Tunnel surface point cloud reconstruction method based on expansion projection and region growth
Through the method based on unfolding projection and region growth, the problems of high computational complexity and inaccurate surface reconstruction in the point cloud reconstruction of ultra-under-digging tunnel are solved, and the accurate and efficient reconstruction of the point cloud of tunnel is achieved, which is suitable for complex tunnel surface treatment. .
Patent Information
- Application Number
- CN202410976763.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-21
- Publication Date
- 2024-11-19
AI Technical Summary
The prior art faces problems of high computational complexity, inaccurate surface reconstruction and serious reflection interference when dealing with point clouds of ultra-under-digging tunnels, especially when dealing with large amounts of point clouds, it is difficult to effectively reconstruct the complex tunnel surface morphology.
Using a method based on expansion projection and region growth, the point cloud is expanded through tunnel design contour lines, divided flat and non-flat areas, and the grid topology is reconstructed using the Delaunay triangular grid generation and region growth algorithm, combining feature point filtering and candidate points Acceptability, realizes accurate reconstruction of tunnel surface point clouds.
Reduces the calculation amount, improves reconstruction efficiency and accuracy, and can effectively handle complex tunnel surfaces, ensuring the quality and authenticity of the reconstruction grid.
Smart Images

Figure CN118982630A_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the technical field of tunnel engineering, relates to point cloud surface reconstruction in tunnel engineering, and in particular to a tunnel surface point cloud reconstruction method based on unfolding projection and regional growth. Background Art
[0002] The drilling and blasting method is still the mainstream construction method for mountain tunnel construction, but it cannot avoid the problem of over-break and under-break. Tunnels with over-break and under-break problems have complex surface morphology and many irregular potholes. Therefore, accurate and efficient reconstruction of the over-break and under-break tunnel surface is crucial for subsequent over-break and under-break calculations.
[0003] In recent years, 3D laser scanning technology has developed rapidly. Laser scanners can quickly obtain point cloud data on the surface of 3D objects. These point cloud data are huge and discretely distributed. Surface reconstruction is based on point clouds, obtaining reasonable topological relationships between point clouds, and restoring the continuous surface shape of 3D objects. In tunnel engineering, the ground and elliptical structure of the tunnel will cause serious reflection interference to the laser radar scanning, forming dense interference point clouds.
[0004] There are many methods for constructing surfaces, which can be divided into two categories: 1) implicit surface reconstruction methods, mainly including moving least squares method, B-spline surface, Poisson surface method, radial function method, etc. Such methods can obtain a strict surface, but the amount of calculation is large, and the generated surface does not necessarily pass through the actual point cloud, and the original surface information may be lost; 2) explicit surface reconstruction methods, the explicit method is divided into methods based on three-dimensional Delaunay triangulation and methods based on region growing. The former mainly includes Crust algorithm, Robust Cocone algorithm, etc.; such methods first divide the three-dimensional point cloud into tetrahedrons, and then delete redundant tetrahedrons to obtain the reconstructed surface, but the computational complexity is too high when processing a large number of point clouds; the latter mainly includes BPA algorithm, IPD algorithm, bounding sphere algorithm, etc. This method starts from the seed triangle, selects candidate points according to certain criteria to form a new triangle, and repeats the above process until a complete surface is formed. The idea of this method is simple, but it needs to screen and judge multiple candidate points. The screening criteria have a great influence on the quality of mesh reconstruction, and the amount of calculation is also high. Summary of the invention
[0005] In view of the above-mentioned deficiencies in the prior art, the present invention provides a tunnel surface point cloud reconstruction method based on unfolding projection and region growing, which can accurately reconstruct the tunnel surface point cloud topological structure.
[0006] In order to achieve the above-mentioned object of the invention, the present invention adopts the following technical solutions to achieve it.
[0007] The tunnel surface point cloud reconstruction method based on unfolding projection and region growing provided by the present invention comprises the following steps:
[0008] S1 unfolds the scanned tunnel point cloud based on the tunnel design contour line;
[0009] S2 performs plane grid division based on the unfolded point cloud, and divides the grid into flat areas and non-flat areas according to the curvature;
[0010] S3 uses projection-based Delaunay triangulated meshes for flat areas to generate the topological structure of the unfolded point cloud mesh;
[0011] S4 uses the regional growing algorithm to reconstruct the mesh of the non-flat area; specifically, the Delaunay triangular mesh boundary generated in step S3 is used as the growing edge, starting from the growing edge, a triangular mesh is formed based on the selected candidate points and the growing edge, and the growing edge set is updated at the same time, and this step operation is repeated until the mesh reconstruction of the non-flat area is completed, and the mesh topology structure of the unfolded point cloud is obtained;
[0012] S5 projects the point cloud mesh topological structure after unfolding the flat area and the non-flat area into the point cloud space before unfolding to obtain the surface of the tunnel point cloud reconstruction.
[0013] In the above step S1, the tunnel is scanned by a laser scanner to obtain a tunnel point cloud, and each point cloud is coded and marked.
[0014] Then the tunnel point cloud is unfolded; the tunnel surface unfolding method is to align the actual point cloud of the tunnel with the theoretical model according to the central axis after the central axis is extracted from the tunnel scanning point cloud, and unfold the tunnel point cloud based on the tunnel design contour line. The point coordinate conversion formula is as follows:
[0015]
[0016] Among them, (x i ,y i ,z i ) is the coordinate in the tunnel theory model (i.e., the expanded point cloud space); (X i ,Y i ) is the coordinate of the point before expansion; f(x) is the tunnel design contour equation, f′(x) is the derivative of f(x) with respect to x; p i is the i-th point; mindis(f(x),p i ) is the shortest distance from the ith point to the tunnel design contour.
[0017] In the above step S2, the plane grid division method is to project the expanded tunnel surface point cloud onto the XOY plane (tunnel ground), that is, to set the Z coordinates of all points to 0, and then use a plane rectangle to surround the projected point cloud, and the minimum and maximum values of the rectangle in the X-axis direction are x and y respectively. min 、x max , the minimum and maximum values in the Y-axis direction are y min ,y max , set the appropriate grid side length d according to the point cloud density x d y , the number of grids in the X and Y directions is as follows:
[0018]
[0019] Among them, D x =x max -x min ;D y =y max -y min ;
[0020] For all points in the point cloud, their index numbers are calculated using the following formula, and points with the same index number are grouped into the same grid.
[0021] index = Vu + v;
[0022] Where u=(x i -x min ) / d x ; v = (y i -y min ) / d y , x i ,y i are the X and Y coordinates of the i-th point in the point cloud, respectively.
[0023] After the plane grid division, the point cloud is divided into multiple grids of the same size. The surface variation of the point cloud in each grid is different. The covariance matrix is constructed for the point cloud in the grid. By calculating the eigenvalue of this matrix, the variation of the grid point cloud in three directions is obtained, as shown in the following formula:
[0024]
[0025] in, It is the kth point in the grid minus the average value of all points in the grid, and is a row vector consisting of 3 elements. x k ,y k 、z k For point p k X, Y, Z coordinates of It is the average value of the coordinates of all points in the grid.
[0026] The covariance matrix is a third-order symmetric semi-positive definite matrix with three eigenvalues λ 0 <λ 1 <λ 2 ,λ 0 Represents the change of the point cloud in the normal direction, so the curvature φ=λ 0 / (λ 0 +λ 1 +λ 2 ) describes the complexity of the features in the grid; if φ is greater than φ threshhold This indicates that the curvature of the grid changes greatly, and the grid is divided into a non-flat area; if φ is not greater than φ threshhold This indicates that the curvature of the grid changes little, and the grid is divided into a flat area; φ threshhold is the curvature threshold set.
[0027] In the above step S3, the Delaunay triangulated mesh based on the unfolded projection is generated by further refining the grid of the flat area into several small grids based on the first grid division, taking any point in each small grid to form a point set cloud1, and the remaining point cloud is cloud2. First, perform a three-dimensional Delaunay mesh reconstruction on cloud1 to obtain a Delaunay triangulated mesh; for each point in cloud2, find the triangle where the projection of each point in cloud2 is located through its projection on the XOY plane with the triangulated mesh generated by cloud1; then, in the unfolded point cloud space, combine the distance from the point to the corresponding triangle and the minimum value of the side length formed by the point and the triangle vertex to jointly determine whether the point is retained, as shown in the following formula:
[0028] B i =m×d ai +n×d bi ;
[0029] Among them, d ai is the distance from the i-th point to the corresponding triangle. The larger the distance, the greater the influence of the point on the accuracy of the reconstructed surface features. Let the vertex of the triangle be v 1 、v 2 and v 3 , the point is p i , d bi For edge The smaller the value is, the more uniform the generated triangular mesh will be. ai The weight of bi The weight of , m+n=1.
[0030] B iIf the value is greater than the set threshold, it means that the point is an important feature point and needs to be included in the mesh reconstruction of the point cloud. The feature point is connected to the three vertices of the corresponding triangular mesh to continue to construct the triangular mesh. Points that are not greater than the set threshold are deleted.
[0031] In the above step S4, the region growing algorithm grid reconstruction algorithm is to use the boundaries of these grids as growth edges after gridding the flat area, and based on the point screening principle, continuously select new points and growth edges to form new triangles, and update the growth edge set at the same time. As long as a reasonable point screening criterion is selected, the complex tunnel surface can be reconstructed.
[0032] When selecting candidate points, the flat area is first meshed. By judging whether each edge is repeated, the edge can be divided into an inner edge and a growing edge (the non-repeated edge is the growing edge). Each growing edge corresponds to a triangle. Let the two endpoints corresponding to the growing edge be V 1 and V 2 , the opposite vertex of the growing edge is V 3 .
[0033] The point screening principle is: first use kd tree to search V 1 and V 2 Points in the neighborhood of the two points, and the searched points and V 1 ,V 2 ,V 3 Project it onto the XOY plane and use the cross product to filter out the points outside the growing edge, as follows:
[0034]
[0035] Among them, p i is the i-th point in the point set, if flag i >0 or flag i =0, indicating point p i With V 3 On the growing side The same side or point p i With growth edge Collinear, does not meet the requirements; if flag i <0, indicating point p i With V 3 On the growing side The points on the opposite side of the growing edge meet the requirements; all points on the opposite side of the growing edge constitute candidate points.
[0036] Based on the minimum inner angle of the triangle to be grown, the radius of the circumscribed circle, the maximum side length, and the angle between the triangle to be grown and the generated triangle plane, a parameter acceptability is customized as the standard for selecting the appropriate triangle face (ie, triangle plane).
[0037]
[0038] Among them, α i is the minimum interior angle of the triangle to be generated; r min is the minimum radius of the circumscribed circle of all candidate points that form a triangle with the growing edge; r i e is the radius of the circumscribed circle of the triangle formed by the i-th candidate point and the growing edge; i is the maximum side length of the triangle to be generated; e grow is the length of the growing edge; β i is the angle between the generated triangle and the triangle to be generated; a, b, c, d are weights respectively, a+b+c+d=1.
[0039] After calculating the acceptability of each candidate point, sort them from large to small and make grid non-overlapping judgments in turn. If the grid non-overlapping principle is met, this point will be used as a feature point and will be involved in the grid reconstruction of the point cloud. The feature point will be retained and connected to the two end points of the growing edge to generate a triangular mesh. If it is not met, the next point will be selected for judgment until all candidate points have been judged, and then the next generated edge will be selected to continue grid reconstruction.
[0040] To determine whether the meshes overlap, we need to determine whether the edge length of the triangle to be generated intersects with the edge length of the existing triangle. Let the endpoints of the two line segments be P and 1 , P 2 and P 3 , P 4 , determine the line segment p by the following formula 1 p 2 、p 3 p 4 Intersect:
[0041]
[0042] In the formula, If the above equation is satisfied, it proves that the two line segments have no intersection.
[0043] In the above step S5, the method of projecting the expanded point cloud mesh topological structure into the point cloud space before expansion is to utilize the characteristic that the coding marks of each point in the point cloud before and after expansion remain unchanged, apply the plane mesh topological structure to the point cloud before expansion, and realize the spatial transformation of the point cloud topological structure.
[0044] The working principle of the above-mentioned tunnel surface point cloud reconstruction method based on unfolding projection and region growing is as follows: first, the point cloud tunnel surface is unfolded, and the point cloud surface is divided into flat areas and non-flat areas using a grid, and plane projection and region growing algorithms are used for different areas. Then, appropriate points are selected in the unfolded tunnel point cloud to construct a triangular mesh. Finally, the unfolded point cloud mesh topology is projected back to the point cloud space before unfolding to achieve tunnel surface reconstruction.
[0045] Compared with the prior art, the tunnel surface point cloud reconstruction method based on unfolding projection and region growing provided by the present invention has the following beneficial effects:
[0046] (1) The present invention unfolds the tunnel surface from the beginning, which reduces the amount of calculation. At the same time, it inherits the efficiency of plane projection and the accuracy of region growing algorithm, and can accurately reconstruct the topological structure of the tunnel surface point cloud;
[0047] (2) In the generation of Delaunay triangulated mesh based on plane projection, the present invention selects feature points by combining the distance from the point to the triangle where it is located and the maximum value of the angle formed by the point and the vertex of the triangle, thereby improving the efficiency of mesh reconstruction and ensuring the accuracy of mesh reconstruction in flat areas;
[0048] (3) In the process of mesh reconstruction by region growing algorithm, the present invention selects feature points by using the acceptability of customized candidate points, thereby further improving the reconstruction efficiency while effectively reconstructing the non-flat area;
[0049] (4) The multi-parameter setting of the present invention helps to adjust the mesh quality according to different reconstruction requirements, thereby ensuring that the reconstructed tunnel surface is more realistic and reasonable. BRIEF DESCRIPTION OF THE DRAWINGS
[0050] Figure 1 It is a flow chart of a method for reconstructing a tunnel surface point cloud based on unfolding projection and region growing according to an embodiment of the present invention;
[0051] Figure 2 This is a schematic diagram of the tunnel point cloud expansion;
[0052] Figure 3 Generate schematic diagram for planar projected delaynay grid;
[0053] Figure 4 Schematic diagram for screening candidate points for region growing;
[0054] Figure 5 It is an iterative process of mesh region growing;
[0055] Figure 6Schematic diagram of the tunnel surface point cloud reconstruction process; (a) is the tunnel scanning point cloud, (b) is the unfolded point cloud; (d) is the reconstructed tunnel surface. DETAILED DESCRIPTION
[0056] The specific implementation modes of the present invention are described below so that those skilled in the art can understand the present invention. However, it should be clear that the present invention is not limited to the scope of the specific implementation modes. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the attached claims, these changes are obvious, and all inventions and creations utilizing the concept of the present invention are protected.
[0057] Example 1
[0058] This embodiment provides a tunnel surface point cloud reconstruction method based on unfolding projection and region growing. Figure 1 As shown, the following steps are included:
[0059] S1 unfolds the tunnel point cloud based on the tunnel design contour line.
[0060] First, the tunnel is scanned by a laser scanner to obtain the tunnel point cloud, and each point in the point cloud is coded and marked.
[0061] Then the tunnel point cloud is unfolded; the tunnel surface unfolding method is as follows: Figure 2 As shown in the figure, after the central axis of the tunnel scanning point cloud is extracted, the actual point cloud of the tunnel is aligned with the theoretical model according to the central axis, and the tunnel point cloud is unfolded based on the tunnel design contour line. The point coordinate conversion formula is as follows:
[0062]
[0063] Among them, (x i ,y i ,z i ) is the coordinate in the tunnel theory model (i.e., the expanded point cloud space); (X i ,Y i ) is the coordinate of the point before expansion; f(x) is the tunnel design contour equation, f′(x) is the derivative of f(x) with respect to x; p i is the i-th point; mindis(f(x),p i ) is the shortest distance from the ith point to the tunnel design contour.
[0064] S2 performs plane grid division based on the unfolded point cloud, and divides the grid into flat areas and non-flat areas according to the curvature.
[0065] The plane grid division method is to project the expanded tunnel surface point cloud onto the XOY plane, that is, to set the Z coordinates of all points to 0, and then use a plane rectangle to surround the projected point cloud, with the minimum and maximum values of the rectangle in the X-axis direction being x and y respectively. min 、x max , the minimum and maximum values in the Y-axis direction are y min ,y max , set the appropriate grid side length d according to the point cloud density x d y , the number of grids in the X and Y directions is as follows:
[0066]
[0067] Among them, D x =x max -x min ;D y =y max -y min ;
[0068] For all points in the point cloud, use the following formula to calculate their index numbers, and assign points with the same index number to the same grid:
[0069] index = Vu + v;
[0070] Where u=(x i -x min ) / d x ; v = (y i -y min ) / d y , x i ,y i are the X and Y coordinates of the i-th point in the point cloud, respectively.
[0071] After the plane grid division, the point cloud is divided into multiple grids of the same size. The surface variation of the point cloud in each grid is different. The covariance matrix is constructed for the point cloud in the grid. By calculating the eigenvalue of this matrix, the variation of the grid point cloud in three directions is obtained, as shown in the following formula:
[0072]
[0073] in, Represents the kth point in the grid minus the average value of all points in the grid, which is a row vector consisting of 3 elements x k ,y k 、z k For point p k X, Y, Z coordinates of It is the average value of the coordinates of all points in the grid.
[0074] The covariance matrix is a third-order symmetric semi-positive definite matrix with three eigenvalues λ 0 <λ 1 <λ 2 ,λ 0 Represents the change of the point cloud in the normal direction, so the curvature φ = λ 0 / (λ 0 +λ 1 +λ 2 ) describes the complexity of features within the grid.
[0075] If φ is greater than φ threshhold This indicates that the curvature of the grid changes greatly, and the grid is divided into a non-flat area; if φ is not greater than φ threshhold This indicates that the curvature of the grid changes little, and the grid is divided into a flat area; φ threshhold is the curvature threshold set.
[0076] S3 uses projection-based Delaunay triangulated meshes for flat areas to generate the mesh topology of the unfolded point cloud.
[0077] The Delaunay triangulated mesh based on the unfolded projection is generated as follows: Figure 3 As shown in the figure, based on the first grid division, the grid in the flat area is further refined into several small grids, and any point in each small grid is selected to form the point set cloud1, and the remaining point cloud is taken as cloud2.
[0078] First, perform 3D Delaunay mesh reconstruction on cloud1 to obtain Delaunay triangular mesh. For each point in cloud2, find the triangle where the projection of each point in cloud2 is located by projecting it on the XOY plane with the triangular mesh generated by cloud1; then, in the expanded point cloud space, combine the distance from the point to the corresponding triangle and the minimum length of the side formed by the point and the vertex of the triangle to determine whether the point should be retained, as shown in the following formula.
[0079] B i =m×d ai +n×d bi ;
[0080] Among them, d ai is the distance from the i-th point to the corresponding triangle. The larger the distance, the greater the influence of the point on the accuracy of the reconstructed surface features. Let the vertex of the corresponding triangle be v 1 、v 2 and v 3 , the point is p i , d bi For edge The smaller the value is, the more uniform the generated triangular mesh will be.ai The weight of bi The weight of B is m+n=1. i If the value is greater than the set threshold, it means that the point is an important feature point and needs to be included in the mesh reconstruction of the point cloud. i If the value is greater than the set threshold, it means that the point is an important feature point and needs to be included in the mesh reconstruction of the point cloud. The feature point is connected to the three vertices of the corresponding triangular mesh to continue to construct the triangular mesh. Points that are not greater than the set threshold are deleted.
[0081] S4 uses the regional growing algorithm to reconstruct the mesh of the non-flat area; specifically: the Delaunay triangular mesh boundary generated in step S3 is used as the growing edge, starting from the growing edge, a triangular mesh is formed based on the selected candidate points and the growing edge, and the growing edge set is updated at the same time. This step is repeated until the mesh reconstruction of the non-flat area is completed, and the expanded point cloud mesh topology structure is obtained.
[0082] The region growing algorithm grid reconstruction algorithm is as follows: Figure 4 As shown in the figure, after the flat area is gridded, the boundaries of these grids are used as growth edges. Starting from the growth edges, based on the point screening principle, new points are continuously selected to form new triangles with the growth edges, and the growth edge set is updated at the same time. As long as a reasonable point screening criterion is selected, the complex tunnel surface can be reconstructed.
[0083] When selecting candidate points, the flat area is first meshed. By judging whether each edge is repeated, the edge can be divided into an inner edge and a growing edge (the non-repeated edge is the growing edge). Each growing edge corresponds to a triangle. Let the two endpoints corresponding to the growing edge be V 1 and V 2 , the opposite vertex of the growing edge is V 3 .
[0084] The point screening principle is: first use k-dtree (using the conventional method disclosed in the art) to search V 1 and V 2 Points in the neighborhood of the two points, and the searched points and V 1 ,V 2 ,V 3 Project it onto the XOY plane and use the cross product to filter out the points outside the growing edge, as follows:
[0085]
[0086] Among them, p i is the i-th point in the point set, if flag i >0 or flag i =0, indicating point p i With V 3 On the growing side The same side or point p i With growth edge Collinear, does not meet the requirements; if flag i <0, indicating point p i With V 3 On the growing side The opposite side meets the requirements.
[0087] Based on the minimum inner angle of the triangle to be grown, the radius of the circumscribed circle, the maximum side length, and the angle between the triangle to be grown and the generated triangle plane, a parameter acceptability is customized as the standard for selecting the appropriate triangle face (ie, triangle plane).
[0088]
[0089] Among them, α i is the minimum interior angle of the triangle to be generated, r min is the minimum radius of the circumscribed circle of all candidate points that form a triangle with the growing edge, r i is the radius of the circumscribed circle of the triangle formed by the i-th candidate point and the growing edge, e i is the maximum side length of the triangle to be generated, e grow is the length of the growing edge, β i is the angle between the generated triangle and the triangle to be generated, a, b, c, d are weights respectively, a+b+c+d=1.
[0090] After calculating the acceptability of each candidate point, sort them from large to small and make grid non-overlapping judgments in turn. If the grid non-overlapping principle is met, this point will be used as a feature point and will be involved in the grid reconstruction of the point cloud. The feature point will be retained and connected to the two end points of the growing edge to generate a triangular mesh. If it is not met, the next point will be selected for judgment until all candidate points have been judged, and then the next generated edge will be selected to continue grid reconstruction.
[0091] To determine whether the meshes overlap, we need to determine whether the edge length of the triangle to be generated intersects with the edge length of the existing triangle. Let the endpoints of the two line segments be P and 1 , P 2 and P 3 , P 4 , determine the line segment p by the following formula 1 p 2 、p 3 p 4 Intersect:
[0092]
[0093] In the formula, If the above equation is satisfied, it proves that the two line segments have no intersection. Otherwise, the two line segments have a focus, indicating that the grids overlap and the corresponding point is not a feature point. The next point is selected for judgment. The tunnel surface grid region growth iterative process used in this embodiment is as follows: Figure 5 shown.
[0094] S5 projects the point cloud mesh topological structure after unfolding the flat area and the non-flat area into the point cloud space before unfolding to obtain the surface of the tunnel point cloud reconstruction.
[0095] Taking advantage of the fact that the coding marks of each point in the point cloud before and after expansion remain unchanged, the plane grid topological structure is applied to the point cloud before expansion to achieve the spatial transformation of the point cloud topological structure.
[0096] Figure 6 (a) shows the tunnel point cloud obtained by scanning with a laser scanner; the expanded point cloud is obtained by the previous step S1, such as Figure 6 (b); then follow the steps S2-S5 given above to obtain the reconstructed tunnel surface, as shown in Figure 6 (c) as shown.
[0097] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0098] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.
[0099] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.
[0100] The present invention uses specific embodiments to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea. At the same time, for those skilled in the art, according to the idea of the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.
[0101] Those skilled in the art will appreciate that the embodiments described herein are intended to help readers understand the principles of the present invention, and should be understood that the protection scope of the present invention is not limited to such specific statements and embodiments. Those skilled in the art can make various other specific variations and combinations that do not deviate from the essence of the present invention based on the technical revelations disclosed by the present invention, and these variations and combinations are still within the protection scope of the present invention.
Claims
1. A tunnel surface point cloud reconstruction method based on unfolding projection and region growing, It is characterized in that The following steps are involved: S1 unfolds the scanned tunnel point cloud based on the tunnel design contour line; S2 performs plane grid division based on the unfolded point cloud, and divides the grid into flat areas and non-flat areas according to the curvature; S3 uses projection-based Delaunay triangulated meshes for flat areas to generate the topological structure of the unfolded point cloud mesh; S4 uses the regional growing algorithm to reconstruct the mesh of the non-flat area; specifically, the Delaunay triangular mesh boundary generated in step S3 is used as the growing edge, starting from the growing edge, a triangular mesh is formed based on the selected candidate points and the growing edge, and the growing edge set is updated at the same time, and this step operation is repeated until the mesh reconstruction of the non-flat area is completed, and the mesh topology structure of the unfolded point cloud is obtained; S5 projects the point cloud mesh topological structure after unfolding the flat area and the non-flat area into the point cloud space before unfolding to obtain the surface of the tunnel point cloud reconstruction.
2. The tunnel surface point cloud reconstruction method based on unfolding projection and region growing according to claim 1, It is characterized in that In step S1, the tunnel surface development method is to align the actual point cloud of the tunnel with the theoretical model according to the central axis after extracting the central axis of the tunnel scanning point cloud, and to develop the tunnel point cloud based on the tunnel design contour line. The point coordinate conversion formula is as follows: Among them, (x i ,y i ,z i ) is the coordinate in the tunnel theory model; (X i ,Y i ) is the coordinate of the point before expansion; f(x) is the tunnel design contour equation, f′(x) is the derivative of f(x) with respect to x; p i is the i-th point; mindis(f(x),p i ) is the shortest distance from the ith point to the tunnel design contour.
3. The tunnel surface point cloud reconstruction method based on unfolding projection and region growing according to claim 1, It is characterized in that In step S2, the plane grid division method is to project the expanded tunnel surface point cloud onto the XOY plane, that is, to set the Z coordinates of all points to 0, and to surround the projected point cloud with a plane rectangle, the minimum and maximum values of which are x and y in the X-axis direction, respectively. min 、x max , the minimum and maximum values in the Y-axis direction are y min ,y max , set the grid side length d according to the point cloud density x ,d y , the number of grids in the X and Y directions is as follows: where D x = x max - x min ; D y = y max - y min ; For all points in the point cloud, use the following formula to calculate their index numbers, and assign points with the same index number to the same grid; index = Vu + v; Where u=(x i -x min ) / d x ; v = (y i -y min ) / d y , x i ,y i are the X and Y coordinates of the i-th point in the point cloud respectively; Construct the covariance matrix for the point cloud in the grid as follows: in, It is the kth point in the grid minus the average value of all points in the grid, and is a row vector consisting of 3 elements. x k ,y k 、z k For point p k X, Y, Z coordinates of is the average value of the coordinates of all points in the grid; The three eigenvalues λ of the covariance matrix 0 <λ 1 <λ 2 ,λ 0 Represents the change of the point cloud in the normal direction, with curvature φ = λ 0 / (λ 0 +λ 1 +λ 2 ) describes the complexity of the features in the grid; if φ is greater than φ threshhold , divide the grid into non-flat areas; if φ is not greater than φ threshhold , dividing the grid into flat areas; φ threshhold is the curvature threshold set.
4. The tunnel surface point cloud reconstruction method based on unfolding projection and region growing according to claim 1, It is characterized in that In step S3, the grid in the flat area is further refined into several small grids, and any point in each small grid is selected to form a point set cloud1, and the remaining point cloud is used as cloud2; cloud1 is first reconstructed into a three-dimensional Delaunay grid to obtain a Delaunay triangulated grid.
5. The tunnel surface point cloud reconstruction method based on unfolding projection and region growing according to claim 4, It is characterized in that For each point in cloud2, find the triangle where the projection of each point in cloud2 is located by projecting it and the triangular mesh generated by cloud1 on the XOY plane; then in the expanded point cloud space, combine the distance from the point to the corresponding triangle and the minimum length of the side formed by the point and the vertex of the triangle to determine whether the point should be retained, as shown in the following formula: B i =m×d ai +n×d bi ; Among them, d ai is the distance from the i-th point to the corresponding triangle; let the vertex of the triangle be v 1 、v 2 and v 3 , the point is p i , d bi For edge The minimum value among d ai The weight of bi The weight of B is m+n=1; i Points larger than the set threshold are taken as feature points, and the feature points are connected to the three vertices of the corresponding triangular mesh to continue to construct the triangular mesh; points not larger than the set threshold are deleted.
6. The tunnel surface point cloud reconstruction method based on unfolding projection and region growing according to any one of claims 1 to 5, It is characterized in that In step S4, when selecting candidate points, the flat area is first meshed. By judging whether each edge is repeated, the edge is divided into an inner edge and a growing edge. Each growing edge corresponds to a triangle. Let the two endpoints corresponding to the growing edge be V 1 and V 2 , the opposite vertex of the growing edge is V 3 ; First use kd tree to search for V 1 and V 2 Points in the neighborhood of the two points, and the searched points and V 1 ,V 2 ,V 3 Project it onto the XOY plane and use the cross product to filter out the points outside the growing edge, as follows: Among them, p i is the i-th point in the point set, if flag i >0 or flag i =0, indicating point p i With V 3 On the growing side The same side or point p i With growth edge Collinear, does not meet the requirements; if flag i <0, indicating point p i With V 3 On the growing side The points on the opposite side of the growing edge meet the requirements; all points on the opposite side of the growing edge constitute candidate points.
7. The tunnel surface point cloud reconstruction method based on unfolding projection and region growing according to claim 6, It is characterized in that Customize a parameter acceptability as a criterion for selecting appropriate triangles: Among them, α i is the minimum interior angle of the triangle to be generated; r min is the minimum radius of the circumscribed circle of all candidate points that form a triangle with the growing edge; r i e is the radius of the circumscribed circle of the triangle formed by the i-th candidate point and the growing edge; i is the maximum side length of the triangle to be generated; e grow is the length of the growing edge; β i is the angle between the generated triangle patch and the triangle patch to be generated; a, b, c, d are weights respectively, a+b+c+d=1; After calculating the acceptability of each candidate point, sort them from large to small and make grid non-overlapping judgments in turn. If the grid non-overlapping principle is met, this point is used as a feature point and connected to the two end points of the growing edge to generate a triangular mesh; if it is not met, the next point is selected for judgment.
8. The tunnel surface point cloud reconstruction method based on unfolding projection and region growing according to claim 7, It is characterized in that To determine whether the meshes overlap, we need to determine whether the edge length of the triangle to be generated intersects with the edge length of the existing triangle. Let the endpoints of the two line segments be P and 1 , P 2 and P 3 , P 4 , determine the line segment p by the following formula 1 p 2 、p 3 p 4 Intersect: In the formula, If the above equation is satisfied, it proves that the two line segments have no intersection.
