Method, system and device for constructing a digital twin of a buried space and storage medium

Abstract

The application discloses a buried space digital twin construction method, system, device and storage medium. The method comprises the following steps: collecting point cloud data sequences of a target buried space; mapping and matching each point cloud data in the target buried space with each grid unit in a prior grid model, and determining the point cloud mapping state of each grid unit in the prior grid model; determining each deformation space region of the target buried space that is deformed compared with an original three-dimensional space according to the space semantic information and the point cloud mapping state corresponding to each grid unit; matching and registering each deformation space region with an original space structure component of a prior three-dimensional model, and obtaining the deformation characteristics of each deformation space region compared with the matched original space structure component; and reconstructing each original space structure component matched in the prior three-dimensional model according to the geometric contour and the deformation characteristics of each deformation space region, so as to obtain a digital twin corresponding to the target buried space.

Description

Technical Field

[0001] This application relates to the field of computer technology, specifically to a method, system, device, and storage medium for constructing a digital twin of a buried space. Background Technology

[0002] Following disasters such as earthquakes and mine accidents, rapidly and accurately determining the real-time structural status of buried spaces is crucial for developing rescue plans and assessing risks. Digital twin technology, as a key means of connecting the physical world and digital space, has enormous application potential in such emergency scenarios. However, achieving efficient and accurate dynamic reconstruction of digital twins in the unique and complex environment of buried spaces presents numerous technical challenges.

[0003] Therefore, how to efficiently and accurately reconstruct the digital twin of buried space in a dynamic manner, so as to quickly respond to changes on site, accurately reflect the spatial structure, and automatically identify risk areas in emergency rescue of buried space, thereby improving rescue efficiency and safety, has become an urgent problem to be solved. Summary of the Invention

[0004] This application provides a method, system, device, and storage medium for constructing a digital twin of a buried space, in order to solve the problem of how to efficiently and accurately reconstruct the digital twin of a buried space dynamically.

[0005] In a first aspect, this application provides a method for constructing a digital twin of a buried space, the method comprising: Collect point cloud data sequences of the target burial space, wherein the point cloud data sequences include point cloud data of multiple spatial location points in the burial space; The point cloud data in the target burial space are mapped and matched with each grid cell in the prior grid model corresponding to the target burial space to determine the point cloud mapping state of each grid cell in the prior grid model; wherein, the prior grid model is obtained by rasterizing the original three-dimensional space based on the prior three-dimensional space corresponding to the target burial space. Based on the spatial semantic information and point cloud mapping state corresponding to each grid unit, the deformed spatial regions in which the target burial space has deformed compared to the original three-dimensional space are determined. The spatial semantic information corresponding to the grid unit is used to identify the type of spatial object corresponding to the grid unit. Each of the deformable spatial regions is matched and registered with the original spatial structure components of the prior 3D model, and the deformation characteristics of each of the deformable spatial regions compared with the matched original spatial structure components are obtained. Based on the geometric contours and deformation characteristics of each of the deformable spatial regions, the original spatial structure components matched in the prior three-dimensional model are reconstructed to obtain the digital twin corresponding to the target burial space.

[0006] In some embodiments, the prior grid model is constructed in advance in the following manner: The original three-dimensional space is rasterized based on the prior three-dimensional model of the original three-dimensional space to obtain the prior raster model corresponding to the original three-dimensional space. The prior grid model includes multiple grid cells, each of which includes the position coordinates of the grid center point and spatial semantic information.

[0007] In some embodiments, the step of mapping and matching each point cloud data in the target burial space with each grid cell in the prior grid model corresponding to the target burial space, and determining the point cloud mapping state of each grid cell in the prior grid model, includes: Align the point cloud data in the target buried space with the position coordinates according to the position coordinate system of the prior grid model; Identify the initial semantic information corresponding to each point cloud data in the target burial space, and spatially align the initial semantic information corresponding to each point cloud data with the spatial semantic information of each grid cell. Based on the position coordinates of each point cloud data after alignment and the spatial range conditions corresponding to each grid cell, the grid cells that each point cloud data is mapped to and matched are determined; wherein, if the position coordinates of the point cloud data satisfy the spatial range conditions corresponding to the grid cell, then the point cloud data is mapped and matched with the grid cell. Generate a point cloud mapping state for each grid cell, wherein the point cloud mapping state indicates whether there is at least one mapping match point cloud data in the grid cell space.

[0008] In some embodiments, determining the deformed spatial regions of the target burial space that have deformed compared to the original three-dimensional space based on the spatial semantic information and point cloud mapping state corresponding to each grid unit includes: Spatial semantic information is represented as cavity object type, and the point cloud mapping state represents the raster unit of the point cloud data that has mapping, which is divided into an intrusive deformation raster unit set. Spatial semantic information is represented as a structural object type, and the point cloud mapping state represents the raster units of point cloud data that do not have mapping, which are divided into a set of missing deformable raster units. Each invasive deformable grid cell in the set of invasive deformable grid cells is spatially clustered according to the adjacency topology of each invasive deformable grid cell to obtain at least one invasive deformable spatial region. Each missing deformation raster unit in the set of missing deformation raster units is spatially clustered according to the adjacency topology of each missing deformation raster unit to obtain at least one missing deformation spatial region.

[0009] In some embodiments, the matching and registration of each of the deformable spatial regions with the original spatial structural components of the prior three-dimensional model includes: For each of the deformation space regions, the position coordinates of the grid center point of each grid cell in the deformation space region are aligned according to the local engineering coordinate system of the prior 3D model; Based on the grid center point coordinates and spatial semantic information of each grid unit in the deformable spatial region, the corresponding matching original spatial structure component is searched in the prior 3D model.

[0010] In some embodiments, obtaining the deformation characteristics of each of the deformed spatial regions relative to the matched original spatial structural member includes: For each of the deformation space regions, the coordinates of the point cloud data in the deformation space region are transformed to the coordinate system of the original spatial structure component; Based on the point cloud data in the deformation space region and the surface position coordinates of the corresponding matching original spatial structure components, the deformation displacement parameters and deformation volume parameters of the deformation space region are calculated, and the deformation features include the deformation displacement parameters and deformation volume parameters.

[0011] In some embodiments, each of the deformation space regions includes an intrusive deformation space region and / or a missing deformation space region. The step of reconstructing each original spatial structure component matched in the prior 3D model based on the geometric contour and deformation characteristics of each of the deformation space regions to obtain a digital twin corresponding to the target burial space includes: For an intrusive deformation space region, based on the geometric contour of the intrusive deformation region, a local mesh Boolean subtraction operation is performed on the original spatial structure component matched in the prior 3D model to remove the structural part of the original spatial structure component corresponding to the intrusive deformation space region. Based on the deformation characteristics, spatial distribution of point cloud data, and surface characteristics within the invasive deformation space region, a solid structure surface for characterizing the invasive entity is generated in the excised location region. For a missing deformation space region, based on the geometric contour of the missing deformation space region, a local mesh deletion operation is performed on the original spatial structure component matched in the prior 3D model to remove the structural part of the original spatial structure component corresponding to the missing deformation space region; If the deleted region exposes a new cavity space, the boundary mesh of the new cavity space is generated based on the spatial semantic information and adjacency topology of its adjacent grid cells. A digital twin corresponding to the target burial space is obtained.

[0012] In some embodiments, the point cloud mapping state of each grid cell is a first mapping state or a second mapping state. The first mapping state indicates that the grid cell has at least one point cloud data with a mapping match, and the second mapping state indicates that the grid cell does not have any point cloud data with a mapping match. After determining the point cloud mapping state of each grid cell in the prior grid model, the method further includes: Perform point cloud mapping validity detection on the raster cells whose point cloud mapping state is the first mapping state, and identify the raster cells with invalid mapping and the raster cells with valid mapping. Update the point cloud mapping state of invalidally mapped raster cells to the second mapping state, while keeping the point cloud mapping state of validly mapped raster cells unchanged.

[0013] In some embodiments, the step of performing point cloud mapping validity detection on raster cells whose point cloud mapping state is a first mapping state, and identifying raster cells with invalid mapping and raster cells with valid mapping, includes: Iterate through each grid cell whose corresponding point cloud mapping state is the first mapping state, and determine whether the amount of point cloud data matched by the currently traversed grid cell mapping has reached the validity detection threshold. If the amount of point cloud data mapped and matched by the currently traversed raster cell does not reach the validity detection threshold, the currently traversed raster cell will be determined as an invalid mapped raster cell. If the amount of point cloud data mapped and matched by the currently traversed raster cell reaches the validity detection threshold, check the point cloud mapping status of other raster cells adjacent to the currently traversed raster cell. If at least one other grid cell in the adjacent grid cells has a point cloud mapping state of first mapping state, the following steps are performed: inputting the point cloud data volume of the currently traversed grid cell mapping matching, whether there are adjacent other grid cells with a cloud mapping state of first mapping state, and whether it is a spatial boundary grid cell into the preset data classification model to perform abnormal data classification prediction. If the point cloud mapping state of all adjacent raster cells is the second mapping state, and the currently traversed raster cell is not a spatial boundary raster cell, the currently traversed raster cell is determined as an invalid mapping raster cell. If the point cloud mapping state of the adjacent other grid cells is the second mapping state, and the currently traversed grid cell is a spatial boundary grid cell, the following steps are performed: inputting the point cloud data volume of the currently traversed grid cell mapping, whether there are adjacent other grid cells with the point cloud mapping state of the first mapping state, and whether it is a spatial boundary grid cell into the preset data classification model to perform abnormal data classification prediction. If the data classification model outputs abnormal data, the currently traversed raster cell will be determined as an invalid mapped raster cell; If the data classification model outputs valid data, the currently traversed raster cell is determined as a validly mapped raster cell.

[0014] Secondly, this application provides a system for constructing a digital twin of a buried space, the system comprising: The point cloud acquisition module is used to acquire point cloud data sequences of the target burial space, wherein the point cloud data sequences include point cloud data of multiple spatial location points in the burial space; The point cloud mapping module is used to map and match each point cloud data in the target burial space with each grid cell in the prior grid model corresponding to the target burial space, and determine the point cloud mapping state of each grid cell in the prior grid model; wherein, the prior grid model is obtained by rasterizing the original three-dimensional space based on the prior three-dimensional space corresponding to the target burial space. The deformation recognition module is used to determine the deformation spatial regions in the target buried space that have deformed compared to the original three-dimensional space based on the spatial semantic information and point cloud mapping state corresponding to each grid unit. The spatial semantic information corresponding to the grid unit is used to identify the type of spatial object corresponding to the grid unit. The deformation feature extraction module is used to match and register each of the deformation spatial regions with the original spatial structure components of the prior 3D model, and to obtain the deformation features of each of the deformation spatial regions relative to the matched original spatial structure components. The reconstruction module is used to reconstruct each original spatial structure component matched in the prior three-dimensional model according to the geometric contour and deformation characteristics of each of the deformable spatial regions, so as to obtain a digital twin corresponding to the target burial space.

[0015] Thirdly, this application provides an electronic device, including: a memory and a processor, which are communicatively connected to each other. The memory stores computer instructions, and the processor executes the computer instructions to perform the method for constructing a digital twin of a buried space as described in the first aspect or any corresponding embodiment.

[0016] Fourthly, this application provides a computer-readable storage medium storing computer instructions for causing a computer to execute the method for constructing a digital twin of a buried space as described in the first aspect or any corresponding embodiment.

[0017] According to the method for constructing a digital twin of a buried space in this application, based on the mapping relationship between discrete point cloud data in the buried space and grid cells in the prior grid model, the deformed areas in the buried space are automatically and quickly identified. Based on the deformation features, the prior 3D model of the buried space is updated and corrected rapidly and incrementally, thereby realizing rapid and accurate reconstruction of the digital twin of the buried space and improving the efficiency and automation of the reconstruction of the digital twin of the buried space. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the specific embodiments of this application or the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0019] Figure 1 A flowchart illustrating a method for constructing a digital twin of a buried space, provided in an embodiment of this application; Figure 2 This is a flowchart illustrating a method for validating point cloud mappings in an embodiment of this application. Figure 3 A structural block diagram of a digital twin construction system for buried space provided in an embodiment of this application; Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0021] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0022] In related technologies, the three-dimensional reconstruction of buried space usually starts from scratch by processing spatial point cloud data, performing feature extraction, matching and model building. This is computationally intensive and time-consuming. In disaster emergency scenarios, this time-consuming method is difficult to support the minute-level or even second-level update requirements of rescue command for the on-site situation.

[0023] Therefore, how to efficiently and accurately reconstruct the digital twin of buried space in a dynamic manner, so as to quickly respond to changes on site, accurately reflect the spatial structure, and automatically identify risk areas in emergency rescue of buried space, thereby improving rescue efficiency and safety, has become an urgent problem to be solved.

[0024] To effectively address the technical problems existing in the aforementioned related technologies, this application provides a method for constructing a digital twin of a buried space.

[0025] According to an embodiment of this application, a method for constructing a digital twin of a buried space is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0026] This application provides a method for constructing a digital twin of a buried space, which can be applied to terminal devices or servers. Figure 1 This application provides a flowchart illustrating a method for constructing a digital twin of a buried space, as shown in the embodiments below. Figure 1 As shown, the method for constructing a digital twin of a buried space includes the following steps: Step S101: Collect point cloud data sequence of the target burial space. The point cloud data sequence includes point cloud data of multiple spatial location points in the burial space.

[0027] Buried space refers to collapsed, buried, or enclosed spaces formed after disasters such as earthquakes, mine accidents, and building collapses. These include, but are not limited to, physical spaces such as mine roadways, underground chambers, building rooms, and tunnels where there is structural deformation, collapse intrusion, or missing components.

[0028] At rescue sites of buried spaces caused by disasters such as earthquakes, mine accidents, and building collapses, photoelectric scanning equipment (such as LiDAR and 3D laser scanners) is used to perform real-time 3D scanning of the buried space to obtain a continuous 3D point cloud data stream containing time information and spatial coordinate information. This data is then converted into a standardized data format that can be parsed, transmitted, and processed by the system, such as a point cloud data sequence containing timestamps and spatial coordinates. The point cloud data sequence is an ordered data set composed of a large number of 3D spatial points arranged according to the time of acquisition.

[0029] The point cloud data sequence includes point cloud data of multiple spatial locations in the burial space. The spatial locations are surface measurement points actually sampled by photoelectric scanning equipment in the burial space. Each point carries unique three-dimensional spatial coordinates and is the smallest unit that constitutes the point cloud. The point cloud data is the structured data corresponding to a single spatial location point, which includes at least: timestamp, three-dimensional spatial coordinates, and may also include reflection intensity and RGB color values.

[0030] For example, in a rescue scenario involving a mine flooding accident, rescuers use a 3D laser scanner to scan the accident tunnel. The scanner collects up to 2 million points per second, generating high-density point cloud data. This raw point cloud data is read in real time through the device's SDK interface and then encapsulated and transmitted using a serialization format via an edge computing server set up at the rescue site. This forms a point cloud data stream with timestamps (such as 20240126T103000.000Z) and spatial coordinates, which is then transmitted to the command center.

[0031] Step S102: Map and match each point cloud data in the target burial space with each grid cell in the prior grid model corresponding to the target burial space to determine the point cloud mapping state of each grid cell in the prior grid model; wherein, the prior grid model is obtained by rasterizing the original three-dimensional space based on the prior three-dimensional model corresponding to the target burial space.

[0032] The real-time point cloud data of the target buried space, after being collected and serialized, is mapped and matched with the grid cells in the pre-constructed prior grid model through coordinate alignment, spatial semantic alignment, and spatial attribution determination. This process maps and matches the disordered and massive point cloud data to structured grid cells in batches, and marks each grid cell with a binary state indicating whether it has been hit by the point cloud mapping. Finally, the point cloud mapping state of each grid cell is generated, providing standardized input for subsequent deformation recognition.

[0033] The prior grid model is a structured discrete model obtained by regularly dividing the original three-dimensional space into regular three-dimensional grids based on the prior three-dimensional model of the target buried space before the disaster (such as BIM model, engineering design drawings, high-precision historical survey data). Among them, grid subdivision refers to uniformly dividing the continuous three-dimensional space into cubes with preset fixed side lengths, discretizing the continuous space into a set of regular and computable grid units, forming the basic voxels of digital space. The grid unit is a standard cubic unit formed after gridding, which has a unique identifier ID, center point coordinates, side length, spatial semantics, and adjacency topology relationship, and is the smallest unit of mapping and calculation.

[0034] The original three-dimensional space is the complete, uncollapsed, and undeformed continuous physical space where the target buried space was located before the disaster occurred; the prior three-dimensional model is the precise three-dimensional prior knowledge of the target buried space obtained before the disaster, including Building Information Modeling (BIM), engineering CAD drawings, high-precision surveying models, etc., which is the original basis for rasterization processing.

[0035] Step S103: Based on the spatial semantic information and point cloud mapping status corresponding to each grid cell, determine the deformed spatial regions in the target buried space that have deformed compared to the original three-dimensional space. The spatial semantic information corresponding to the grid cell is used to identify the type of spatial object corresponding to the grid cell.

[0036] Based on the point cloud mapping state of each grid cell obtained in step S102, and combined with the predefined spatial semantic information of each grid cell, candidate grid cells with invasive deformation and candidate grid cells with missing deformation are automatically selected according to the preset dual-channel discrimination criteria. Then, based on the three-dimensional adjacency topology of the grid cells, a spatial clustering algorithm is used to merge the discrete candidate grid cells into continuous deformation spatial regions, and the boundary features of each deformation spatial region are accurately extracted to complete the automatic identification and positioning of burial spatial deformation.

[0037] Among them, spatial semantic information, also known as spatial semantic code (spCode), is used to uniquely identify the original spatial object type corresponding to the raster unit. It is the core basis for distinguishing between the cavity type and the solid structure type of spatial objects in the original three-dimensional space, and is pre-assigned by the prior raster model.

[0038] Spatial object types are predefined classifications of spatial objects in the prior raster model, mainly divided into the following types: Cavity type: CORRIDOR (tunnel / passage), ROOM (room / chamber), CHAMBER (equipment chamber), etc. The original state of this type of spatial object is cavity; Solid structure classes: CONCRETE_WALL (concrete load-bearing wall), PILLAR (support column / mine pillar), SUPPORT (support), FLOOR (floor slab), ROOF (roof slab), etc. The original state of this type of spatial object is a solid structure; Unknown class: UNKNOWN, will be treated as a cavity by default.

[0039] Deformation space regions are areas where the structure of the buried target space has changed compared to the original three-dimensional space before the disaster, and are divided into two categories: Intrusive deformation space region: The area where the original cavity is invaded and occupied by collapsed bodies and deposits; Missing deformation space region: The region where the original solid structure has collapsed, broken, or disappeared.

[0040] The three-dimensional adjacency topology of a grid cell refers to the index relationship between adjacent grid cells in 26 directions (up, down, left, right, front, back) in three-dimensional space. It is the basis for spatial clustering and boundary extraction.

[0041] Step S104: Match and register each deformable spatial region with the original spatial structure component of the prior 3D model, and obtain the deformation characteristics of each deformable spatial region relative to the matched original spatial structure component.

[0042] Based on the unified coordinate system transformation, the various deformable spatial regions identified in step S103 are precisely matched and registered with the corresponding original spatial structural components in the prior 3D model through coordinate transformation and geometric alignment, establishing a one-to-one correspondence between the deformable spatial regions and the original spatial structural components. On the basis of registration, deformation quantification analysis is performed on the deformable spatial regions to calculate and extract key deformation features such as displacement, volume, range, and boundary of the region relative to the original spatial structural components, providing accurate geometric positioning and quantitative data support for subsequent digital twin fusion and reconstruction.

[0043] Among them, the original spatial structure components are standardized component units in the prior three-dimensional model that correspond to the actual physical structure, such as alleyway components, wall components, pillar components, roof components, room components, etc., which are the reference objects for registration and comparison.

[0044] Matching and registration is the process of geometrically aligning the three-dimensional coordinates, spatial position, and geometric contour of the deformed spatial region with the corresponding original spatial structural components of the prior three-dimensional model, so that the two are in the same coordinate system and under the same spatial reference.

[0045] Deformation features are a set of parameters used to quantitatively describe the structural changes that occur in a deformed spatial region relative to the original spatial structural components. These features include geometric features, volumetric features, displacement features, and boundary features.

[0046] Step S105: Based on the geometric contours and deformation characteristics of each deformation space region, reconstruct each original spatial structure component matched in the prior 3D model to obtain the digital twin corresponding to the target burial space.

[0047] Based on the geometric contour and quantified deformation characteristics of the deformation space region obtained in step S104, model baking technology is used to perform local mesh calculations and incremental updates on the original spatial structure components that match the prior 3D model. Differential reconstruction processing is performed on the intrusive deformation space region and the missing deformation space region respectively, while retaining the structurally intact components. Finally, a digital twin that can truly reflect the real-time structural state of the buried space is generated.

[0048] Among them, the geometric contour of the deformation space region is the three-dimensional spatial closed range enclosed by the boundary grid of the deformation space region. It is used to define the local area of ​​model reconstruction and serves as the boundary basis for mesh operations.

[0049] Model baking is an incremental update technique that deeply integrates deformation features with the original 3D model, performs local modifications, and reconstructs the mesh, achieving local modifications rather than global reconstruction.

[0050] Local mesh operations are precise mesh modification operations performed on deformable spatial regions, including local mesh Boolean subtraction operations, local mesh deletion operations, and generation of new surfaces for solid structures.

[0051] A digital twin is a digital 3D model that is geometrically, semantically, and state-consistent with the physical space of the burial site in real time. It can be directly used for rescue situation awareness, risk assessment, and accessibility analysis.

[0052] According to the embodiment of this application, the method for constructing a digital twin of a buried space automatically and quickly identifies deformed areas in the buried space based on the mapping relationship between discrete point cloud data in the buried space and grid cells in the prior grid model. Based on the deformation features, the method rapidly and incrementally updates and corrects the pre-constructed prior three-dimensional model of the buried space, thereby achieving rapid and accurate reconstruction of the digital twin of the buried space and improving the efficiency and automation of the reconstruction of the digital twin of the buried space.

[0053] In some embodiments, the prior grid model is constructed in advance by rasterizing the original three-dimensional space based on the prior three-dimensional model of the original three-dimensional space to obtain the prior grid model corresponding to the original three-dimensional space; wherein, the prior grid model includes multiple grid cells, and each grid cell includes the position coordinates of the grid center point and spatial semantic information.

[0054] Specifically, before constructing the digital twin of the buried space, the original three-dimensional space is rasterized based on the prior three-dimensional model of the original three-dimensional space to establish the raster unit structure of the original three-dimensional space and obtain the prior raster model.

[0055] Starting with a precise pre-disaster 3D model (such as Building Information Modeling (BIM), engineering design drawings, or high-precision historical survey data), the complex 3D structure in the original 3D space of the target buried space is first projected to establish a unified coordinate reference and calculation plane. Based on this, a pre-defined fixed-scale (e.g., default value of 1 meter) regular grid is used to systematically and uniformly subdivide the continuous physical space in the original 3D space—a process known as rasterization. This discretizes the continuous physical space into a series of standard cubic unit grids with regular geometric shapes, forming the basic voxels of the digital space, called raster units.

[0056] The system allows for dynamic and flexible adjustment of the spatial scale of grid cells based on the specific accuracy requirements of the simulation and the actual availability of computing resources. When the simulation task focuses on macroscopic situational analysis or requires rapid extrapolation, a larger grid scale can be selected. By reducing the model's resolution, the total number of grid cells can be significantly reduced, thereby greatly improving computational efficiency and meeting the needs of real-time or near-real-time simulations. Conversely, when the simulation task requires detailed analysis of key local areas (such as accurate throughput calculations or signal masking effect simulations), a smaller grid scale can be used. This increases the model's granularity to obtain more detailed local environmental information, ensuring the fidelity of the simulation results.

[0057] The basic data structure of a raster cell includes, but is not limited to: spatial semantic code (spCode), geometric attributes, and physical attributes.

[0058] In some embodiments, the spatial semantic code (spCode) is used to identify the type of spatial object represented by the raster unit. The system predefines the following basic spatial object types and supports extensions based on actual scenarios: Cavity-type: CORRIDOR, indicating a tunnel / passage; ROOM, indicating a chamber / room; CHAMBER, indicating an equipment chamber or special function space; Solid structure classes: CONCRETE_WALL, representing a concrete load-bearing wall; PILLAR, representing a pillar / mine pillar; SUPPORT, representing a support structure; FLOOR, representing a floor slab / ground; ROOF, representing a roof slab / vault. Unknown class: UNKNOWN, indicates an unknown type of space object, which is treated as a cavity by default.

[0059] In some embodiments, each grid cell records, but is not limited to, the following basic geometric attributes: grid identifier, size, center point location coordinates, adjacency topology, etc., for spatial positioning and topology construction.

[0060] Among them, the grid identifier ID is a unique identifier for the grid cell; the size is the side length of the grid cell, which is a cubic grid and the unit is meters; the center point position coordinates cpos(w, m, r) are the spatial position coordinates of the center point of the grid cell, which are defined using the world coordinate system (such as WGS84) that is consistent with the prior 3D model; the adjacency topology relations neighbors are the set of adjacent grid indexes of the grid cell, which records its adjacent grid cells in 26 directions (up, down, left, right, front, back) for spatial topology analysis.

[0061] In some embodiments, each grid cell records, but is not limited to, the following physical properties: water content coefficient, medium coefficient, and surface reflectivity coefficient.

[0062] Among them, the water content coefficient (waterCoeff) is set according to the moisture conditions in the actual scenario. The system constructs the corresponding elements based on this coefficient. In addition to providing visual condition settings, the water content also provides a basis parameter for electromagnetic signal attenuation calculation.

[0063] Medium Coefficient: The medium coefficient is set according to the burial conditions of the material in the real scenario. The medium coefficient mainly reflects the characteristics of the main components of the medium. In addition to providing visual condition settings, the medium coefficient also provides the basis parameters for the simulation calculation of electromagnetic signal absorption and reflection.

[0064] Surface reflectivity coefficient (reflectivityCoeff): This parameter sets the surface reflectivity coefficient based on real-world surface conditions. The system generates ground obstacles and energy attenuation based on this coefficient. Besides providing visualization settings, surface reflectivity also provides a calculation basis for simulating the communication quality of ground targets.

[0065] The aforementioned physical properties can be pre-assigned using prior knowledge, or dynamically updated during real-time scanning by combining multi-dimensional features of the point cloud (such as reflection intensity and color).

[0066] In some embodiments, after establishing grid cells in the original three-dimensional space, the basic data structure of the grid cells can be represented as follows:

[0067] Wherein, ID represents the raster identifier of the raster cell; size represents the side length of the raster cell; cpos(w,m,r) represents the coordinates of the center point of the raster cell, which is usually expressed in the world coordinate system WG84; neighbors represents the adjacency topology of the raster cell, that is, the set of adjacent raster indices of the adjacent raster cells of the raster cell; spCode represents the spatial semantic information corresponding to the raster cell, also known as the spatial semantic code.

[0068] In some embodiments, after establishing grid cells in the original three-dimensional space, the adjacency topology of each grid cell in the original three-dimensional space is established, and its adjacent grid cells in 26 directions (up, down, left, right, front, and back) are determined. The adjacency topology of the grid cells can be represented as follows:

[0069] in, This represents the adjacent raster indices of the top 9 raster cells of the current raster cell. This represents the adjacent raster indexes of the current raster cell and its eight adjacent raster cells in the same horizontal space. This indicates the adjacent raster index of the 9 raster cells that are next to the bottom of the current raster cell.

[0070] In some embodiments, the boundary representation of a spatial volume can be established based on the adjacency topology of each grid cell in the prior grid model. That is, a grid cell with effective adjacent grid cells in all directions can be regarded as an empty volume grid cell, i.e., a grid cell inside the space. A grid cell with some adjacent grid cells missing in at least one direction can be regarded as a spatial boundary grid cell. The boundary relationship can be extracted based on the adjacent grid index of the missing grid cell.

[0071] In some embodiments, in step S102 above, mapping and matching each point cloud data in the target burial space with each grid cell in the prior grid model corresponding to the target burial space to determine the point cloud mapping state of each grid cell in the prior grid model includes: aligning the position coordinates of each point cloud data in the target burial space according to the position coordinate system of the prior grid model; identifying the initial semantic information corresponding to each point cloud data in the target burial space, and spatially aligning the initial semantic information corresponding to each point cloud data with the spatial semantic information of each grid cell; determining the grid cell to which each point cloud data is mapped and matched according to the position coordinates of each point cloud data after alignment and the spatial range conditions corresponding to each grid cell; wherein, if the position coordinates of the point cloud data satisfy the spatial range conditions corresponding to the grid cell, then the point cloud data is mapped and matched with the grid cell; generating the point cloud mapping state of each grid cell, the point cloud mapping state indicating whether there is at least one mapped and matched point cloud data in the grid cell space.

[0072] To ensure that point cloud data can be accurately mapped and matched with the grid cells of the prior grid model in a unified spatial reference coordinate system, two key technical steps, coordinate alignment and coordinate correction, need to be completed sequentially.

[0073] Specifically, the world coordinate system (such as WGS84) used by the prior raster model is read, and the original position coordinate information in the real-time point cloud data is parsed. Through feature point matching or coarse registration methods, the point cloud data is initially transformed to the position coordinate system of the prior raster model, so as to achieve the initial coordinate alignment of the two on the same spatial reference. On the basis of the initial coordinate alignment, a high-precision registration algorithm (such as the Iterative Closest Point (ICP) algorithm) is further used to perform a fine rigid transformation on the position coordinates of the point cloud data, and the translation and rotation parameters are optimized so that the geometric structure of the point cloud data and the prior raster model achieves the highest degree of geometric fit.

[0074] For example, the WGS84 coordinate system used by the prior grid model is read. During the coordinate alignment stage, three feature points (such as two permanent measurement control points on the tunnel wall and the center point of a ventilation duct interface) that exist in both the point cloud data and the prior grid model are automatically identified. Based on these feature points, the least squares method is used to perform preliminary coordinate transformation. During the fine correction stage, the Iterative Closest Point (ICP) algorithm is used to iteratively register the point cloud data with the grid center point on the tunnel surface in the prior grid model as the target. The maximum number of iterations is set to 50. The iteration stops when the registration error is less than 0.005 meters, and finally, a high-precision spatial geometric fit between the point cloud and the prior grid model is achieved.

[0075] After completing position coordinate alignment and rigidity correction, the point cloud data is spatially semantically aligned with the grid cells of the prior grid model.

[0076] Specifically, the preliminary semantic information contained in real-time point cloud data is analyzed, such as different materials distinguished by the reflection intensity and color of the point cloud, and mapped and associated with the rich spatial semantic codes (such as "lane", "load-bearing wall", "equipment area" etc.) predefined by each grid unit in the prior grid model.

[0077] For example, the system pre-defines mapping rules between the reflection intensity and color of point cloud data and initial semantic information. For instance, a point cloud with a reflection intensity in the range of [8000, 10000] and a gray color is defined as having initial semantic information of a concrete structure, while a point cloud with lower reflection intensity and a darker color is defined as having initial semantic information of accumulated coal slag. By parsing the RGB color values ​​and reflection intensity in the point cloud data, and according to the system's pre-defined rules, the initial semantic information of each point cloud data is identified and automatically associated with and matched against the predefined spatial semantic code (spCode, such as CONCRETE_WALL or COAL_PILE) of the corresponding raster cell in the prior raster model, thus completing spatial semantic alignment.

[0078] It is important to note that when the raster resolution is set too high, causing a single raster cell to cover multiple spatial objects with different semantics, the dominant semantic principle or the highest priority semantic inheritance mechanism can be used to determine the unique spatial semantic code (spCode) of that raster cell. Specifically, the area dominance method calculates which semantic type (e.g., "lane," "room," "equipment area") occupies the largest volume or projected area within the raster area, and assigns that semantic to the entire raster. The function priority method, on the other hand, prioritizes semantic types based on their function (e.g., "load-bearing wall" has higher priority than "ordinary partition wall"), inheriting the spatial semantic with the highest priority.

[0079] For each point cloud data P(x,y,z) that has undergone coordinate alignment and spatial semantic alignment, its corresponding raster cell is determined based on its position coordinates in a unified world coordinate system. The determination logic is based on the spatial extent conditions corresponding to the raster cell.

[0080] If the position coordinates (x, y, z) of a point cloud data P(x, y, z) satisfy the spatial extent condition of a certain raster cell Grid(size, cpos(w, m, r)), then the point cloud is determined to belong to that raster cell. The spatial extent condition of the raster cell can be expressed as:

[0081] Where Grid represents the grid identifier ID of the grid cell, size represents the side length scale of the grid cell, cpos(w,m,r) represents the center point coordinates of the grid cell in the world coordinate system, and (x,y,z) represents the position coordinates of the point cloud data P after transformation in the world coordinate system.

[0082] For example, assuming the side length (size) of each grid cell in the prior grid model is set to 1 meter, and the position coordinates of a point cloud P are (105.5, 38.2, -5.1), the system traverses each grid cell and finds that the position coordinates of point cloud P satisfy the spatial range condition of grid cell Grid_001 (the center point position coordinates cpos are (105.5, 38.2, -5.0)), that is: 105.5 - 0.5 ≤ 105.5 ≤ 105.5 + 0.5 38.2 - 0.5 ≤ 38.2 ≤ 38.2 + 0.5 -5.0 - 0.5 ≤ -5.1 ≤ -5.0 + 0.5 Therefore, point cloud P is mapped and matched to grid cell Grid_001, the mapping matching relationship between the point cloud and the grid cell is recorded, and the mapped point cloud count is updated within the data structure of the grid cell. For anomalous point clouds whose position coordinates exceed the entire prior grid model range, they can be logged and ignored, and not involved in subsequent processing.

[0083] By traversing all real-time point cloud data, the affiliation determination of the above-mentioned grid cells is completed, thereby marking the point cloud mapping state of each grid cell in the entire prior grid model.

[0084] The point cloud mapping state of the raster cell is a Boolean label, indicating whether there is at least one real-time scanned point cloud data within the spatial range represented by the raster cell, reflecting the spatial distribution of the real-time scanned point cloud data in the prior raster model.

[0085] The point cloud mapping state of each grid cell is either a first mapping state or a second mapping state. The first mapping state indicates that there is at least one mapping match point cloud data in the grid cell, and the second mapping state indicates that there is no mapping match point cloud data in the grid cell. The first mapping state can be represented by "True", and the second mapping state can be represented by "False".

[0086] When there is at least one point cloud data with a mapping match within the spatial range of a raster cell, its point cloud mapping status is marked as True (first mapping status); when there is no real-time point cloud data with a mapping match within the spatial range of a raster cell, its point cloud mapping status is marked as False (second mapping status).

[0087] After completing the point cloud mapping status marking of all grid cells, a grid point cloud mapping status table is generated, as shown in Table 1. Table 1 records the unique identifier ID, center point coordinates, spatial semantic code, and point cloud mapping status of each grid cell.

[0088] Table 1

[0089] As shown in Table 1, the original spatial semantic code of grid cell Grid_001 is CORRIDOR (tunnel cavity), and the point cloud mapping state is False, indicating that the cavity location is not covered by point cloud and has not undergone deformation; the original spatial semantic code of grid cell Grid_002 is CORRIDOR (tunnel cavity), and the point cloud mapping state is True, indicating that the cavity location is currently covered by point cloud and may have undergone deformation; the original spatial semantic code of grid cell Grid_003 is CONCRETE_WALL (load-bearing wall), and the point cloud mapping state is True, indicating that the wall location still has point cloud coverage, belongs to the expected static structure, and has not undergone deformation; the original spatial semantic code of grid cell Grid_004 is CONCRETE_WALL, and the point cloud mapping state is False, indicating that the original wall location is currently not covered by point cloud, and the wall may be missing or collapsed.

[0090] For example, at the rescue site of a mine flooding accident, the system completes the traversal mapping of approximately 2 million point cloud data obtained from a single scan, mapping each point cloud data to the corresponding raster cell, and generating the raster point cloud mapping status table as shown in Table 1.

[0091] For example, statistics show that there are 156 point cloud data mapping matches within raster cell Grid_002 (center point coordinates (105.5, 38.2, -4.0), spatial semantic code CORRIDOR), so its point cloud mapping status is marked as True; there are no point cloud data mappings within raster cell Grid_001 (center point coordinates (105.5, 38.2, -5.0), spatial semantic code CORRIDOR), so its point cloud mapping status is marked as False; There are 89 point cloud data mapping matches in grid cell Grid_003 (center point coordinates (105.5, 38.2, -3.0), spatial semantic code CONCRETE_WALL), so its point cloud mapping status is marked as True; there are no point cloud data mapping matches in raster cell Grid_004 (center point coordinates (105.5, 38.2, -2.0), spatial semantic code CONCRETE_WALL), so its point cloud mapping status is marked as False.

[0092] After completing the mapping and matching of all point cloud data, it was found that a total of 1250 raster units in this scan had a point cloud mapping state of True. These raster units exhibit different distribution characteristics in three-dimensional space: among them, the spatial semantic code of the raster units with a point cloud mapping state of True is CORRIDOR (tunnel area), which is mainly concentrated in the middle section of the tunnel; the spatial semantic code of the raster units with a point cloud mapping state of True is CONCRETE_WALL (load-bearing wall area), which is distributed on the walls on both sides of the tunnel.

[0093] In some embodiments, in step S103 above, determining the deformed spatial regions of the target burial space that have deformed compared to the original three-dimensional space based on the spatial semantic information and point cloud mapping state corresponding to each grid unit includes: classifying grid units whose spatial semantic information is represented as cavity object type and whose point cloud mapping state indicates the existence of mapped point cloud data into an intrusive deformable grid unit set; classifying grid units whose spatial semantic information is represented as structural object type and whose point cloud mapping state indicates the absence of mapped point cloud data into a missing deformable grid unit set; spatially clustering each intrusive deformable grid unit in the intrusive deformable grid unit set according to the adjacency topology relationship of each intrusive deformable grid unit to obtain at least one intrusive deformable spatial region; and spatially clustering each missing deformable grid unit in the missing deformable grid unit set according to the adjacency topology relationship of each missing deformable grid unit to obtain at least one missing deformable spatial region.

[0094] Specifically, based on the generated raster point cloud mapping state table, all raster units are traversed for semantic filtering, and the raster units are classified into cavity object type, structure object type, and other types according to semantics. Among them, cavity object type refers to raster units whose spatial semantic code (spCode) belongs to CORRIDOR, ROOM, CHAMBER, GALLERY, etc., representing expected cavity spaces; structure object type refers to raster units whose spatial semantic code (spCode) belongs to CONCRETE_WALL, PILLAR, SUPPORT, FLOOR, ROOF, etc., representing expected solid structures; other types of raster units refer to raster units whose spatial semantic code (spCode) is not explicitly classified, and are treated as cavity type by default, or inferred from the semantics of adjacent raster units.

[0095] For a raster cell whose spatial object type is cavity object, if its point cloud mapping state is True (first mapping state), the raster cell is determined to be an intrusive deformation raster cell and is assigned to the intrusive deformation raster cell set; if its point cloud mapping state is False (second mapping state), the raster cell is cavity preserved.

[0096] For a raster cell whose spatial object type is a structure object, if its point cloud mapping state is True (first mapping state), the raster cell is determined to be a static structure with a complete structure; if its point cloud mapping state is False (second mapping state), the raster cell is determined to be a missing deformation raster cell and is assigned to the missing deformation raster cell set.

[0097] After semantic filtering, the following two types of raster unit sets are extracted: 1) Intrusive deformation raster cell set: raster cells whose semantics are cavity class and whose point cloud mapping state is the first mapping state True; 2) Set of missing deformable raster cells: raster cells whose semantics are structural and whose point cloud mapping state is the second mapping state False.

[0098] For example, in a mine roadway collapse accident, after the system generates a raster point cloud mapping status table, the table displays: raster cell Grid_002 (spatial semantic code is CORRIDOR, point cloud mapping status is True), raster cell Grid_003 (spatial semantic code is CONCRETE_WALL, point cloud mapping status is True), and raster cell Grid_004 (spatial semantic code is CONCRETE_WALL, point cloud mapping status is False).

[0099] Semantic filtering was performed on each grid cell based on the grid point cloud mapping status table. According to the filtering rules, grid cell Grid_002 belongs to the cavity class and its point cloud mapping status is True, so it is classified into the intrusive deformation grid cell set. Grid cell Grid_003 belongs to the structure class and its point cloud mapping status is True, so it is marked as a static structure with a sound structure. Grid cell Grid_004 belongs to the structure class and its point cloud mapping status is False, so it is classified into the missing deformation grid cell set. In the end, a total of 850 intrusive deformation grid cells were identified (mainly distributed in the middle section of the tunnel), 120 missing deformation grid cells were identified (mainly distributed on the walls on both sides of the tunnel), and 430 grid cells were marked as structurally sound.

[0100] After extracting the sets of invasive deformable raster cells and the sets of missing deformable raster cells, independent cluster analysis was performed on the two types of deformable raster cell sets (invasive deformable raster cell sets and missing deformable raster cell sets).

[0101] In some embodiments, a region growing algorithm is used to spatially cluster the grid cells in each set, and based on the pre-built adjacency topology of the grid cells, spatially adjacent grid cells are merged into a continuously deformable spatial region.

[0102] Specifically, for each set of deformable raster cells, an unvisited raster cell is selected from the set as a seed point; the 26 adjacent raster cells of the raster cell are traversed, and adjacent raster cells belonging to the same set are merged into the current deformable space region until the current deformable space region can no longer be expanded; the current deformable space region is marked as a complete deformable space region, and the next unvisited seed point is processed. The above process is repeated to determine the next complete deformable space region.

[0103] After clustering, a boundary analysis algorithm is used to identify the boundary raster between each deformed spatial region and the normal region, accurately extracting the deformation boundary features. For invasive deformed spatial regions, the spatial boundary raster cell is defined as: a raster cell belonging to the region and having at least one adjacent raster cell that does not belong to the set of invasive deformed raster cells; for missing deformed spatial regions, the spatial boundary raster cell is defined as: a raster cell belonging to the region and having at least one adjacent raster cell that does not belong to the set of missing deformed raster cells.

[0104] Each deformable spatial region inherits the spatial semantic information of the corresponding raster unit, forming a set of deformable spatial regions with semantic labels. For each deformable spatial region, the following information is recorded: region type (intrusive / missing), a list of raster IDs of the included raster units, a list of raster IDs of the spatial boundary raster units, region volume (number of included raster units × volume of a single raster unit), and the associated original spatial semantic code.

[0105] For example, in a mine roadway collapse accident, 850 extracted intrusive deformation grid cells were subjected to region growing clustering. First, using grid cell Grid_1001 as the seed point, the region growing algorithm was initiated, traversing its 26 adjacent grid cells. It was found that adjacent grid cells such as Grid_1005 and Grid_1006 also belonged to the intrusive deformation grid cell set, and Grid cells Grid_1001, Grid_1005, and Grid_1006 were merged into the same deformation space region. This algorithm continued to iterate, eventually forming an intrusive deformation space region containing these 800 grid cells, which was labeled as Deformation_Zone_In_1. After determining the deformation space region, the spatial boundary raster cells of the region were analyzed. It was found that the semantics of the adjacent raster cells on the east side of the region were cavity type (such as tunnel) and the point cloud mapping state was False. Therefore, these raster cells were identified as the spatial boundary raster cells of the deformation space region. Finally, the semantic label of the deformation space region Deformation_Zone_In_1 was assigned as "tunnel intrusion collapse zone".

[0106] In some embodiments, in step S104 above, matching and registering each deformable space region with the original spatial structure component of the prior three-dimensional model includes: for each deformable space region, aligning the position coordinates of the grid center point of each grid cell in the deformable space region according to the local engineering coordinate system of the prior three-dimensional model; and searching for the corresponding matching original spatial structure component in the prior three-dimensional model based on the position coordinates of the grid center point of each grid cell in the deformable space region and the spatial semantic information.

[0107] Specifically, the system reads the world coordinate system (such as WGS84) used by the prior grid model and the local engineering coordinate system used by the prior 3D model (BIM / CAD model); using preset coordinate transformation parameters, or by calculating coordinate transformation parameters through common feature point matching, the coordinates of the center point of each grid cell in the current deformation space region are accurately transformed to the local engineering coordinate system of the prior 3D model, achieving millimeter-level geometric alignment. Based on the spatial semantic codes inherited by the deformation space region (such as CORRIDOR, CONCRETE_WALL, ROOM, PILLAR, etc.), the system automatically retrieves and locates the original spatial structure components of the same type and position in the prior 3D model; the geometric range and spatial boundary grid cells of the deformation space region are geometrically matched with the geometric contours of the original spatial structure components to ensure that the deformation space region and the original spatial structure components correspond completely in spatial position; millimeter-level registration is achieved through iterative optimization to ensure that there is no offset or misalignment between the deformation space region and the original spatial structure components.

[0108] In some embodiments, in step S104 above, obtaining the deformation characteristics of each deformation space region relative to the matching original spatial structure component includes: for each deformation space region, transforming the coordinates of the point cloud data in the deformation space region to the coordinate system where the original spatial structure component is located; calculating the deformation displacement parameter and deformation volume parameter of the deformation space region based on the point cloud data in the deformation space region and the surface position coordinates of the corresponding matching original spatial structure component, wherein the deformation characteristics include, but are not limited to, the deformation displacement parameter and deformation volume parameter.

[0109] After matching and registering each deformable spatial region with the original spatial structural components of the prior 3D model, deformation scaling calculations are performed for different types of deformable spatial regions: 1) For intrusive deformation spatial regions: calculate the deformation displacement (by analyzing the average intrusion distance of point cloud data relative to the surface of the original spatial structure component), the intrusion volume (the total volume of the grid within the region), the maximum intrusion depth, etc. 2) For missing deformation space regions: calculate the missing volume (total volume of the grid within the region), the missing range (boundary length, area, etc.), and the size of the cavity exposed in the missing part.

[0110] For example, the identified deformation zone (Deformation_Zone_In_1) is registered with the original mine BIM model. First, using a preset coordinate transformation matrix, the center point coordinates of each grid cell in the deformation zone (based on WGS84) are accurately transformed to the local engineering coordinate system of the BIM model, achieving millimeter-level geometric alignment. Next, based on the semantic tag "CORRIDOR" of the grid cells in the deformation zone, the corresponding tunnel component (ID: BIM_Corridor_Section_5) is automatically located in the BIM model. By comparing the mapped and matched point cloud data in the deformation zone with the corresponding original spatial structure component surface in the BIM model, the average inward encroachment (deformation displacement) of the deformation zone relative to the corresponding tunnel component surface in the BIM model is calculated to be approximately 0.8 meters, and the total volume of the collapsed object is estimated to be approximately 85 cubic meters (i.e., the volume of 85 1m³ grid cells).

[0111] In some embodiments, after matching and registering each deformable spatial region with the original spatial structure components of the prior 3D model and obtaining the deformation features, a standardized registration result is generated for each deformable spatial region, including but not limited to: the corresponding original spatial structure component ID; the type of deformable region (intrusive / missing); the amount of deformation displacement, the volume of encroachment / missing; the boundary geometric data; and the spatial semantic label.

[0112] In some embodiments, each deformation space region includes an intrusive deformation space region and / or a missing deformation space region. In step S105 above, based on the geometric contour and deformation characteristics of each deformation space region, the original spatial structure components matched in the prior 3D model are reconstructed to obtain a digital twin corresponding to the target buried space. This includes: for intrusive deformation space regions, based on the geometric contour of the intrusive deformation region, performing a local mesh Boolean subtraction operation on the original spatial structure components matched in the prior 3D model to remove the structural portion of the original spatial structure component corresponding to the intrusive deformation space region; based on the intrusive deformation space region... The deformation features, spatial distribution of point cloud data, and surface features within the deformation region are used to generate a solid structural surface representing the intruding entity at the cut-off location. For missing deformation spatial regions, based on the geometric contour of the missing deformation spatial region, a local mesh deletion operation is performed on the original spatial structural components matched in the prior 3D model to remove the structural parts of the original spatial structural components corresponding to the missing deformation spatial region. If the deleted region exposes a new cavity space, a new boundary mesh for the cavity space is generated based on the spatial semantic information and adjacency topology of adjacent grid cells. The digital twin corresponding to the target buried space is obtained.

[0113] Specifically, the geometric contours, deformation types, deformation characteristics, and original spatial structural component information that have been matched and registered with the prior 3D model for each deformation spatial region output in step S104 are obtained to determine the components to be reconstructed and the operation range. The 3D spatial range enclosed by spatial boundary grid cells of each deformation spatial region is used as the local operation boundary for model reconstruction, and reconstruction processing is performed on the local component regions where deformation occurs. At the same time, it is confirmed that the deformation spatial region and the prior 3D model are in a unified world coordinate system to ensure that the reconstruction position corresponds accurately with the physical space on site, providing accurate geometric boundaries and data foundation for subsequent differentiated local mesh reconstruction.

[0114] For intrusive deformation space regions, a local mesh Boolean subtraction operation is performed within the geometric contour range of the original spatial structural components of the corresponding cavity class to remove the space occupied by the collapsed body and deposits from the original cavity model. Based on deformation characteristics (such as deformation displacement and deformation volume), the spatial distribution and surface features of point cloud data within the deformation space region, a new solid structure surface is generated at the cut-off position to characterize the geometric shape of the intrusive entity on site, i.e., the geometric shape of the intrusive deformation space region. Furthermore, based on the reflection intensity, color, and point cloud density distribution of the point cloud data within the intrusive deformation space region, the newly generated solid structure surface can be assigned corresponding material properties such as gravel accumulation, coal slag accumulation, and silt deposition to restore the real surface features of the burial site.

[0115] For missing deformation spatial regions, a local mesh deletion operation is performed within the geometric contour range corresponding to the original spatial structural components to remove structural parts that have collapsed, been damaged, or disappeared from the model, i.e., the structural parts corresponding to the missing deformation spatial regions. If the deleted region exposes a new cavity space, a new boundary mesh for the cavity space is automatically generated based on the spatial semantic information and three-dimensional adjacency topology of the adjacent grid cells of the new cavity space, ensuring the integrity and continuity of the reconstructed spatial structure.

[0116] Among them, the deletion of the region exposes a new cavity space. This means that in the prior 3D model, after performing local mesh deletion on the original spatial structural components (such as concrete walls, columns, and roof slabs) corresponding to the missing deformation space region, the internal space that was originally obscured and closed by the entity is opened up, forming a new exposed and passable cavity region. The spatial semantics of the new cavity space can be inherited based on the spatial semantic information of the adjacent grid cells or default to the cavity class.

[0117] The adjacent grid cells of the new cavity space refer to the grid cells that are adjacent to the newly exposed cavity space and have not been deleted (i.e., still retained in the prior 3D model).

[0118] For raster cells whose semantics are entity structures and whose point cloud mapping state is True, i.e., the original spatial structural components corresponding to the raster cells with intact structures, their original geometric shape, spatial semantics and physical properties in the prior 3D model are kept unchanged. No mesh modification, deletion or update operations are performed to maintain the overall structural stability of the model and reduce redundant calculations.

[0119] In some embodiments, during the reconstruction process, the physical property parameters such as the medium coefficient, water content coefficient, and surface reflectivity coefficient of the corresponding grid cells are dynamically updated based on the point cloud density distribution, reflection intensity, and on-site environmental characteristics in the deformed space region, providing reliable data support for subsequent electromagnetic signal simulation, communication quality analysis, and rescue passability assessment.

[0120] The deformable space components that have undergone partial reconstruction, the updated physical property parameters, and the unmodified structural components are integrated to form a complete three-dimensional digital model. This results in a digital twin that can accurately reflect the real-time structural status, deformation location, boundary range, and physical properties of the target buried space, which can be used for emergency rescue command, on-site situational awareness, and risk assessment.

[0121] For example, model baking technology is used for fusion and reconstruction. First, based on the geometric contour of the intrusive deformation zone Deformation_Zone_In_1 (defined by the boundary grid), a local mesh Boolean subtraction operation is performed on the corresponding original spatial structural component BIM_Corridor_Section_5 of the original BIM model. This "cuts out" the part representing the collapsed entity from the original tunnel cavity, and the surface of the collapsed entity is generated at the cut-off location based on the deformation characteristics. Then, based on the average reflection intensity of the point cloud data within the deformation zone, the newly added surface of the collapsed entity is assigned corresponding material properties (such as "rubble accumulation"). Finally, the updated digital twin is output. This model clearly shows that there is a collapsed entity approximately 100 meters long and 85 cubic meters in volume within the BIM_Corridor_Section_5 tunnel, achieving an accurate restoration of the actual disaster site.

[0122] In some embodiments, due to the technical limitations of laser scanning, some abnormal point clouds may exist. These abnormal point clouds are identified and processed to avoid identification errors. Since the entity shape of an object in physical space is continuous, during the construction process, methods such as adjacency detection, sparsity detection, and artificial intelligence can be used to detect the validity of point cloud mapping, so as to achieve intelligent filtering and optimization of point cloud data and improve the accuracy of reconstruction. After determining the point cloud mapping state of each grid cell in the prior grid model, that is, after the above step S102, the method further includes: performing point cloud mapping validity detection on the grid cells with the point cloud mapping state of the first mapping state, identifying grid cells with invalid mapping and grid cells with valid mapping; updating the point cloud mapping state of the grid cells with invalid mapping to the second mapping state, while keeping the point cloud mapping state of the grid cells with valid mapping unchanged.

[0123] Figure 2 This is a flowchart illustrating a method for validating point cloud mappings in an embodiment of this application. Figure 2 As shown, the above-mentioned point cloud mapping validity detection is performed on the raster cells whose point cloud mapping state is the first mapping state, identifying raster cells with invalid mapping and raster cells with valid mapping, including: Step S201: Traverse each corresponding raster cell whose point cloud mapping state is the first mapping state; Extract all raster cells containing at least one point cloud data within the space, i.e. raster cells whose point cloud mapping state is the first mapping state, to form a set of raster cells whose point cloud mapping validity needs to be identified.

[0124] For example, in the rescue of a mine flooding accident, a raster point cloud mapping status table containing 50,000 raster cells was generated. Among them, the point cloud mapping status of 1,200 raster cells was marked as "True" (first mapping status). These 1,200 raster cells were extracted to form a set of raster cells whose point cloud mapping validity was to be identified.

[0125] Using an iterative loop, each grid cell in the set is sequentially traversed. For the currently traversed grid, one or more of the following discrimination condition checks based on preset rules are performed.

[0126] Step S202: Determine whether the amount of point cloud data mapped and matched by the currently traversed raster cells has reached the validity detection threshold. If yes, proceed to step S204; otherwise, proceed to step S203.

[0127] For each raster cell marked as the first mapping state (True), the amount of real-time point cloud data contained within the spatial range of that raster cell is counted.

[0128] The amount of real-time point cloud data contained within the spatial range of the grid cell is compared with a preset validity detection threshold. This threshold is an empirical or calculated value determined based on factors such as the accuracy of the scanning device, the level of environmental noise, and the required model reconstruction accuracy.

[0129] If the amount of point cloud data mapped and matched by the currently traversed raster cell reaches or exceeds the validity detection threshold, it means that the amount of point cloud data contained in the raster cell is sufficient and the data reliability is high. It can be initially determined to be valid data, so the discrimination in step S204 can continue.

[0130] If it is determined that the amount of point cloud data mapped and matched by the currently traversed grid cell does not reach the validity detection threshold, it means that the point cloud data in the grid cell is too sparse and is very likely to be an abnormal point cloud (outlier) caused by scanning noise, dust interference or accidental reflection. Such data has low reliability. Therefore, step S203 is executed to determine the currently traversed grid cell as an invalid mapped grid cell and delete the currently traversed grid cell from the grid cell set.

[0131] For example, the preset validity detection threshold for point cloud data within a single grid is 50. When traversing to the grid cell Grid_2050, it is found that it contains only 8 point cloud data. Since 8 < 50, the validity detection threshold is not met. Therefore, the point cloud data in this grid cell is determined to be unreliable and may be caused by dust interference. Thus, this grid cell is directly deleted from the grid cell set.

[0132] Step S203: If the amount of point cloud data mapped and matched by the currently traversed raster cell does not reach the validity detection threshold, the currently traversed raster cell is determined as an invalid mapped raster cell. Step S204: If the amount of point cloud data mapped and matched by the currently traversed grid cell reaches the validity detection threshold, check the point cloud mapping status of other grid cells adjacent to the currently traversed grid cell, and determine whether there are other adjacent grid cells with the point cloud mapping status as the first mapping status. If the amount of point cloud data matched by the currently traversed raster cell reaches the validity detection threshold, check the point cloud mapping status of other raster cells adjacent to the currently traversed raster cell, that is, check whether the point cloud mapping status of these other adjacent raster cells is True.

[0133] If at least one of the 26 adjacent raster cells in the currently traversed raster cell has a point cloud mapping state of True (i.e., it is also hit by the point cloud mapping), then it is determined that the currently traversed raster cell is not isolated in space. This is consistent with the law that physical entities have continuity. Therefore, the point cloud data in the currently traversed raster cell is very likely to be valid data.

[0134] If the point cloud mapping state of all adjacent grid cells of the currently traversed grid cell is the second mapping state False (i.e., none of them are mapped by the point cloud), then the currently traversed grid cell is determined to be an "isolated point" in space. This seriously violates the principle of entity continuity. Therefore, the point cloud data in the currently traversed grid cell is very likely to be an anomaly point caused by noise or interference.

[0135] For example, the grid Grid_3050 has 200 point clouds. After passing the threshold check, its 26 adjacent grids were searched. It was found that the Boolean state of all adjacent grids was "False" (cavity). This indicates that Grid_3050 is a completely isolated point cloud cluster, which violates the principle of spatial continuity. It is determined to be an outlier (such as caused by accidental reflection from a suspended cable), so it is removed from the set.

[0136] Step S205: If at least one other grid cell in the adjacent grid cells has a point cloud mapping state of the first mapping state, proceed to step S208. If at least one of the 26 adjacent raster cells in the currently traversed raster cell has a point cloud mapping state of True (i.e., it is also hit by the point cloud mapping), then it is determined that the currently traversed raster cell is not isolated in space. This is consistent with the law that physical entities are continuous. Therefore, the point cloud data in the currently traversed raster cell is likely to be valid data. Therefore, the process jumps to step S208 for further judgment.

[0137] Step S206: If the point cloud mapping state of other adjacent raster cells is the second mapping state, and the currently traversed raster cell is not a spatial boundary raster cell, the currently traversed raster cell is determined as an invalid mapping raster cell. If the point cloud mapping state of all adjacent raster cells is the second mapping state, check whether the currently traversed raster cell has been pre-identified and marked as a spatial boundary raster cell. A spatial boundary raster cell is a boundary representation of a spatial volume pre-established based on the adjacency topology of raster cells. That is, a raster with effective adjacency in all directions can be regarded as an empty volume raster cell, and a raster with partially missing adjacency is regarded as a spatial boundary raster cell. By analyzing the adjacency topology of each raster cell, the raster cells located at the boundary between the solid and the cavity are automatically identified and marked as spatial boundary raster cells.

[0138] If the point cloud mapping states of all adjacent raster cells are in the second mapping state, and the currently traversed raster cell is not a spatial boundary raster cell, the point cloud data in the currently traversed raster cell is determined to be an abnormal point, and the currently traversed raster cell is determined to be an invalid mapping raster cell.

[0139] Step S207: If the point cloud mapping state of other adjacent grid cells is the second mapping state, and the currently traversed grid cell is a spatial boundary grid cell, proceed to step S208. If the point cloud mapping states of all adjacent raster cells are in the second mapping state, and the currently traversed raster cell is a spatial boundary raster cell, it means that even if some of its adjacent raster cells are not mapped by point cloud (i.e., in a cavity state), it still meets the boundary characteristics of physical space. Therefore, the point cloud data in the currently traversed raster cell is very likely to be valid boundary data, and the process jumps to step S208 for further judgment.

[0140] Step S208: Input the features of the point cloud data volume of the currently traversed grid cell mapping, whether there are other adjacent grid cells with the point cloud mapping state of the first mapping state, and whether it is a spatial boundary grid cell into the preset data classification model to perform abnormal data classification prediction. A data classification model refers to a pre-trained artificial intelligence classification model used for intelligent identification of point cloud anomalies.

[0141] The various features of the currently traversed raster cells are combined into a feature vector, which is then used as input to the data classification model. These features typically include, but are not limited to: The amount of point cloud data mapped and matched by the currently traversed raster cell: the point cloud density within the currently traversed raster cell; Features indicating whether there are other adjacent raster cells with a point cloud mapping state of the first mapping state: the situation where the point cloud hits other raster cells adjacent to the currently traversed raster cell; used to evaluate spatial continuity. This is a feature that indicates whether the currently traversed raster cell is a spatial boundary raster cell.

[0142] A pre-trained data classification model (such as a model based on random forest or neural network) analyzes the feature vector and outputs a judgment result (such as "abnormal data" or "valid data") and the corresponding confidence score.

[0143] Step S209: If the data classification model outputs abnormal data, the currently traversed raster cell is determined as an invalid mapped raster cell. Step S210: If the data classification model outputs valid data, the currently traversed raster cell is determined as a validly mapped raster cell.

[0144] If the data classification model outputs valid data, the currently traversed raster cell is determined as a validly mapped raster cell, and the traversal continues to the next raster cell in the raster cell set.

[0145] Through the above point cloud mapping validity detection, the set of validly mapped raster cells is finally selected. This set contains all raster cells that have been verified by multi-level filtering and confirmed to be mapped by reliable point cloud data. In the subsequent digital twin reconstruction process, the following local features are extracted from each validly mapped raster cell to form the data foundation required for reconstruction, including but not limited to the following local features: geometric features and semantic features.

[0146] Geometric features include, but are not limited to: 1) Precise spatial location: Based on the center point coordinates of the grid cells, determine the precise three-dimensional positioning of the deformable spatial region; 2) Local surface normal vector: By analyzing the distribution of point cloud within a grid cell, the orientation (normal vector) of the local surface it represents is calculated, which is used to reconstruct a smooth and physically consistent geometric surface. 3) Point cloud density distribution: used as a basis for assessing the "solidity" of the deformed entity or the surface roughness.

[0147] Semantic features include, but are not limited to: 1) Inherited spatial semantic codes: For example, whether the grid cell represents a deformed portion of a "lane", "room", or "equipment area"; 2) Associated physical parameters: such as the water content coefficient and dielectric coefficient corresponding to the grid cell. These parameters will provide input for subsequent applications such as communication simulation and structural analysis.

[0148] For example, after filtering the point cloud data of all raster cells, only 950 of the initial 1200 raster cells with the first mapping state are retained, forming a high-quality set of effectively mapped raster cells. During the digital twin reconstruction process, local feature reconstruction is performed based on this set. For instance, it is found that raster cell Grid_4050 and its adjacent effectively mapped raster cells together constitute the boundary of a collapsed body. By calculating the normal vectors of the point clouds within these raster cells, the tilt angle and roughness of the collapsed body surface can be accurately reconstructed, providing high-fidelity geometric and physical parameters for subsequent passability analysis and risk assessment.

[0149] According to the method for constructing a digital twin of buried space in the embodiments of this application, the continuous space is discretized into standard grid units through the above-mentioned prior grid model model, so that the complex spatial analysis is transformed into parallel computation of discrete grids, providing a structured data foundation for subsequent rapid point cloud mapping and deformation recognition; the grid scale can be dynamically adjusted according to the simulation accuracy requirements, realizing the optimal balance between computational efficiency and model fineness, and can adapt to different application scenarios from macroscopic deduction to fine analysis.

[0150] By establishing a point cloud-to-raster mapping mechanism and a semantic-based rapid deformation identification criterion, a standardized process for batch mapping real-time point cloud data to prior raster models was established. A dual-channel deformation discrimination criterion to distinguish between invasive and missing deformations was proposed, transforming the complex problem of 3D point cloud registration and model comparison into a rapid logical judgment of the state of discrete raster cells. This significantly improves computational efficiency and meets the real-time requirements for minute-level situation updates in disaster emergency response. This mechanism requires no manual intervention and can automatically and in batches locate all deformed spatial regions, overcoming the shortcomings of related technologies that rely on manual interpretation, leading to delays and subjectivity.

[0151] A multi-level intelligent point cloud filtering algorithm, employing rule-based pre-screening and AI arbitration, performs preliminary screening through multiple rules, including data volume threshold judgment, adjacency relationship point cloud hit detection, and boundary grid recognition. Then, an AI model is introduced for final point cloud mapping validity arbitration. Compared to single filtering methods, this approach more accurately identifies and removes anomalies caused by dust, occlusion, etc., while effectively protecting the boundary details of true deformation. This provides high-quality, highly reliable input data for subsequent deformation recognition and model reconstruction, fundamentally avoiding the risk of data noise misleading rescue decisions.

[0152] By integrating geometric and semantic automatic cavity boundary extraction with a digital twin incremental update mechanism, the system automatically identifies boundary grids and empty grids using the adjacency topology of the grid, thereby achieving automatic extraction of cavity boundaries. Furthermore, by employing model baking technology, deformation features are incrementally updated into the prior 3D model to reconstruct the digital twin. This allows for accurate extraction of cavity boundaries without manual intervention, providing a direct and reliable basis for accessibility analysis and risk assessment during rescue operations.

[0153] This invention also provides a system for constructing a digital twin of a buried space, which is used to implement the above embodiments and preferred embodiments; details already described will not be repeated. As used below, the term "module" can refer to a combination of software and / or hardware that performs a predetermined function. Although the systems described in the following embodiments are preferably implemented in software, hardware implementations, or a combination of software and hardware, are also possible and contemplated.

[0154] Figure 3 This is a structural block diagram of a digital twin construction system for buried space provided in an embodiment of this application. The embodiment of this application provides a digital twin construction system for buried space, such as... Figure 3 As shown, the system for constructing a digital twin of a buried space includes: The point cloud acquisition module 301 is used to acquire point cloud data sequences of the target burial space. The point cloud data sequences include point cloud data of multiple spatial location points in the burial space. The point cloud mapping module 302 is used to map and match each point cloud data in the target burial space with each grid cell in the prior grid model corresponding to the target burial space, and determine the point cloud mapping state of each grid cell in the prior grid model; wherein, the prior grid model is obtained by rasterizing the original three-dimensional space based on the prior three-dimensional space corresponding to the target burial space. The deformation recognition module 303 is used to determine the deformation space regions in which the target burial space has deformed compared to the original three-dimensional space based on the spatial semantic information and point cloud mapping status corresponding to each grid unit. The spatial semantic information corresponding to the grid unit is used to identify the type of spatial object corresponding to the grid unit. The deformation feature extraction module 304 is used to match and register each deformation spatial region with the original spatial structure component of the prior 3D model, and to obtain the deformation features of each deformation spatial region relative to the matched original spatial structure component. The reconstruction module 305 is used to reconstruct each original spatial structure component matched in the prior 3D model according to the geometric contour and deformation characteristics of each deformation spatial region, so as to obtain the digital twin corresponding to the target burial space.

[0155] In some embodiments, the system further includes a prior raster modeling module, which is used to rasterize the original three-dimensional space based on a prior three-dimensional model of the original three-dimensional space to obtain a prior raster model corresponding to the original three-dimensional space; wherein, the prior raster model includes multiple raster units, and each raster unit includes the position coordinates of the raster center point and spatial semantic information.

[0156] In some embodiments, the system further includes a point cloud intelligent filtering module, which, after determining the point cloud mapping state of each grid cell in the prior grid model, performs point cloud mapping validity detection on the grid cells with the point cloud mapping state of the first mapping state, identifies grid cells with invalid mapping and grid cells with valid mapping; updates the point cloud mapping state of the grid cells with invalid mapping to the second mapping state, and keeps the point cloud mapping state of the grid cells with valid mapping unchanged.

[0157] In some embodiments, the system further includes a system data management module, which is responsible for the unified scheduling and management of the entire workflow from data input to result output. This module coordinates the execution order and data interaction of various modules such as prior raster modeling, real-time point cloud mapping, deformation recognition and reconstruction, and centrally manages all data assets.

[0158] The buried space digital twin construction system provided in this embodiment of the invention can execute the buried space digital twin construction method provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the method. Further functional descriptions of the above modules and units are the same as in the corresponding embodiments described above, and will not be repeated here.

[0159] Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0160] The following is a detailed reference. Figure 4 This diagram illustrates a suitable structural schematic for implementing the electronic device described in the embodiments of this application. The electronic device may include a processor (e.g., a central processing unit, graphics processor, etc.) 401, which can perform various appropriate actions and processes based on a program stored in read-only memory (ROM) 402 or a program loaded from memory 408 into random access memory (RAM) 403. RAM 403 also stores various programs and data required for the operation of the electronic device. The processor 401, ROM 402, and RAM 403 are interconnected via bus 404. Input / output (I / O) interface 405 is also connected to bus 404.

[0161] Typically, the following devices can be connected to I / O interface 405: input devices 406 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 407 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; memory devices 408 including, for example, magnetic tapes, hard disks, etc.; and communication devices 409. Communication device 409 allows electronic devices to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 4 Electronic devices with various devices are shown, but it should be understood that it is not required to implement or have all of the devices shown, and more or fewer devices may be implemented or have instead.

[0162] Specifically, according to embodiments of this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this application include a computer program product comprising a computer program carried on a non-transitory computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device 409, or installed from a memory 408, or installed from a ROM 402. When the computer program is executed by the processor 401, it performs the functions defined in the method for constructing a digital twin of a buried space according to embodiments of this application.

[0163] Figure 4 The electronic device shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.

[0164] This application also provides a computer-readable storage medium. The methods described in this application can be implemented in hardware or firmware, or implemented as recordable on a storage medium, or implemented as computer code downloaded over a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and then stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code. When the software or computer code is accessed and executed by the computer, processor, or hardware, the method for constructing a digital twin of a buried space shown in the above embodiments is implemented.

[0165] A portion of this application can be applied as a computer program product, such as computer program instructions, which, when executed by a computer, can invoke or provide the methods and / or technical solutions according to this application through the operation of the computer. Those skilled in the art will understand that the forms in which computer program instructions exist in a computer-readable medium include, but are not limited to, source files, executable files, installation package files, etc. Correspondingly, the ways in which computer program instructions are executed by a computer include, but are not limited to: the computer directly executing the instructions, or the computer compiling the instructions and then executing the corresponding compiled program, or the computer reading and executing the instructions, or the computer reading and installing the instructions and then executing the corresponding installed program. Here, the computer-readable medium can be any available computer-readable storage medium or communication medium accessible to a computer.

[0166] Although embodiments of this application have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of this application, and all such modifications and variations fall within the scope defined by the appended claims.

Claims

1. A method for constructing a digital twin of a buried space, characterized in that, The method includes: Collect point cloud data sequences of the target burial space, wherein the point cloud data sequences include point cloud data of multiple spatial location points in the burial space; The point cloud data in the target burial space are mapped and matched with each grid cell in the prior grid model corresponding to the target burial space to determine the point cloud mapping state of each grid cell in the prior grid model; wherein, the prior grid model is obtained by rasterizing the original three-dimensional space based on the prior three-dimensional space corresponding to the target burial space. Based on the spatial semantic information and point cloud mapping state corresponding to each grid unit, the deformed spatial regions in which the target burial space has deformed compared to the original three-dimensional space are determined. The spatial semantic information corresponding to the grid unit is used to identify the type of spatial object corresponding to the grid unit. Each of the deformable spatial regions is matched and registered with the original spatial structure components of the prior 3D model, and the deformation characteristics of each of the deformable spatial regions compared with the matched original spatial structure components are obtained. Based on the geometric contours and deformation characteristics of each of the deformable spatial regions, the original spatial structure components matched in the prior three-dimensional model are reconstructed to obtain the digital twin corresponding to the target burial space; The step of determining the deformed spatial regions of the target burial space that have undergone deformation compared to the original three-dimensional space based on the spatial semantic information and point cloud mapping state corresponding to each grid unit includes: Spatial semantic information is represented as cavity object type, and the point cloud mapping state represents the raster unit of the point cloud data that has mapping, which is divided into an intrusive deformation raster unit set. Spatial semantic information is represented as a structural object type, and the point cloud mapping state represents the raster units of point cloud data that do not have mapping, which are divided into a set of missing deformable raster units. Each invasive deformable grid cell in the set of invasive deformable grid cells is spatially clustered according to the adjacency topology of each invasive deformable grid cell to obtain at least one invasive deformable spatial region. Each missing deformation raster unit in the set of missing deformation raster units is spatially clustered according to the adjacency topology of each missing deformation raster unit to obtain at least one missing deformation spatial region.

2. The method according to claim 1, characterized in that, The prior grid model is constructed in advance in the following way: The original three-dimensional space is rasterized based on the prior three-dimensional model of the original three-dimensional space to obtain the prior raster model corresponding to the original three-dimensional space. The prior grid model includes multiple grid cells, each of which includes the position coordinates of the grid center point and spatial semantic information.

3. The method according to claim 1, characterized in that, The step of mapping and matching each point cloud data in the target burial space with each grid cell in the prior grid model corresponding to the target burial space to determine the point cloud mapping state of each grid cell in the prior grid model includes: Align the point cloud data in the target buried space with the position coordinates according to the position coordinate system of the prior grid model; Identify the initial semantic information corresponding to each point cloud data in the target burial space, and spatially align the initial semantic information corresponding to each point cloud data with the spatial semantic information of each grid cell. Based on the position coordinates of each point cloud data after alignment and the spatial range conditions corresponding to each grid cell, the grid cells that each point cloud data is mapped to and matched are determined; wherein, if the position coordinates of the point cloud data satisfy the spatial range conditions corresponding to the grid cell, then the point cloud data is mapped and matched with the grid cell. Generate a point cloud mapping state for each grid cell, wherein the point cloud mapping state indicates whether there is at least one mapping match point cloud data in the grid cell space.

4. The method according to claim 1, characterized in that, The process of matching and registering each of the deformable spatial regions with the original spatial structural components of the prior 3D model includes: For each of the deformation space regions, the position coordinates of the grid center point of each grid cell in the deformation space region are aligned according to the local engineering coordinate system of the prior 3D model; Based on the grid center point coordinates and spatial semantic information of each grid unit in the deformable spatial region, the corresponding matching original spatial structure component is searched in the prior 3D model.

5. The method according to claim 1, characterized in that, The step of obtaining the deformation characteristics of each of the deformed spatial regions relative to the matched original spatial structural components includes: For each of the deformation space regions, the coordinates of the point cloud data in the deformation space region are transformed to the coordinate system of the original spatial structure component; Based on the point cloud data in the deformation space region and the surface position coordinates of the corresponding matching original spatial structure components, the deformation displacement parameters and deformation volume parameters of the deformation space region are calculated, and the deformation features include the deformation displacement parameters and deformation volume parameters.

6. The method according to claim 1, characterized in that, Each of the aforementioned deformation space regions includes an intrusive deformation space region and / or a missing deformation space region. The step of reconstructing each original spatial structure component matched in the prior 3D model based on the geometric contour and deformation characteristics of each of the aforementioned deformation space regions to obtain a digital twin corresponding to the target burial space includes: For an intrusive deformation space region, based on the geometric contour of the intrusive deformation region, a local mesh Boolean subtraction operation is performed on the original spatial structure component matched in the prior 3D model to remove the structural part of the original spatial structure component corresponding to the intrusive deformation space region. Based on the deformation characteristics, spatial distribution of point cloud data, and surface characteristics within the invasive deformation space region, a solid structure surface for characterizing the invasive entity is generated in the excised location region. For a missing deformation space region, based on the geometric contour of the missing deformation space region, a local mesh deletion operation is performed on the original spatial structure component matched in the prior 3D model to remove the structural part of the original spatial structure component corresponding to the missing deformation space region; If the deleted region exposes a new cavity space, the boundary mesh of the new cavity space is generated based on the spatial semantic information and adjacency topology of its adjacent grid cells. A digital twin corresponding to the target burial space is obtained.

7. The method according to claim 1, characterized in that, The point cloud mapping state of each grid cell is either a first mapping state or a second mapping state. The first mapping state indicates that the grid cell has at least one point cloud data with a mapping match, and the second mapping state indicates that the grid cell does not have any point cloud data with a mapping match. After determining the point cloud mapping state of each grid cell in the prior grid model, the method further includes: Perform point cloud mapping validity detection on the raster cells whose point cloud mapping state is the first mapping state, and identify the raster cells with invalid mapping and the raster cells with valid mapping. Update the point cloud mapping state of invalidally mapped raster cells to the second mapping state, while keeping the point cloud mapping state of validly mapped raster cells unchanged.

8. The method according to claim 7, characterized in that, The step of performing point cloud mapping validity detection on raster cells whose point cloud mapping state is the first mapping state, and identifying raster cells with invalid mapping and raster cells with valid mapping, includes: Iterate through each grid cell whose corresponding point cloud mapping state is the first mapping state, and determine whether the amount of point cloud data matched by the currently traversed grid cell mapping has reached the validity detection threshold. If the amount of point cloud data mapped and matched by the currently traversed raster cell does not reach the validity detection threshold, the currently traversed raster cell will be determined as an invalid mapped raster cell. If the amount of point cloud data mapped and matched by the currently traversed raster cell reaches the validity detection threshold, check the point cloud mapping status of other raster cells adjacent to the currently traversed raster cell. If at least one other grid cell in the adjacent grid cells has a point cloud mapping state of first mapping state, the following steps are performed: inputting the point cloud data volume of the currently traversed grid cell mapping matching, whether there are adjacent other grid cells with a cloud mapping state of first mapping state, and whether it is a spatial boundary grid cell into the preset data classification model to perform abnormal data classification prediction. If the point cloud mapping state of all adjacent raster cells is the second mapping state, and the currently traversed raster cell is not a spatial boundary raster cell, the currently traversed raster cell is determined as an invalid mapping raster cell. If the point cloud mapping state of the adjacent other grid cells is the second mapping state, and the currently traversed grid cell is a spatial boundary grid cell, the following steps are performed: inputting the point cloud data volume of the currently traversed grid cell mapping, whether there are adjacent other grid cells with the point cloud mapping state of the first mapping state, and whether it is a spatial boundary grid cell into the preset data classification model to perform abnormal data classification prediction. If the data classification model outputs abnormal data, the currently traversed raster cell will be determined as an invalid mapped raster cell; If the data classification model outputs valid data, the currently traversed raster cell is determined as a validly mapped raster cell.

9. A system for constructing a digital twin of a buried space, characterized in that, The system includes: The point cloud acquisition module is used to acquire point cloud data sequences of the target burial space, wherein the point cloud data sequences include point cloud data of multiple spatial location points in the burial space; The point cloud mapping module is used to map and match each point cloud data in the target burial space with each grid cell in the prior grid model corresponding to the target burial space, and determine the point cloud mapping state of each grid cell in the prior grid model; wherein, the prior grid model is obtained by rasterizing the original three-dimensional space based on the prior three-dimensional space corresponding to the target burial space. The deformation recognition module is used to determine the deformation spatial regions in the target buried space that have deformed compared to the original three-dimensional space based on the spatial semantic information and point cloud mapping state corresponding to each grid unit. The spatial semantic information corresponding to the grid unit is used to identify the type of spatial object corresponding to the grid unit. The deformation feature extraction module is used to match and register each of the deformation spatial regions with the original spatial structure components of the prior 3D model, and to obtain the deformation features of each of the deformation spatial regions relative to the matched original spatial structure components. The reconstruction module is used to reconstruct each original spatial structure component matched in the prior three-dimensional model according to the geometric contour and deformation characteristics of each of the deformation spatial regions, so as to obtain the digital twin corresponding to the target burial space. The deformation recognition module is used to: represent spatial semantic information as cavity object types and point cloud mapping states as point cloud data with mapping, and divide them into a set of invasive deformation raster units; represent spatial semantic information as structural object types and point cloud mapping states as point cloud data without mapping, and divide them into a set of missing deformation raster units; spatially cluster each invasive deformation raster unit in the set of invasive deformation raster units according to the adjacency topology of each invasive deformation raster unit to obtain at least one invasive deformation spatial region; spatially cluster each missing deformation raster unit in the set of missing deformation raster units according to the adjacency topology of each missing deformation raster unit to obtain at least one missing deformation spatial region.

10. An electronic device, characterized in that, include: A memory and a processor are communicatively connected, the memory stores computer instructions, and the processor executes the computer instructions to perform the method for constructing a digital twin of a buried space as described in any one of claims 1 to 8.

11. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing a computer to execute the method for constructing a digital twin of a buried space as described in any one of claims 1 to 8.

Images