structure, including the spatial positions of tracks, pipelines, and internal platforms. These levels of detail allow for precise analysis and visualization of the tunnel, accurately reflecting the true physical environment of the tunnel and providing strong support for tunnel construction management and operation optimization.

[0194] A multi-shape tunnel three-dimensional modeling system based on laser point cloud, the system includes: a data loading and display module, a tunnel point cloud denoising and thinning module, a tunnel point cloud texture flattening module, and a tunnel point cloud three-dimensional reconstruction module;

[0195] The data loading and display module, the tunnel point cloud denoising and thinning module, the tunnel point cloud texture flattening module, and the tunnel point cloud three-dimensional reconstruction module are connected in sequence.

[0196] The data loading and display module is used to load and parse point cloud data, three-dimensional models, and two-dimensional images;

[0197] The tunnel point cloud denoising and thinning module is used to preprocess the point cloud data;

[0198] The tunnel point cloud texture flattening module is used to calculate the minimum circumscribed circle of the tunnel section; determine the projection cylinder of the tunnel annulus; project the point cloud of the tunnel annulus onto the surface of the projection cylinder through the minimum circumscribed circle;

[0199] The tunnel point cloud texture flattening module is also used to generate a two-dimensional intensity image of the inner surface of the tunnel lining according to the projection cylinder and the point cloud intensity value of the tunnel annulus;

[0200] The tunnel point cloud texture flattening module is also used to standardize the initial section point set of the tunnel section to obtain a standard section; correct the two-dimensional intensity image according to the minimum circumscribed circle and the standard section;

[0201] The tunnel point cloud three-dimensional reconstruction module is used to model the preprocessed point cloud data based on an improved Poisson reconstruction algorithm with adaptive threshold segmentation to obtain a tunnel three-dimensional mesh model;

[0202] The tunnel point cloud three-dimensional reconstruction module is also used to generate a tunnel three-dimensional texture model by using automatic texture mapping technology, combining the tunnel three-dimensional mesh model and the corrected two-dimensional intensity image.

[0203] As an embodiment, the tunnel point cloud three-dimensional reconstruction module uses the Poisson reconstruction algorithm to generate a three-dimensional model of the tunnel, and applies the pseudo-plane segmentation algorithm to optimize the model and remove the pseudo-planes that do not belong to the tunnel. In addition, it is also responsible for automatic texture mapping, using the flattened two-dimensional image as the texture of the model, and performing quantitative analysis of the model accuracy.

[0204] The above are only exemplary embodiments of the present disclosure, and the scope of the present disclosure cannot be limited thereby. That is, any equivalent changes and modifications made in accordance with the teachings of the present disclosure still fall within the scope covered by the present disclosure.

[0205] This application is intended to cover any variations, uses, or adaptations of the present disclosure, which follow the general principles of the present disclosure and include known common knowledge or conventional technical means in the technical field not recorded in the present disclosure. The description and examples are only regarded as exemplary, and the scope and spirit of the present disclosure are defined by the claims.

Claims

1. A method for three-dimensional modeling of multi-shape tunnels based on laser point cloud, characterized in that: The method comprises the following steps: S1: Obtain point cloud data of tunnels of arbitrary shapes and perform preprocessing; S2: Extract tunnel rings and tunnel sections segmentally from the preprocessed point cloud data; S3: Calculate the minimum outer enclosing circle of the tunnel section; determine the projection cylinder of the tunnel ring; project the point cloud of the tunnel ring onto the surface of the projection cylinder through the minimum outer enclosing circle; S4: Generate a two-dimensional intensity image of the inner surface of the tunnel lining according to the point cloud intensity values ​​of the projection cylinder and the tunnel ring; S5: standardize the initial cross-section point set of the tunnel section to obtain a standard cross-section; correct the two-dimensional intensity image according to the minimum outer enclosing circle and the standard cross-section; S6: Based on the improved Poisson reconstruction algorithm of adaptive threshold segmentation, the pre-processed point cloud data is modeled to obtain the 3D mesh model of the tunnel; S7: Generate a 3D texture model of the tunnel by combining the 3D mesh model of the tunnel and the corrected 2D intensity image using automatic texture mapping technology.

2. The method for three-dimensional modeling of multi-shape tunnels based on laser point cloud according to claim 1, characterized in that: Step S1 includes: Use mobile laser scanners to conduct on-site 3D laser scanning operations at the tunnel site to obtain point cloud data of tunnels of any shape; CloudCompare point cloud processing software was used to delete noise points inside the tunnel and outliers outside the tunnel in the point cloud data.

3. The method for three-dimensional modeling of multi-shape tunnels based on laser point cloud according to claim 1, characterized in that: Step S2 includes: S21: performing coordinate conversion on the pre-processed point cloud data, extracting the preset mileage interval along the x-axis, obtaining the tunnel ring within the mileage interval and unifying the mileage value; S22: Select a point p(x, y) from the two-dimensional point set on the YOZ plane on the segmented tunnel ring; Preset the radius R of a circle, and search for the set Q of all points within the distance 2R from point p(x,y) among the remaining points in the two-dimensional point set; S23: Select a point p1(x1, y1) from the set Q, and calculate the coordinates of the two circle centers p2(x2, y2) and p3(x3, y3) according to the coordinate values ​​of the points p and p1 and the radius R, combined with the back intersection method; remove the point p1 from the set Q, and then calculate the distances from the other points in the set Q to the circle centers p2 and p3; S24: If the distances from other points in the set Q to the two centers are all greater than R, then point p is determined to be a boundary point; if the distances from other points in the set Q to the two centers are not all greater than R, then point p is determined not to be a boundary point, i.e., a non-boundary point; S25: traverse the remaining points in set Q in sequence, take the remaining points traversed as points p(x, y), repeat steps S23 and S24, and determine the boundary points in set Q; S26: Determine the tunnel section extracted from each segmented point cloud according to the boundary points.

4. The method for three-dimensional modeling of multi-shape tunnels based on laser point cloud according to claim 1, characterized in that: Step S3 includes: S31: Set the coordinates of the first point P1 in the initial cross-sectional point set of the tunnel section as the center of the minimum outer enclosing circle, set the radius to a minimum value, and record the circle as C1; S32: Take a second point P2 from the initial cross-section point set. If P2 is not in the circle C1, update the circle C1 with P1 and P2 as the diameter and record it as C2. S33: Take the third point P3 from the initial cross-section point set. If P3 is within the circle C2, the circle C2 is not updated. Otherwise, a new circle C3 is constructed with the coordinates of the three points P1, P2, and P3. The coordinates of the three points P1, P2, and P3 are on the circle C3. S34: Traverse the remaining points P in the initial cross-section point set in order i , if the remaining point P i In circle C j If j≤i, circle C is not updated. j ; If the remaining point P i Not in circle C j If P i As the boundary point, update the circle C j C j+1 , so that C j+1 Contains all the points traversed before; S35: after the traversal of the remaining points in the initial section point set is completed, the center and radius parameters of the minimum outer enclosing circle of the initial section point set of the tunnel section are obtained according to the coordinates of the three points and the equation of the circle; S36: taking the point cloud mileage as the height of the projection cylinder, the specific steps include: setting the height of the projection cylinder of the tunnel ring belt to the difference between the maximum value and the minimum value of the x coordinate in the point cloud data of the tunnel ring belt; the projection cylinder is the minimum outer enclosing cylinder; S37: Based on the projection cylinder, traverse all points of the tunnel ring, and set the X coordinate of any point in the tunnel mileage direction as x i , the clockwise angle β with the coordinate axis y i , then the coordinates of the point after coordinate mapping are (β i ,x i ), that is, converting the three-dimensional coordinates of all points in the tunnel ring into two-dimensional coordinates; β i The calculation formula is as follows: Where O(y0,z0) is the center of the smallest outer enclosing circle; y i ,z i Indicates the Y and Z coordinates of any point in the tunnel mileage direction.

5. The method for three-dimensional modeling of multi-shape tunnels based on laser point cloud according to claim 1, characterized in that: Step S4 includes: S41: unfolding the three-dimensional projection cylinder along the center line of the track to obtain a flattened image; the flattened image is a two-dimensional rectangle; S42: Searching for the point cloud corresponding to each pixel in the flattened image; S43: when the point cloud of the tunnel ring only contains reflection intensity information, calculating the average reflection intensity of the points within each pixel; S44: determining the color value C of the pixel by the average value of the reflection intensity of the points within each pixel, that is, obtaining the color values ​​of all pixels of the flattened image; S45: Generate a two-dimensional intensity image of the inner surface of the tunnel lining by flattening the color values ​​of all pixels of the image.

6. The method for three-dimensional modeling of multi-shape tunnels based on laser point cloud according to claim 1, characterized in that: Step S5 includes: S51: sorting the initial cross-section point set of the tunnel cross-section in the form of two-dimensional coordinates according to the clockwise angle between the point set and the y-axis; S52: cyclically determine whether there is a point in each preset angle interval. If there are multiple points, select the point whose angle value is closest to the midpoint value of the angle interval as the standard point of the standard section in the angle interval; if there is no point in the angle interval, perform smoothing and point filling to obtain the final standard section; S53: calculating a correction ratio of each sub-image in the two-dimensional intensity image according to the straight line length and arc length of two points in the same angle interval between the standard section and the minimum outer enclosing circle; S54: Correcting the two-dimensional intensity image according to the image correction ratio in combination with cubic spline interpolation.

7. The method for three-dimensional modeling of multi-shape tunnels based on laser point cloud according to claim 1, characterized in that: Step S6 includes: S61a: Determine the boundaries of the tunnel and the indicator function χ M , set the outside of the tunnel to 0 and the inside to 1, and calculate each point P in the point cloud data i The indicator function χ M (P i ); S61b: Calculate the normal vector of each point in the tunnel x n =-x,y n =0, z n =-z; S61c: Use the indicator function χ after smoothing filtering M (P i ) and smoothing function, calculate the gradient where F p (q0) represents a smoothing function, and q0 represents any point on the surface to be reconstructed; Representing Boundaries Normal vector of point p; p represents the boundary A point on S61d: Divide the surface of the tunnel into small blocks P s , replace the surface integral with the sum of the integrals of the interior points in each small block, and the calculation formula satisfies where F s.p (q) represents the function value at point p on the surface patch s for point q; q represents a specific point where the gradient needs to be calculated during the Poisson reconstruction process; represents the dot product of the normal vector of the surface patch s and the normal vector of the point p; represents the gradient vector at point q; S61e: Generate the Poisson model of the tunnel through the Marching Cube algorithm; S61f: Calculate the distribution probability of model triangles with the same perimeter in the Poisson model; S61g: The vertices of the triangles with the same perimeter and the largest distribution probability are used as sampling points; Assume that the number of input point clouds of preprocessed point cloud data is k, and the set of input point clouds P = {p1, p2, ..., p k }, the coordinates of the input point cloud are (x p ,y p ,z p ); The sampling point coordinates are (x q ,y q ,z q ), the set of sampling points Q = {q1,q2,…,q m }, calculate the Euclidean distance d between the sampling point and the sampling point of the input point cloud pq ; S61h: With the vertices of the triangle patch as rows and the input point cloud as columns, the Euclidean distance d from the vertex of the triangle patch to the input point cloud is calculated. pq Sort in ascending order, and the density of the vertices of the Poisson model is ρ′; S61i: Round down ρ′ to get [ρ′], and take d pq The [ρ′]th row of , the [ρ′]th row represents the distance from the [ρ′]th vertex to the input point cloud S61j: Traverse all vertices of the Poisson model and the input point cloud and calculate the corresponding distance between the two. If the distance value is greater than If the distance value is less than or equal to Mark the vertex as 1; S61k: Eliminate all triangles marked as 0 to obtain a non-closed tunnel 3D mesh model.

8. The method for three-dimensional modeling of multi-shape tunnels based on laser point cloud according to claim 1, characterized in that: Step S7 includes: S71: determining the pixel units where the vertices of the three-dimensional mesh model of the tunnel are located according to the point cloud flattening process; normalizing the position coordinates of the model vertices to obtain texture mapping coordinates; S72: constructing a three-dimensional texture model of the tunnel according to the three-dimensional mesh model of the tunnel, the two-dimensional intensity image, and the texture map coordinates.

9. A laser point cloud-based 3D modeling system for multi-shape tunnels, used to implement a laser point cloud-based 3D modeling method for multi-shape tunnels as claimed in any one of claims 1 to 8, characterized in that: The system includes: a data loading and display module, a tunnel point cloud denoising and simplification module, a tunnel point cloud texture flattening module and a tunnel point cloud 3D reconstruction module; The data loading and display module, the tunnel point cloud denoising and simplification module, the tunnel point cloud texture flattening module and the tunnel point cloud three-dimensional reconstruction modeling are sequentially connected; Data loading and display module, used to load and analyze point cloud data, 3D models, and 2D images; Tunnel point cloud denoising and streamlining module, used to pre-process point cloud data; The tunnel point cloud texture flattening module is used to calculate the minimum outer bounding circle of the tunnel section; determine the projection cylinder of the tunnel ring belt; project the point cloud of the tunnel ring belt onto the surface of the projection cylinder through the minimum outer bounding circle; The tunnel point cloud texture flattening module is also used to generate a two-dimensional intensity image of the inner surface of the tunnel lining based on the point cloud intensity values ​​of the projected cylinder and the tunnel annulus; The tunnel point cloud texture flattening module is also used to standardize the initial section point set of the tunnel section to obtain the standard section; correct the two-dimensional intensity image according to the minimum outer enclosing circle and the standard section; The tunnel point cloud 3D reconstruction module is used to model the pre-processed point cloud data using an improved Poisson reconstruction algorithm based on adaptive threshold segmentation to obtain a 3D mesh model of the tunnel; The tunnel point cloud 3D reconstruction module is also used to generate a tunnel 3D texture model by combining the tunnel 3D mesh model and the corrected 2D intensity image using automatic texture mapping technology.

Images