Double-layer fast indexing method and system for massive point cloud data of underground caverns

Abstract

The present invention discloses a dual-layer rapid indexing method and system for massive point cloud data of underground caverns, comprising: sampling and processing the acquired point cloud data of the underground cavern to obtain an initial model of the cavern point cloud; obtaining the center point of the key characteristic section of the cavern according to the initial point cloud model, and fitting the unit vector of the cavern axis; constructing a top-level segmented index structure of the cavern point cloud data along the axial direction of the cavern based on the fitted cavern axis vector; for each point cloud subset after the top-level segmentation, constructing a virtual octree index structure, and recording it in a node file, to obtain a dual index of the underground cavern point cloud data. The present invention provides a solution for dual-layer indexing, which performs segmented decomposition of the underground cavern engineering point cloud data along the cavern axis, and then constructs a virtual octree index for each segmented point set. With the support of the dual-layer indexing structure, the balance and query efficiency of the point cloud spatial index nodes are effectively improved.

Description

Technical Field

[0001] The present invention relates to fields such as point cloud data processing technology, and specifically to a dual-layer rapid indexing management method and system for massive point cloud data of underground caverns. The method constructs a rapid spatial index for massive point cloud data based on the spatial distribution characteristics of underground cavern projects. Background Art

[0002] Underground cavern projects are safe, convenient, environmentally friendly, and minimally impacted by the natural environment and human activities. They are increasingly being used in transportation, water conservancy, energy, and other fields, particularly in mountainous and hilly areas with challenging surface environments. Underground cavern projects are playing an increasingly important role in economic and social development. However, underground cavern projects are characterized by their large scale, complex structures, and complex interactions between surrounding rock and structure, making their construction, operation, and maintenance challenging. Comprehensive monitoring of underground cavern projects is a crucial means of ensuring project safety throughout their entire lifecycle, from construction to operation and maintenance.

[0003] In recent years, with the improvement of the accuracy of 3D laser scanners and the portability and popularization of equipment applications, its advantages in comprehensively acquiring 3D point cloud data of the surface of objects, and its characteristics of fast measurement speed, high accuracy and strong anti-interference, have gradually been promoted and applied in the construction and operation and maintenance of tunnel projects. The ultra-large-scale digital 3D of underground cavern projects has gradually become a reality.

[0004] However, the large-scale project and ultra-high-density laser scanning accuracy also create immense pressure to process massive amounts of point cloud data. Underground cavern projects often span long distances, with relatively small, narrow cavern diameters. The resulting point cloud data on the cavern walls is characterized by uneven spatial distribution and high data redundancy. The efficient and effective indexing, organization, management, and dynamic scheduling of massive amounts of data from underground cavern projects directly impacts the efficiency of point cloud data processing, including denoising and reconstruction.

[0005] Currently, common point cloud data spatial indexing methods include single indexing methods such as grid indexing, quadtree indexing, octree indexing, and KD tree indexing, as well as derived hybrid indexing methods. However, single point cloud indexing methods are difficult to adapt to the spatial distribution characteristics of underground cavern engineering point clouds. The constructed point cloud index tree is prone to problems such as poor node balance and excessive index tree depth. Although hybrid indexing methods improve the effect of single index construction, they do not consider the posture characteristics of cavern point clouds and are not well adapted to subsequent processing requirements such as cavern point cloud segmentation, cross-section extraction, and orthographic projection expansion. As a result, they are unable to effectively support the organization and management of underground cavern engineering point cloud data. Summary of the Invention

[0006] In view of the shortcomings of the existing technology, the purpose of the present invention is to propose a two-layer fast index management method for massive point cloud data of underground caverns, redistribute the point cloud data under the guidance of the cavern point cloud axis, realize the fast spatial index construction of massive point cloud data of underground cavern engineering, improve the balance of the index tree, optimize the index memory management, improve the data indexing efficiency, and provide effective index support for cavern point cloud post-processing.

[0007] To achieve the above-mentioned object, the present invention provides the following technical solution: a two-layer fast indexing method for massive point cloud data of underground caverns, characterized by comprising the following steps:

[0008] Step S1: Read the point cloud data file and create an initial sampling model of the underground cavern point cloud;

[0009] Step S2: Based on the sampled cave point cloud data small model, obtain the center point startC of the key feature start and end sections of the cave i (sx i ,sy i ,sz i ) and endC i (ex i ,ey i ,ez i ), fitting the unit vector a of the cave axis i (x i ,y i ,z i ), set the axis segment distance d;

[0010] Step S3: traverse and read the point cloud data, map the points to the direction of the cavern axis based on the fitted cavern axis unit vector, and construct the top-level segmented index structure of the cavern point cloud data along the direction of the cavern;

[0011] Step S4: Perform temporary coordinate transformation on each point cloud subset after the top-level segmentation according to the axis vector information in step S2; then, based on the regular grid subdivision, the local virtual octree index structure is constructed from the bottom up through the spatial index merging algorithm of grid refinement subdivision-layer-by-layer grid merging. The original point cloud data is not modified and is recorded in the node file, completing the double-layer index construction of the underground cavern point cloud data.

[0012] Specifically, the small model of the underground cavern engineering after point cloud sampling is used to efficiently and quickly obtain the cavern axis information, and the massive point cloud data is reorganized along the cavern axis to construct a two-layer index structure: among them, the top-level segmented index is divided according to the distance along the cavern axis from the starting section to establish a linear segmented index; secondly, the bottom-level virtual octree index structure refers to the spatial index merging construction algorithm of grid refinement-layer-by-layer grid merging without changing the original point cloud data, and the octree index construction is completed based on the temporary coordinate transformation of the point cloud data based on the axis vector information.

[0013] Specifically, step S1 includes:

[0014] Step S101: Read the selected point cloud data file and obtain the number of points NumPts contained in the point cloud file;

[0015] Step S102: According to the number of point clouds that the computer can load and render at one time, set the number of initial sampling models of underground cavern point clouds NPts, and calculate the initial sampling rate based on the point cloud information obtained in step S101 and sampling step size

[0016] Using file memory mapping, large point cloud files are read in sequentially, and each point cloud file is randomly sampled with sampleStep as the sampling interval to create an initial prototype model of the underground cavern, which is then rendered and visualized.

[0017] Specifically, step S2 includes:

[0018] Step S201: Based on the point cloud sampling model created in step S1, create the key section where the axial change of the cavern occurs, and obtain the corresponding key starting and ending section center point startC i (sx i ,sy i ,sz i ) and endC i (ex i ,ey i ,ez i );

[0019] Step S202: Calculate the axial unit vector a between different key sections based on the coordinates of the center point of the key section of the cavern in step S201. i (x i ,y i ,z i ) and axial length l i , calculate the number of segments of the cave point cloud model along the axis Where d is the segment distance along the axis of the cavern.

[0020] Specifically, a compact segmented point cloud bounding box is obtained by temporarily transforming the massive point cloud data of the underground cavern and redistributing it along the cavern axis. Step S3 is specifically as follows:

[0021] For points in the point cloud Calculate the corresponding starting section center point startC i To point P j The vector v j , P j The top segment index is At the same time, in the process of traversing the point cloud, the original data is not changed, and the point set is temporarily transformed to obtain the point Make the coordinate system z-axis consistent with the tunnel axis direction. The coordinate transformation formula is as follows:

[0022]

[0023] where α is the vector a i The angle between the projection on the xoz plane and the z axis, β is the vector a i The angle between the xoz plane and the minimum and maximum vertices of each segment bounding box (tMinPt k ,tMaxPt k );

[0024] Specifically, an octree index merging construction method of grid refinement and grid merging layer by layer is used to solve the node conflict problem when merging block index structures in a large-scale point cloud data block processing scenario. The step S4 specifically includes:

[0025] Step S401: According to the three-dimensional grid method, the maximum layer depth of the virtual octree is set to maxDepth, a regular grid index is established for each segment point set determined in S3, and the grid index is encoded for subsequent grid merging operations;

[0026] Step S402: merging grids according to the grid codes established in step S401 to establish a virtual octree index structure of the underground cavern engineering point cloud.

[0027] Specifically, in step S401, the minimum and maximum vertices (tMinPt k ,tMaxPt k ) Calculate the bounding box size of each segmented point cloud k (l k ,w k ,h k ), the maximum number of segments of each bounding box maxlength = 2 maxDepth and the minimum cell size minSize;

[0028] Traverse the points P in each segmented point cloud set j , after the temporary coordinate transformation Determine the point P j The three-dimensional grid index (i, j, k) of the point P in the outer bounding box of the segmented point set;

[0029]

[0030]

[0031]

[0032] Then, perform grid coding on the calculated three-dimensional grid index (i, j, k).

[0033] Specifically, the coding rule is as follows: Convert the grid index (i, j, k) into a binary index representation form (i 0B , j 0B , k 0B ):

[0034]

[0035] where, x l , y l , z l ∈{0,1}, 1 < l ≤ maxDepth,; Finally, encode the grid in the "z - y - x" three-axis order from top to bottom according to the octree depth, and the calculation formula is as follows,

[0036]

[0037] [[ID=4�]]where, code (maxdepth) is the regular grid encoding of the maximum subdivision depth. [[ID=^{50}]]

[0038] Specifically, in step S402, the point cloud data encoding of each segmented point set is sorted in ascending order according to the grid encoding, and the point cloud information contained in each grid at each depth, that is, the number of points in the grid and the index of the contained points, is counted layer by layer from top to bottom. The grid encoding conversion formula for the specified depth depth is as follows:

[0039] code (depth) = code (maxdepth) >>(3 * (maxdepth - depth))

[0040] where, code (depth) is the regular grid encoding of the specified subdivision depth converted from the encoding code (maxdepth) based on the maximum subdivision depth; The bottom regular grid is in accordance with the same code (depth)Merge layer by layer. When the number of points in the grid is less than the number threshold or reaches the maximum depth, the point data in the grid is output as a leaf node to complete the construction of the virtual octree of the segmented point set.

[0041] The present invention also provides a dual-layer rapid indexing system for massive point cloud data of underground caverns, comprising:

[0042] Module 1: It is used to read point cloud data files and create an initial sampling model of underground cavern point clouds;

[0043] Module 2: It is used to obtain the center point startC of the key feature of the cave based on the sampled cave point cloud data small model i (sx i ,sy i ,sz i ) and endC i (ex i ,ey i ,ez i ), fitting the unit vector a of the cave axis i (x i ,y i ,z i ), set the axis segment distance d;

[0044] Module 3: It is used to traverse and read point cloud data, map points to the direction of the cavern axis based on the fitted cavern axis unit vector, and construct the top-level segmented index structure of the cavern point cloud data along the direction of the cavern;

[0045] Module 4: It is used to perform temporary coordinate transformation on each point cloud subset after the top-level segmentation. Based on the regular grid subdivision based on the axis vector information of module 2, the local virtual octree index structure is constructed from the bottom up through the spatial index merging algorithm of grid refinement subdivision and layer-by-layer grid merging. The original point cloud data is not modified and is recorded in the node file, completing the double-layer index construction of the underground cavern point cloud data.

[0046] Beneficial effects of the present invention:

[0047] The present invention aims to address the problem that the single point cloud indexing method in the existing technology is difficult to adapt to the spatial distribution characteristics of underground cavern engineering point clouds, and the constructed point cloud index tree has poor balance. It effectively supports the organization and management of underground cavern engineering point cloud data and provides a two-layer fast indexing management method for massive point cloud data of underground caverns. It proposes to use the axis information of the underground cavern point cloud model to perform top-level axial segmentation and bottom-level virtual octree double-layer spatial index construction, avoiding the node conflict problem in the index file merging process after block processing in large-scale point cloud data processing scenarios, improving the organization balance and retrieval efficiency of massive point cloud data of underground cavern projects, and providing direct indexing support for subsequent processing tasks such as segmentation, section extraction, and orthographic projection expansion around the cavern axis based on cavern point cloud data. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the examples of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for the embodiments or the description of the prior art.

[0049] Figure 1 This is a schematic diagram of the indexing of massive point cloud data for underground cavern engineering according to the present invention;

[0050] Figure 2 A flowchart for constructing a double-layer index for an underground cavern engineering point cloud model provided by the present invention;

[0051] Figure 3 Schematic diagram comparing the "top-down, layer-by-layer refinement" octree construction method and the "grid refinement subdivision-layer-by-layer grid merging" bottom-up merging octree construction method used in the present invention. DETAILED DESCRIPTION

[0052] To make the purpose, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians of the present invention without making creative efforts are within the scope of protection of the present invention.

[0053] Example 1

[0054] This example constructs a dual-layer index structure for a tunnel point cloud data set and dynamically visualizes massive point cloud data based on internal and external memory scheduling. Obviously, this example is only intended to further illustrate the technical details and promote the application of the present invention. Specifically, the following steps are included:

[0055] Step S1: Read the point cloud data file and create an initial sampling model of the underground cavern point cloud. This includes:

[0056] Step S101: Read the selected point cloud data file and obtain the number of points NumPts contained in the point cloud file.

[0057] Among them, for files with point cloud file header information, the number of points in the file header information is interpreted, and for files containing only point cloud data information, the number of points in the point cloud is obtained by traversing all point information;

[0058] Step S102: According to the number of point clouds that the computer can load and render at one time, set the number of initial sampling models of underground cavern point clouds NPts. Calculate the initial sampling rate based on the point cloud information obtained in step S101. and sampling step size The point cloud files are read in sequentially, and each point cloud file is randomly sampled with sampleStep as the sampling interval to create an initial prototype model of the underground cavern, and the model is rendered and visualized.

[0059] In this example, NPts is set to 10000000.

[0060] Step S2: Obtain the center point of the key characteristic section of the cavern and fit the cavern axis vector. This includes:

[0061] Step S201: Based on the point cloud sampling model created in step S1, create the key section where the axial change of the cavern occurs, and obtain the corresponding key starting and ending section center point startC i (sx i ,sy i ,sz i ) and endC i (ex i ,ey i ,ez i );

[0062] The geometric information of the axial mutation of the cave point cloud model can be obtained by point cloud projection or manual interactive point selection. In this example, the cross-section center point of the key cross section of the point cloud sampling model established in S1 is selected by manual interaction.

[0063] Step S202: Calculate the axial unit vector a between different key sections based on the coordinates of the center point of the key section of the cavern in step S201. i (x i ,y i ,z i ) and axial length l i Set the segment distance d along the cavern axis and calculate the number of segments of the cavern point cloud model along the axis.

[0064] The segment distance d of the cavern axis is set to the cavern diameter in the engineering design data.

[0065] Step S3: Based on the fitted cavern axis vector, construct the top-level segmented index structure of the cavern point cloud data along the direction of the cavern. Specifically including:

[0066] Traverse and read the point cloud data, map the points to the direction of the cavern axis according to the coordinates of the points, and classify them into the corresponding cavern axis segments to build the segment index in the top-level index structure.

[0067] Specifically, traverse the points in the point cloud Calculate the corresponding starting section center point startC i To point P j The vector v j =(p x j -sx i ,p y j -sy i ,p z j -sz i ), click P j The top segment index

[0068] Furthermore, during the traversal of the point cloud, temporary coordinate transformation is required for the point set so that the z-axis of the coordinate system is consistent with the direction of the tunnel axis. The coordinate transformation formula is as follows, where α is the vector a i The angle between the projection on the xoz plane and the z axis, β is the vector a i The angle between the xoz plane and the bounding box of each segment point set is determined during the traversal process (tMinPt k ,tMaxPt k ).

[0069]

[0070] Step S4: For each point cloud subset after top-level segmentation, a virtual octree index structure is constructed and recorded in the node file, completing the task of establishing a double-layer index structure for underground cavern point cloud data; including:

[0071] Step S401: According to the three-dimensional grid method, the maximum layer depth maxDepth of the virtual octree is set, a grid index is established for each segment point set determined in S3, and the grid index is encoded;

[0072] Specifically, in step S401, the outer bounding box size of each segmented point cloud is calculated based on the minimum and maximum vertex coordinates of the outer bounding box of the segmented point cloud. k (lk , w k , h k ), the maximum number of segments of the outer box maxlength = 2 maxDepth and the minimum cell size minSizeIn this example, maxDepth = 8, maxlength = 2 8 = 256; where:

[0073]

[0074]

[0075] Specifically, in step S401, traverse the points P in each segmented point cloud j , perform a temporary conversion on it to obtain the point Calculate the three-dimensional grid index (i, j, k) of the point P j in the outer bounding box of the segmented point set;

[0076]

[0077] Furthermore, perform grid encoding on the calculated three-dimensional grid index (i, j, k). The encoding rule is as follows: Convert the grid index (i, j, k) into a binary index representation form (i 0B , j 0B , k 0B ):

[0078]

[0079] where, x l , y l , z l ∈ {0, 1}, 1 < l ≤ maxDepth,; Finally, encode the grid in the "z - y - x" three-axis order from top to bottom according to the octree depth. The calculation formula is as follows, code (maxdepth) is the regular grid encoding of the maximum dissection depth

[0080]

[0081] Step S402: According to the grid encoding established in step S401, perform grid merging to establish a virtual octree index structure for the point cloud of the underground chamber project

[0082] Specifically, in step S402, sort the point cloud data encoding of each segmented point set in ascending order according to the grid encoding, and layer by layer from top to bottom, count the point cloud information contained in each depth grid, that is, the number of points in the grid and the index of the included points. The conversion formula for the grid encoding at the specified depth depth is as follows:

[0083] code(depth) =code (maxdepth) >>(3*(maxdepth-depth))

[0084] Among them, code (depth) It is based on the maximum segmentation depth code (maxdepth) The converted regular grid code of the specified subdivision depth is converted into a regular grid code; the underlying regular grid is converted into a regular grid code according to the same code (depth) Merge layer by layer. When the number of points in the grid is less than the number threshold or reaches the maximum depth, the point data in the grid is output as a leaf node to complete the construction of the virtual octree of the segmented point set.

[0085] On this basis, based on the segmented octree index results and layer sampling, an LOD model is established, and the massive point cloud data of underground caverns is dynamically displayed through internal and external memory scheduling technology; specifically,

[0086] Step S501: According to the computer's rendering capability, the maximum level of the LOD model, max_LodLevel, is set. The segmented point sets along the axial segmentation are sampled and independently constructed into LOD models of different levels, and the LOD models are output to a designated location in the external memory.

[0087] Step S502: Dynamically read the LOD model based on viewpoint transformation and dynamic scheduling of internal and external memory, realize dynamic visualization of massive point cloud data, and reduce memory usage during the point cloud data visualization process.

[0088] Example 2

[0089] This embodiment provides a two-layer fast indexing system for massive point cloud data of underground caverns, including:

[0090] Module 1: It is used to read point cloud data files and create an initial sampling model of underground cavern point clouds;

[0091] Module 2: It is used to obtain the center point startC of the key feature of the cave based on the sampled cave point cloud data small model i (sx i ,sy i ,sz i ) and endC i (ex i ,ey i ,ez i ), fitting the unit vector a of the cave axis i (x i ,y i ,z i ), set the axis segment distance d;

[0092] Module 3: It is used to traverse and read point cloud data, map points to the direction of the cavern axis based on the fitted cavern axis unit vector, and construct the top-level segmented index structure of the cavern point cloud data along the direction of the cavern;

[0093] Module 4: It is used to perform temporary coordinate transformation on each point cloud subset after the top-level segmentation. Based on the regular grid subdivision based on the axis vector information of module 2, the local virtual octree index structure is constructed from the bottom up through the spatial index merging algorithm of grid refinement subdivision and layer-by-layer grid merging. The original point cloud data is not modified and is recorded in the node file, completing the double-layer index construction of the underground cavern point cloud data.

[0094] Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0095] Obviously, those skilled in the art may make various changes and modifications to the embodiments of the present invention without departing from the spirit and scope of the embodiments of the present invention. Thus, if such changes and modifications of the embodiments of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is intended to include such changes and modifications.

[0096] Other parts not described in detail are prior art.

Claims

1. A two-layer fast indexing method for massive point cloud data of underground caverns, characterized by: The steps include: Step S1: Read the point cloud data file and create an initial sampling model of the underground cavern point cloud; Step S2: Based on the sampled cave point cloud data small model, obtain the center point startC of the key feature start and end sections of the cave i ( sx i , sy i , sz i ) and endC i ( ex i , ey i , ez i ), fitting the unit vector of the cave axis a i ( x i , y i , z i ), set the axis segment distance d ; Step S3: traverse and read the point cloud data, map the points to the direction of the cavern axis based on the fitted cavern axis unit vector, and construct a top-level segmented index structure of the cavern point cloud data along the direction of the cavern; obtain a compact segmented point cloud bounding box by temporarily transforming the massive point cloud data of the underground cavern and redistributing it along the cavern axis. Step S3 is specifically as follows: For point P in the point cloud j ( p x j , p y j , p z j ), calculate the corresponding starting section center point startC i To point P j Vector v j , P j The top segment index is ; At the same time, in the process of traversing the point cloud, the original data is not changed, and the point set is temporarily transformed to obtain the point p ( , , ), so that the z-axis of the coordinate system is consistent with the direction of the tunnel axis. The coordinate transformation formula is as follows: in α is a vector a i With xoz The angle between the projection of the plane and the z-axis, β is a vector a i With xoz The angle of the plane, during the traversal process, determines the minimum and maximum vertices of each segment bounding box (tMinPt k , tMaxPt k ); Step S4: Perform temporary coordinate transformation on each point cloud subset after the top-level segmentation according to the axis vector information in step S2; then, based on the regular grid subdivision, the local virtual octree index structure is constructed from the bottom up through the spatial index merging algorithm of grid refinement subdivision-layer-by-layer grid merging. The original point cloud data is not modified and is recorded in the node file, completing the double-layer index construction of the underground cavern point cloud data.

2. The dual-layer fast indexing method for massive point cloud data of underground caverns according to claim 1 is characterized in that: The small model of the underground cavern project after point cloud sampling is used to efficiently and quickly obtain the cavern axis information, reorganize the massive point cloud data along the cavern axis, and construct a two-layer index structure: among them, the top-level segmented index is divided according to the distance along the cavern axis from the starting section to establish a linear segmented index; secondly, the bottom-level virtual octree index structure refers to the spatial index merging construction algorithm through grid refinement subdivision and layer-by-layer grid merging, without changing the original point cloud data, and completing the octree index construction based on the temporary coordinate transformation of the point cloud data based on the axis vector information.

3. The dual-layer fast indexing method for massive point cloud data of underground caverns according to claim 1 is characterized in that: Step S1 includes: Step S101: Read the selected point cloud data file and obtain the number of points in the point cloud file NumPts ; Step S102: Set the number of initial sampling models of underground cavern point clouds according to the number of point clouds that the computer can load and render at one time. NPts , calculate the initial sampling rate based on the point cloud information obtained in step S101 and sampling step size ; Using file memory mapping, large point cloud files are read in sequence, and each point cloud file is mapped to sampleStep Random sampling was performed for the sampling intervals, an initial prototype model of the underground chamber was created, and the model was rendered for visualization.

4. The dual-layer fast indexing method for massive point cloud data of underground caverns according to claim 1 is characterized in that: Step S2 includes: Step S201: Based on the point cloud sampling model created in step S1, create the key section where the axial change of the cavern occurs, and obtain the corresponding key starting and ending section center point startC i ( sx i , sy i , sz i ) and endC i ( ex i , ey i , ez i ); Step S202: Calculate the axial unit vectors of the cavern between different key sections based on the coordinates of the center points of the key sections in step S201. a i ( x i , y i , z i ) and axial length l i , calculate the number of segments of the cave point cloud model along the axis ,in d It is the segment distance along the axis of the cave.

5. The dual-layer fast indexing method for massive point cloud data of underground caverns according to claim 1 is characterized in that: The octree index merging construction method of grid refinement and layer-by-layer grid merging is used to solve the node conflict problem when merging block index structures in the large-scale point cloud data block processing scenario. The step S4 specifically includes: Step S401: According to the three-dimensional grid method, set the maximum layer depth of the virtual octree maxDepth , establish a regular grid index for each segment point set determined in S3, and encode the grid index for subsequent grid merging operations; Step S402: merging grids according to the grid codes established in step S401 to establish a virtual octree index structure of the underground cavern engineering point cloud.

6. The dual-layer fast indexing method for massive point cloud data of underground caverns according to claim 5 is characterized in that: In step S401, the minimum and maximum vertices (tMinPt k , tMaxPt k ) Calculate the bounding box size of each segmented point cloud Size k ( l k , w k , h k ), the maximum number of segments per bounding box maxlength =2 maxDepth and the minimum cell size minSize ; Traverse the point P in each segmented point cloud set j , after temporary coordinate transformation p ( , , ) Determine point P j The 3D grid index in the bounding box outside the segment point set ( i, j, k ); Then, the calculated three-dimensional grid index ( i, j, k ) for grid coding.

7. The dual-layer fast indexing method for massive point cloud data of underground caverns according to claim 6 is characterized in that: The encoding rules are as follows: grid index ( i, j, k ) is converted to a binary index representation ( i 0B , j 0B , k 0B ): in, ,1< l ≤ maxDepth , and finally from top to bottom according to the depth of the octree" z - y - x "The three axes are sequentially encoded on the grid, and the calculation formula is as follows: in, Codes the maximum mesh depth for a regular grid.

8. The dual-layer fast indexing method for massive point cloud data of underground caverns according to claim 7 is characterized in that: In step S402, the point cloud data encoding of each segmented point set is sorted from small to large according to the grid code, and the point cloud information contained in the grid of each depth, that is, the number of points in the grid and the index of the contained points, is counted layer by layer from top to bottom. The grid code conversion formula for a specified depth depth is as follows: in, It is based on maximum segmentation depth coding The converted regular grid code of the specified subdivision depth is converted into a regular grid code; the underlying regular grid is encoded in the same Merge layer by layer. When the number of points in the grid is less than the number threshold or reaches the maximum depth, the point data in the grid is output as a leaf node to complete the construction of the virtual octree of the segmented point set.

9. A dual-layer rapid indexing system for massive point cloud data of underground caverns, characterized by: include: Module 1: It is used to read point cloud data files and create an initial sampling model of underground cavern point clouds; Module 2: It is used to obtain the center point startC of the key feature of the cave based on the sampled cave point cloud data small model i ( sx i , sy i , sz i ) and endC i ( ex i , ey i , ez i ), fitting the unit vector of the cave axis a i ( x i , y i , z i ), set the axis segment distance d ; Module 3: It is used to traverse and read point cloud data. Based on the fitted cavern axis unit vector, the points are mapped to the cavern axis direction, and the top-level segmented index structure of the cavern point cloud data is constructed along the cavern direction. By temporarily transforming the massive point cloud data of the underground cavern and redistributing it along the cavern axis, a compact segmented point cloud bounding box is obtained. Specifically: For point P in the point cloud j ( p x j , p y j , p z j ), calculate the corresponding starting section center point startC i To point P j Vector v j , P j The top segment index is ; At the same time, in the process of traversing the point cloud, the original data is not changed, and the point set is temporarily transformed to obtain the point p ( , , ), so that the z-axis of the coordinate system is consistent with the direction of the tunnel axis. The coordinate transformation formula is as follows: in α is a vector a i With xoz The angle between the projection of the plane and the z-axis, β is a vector a i With xoz The angle of the plane, during the traversal process, determines the minimum and maximum vertices of each segment bounding box (tMinPt k , tMaxPt k ); Module 4: It is used to perform temporary coordinate transformation on each point cloud subset after the top-level segmentation. Based on the regular grid subdivision based on the axis vector information of module 2, the local virtual octree index structure is constructed from the bottom up through the spatial index merging algorithm of grid refinement subdivision and layer-by-layer grid merging. The original point cloud data is not modified and is recorded in the node file, completing the double-layer index construction of the underground cavern point cloud data.

Images