Virtual model rendering method, device, system and equipment based on mine site environment information and medium

Abstract

The application provides a virtual model rendering method, device, system, equipment and medium based on mine field environment information, the method comprises the following steps: obtaining mine field environment information, and constructing a semantic grid model by using the mine field environment information; the semantic grid model is split into a plurality of model leaf blocks by using a quadtree standard splitting algorithm, and the simplification probability value of each model leaf block is determined based on the semantics of each model leaf block; the plurality of model leaf blocks are divided into a simplified grid set and a precision grid set according to the size of the simplification probability value; the boundary of the precision grid set is used as a constraint condition, the QEM quadratic error measurement algorithm is used to simplify the simplified grid set and sample the simplified region to obtain simplified region data; the simplified region data and the precision grid set are visualized and rendered to obtain a mine field virtual model. The application can not only shorten the rendering time and improve the rendering efficiency, but also avoid rendering distortion and improve the rendering effect.

Description

Technical Field

[0001] This application relates to the field of model rendering and display, and in particular to a virtual model rendering method, apparatus, system, equipment and medium based on mine environment information. Background Technology

[0002] With the deep integration of intelligent manufacturing and artificial intelligence, digital twin technology is being widely applied across various industries. Digital twin technology features virtual model construction characterized by model visualization, logical controllability, and data computation, making it widely used in industrial intelligent manufacturing and virtual production scenarios. Based on machine vision and computer simulation technologies, digital twin technology will play a crucial role in applications such as 3D scene modeling in mines, physical information systems, and virtual space digital mirroring.

[0003] To facilitate construction personnel's real-time understanding of the mine's construction status from the backend, one common method is to collect real-time environmental information from the mine using a detection terminal, construct a 3D model of the mine based on this information as a digital twin model, and then visualize and render the digital twin model for the construction personnel to view from the backend.

[0004] However, the above method has the following technical problems: In visualization rendering, if high-precision mesh rendering is used for all parts, the number of triangular faces is huge, the real-time rendering frame rate is seriously insufficient, the rendering time is long and the efficiency is low; if the LOD (Level of Detail) simplification algorithm is used for rendering, key engineering features (slope cracks, tunnel support structure) are easily lost. Moreover, the geological entities of the construction mine are irregular curved surfaces and intersecting faults. Mesh division is very easy to produce distorted meshes. Using the simplification algorithm will further lead to rendering distortion and reduce the rendering effect. Summary of the Invention

[0005] In view of the aforementioned problems, this application is made to provide a virtual model rendering method, apparatus, system, device, and medium based on mining environment information that overcomes or at least partially solves the aforementioned problems, including: A virtual model rendering method based on mine environment information, the method comprising: Obtain mine environment information and use the mine environment information to construct a semantic mesh model, wherein the semantic mesh model is a three-dimensional model with added semantic labels to triangular facets; The semantic grid model is split into multiple model leaf blocks using the standard quadtree splitting algorithm, and the simplified probability value of each model leaf block is determined based on the semantics of each model leaf block. Based on the magnitude of the simplified probability value, the multiple model leaf blocks are divided into a simplified grid set and a precision grid set; Using the boundary of the precision grid set as a constraint, the simplified grid set is simplified using the QEM quadratic error metric algorithm, and the simplified region is sampled to obtain simplified region data; A virtual model of the mine is obtained by visual rendering based on the simplified regional data and the precision grid set.

[0006] A virtual model rendering device based on mine environment information, the device comprising: A construction module is used to acquire mine environment information and use the mine environment information to construct a semantic mesh model, wherein the semantic mesh model is a three-dimensional model with added semantic labels to triangular facets; The determination module is used to split the semantic grid model into multiple model leaf blocks using the standard quadtree splitting algorithm, and to determine the simplified probability value of each model leaf block based on the semantics of each model leaf block; A partitioning module is used to divide multiple model leaf blocks into a simplified grid set and a precision grid set according to the magnitude of the simplified probability value; The simplification module is used to simplify the simplified grid set using the boundary of the precision grid set as a constraint condition, and to sample the simplified region to obtain simplified region data. The rendering module is used to perform visualization rendering based on the simplified area data and the precision grid set to obtain a virtual model of the mine.

[0007] A virtual model rendering system based on mine environment information, the system comprising: a rendering processing platform and several visualization terminals; The rendering platform is communicatively connected to several visualization terminals, and the rendering platform is applicable to the virtual model rendering method based on mine environment information as described above.

[0008] An apparatus includes a processor, a memory, and a computer program stored in the memory and capable of running on the processor, wherein the computer program, when executed by the processor, implements the steps of the virtual model rendering method based on mine environment information as described above.

[0009] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the virtual model rendering method based on mine environment information as described above.

[0010] This application has the following advantages: In the embodiments of this application, the application acquires mine environment information and constructs a semantic mesh model using this information; it uses a quadtree standard splitting algorithm to divide the semantic mesh model into multiple model leaf blocks, and determines the simplification probability value of each model leaf block based on its semantics; it divides the multiple model leaf blocks into a simplified mesh set and a precision mesh set according to the magnitude of the simplification probability value; using the boundary of the precision mesh set as a constraint, it uses the QEM quadratic error metric algorithm to simplify the simplified mesh set and samples the simplified region to obtain simplified region data; it performs visualization rendering based on the simplified region data and the precision mesh set to obtain a virtual mine model. This application can simplify some triangular faces, eliminating the need for visualization rendering of a large number of faces, thus shortening rendering time and improving rendering efficiency; moreover, this application uses high-precision mesh faces as constraints for simplification, which can reduce the probability of distorted meshes after simplification, thereby avoiding rendering distortion and improving the rendering effect. Attached Figure Description

[0011] To more clearly illustrate the technical solution of this application, the drawings used in the description of this application will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 This is a flowchart illustrating the steps of a virtual model rendering method based on mine environment information provided in an embodiment of this application; Figure 2 This is a structural block diagram of a virtual model rendering device based on mine environment information provided in one embodiment of this application; Figure 3 This is a structural block diagram of a virtual model rendering system based on mine environment information provided in one embodiment of this application; Figure 4 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0013] To make the objectives, features, and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0014] With the deep integration of intelligent manufacturing and artificial intelligence, digital twin technology is being widely applied across various industries. Digital twin technology features virtual model construction characterized by model visualization, logical controllability, and data computation, making it widely used in industrial intelligent manufacturing and virtual production scenarios. Based on machine vision and computer simulation technologies, digital twin technology will play a crucial role in applications such as 3D scene modeling in mines, physical information systems, and virtual space digital mirroring.

[0015] To facilitate construction personnel's real-time understanding of the mine's construction status from the backend, one common method is to collect real-time environmental information from the mine using a detection terminal, construct a 3D model of the mine based on this information as a digital twin model, and then visualize and render the digital twin model for the construction personnel to view from the backend.

[0016] However, the above method has the following technical problems: In visualization rendering, if high-precision mesh rendering is used for all parts, the number of triangular faces is huge, the real-time rendering frame rate is seriously insufficient, the rendering time is long and the efficiency is low; if the LOD (Level of Detail) simplification algorithm is used for rendering, key engineering features (slope cracks, tunnel support structure) are easily lost. Moreover, the geological entities of the construction mine are irregular curved surfaces and intersecting faults. Mesh division is very easy to produce distorted meshes. Using the simplification algorithm will further lead to rendering distortion and reduce the rendering effect.

[0017] To address the aforementioned issues, the following specific embodiments will provide a detailed description and explanation of a virtual model rendering method, apparatus, system, equipment, and medium based on mine environment information provided in this application.

[0018] To address the issues of low efficiency and high distortion in existing visualization and rendering technologies, this paper refers to... Figure 1 The diagram shows a flowchart of a virtual model rendering method based on mine environment information provided in an embodiment of this application.

[0019] In one embodiment, the virtual model rendering method based on mine environment information is applicable to a rendering processing platform. The rendering processing platform can be located on a computer.

[0020] As an example, the virtual model rendering method based on mine environment information may include: S11. Obtain mine environment information and use the mine environment information to construct a semantic mesh model, wherein the semantic mesh model is a three-dimensional model with added semantic labels to triangular facets.

[0021] In one embodiment, mine environmental information can be obtained from mine environmental data uploaded by technicians or construction workers, or multi-source data can be collected through terminals set up at the mine. Mine environmental information may include spatial geometric data of the mine, as well as engineering attribute data and semantically annotated raw data added by construction workers.

[0022] Next, a three-dimensional simulation model can be constructed using the mine environment information, and semantic labels can be added to the triangular facets of the three-dimensional simulation model to obtain a semantic mesh model.

[0023] By adding semantics, the definition of each triangle can be determined, so that subsequent optimization or simplification can be determined based on the semantics, thereby improving the processing efficiency of subsequent visualization rendering.

[0024] As an example, the step of acquiring mine environment information and using the mine environment information to construct a semantic mesh model may include the following sub-steps: S111. Obtain the mining environment information uploaded by the construction personnel, wherein the mining environment information includes: spatial geometric data, engineering attribute data and semantically labeled raw data.

[0025] S112. The mining environment information is preprocessed to obtain mining processed data, wherein the preprocessing includes noise reduction, coordinate alignment and duplicate point removal.

[0026] S113. A mining grid model is constructed based on the mining data using the Poisson reconstruction algorithm, and a unique digital semantic code is bound to each triangular facet of the mining grid model to obtain a semantic grid model.

[0027] In one embodiment, this application can be applied to open-pit mines, deploying multiple types of mine detection terminals at the front end of the mine: airborne 3D LiDAR, tunnel inspection robot, slope high-definition depth camera, geological borehole sensor terminal, support stress acquisition instrument, and mine positioning base station; the back-end rendering and processing platform receives real-time data streams from the terminals and automatically generates mine triangular semantic meshes with facet semantic tags, solving the pain points of traditional mine meshes having no engineering attributes, many deformed faces, and inability to fuse multi-source data.

[0028] In one embodiment, the mine is divided into open-pit mining area, underground roadways, slope support area, and spoil heap, etc. Heterogeneous data acquisition terminals are deployed in each area. A time-sharing polling mechanism can be adopted to prioritize the acquisition of terminal data from high-risk areas such as slopes and roadways. Offline terminals are marked as disconnected, and the cached data of the previous frame is retained as a backup to avoid data gaps in modeling. Three types of core raw data are received in real time to obtain mine environmental information. Specifically, the mine environmental information may include: spatial geometric data, engineering attribute data, and semantically labeled raw data.

[0029] Spatial geometric data can be a 3D point cloud of a lidar radar (x / y / z coordinates, reflection intensity), pixel depth values ​​of a depth map, or a 2D scan line of a tunnel outline; Engineering attribute data can include support steel frame dimensions, crack length and width, soil and rock stratification and lithology, fault dip angle, and equipment coordinates; The original semantic annotation data can be manual inspection marks or AI visual recognition pre-annotated text (cracks, anchors, steps, soil and rock base).

[0030] Next, the mine environment information can be preprocessed to obtain processed mine data. This preprocessing includes noise reduction, coordinate alignment, and duplicate point removal. Since the collected data may be inconsistent, data time-series alignment can be performed before preprocessing. Using a unified mine UTC timestamp as a benchmark, time interpolation alignment is performed on asynchronously collected data from different terminals. This yields mine environment information containing a unified timestamp, multi-source fused geometric raw data, and unbound engineering semantic raw labels.

[0031] Because of the existence of outliers and noise caused by dust and lighting in lidar; edge distortion in depth cameras; and the use of local coordinate systems by each terminal (robot body coordinates, radar local coordinates), spatial misalignment, direct reconstruction will result in model layering, overlap, and distorted surfaces.

[0032] Preprocessing can perform global unified coordinate benchmark transformation and point cloud hierarchical statistical filtering for noise reduction. For the 2D depth map output by the depth camera, spatial bilateral filtering can be performed. At the same time, invalid duplicate points can be removed. Specifically, spatial grid downsampling (grid size 0.02m) can be set, retaining only one sampling point within the grid, deleting a large number of overlapping and redundant points, compressing the amount of point cloud data, and reducing the computational cost of subsequent grid reconstruction.

[0033] Finally, the mining farm processing data can be output. This mining farm processing data can be a fused point cloud dataset with a unified world coordinate system, noise reduction, and registration. It has no outlier noise, no spatial layering, and includes complete geometric points of faults and tunnels after completion.

[0034] After obtaining the mining site processing data, the screened Poisson reconstruction algorithm can be used, with input fused point cloud and point normal vector; the reconstruction level is set to 8 levels to balance surface smoothness and detail preservation; screen constraints are used to force the fit to sharp geometric boundaries such as cracks and support steel frames to avoid excessive smoothing of engineering features.

[0035] The Poisson algorithm outputs a continuous, closed triangular mesh. It iterates through all faces, recording vertex indices, 3D coordinates, and normal vectors to generate an unclassified, unlabeled basic triangular mesh. It then iterates through all face edges, detecting non-manifold edges (edges associated with 3 or more faces) and free, isolated faces. For non-manifold edges, it splits the local mesh and deletes disconnected, isolated faces, ensuring topological continuity and closure. Finally, it outputs a topologically continuous basic triangular mesh for the mine, containing only geometric information and no semantic labels.

[0036] Mesh only has geometric shape, and computers cannot distinguish engineering objects such as slopes, anchors, cracks, and faults; a three-dimensional semantic segmentation algorithm is needed to automatically match each triangular facet with the category of mining engineering entities, providing semantic basis for subsequent simplified probability calculations.

[0037] Specifically, three-dimensional features can be extracted from all vertices of the basic mesh: point coordinates, surface curvature, angle between the normal vectors of neighboring patches, local point density, and geometric size features; these features are then concatenated to form a 12-dimensional feature vector, which is used as input to the semantic segmentation model.

[0038] Using a mining-specific PointNet segmentation model, the input is the feature vector of the mesh vertex; the model outputs a semantic classification probability array corresponding to each vertex, obtaining a preliminary semantic classification result at the vertex level. Then, based on the semantic classification result, each triangular facet is bound to a unique numerical semantic code, engineering feature weights, and an intermediate mesh indicating whether it crosses a feature boundary.

[0039] Metadata such as semantic encoding, engineering weights, and geological parameters need to be permanently associated with triangular facets and stored as additional attributes of the mesh. This allows for direct reading during subsequent quadtree partitioning, LOD simplification, and rendering stages, eliminating the need for repeated segmentation and calculation. Furthermore, an intermediate semantic mesh can be obtained, with geometric information and semantic metadata bound one-to-one, enabling rapid retrieval of engineering semantics and physical parameters from the facets.

[0040] Optionally, before generating the semantic mesh model, automated full verification and filtering of abnormal faces can be performed to output a compliant and standardized semantic mesh. For example, face semantic integrity verification can be performed by traversing all face indices in the hash mapping table to check whether the semantic encoding field is empty or exceeds the dictionary encoding range; faces without semantic meaning or with illegal encoding are marked and stored in the abnormal face list. Spatial coordinate out-of-bounds verification, secondary verification of mesh topology continuity, batch filtering and deletion of abnormal faces, and unified standardization of semantic metadata can also be performed.

[0041] After completing the above operations, the semantic mesh file can be standardized and encapsulated to output the industry-standard attributed 3D mesh format GLB. The file contains: basic mesh geometric data, patch semantic attribute buffer, semantic encoding dictionary, mine global bounding box parameters, and acquisition timestamp; it supports direct loading and parsing by backend rendering platforms and digital twin frontends to obtain semantic mesh models.

[0042] S12. The semantic grid model is split into multiple model leaf blocks using the standard quadtree splitting algorithm, and the simplified probability value of each model leaf block is determined based on the semantics of each model leaf block.

[0043] The semantic mesh of the mining farm output by S11 often consists of millions to tens of millions of triangular faces. Directly performing LOD simplification on the whole will result in two major pain points in the field: 1. A unified and simplified grid will indiscriminately compress high-value engineering surfaces such as slope cracks and support steel frames, making it difficult for the digital twin back-end inspection to see hidden dangers; 2. The computational cost of a single operation of the overall mesh is extremely high. The computing power of the rendering platform at the edge of the underground or open-pit mine is limited, and real-time model refresh will result in stuttering and a sharp drop in frame rate.

[0044] This step uses the standard quadtree splitting algorithm to divide the giant semantic grid into multiple independent model leaf blocks. The simplified probability value of each model leaf block is quantified by combining the engineering semantics and surface geometric features within the model leaf block. The higher the probability, the more lightweight the region can be, while the lower the probability, the more high-risk engineering features must be preserved with high precision. This provides a quantitative basis for subsequent grid set division.

[0045] In an optional embodiment, the step of splitting the semantic mesh model into multiple model leaf blocks using the quadtree standard splitting algorithm, and determining the simplified probability value of each model leaf block based on the semantics of each model leaf block, may include the following sub-steps: S121. The semantic grid model is split into multiple model leaf blocks using the standard quadtree splitting algorithm.

[0046] S122. Determine the multi-dimensional feature index parameters for each of the model leaf blocks, wherein the multi-dimensional feature index parameters include: semantic weight ratio, average curvature of the surface, patch spatial density, and fault boundary ratio.

[0047] S123. Normalize and weight the multi-dimensional feature index parameters to obtain the simplified probability value of each model leaf block.

[0048] In one embodiment, semantic mesh bounding box normalization can be performed, and a quadtree splitting reference space can be constructed. Specifically, the coordinates of all triangular facet vertices of the semantic mesh model can be extracted, all vertices can be traversed to solve for the complete mesh 3D AABB bounding box, and the quadtree splitting reference space can be constructed using the triangular facet vertex coordinates.

[0049] Then, the standard quadtree splitting algorithm in MATLAB is used to split the semantic grid model into multiple model leaf blocks in the quadtree splitting reference space.

[0050] It should be noted that during the splitting process, blank nodes without faces can be filtered out, and only valid leaf blocks containing the mine entity mesh can be retained for subsequent probability calculations, thus obtaining the model leaf blocks.

[0051] In a specific operation, the number of facets in the leaf node triList can be read. If the number of facets is 0, it can be determined that the facet is a blank area (empty land in the mine, sky, or area without soil or rock), and it can be discarded directly without being added to the linked list.

[0052] After obtaining multiple model leaf blocks, the normalized bounding box and corresponding physical world coordinate extreme values ​​of each model leaf block can be stored synchronously to provide spatial data for subsequent mesh set boundary extraction. Finally, an ordered linked list of all valid mine mesh model leaf blocks can be stored, with each model leaf block carrying a facet index, spatial bounding box, cross-boundary marker, and basic semantic statistics.

[0053] Next, the four-dimensional feature vector of the model leaf block can be initialized, where the feature vector F=[Srate,Cavg,Dtri,Bfault], and the four dimensions are semantic weight ratio, average surface curvature, patch density, and fault boundary patch ratio, respectively.

[0054] When calculating the semantic weight percentage (Srate) of the model leaf block, all faces of the model leaf block can be traversed, and the pre-stored engineering semantic weights of each face (crack 5.0, support 4.5, fault 4.0, ordinary soil 1.0) can be read. All weights are summed to obtain the total semantic weight (Wsum). Dividing this by the total number of faces within the block gives the semantic weight percentage per face, as shown in the following formula: Srate = Wsum / N.

[0055] N is the total number of internal faces of the block.

[0056] In one embodiment, when calculating the average curvature Cavg of the surface for the semantic weight ratio, the average curvature value stored in each triangular facet can be read. The larger the curvature, the richer the details such as slope folds and cracks. The average value of the curvature of all facets is summed and normalized to the [0,1] interval to obtain the average curvature Cavg of the surface.

[0057] In one embodiment, when calculating the patch spatial density Dtri, the two-dimensional bounding box area of ​​the model leaf block can be calculated, as shown in the following formula: Area=(xmax xmin) (ymax ymin); Next, calculate the spatial density of the patches, as shown in the following formula: Dtri = N / Area.

[0058] Higher density means more detailed meshes in that area, but also a greater cost to simplification.

[0059] In one embodiment, when calculating the fault boundary proportion Bfault, the number of facets semantically encoded as faults within the model leaf blocks, Nfault, can be counted, and then calculated according to the fault boundary proportion, as shown in the following formula: Bfault = Nfault / N.

[0060] The geological structure of fault zones is complex, and the higher the proportion of fault zones, the less suitable it is to simplify them significantly.

[0061] In one embodiment, the semantic weight percentage (Srate), the average curvature of the surface (Cavg), the spatial density of the patch (Dtri), and the fault boundary percentage (Bfault) can all be linearly mapped to [0,1] to eliminate differences in the dimensions of different features and ensure fair weighted calculation. The normalized four-dimensional feature vector is stored in the leaf block extended attributes and permanently cached, eliminating the need for repeated statistical calculations.

[0062] Finally, the simplified probability value can be obtained by weighting and summing the semantic weight ratio Srate, the average curvature of the surface Cavg, the spatial density of the patch Dtri, and the fault boundary ratio Bfault.

[0063] Among them, a weighted summation can be performed using fixed weight coefficients, for example, w1=0.4 (semantic weight ratio), w2=0.2 (average curvature), w3=0.25 (patch density), w4=0.15 (fault ratio); the semantic weight coefficient is the highest, giving priority to engineering features such as cracks and supports.

[0064] Specifically, the simplified probability value can be calculated as follows: Pbase=1 (w1*Srate+w2*Cavg+w3*Dtri+w4*Bfault); By subtracting the weighted sum of features from 1, we can achieve the mapping logic that the higher the feature value, the lower the base probability.

[0065] It should be noted that if the model leaf blocks have boundary patches that cross quadtrees, the correction factor k = 0.85 to avoid mesh cracks caused by oversimplification of the boundary blocks; if there are no boundary patches, k = 1.0, and the simplification probability value after correction can be as follows: Pfix=Pbase k.

[0066] Additionally, if the model leaf block contains cracks or support panels, the upper limit of the constraint is enforced: if Pfix>0.3, directly assign Pfix=0.3, and only slight simplification is allowed in high-value areas; there is no upper limit constraint for ordinary soil blocks.

[0067] Simultaneously, the simplified probability values ​​can be normalized, setting them to 0 for values ​​less than 0 and 1 for values ​​greater than 1, ensuring a uniform numerical range and facilitating subsequent threshold division.

[0068] Finally, a unified quantitative indicator can be output: a simplification probability value P∈[0,1]. The higher the value, the more the leaf blocks of the model can be simplified and the number of faces can be reduced; the lower the value, the more the leaf blocks of the model contain high-risk engineering features, and excessive simplification is prohibited. This is adapted to the business logic of prioritizing the retention of hidden danger details in mine safety inspections.

[0069] Optionally, the final simplified probability value P can be stored in the prob field of the leaf block structure to establish a hash mapping table of "leaf block ID - simplified probability" to support fast batch query.

[0070] After the above processing, each model leaf block carries a unique simplified probability value in the range of 0 to 1, and the probability of high-risk engineering areas is forcibly reduced.

[0071] Optionally, after the block division, feature and probability calculations are completed, all intermediate data can be stored in a structured manner to provide a complete and directly readable leaf block dataset for subsequent processing, avoiding repeated grid traversal and feature calculations, thus saving computing power.

[0072] In practice, all fields can be integrated: unique block ID, quadtree level depth, 2D bounding box (normalized + physical coordinates), patch index array, 4D feature vector, simplified probability value P, cross-boundary marker, high-risk semantic marker, and statistical arrays of various semantic patch counts. Then, binary serialization is used to separate the geometric patch index blocks and feature probability value blocks, storing these two types of data separately. Downstream data can be loaded on demand, reducing memory usage. Finally, a global model leaf block index hash library is constructed for storage.

[0073] This step uses a lightweight two-dimensional quadtree for block division to adapt to the low computing power environment of mining farm edge servers and local industrial control computers; moreover, it performs multi-dimensional geometric feature fusion calculation, which can take into account the curvature of the surface, the grid density, the fault distribution, and the partitioning deviation caused by single semantic judgment.

[0074] S13. Divide the multiple model leaf blocks into a simplified grid set and a precision grid set according to the magnitude of the simplified probability value.

[0075] After the above steps, hundreds to thousands of model leaf blocks with simplified probability values ​​can be obtained.

[0076] In an optional embodiment, in order to quickly distinguish between model leaf blocks with high precision requirements and model leaf blocks with low precision requirements, multiple model leaf blocks can be divided into simplified mesh sets and precision mesh sets by using a fixed threshold division method.

[0077] For example, if the simplification probability value is greater than 0.8, the model leaf blocks are classified as simplifiable; conversely, if the simplification probability value is less than or equal to 0.8, they are classified as non-simplifiable model leaf blocks. Finally, the simplifiable model leaf blocks are grouped into a set to obtain a simplified mesh set, and the non-simplifiable model leaf blocks are grouped into a set to obtain a precision mesh set.

[0078] If a fixed threshold is used to divide the area into a single, simplified grid, slope cracks and tunnel support blocks will be included in the simplified grid set. When viewed in the digital twin backend, the details will be blurry, making it impossible to identify safety hazards. The total number of grid patches and the proportion of high-risk areas vary greatly among different mines, and manually fixed thresholds cannot adapt to the changing scenarios. The probability of high risk is generally higher in open-pit mines with large areas of rock and soil, while the probability of high risk is generally lower in underground tunnels with dense support.

[0079] In a preferred embodiment, the division can be performed using an adaptive threshold, and global simplified probability distribution statistics can be performed. The OTSU adaptive threshold solution solves for the critical threshold T of the division.

[0080] Specifically, the leaf block hash library constructed in the above steps can be traversed to read the prob (simplified probability value 0~1) of all valid model leaf blocks, filter out blank leaf blocks, and generate a one-dimensional probability array Pall=[p1,p2,...,pn]; the total number of leaf blocks Nblock is counted simultaneously.

[0081] The probability interval of 0 to 1 is evenly divided into 100 discrete levels (step size 0.01). The number of leaf blocks in each level is counted by traversing Pall, and a probability distribution histogram Hist

[100] is constructed. The histogram records the sample density of each probability interval, reflecting the overall simplification degree distribution of the mine grid.

[0082] Iterate through all candidate thresholds t∈[0,0.01,...,0.99] and execute the OTSU algorithm: Leaf blocks are divided into two categories with t as the boundary: low-probability blocks C0 (P≤t) and high-probability blocks C1 (P>t); calculate the proportions w0 and w1 of the number of leaf blocks in the two categories, and the probability means μ0 and μ1 of the two categories respectively; then calculate the between-class variance σ2=w0*w1*(μ0 - μ1)² 2 ; After traversal, the t corresponding to the maximum between-class variance is taken as the initial critical threshold Totsu.

[0083] Combined with the mine safety specification, the threshold fluctuation range is limited: the lower threshold limit Tmin=0.2, and the upper threshold limit Tmax=0.6; If Totsu<Tmin, force T=Tmin (there are many high-risk areas, and the simplification range is tightened); If Totsu>Tmax, force T=Tmax (there are many gentle rock and soil areas, and the simplification range is relaxed); The final division critical threshold T is obtained, and then division is performed according to the adaptive threshold, so that the simplified mesh set and the precision mesh set can be obtained.

[0084] Subsequently, the probability histogram and the adaptive threshold T can be stored in the global memory cache, and a log is generated to record the threshold and the number of high and low probability blocks for operation and maintenance to monitor the mesh layering effect.

[0085] At the same time, threshold validity check can be performed, which checks that the number of two types of leaf blocks after division cannot be 0: if all are low-probability blocks (no simplified blocks), automatically increase T by 0.1; if all are high-probability blocks (no high-precision blocks), automatically decrease T by 0.1, ensuring that both sets exist.

[0086] In another alternative embodiment, the simplified mesh set and the precision mesh set can be deduplicated. Wherein, for deduplication of the high-precision precision mesh set, the unordered set Trihigh of the precision mesh set can be converted into an ordered array, and an adjacent comparison deduplication algorithm is adopted: after sorting, the array is traversed, adjacent repeated patch indexes are deleted, the non-repetitive high-precision patch array Finalhigh is obtained, and the precision mesh set is obtained.

[0087] For deduplication of the simplified simplified mesh set, the same sorting and deduplication logic can be executed on the simplified mesh set Trilow, the non-repetitive simplified patch array Finallow is output, and the simplified mesh set is obtained.

[0088] Subsequently, the total number of global patches can be checked, specifically, the total number of patches after merging Nall=Length(Finalhigh)+Length(Finallow) is calculated, and compared with the total number of original mesh patches, the difference is controlled within the allowable error (only a small number of abnormal patches are filtered), and a log alarm is triggered if the difference is too large.

[0089] S14. Using the boundary of the precision grid set as a constraint, the simplified grid set is simplified using the QEM quadratic error metric algorithm, and the simplified region is sampled to obtain simplified region data.

[0090] In one embodiment, traditional global LOD simplification does not distinguish boundaries, and a large number of vertices at the intersection of the precision mesh and the simplified mesh are deleted. Subsequent rendering will show cracks, faults, steps, and visual breaks in slope support and tunnel cracks. Moreover, the simplified mesh loses its original semantic attributes and cannot distinguish between soil, support, and cracks on the twin platform to generate color warnings.

[0091] Therefore, the simplified grid set can be simplified by using the boundary of the precision grid set as a constraint and the QEM quadratic error metric algorithm. Then, the simplified region can be sampled to obtain simplified region data.

[0092] Using the boundary of the precision mesh set as a constraint, the closer to the boundary of the precision mesh set, the less simplification is required; the farther away from the boundary of the precision mesh set, the more simplification can be required.

[0093] This simplification method avoids the extensive deletion of vertices at the boundary between the precision mesh and the simplified mesh, thereby reducing the probability of rendering distortion.

[0094] In one embodiment, the step of simplifying the simplified grid set using the boundary of the precision grid set as a constraint, and sampling the simplified region to obtain simplified region data using the QEM quadratic error metric algorithm, may include the following sub-steps: S141. Extract the boundary vertices of the precision mesh set and determine the distance between each triangular facet and the boundary vertex in the simplified mesh set.

[0095] S142. Assign a corresponding boundary weight value to each triangular facet in the simplified mesh set according to the interval distance.

[0096] S143. The QEM quadratic error metric algorithm is used to perform edge folding simplification on the facets of the simplified mesh set based on the boundary weight value, and the simplified region is sampled to obtain simplified region data.

[0097] If all faces of the simplified mesh are simplified at the same cost, a large number of vertices in the boundary region will be folded and deleted, causing a visual separation between the high-precision entity and the surrounding soil and rock. Differentiated protection weights need to be set according to the distance from the boundary of the precision mesh, and the closer to the boundary, the more prohibited the large simplification should be.

[0098] In one embodiment, boundary constraint weights can be initialized first to distinguish between the boundary protection zone and the conventional simplified zone.

[0099] Specifically, the boundary vertex set Vborder, boundary bounding box BBinter, and boundary edge Edgeinter of the precision mesh set can be read. A boundary vertex KD-Tree spatial index can be established through the boundary vertex set Vborder, boundary bounding box BBinter, and boundary edge Edgeinter to achieve fast query of the distance from any face vertex to the boundary (O(logn)).

[0100] In one operation mode, after obtaining the precision mesh set Finalhigh and the simplified mesh set Finallow, the boundary edge Edgeinter, boundary vertex Vborder, and boundary region bounding box of the precision mesh set Finalhigh can be obtained.

[0101] Among them, the interface boundary edge is to collect all the outer boundary edges of the precision mesh set and store them in the boundary edge list Edgeborder. Iterate through each boundary edge of Edgeborder (the boundary between the precision mesh and the simplified mesh), and determine whether there are simplified mesh patches shared by the two vertices of the edge. If there are, it is marked as the interface boundary and stored separately in the interface boundary edge Edgeinter as the core constraint boundary.

[0102] Similarly, we can traverse the boundary edges of the interface, extract the 3D coordinates of all boundary vertices, store them in the boundary vertex array, obtain the boundary vertex Vborder, and uniformly store the original coordinates of the world coordinate system as the basis for resampling distance judgment.

[0103] Based on the extreme values ​​of the Vborder vertices, an AABB bounding box can be generated for the boundary region, resulting in a bounding box. The subsequent LOD simplification only applies boundary protection weights to the faces within this bounding box, reducing computational power consumption.

[0104] Next, two levels of protection distance thresholds can be defined. Based on the actual dimensions of the mine, configure the distance parameters (unit: m): Level 1 strong protection zone: distance from boundary d≤0.5m; Level 2 weak protection zone: 0.5m<distance from boundary d≤2m; Conventional simplified zone: distance from boundary d>2; For fine areas such as roadways and slopes, the threshold can be dynamically reduced, while for open-pit flat rock and soil, the threshold can be increased.

[0105] After setting a distance threshold, you can traverse all triangles in the simplified mesh set Finallow to determine the distance between each triangle in the simplified mesh set and the boundary vertices of the precision mesh set Finalhigh.

[0106] Suppose that for each facet, take 3 vertices, calculate the Euclidean distance dv from each vertex to the nearest boundary vertex using KD-Tree, and take the minimum vertex distance dmin=min(dv1,dv2,dv3) of the facet as the interval distance of the facet boundary.

[0107] After calculating the interval distance, a corresponding boundary weight value can be assigned to each triangular facet within the simplified mesh set according to the interval distance. This boundary weight value is a hierarchical allocation of the boundary protection weight wprotect. The boundary weight values ​​can be as follows: For strong protected areas, the interval distance dmin ≤ 0.5m: wprotect = 5; For weak protection zones, 0.5m < interval distance dmin ≤ 2m: wprotect = 2; For regular areas, with an interval distance dmin > 2m: wprotect = 1.

[0108] Optionally, if there are fault or support-type precision mesh patches in the neighborhood of a patch, the corresponding boundary weight value wprotect is increased by an additional 1.5 to further reduce the boundary simplification intensity and preserve the geological and support details.

[0109] Next, the boundary weight value `wprotect` can be bound to the attribute structure of each facet in the simplified mesh set, and bound one-to-one with the facet semantic encoding, curvature, and index, for real-time reading by the QEM simplification algorithm. This allows the output of a complete simplified mesh facet dataset with hierarchical boundary protection weight labels and a boundary KD-Tree index cache.

[0110] In one embodiment, the QEM quadratic error metric algorithm can be used to perform edge folding simplification on the facets of the simplified mesh region based on the boundary weight value, and the simplified region can be sampled to obtain simplified region data.

[0111] Edge simplification can involve folding and deleting edges; the larger the boundary weight value, the more difficult it is to fold and delete. After deletion, the data can be reprocessed to obtain the cloud point data of each facet within the simplified region, thus obtaining the simplified region data.

[0112] In one embodiment, the step of using the QEM quadratic error metric algorithm to perform edge folding simplification on the patches of the simplified mesh set based on the boundary weight values, and sampling the simplified region to obtain simplified region data, may include the following sub-steps: S1431. Sort the patches of the simplified mesh set according to the boundary weight values ​​to obtain a patch queue; Based on the facet queue, traverse all triangular edges of the simplified mesh set and calculate the total distortion cost of each triangular edge after folding.

[0113] S1432. Fold the edges of the simplified mesh set according to the total distortion cost to obtain the simplified mesh region.

[0114] S1433. After optimizing the simplified patch region to obtain the optimized patch region, the optimized patch region is sampled to obtain simplified region data. The optimization process includes repairing the holes in the simplified patch region, removing overlapping patches in the simplified patch region, and verifying the topology of the simplified patch region.

[0115] Existing technologies only use geometric errors as the standard for ordinary edge folding, and without boundary constraints, they will directly delete key vertices at the intersection. This solution integrates boundary protection weights into the QEM error matrix, achieving "significant simplification of the hinterland and lightweight simplification of the intersection".

[0116] In one embodiment, a simplified mesh vertex error matrix QEM can be initialized. Specifically, for each vertex in the simplified mesh set, an initial 4×4 quadratic error matrix Q is calculated based on the neighborhood patch plane equation. The original error represents the degree of geometric distortion caused by deleting the vertex.

[0117] Next, the boundary weight values ​​are corrected based on the quadratic error matrix Q to obtain the corrected weight values, as shown in the following formula: Qfinal = Q * wprotect; Here, Qfinal is the adjusted weight value.

[0118] The greater the boundary protection weight, the higher the overall error value of the vertices. The algorithm will prioritize vertices with small errors and far from the boundary to perform folding, and automatically avoid boundary vertices.

[0119] Then, the faces of the simplified mesh set are sorted according to the boundary weight values ​​to obtain a face queue. Based on the order of the face queue, all triangular edges of the simplified mesh set are traversed, and the total distortion cost after folding each edge is calculated. The edge, the two end vertices, and the folding cost are stored in a min-heap, and the top of the heap is always the edge with the lowest current cost and the most suitable for deletion.

[0120] At the same time, a global threshold for the number of simplified facets can be set, and the compression ratio can be configured according to the computing power of the rendering platform. For example, open-air rock and soil areas can be compressed to 20%~40% of the original facets; underground minor rock and soil areas can be compressed to 30%~50%; the upper limit of compression for weak protection zones is 70%; and strong protection boundary areas are prohibited from being compressed to below 80%.

[0121] The edge folding iteration is then performed again, and the lowest-cost edge at the top of the heap is extracted to complete the vertex merging, as follows: 1. Merge two vertices to generate a new vertex; 2. Update the neighborhood patch and vertex error matrix; 3. Recalculate the folding cost of the affected edges and update the min-heap; Loop termination condition: The total number of facets reaches the preset compression threshold or the cost of folding the remaining edges in the heap exceeds the distortion limit.

[0122] Optionally, a read-only blacklist of boundary vertices can be established. If any end of the edge to be folded belongs to the Vborder boundary vertex, the edge is skipped directly and will never participate in the folding, thus completely preventing the loss of boundary contours.

[0123] Through the above operations, a lightweight and simplified intermediate mesh can be output, resulting in a simplified patch region. The simplified patch region retains the complete boundary contour, while the surface patches are significantly reduced, and each vertex carries the original semantic label.

[0124] After obtaining the simplified patch region, the simplified patch region can be optimized to obtain the optimized patch region. Then, the optimized patch region is sampled to obtain the simplified region data. The optimization process includes repairing the holes in the simplified patch region, removing overlapping patches in the simplified patch region, and verifying the topology of the simplified patch region.

[0125] Because the simplified mesh is easily damaged by edge folding, single-sided holes, non-manifold edges, and overlapping patches can easily appear at the boundary between the simplified mesh and the precision mesh. This can cause holes, flickering, and model tearing in the twin image, failing to meet the requirements for mine safety visualization. Therefore, holes in the simplified patch area can be repaired to eliminate boundary holes and non-manifold patches.

[0126] Using the BBinter boundary bounding box as the clipping range, only simplified mesh faces within this range are extracted for topology checking, reducing the computational overhead of global traversal. Next, all edges in the boundary region are traversed, and the number of faces associated with each edge is counted. If an edge is associated with 0 faces, it is identified as an isolated free edge and deleted directly; if it is associated with 3 or more faces, it is identified as a non-manifold topology and marked for repair. Simultaneously, isolated triangular faces without adjacent faces are counted and stored in an anomaly list.

[0127] For the detection and repair of micro-holes, the grid boundary loop can be traversed to filter out micro-holes with a perimeter of less than 1.5m (gaps caused by simplification). The algorithm based on minimum area triangle filling is used to automatically generate regular triangular facets to close the holes.

[0128] For local subdivision repair of non-manifold edges, subdivision vertices can be inserted at the midpoint of the non-manifold boundary edges to split overlapping patches, ensuring that each edge is associated with only 1 to 2 patches, thus restoring the manifold topology.

[0129] For overlapping facet fusion and culling, spatial raster comparison can be used to detect duplicate faces with completely overlapping coordinates, retain one of them, and delete redundant overlapping faces to avoid rendering ghosting.

[0130] Optionally, after the optimization operation is completed, a second topology check can also be performed. Specifically, the topological relationship of the boundary edges can be retraced to confirm that there are no holes, no isolated facets, and no non-manifold structures. If the check fails, the holes are filled by iterative subdivision until compliance is achieved.

[0131] After completing the above operations, a lightweight simplified region mesh with continuous topology, no holes or tears, and complete boundary contours can be output.

[0132] Next, data from a lightweight simplified region grid can be collected to obtain simplified region data (including geometric buffers, semantic attribute buffers, and topological boundary metadata).

[0133] Optionally, the simplified mesh geometry, semantics, boundary markers, normal vectors and other types of data need to be packaged in a unified manner and aligned with the high-precision mesh data format to facilitate subsequent GPU buffer merging; at the same time, the original semantics are preserved to support coloring and hazard annotation functions.

[0134] This application removes the bottom-level constraint boundary vertices, completely resolving the issues of high-precision support, fault and surrounding soil and rock crack rendering, and adapting to the core needs of mine safety monitoring. Furthermore, it simplifies and compresses geometry while fully preserving the engineering semantics of the surface patches, supporting functions such as hazard highlighting and data traceability.

[0135] S15. Visualize and render the simplified regional data and the precision grid set to obtain a virtual model of the mine.

[0136] Since the simplified regional data and the precision grid set belong to two separate buffers, direct rendering will result in spatial misalignment and layering; it is necessary to complete vertex alignment based on a unified mine geodetic coordinate system, merge buffers and distinguish rendering layers.

[0137] Next, the visualization rendering engine can be invoked to render the merged mesh, thereby obtaining a virtual model of the mine.

[0138] In one embodiment, the step of performing visualization rendering based on the simplified region data and the precision grid set to obtain a virtual model of the mine may include the following sub-steps: S151. Unify the spatial coordinates of the simplified region data and the precision grid set, and stitch the simplified region data and the precision grid set together according to the spatial coordinates to obtain the model to be displayed.

[0139] S152. Load a visualization texture according to the semantics of each grid of the model to be displayed, and perform visualization rendering on the model to be displayed according to the visualization texture to obtain a virtual model of the mine.

[0140] In one embodiment, a precision mesh and a simplified mesh are spatially aligned and stitched together to construct a complete scene geometry cache.

[0141] In one operation mode, the high-precision patch corresponding to Finalhigh in the semantic mesh model can be read to generate high-precision mesh independent VBO, IBO, and ABO attribute buffers without any simplification or modification.

[0142] Extract the vertex extreme bounding boxes of the simplified regional data and the precision grid set respectively, compare them with the global mine reference GNSS coordinates, and verify the offset error of the two sets of grid coordinates; if there is a millimeter-level offset, correct the vertex coordinates of the simplified grid based on the unified translation matrix.

[0143] Add a layer marker bit to the attribute buffer: 0 = high-precision layer (cracks, support, fault), 1 = lightweight simplified layer (ordinary soil and rock). The GPU rendering pipeline can be scheduled in layers.

[0144] Next, vertex buffer merging and partitioning can be performed. For GPU memory partitioning, the first half of the buffer stores high-precision mesh vertices, and the second half stores simplified mesh vertices. Partition offsets are set to avoid the two sets of data from overwriting each other.

[0145] Simultaneously, the global patch index array can be reconstructed, first arranging all high-precision patch indices, then arranging simplified patch indices, recording the layer start offset subscript, and realizing layered batch drawing.

[0146] By fusing the vertex extrema of two sets of meshes, a complete bounding box for the mining scene is generated, which is used for subsequent frustum culling and camera near-far clipping calculations. This allows the output of a complete geometric cache of the mining scene, complete with spatial alignment, layer marking, and buffer merging, resulting in the model to be displayed.

[0147] Undifferentiated mesh coloring makes it impossible to intuitively distinguish engineering objects such as slope cracks, support structures, and soil and rock foundations. Digital twin platforms require different semantic entities to be matched with exclusive colors, textures, and normal details to achieve visualized highlighting of safety hazards. To address this, layered semantic texture mapping and material coloring configuration can be implemented.

[0148] Specifically, a mining semantic material library can be loaded. This library includes predefined semantic-material mapping rules, which can be as follows: 01 Crack: High-gloss red material, high-intensity normal texture, semi-transparent outline; 02 Support Anchor Bolt / Steel Frame: Metallic gray material, metallic reflective texture; 03 Fault: Yellow-orange layered texture; 04~08 Ordinary Soil and Rock: Basic sand and gravel texture, simplified mesh using low-resolution textures.

[0149] For the precision patch marked 0 in layer, load a 4K high-precision normal map and roughness map to restore the crack bumps and steel frame structure details; enable texture anisotropy filtering to ensure no blurring when viewed up close.

[0150] For the simplified facet marked 1 in layer, a 1K base texture is used, high-precision normal mapping is turned off, and only the base color texture is retained to reduce video memory usage.

[0151] Simultaneously, a GPU fragment shader was written to read the semantic encoding of the facets and match the corresponding RGB colors, transparency, and specular parameters in real time; dynamic flashing highlight logic was added to crack faces for early warning of potential hazards in the mining area. Two sets of texture interpolation were mixed in the facets at the boundary to eliminate abrupt changes in texture color between the high-precision area and the simplified area, resulting in a natural visual transition.

[0152] Finally, all material textures are uploaded to the GPU texture unit in advance, establishing a mapping between texture IDs and semantic codes, eliminating the need to read disk texture files in real time during rendering. This allows for the output of a complete scene rendering resource package that binds layered materials, semantic maps, and hidden highlighting shading logic. Finally, based on this complete scene rendering resource package, a visual rendering can be performed, outputting an interactive virtual mine model.

[0153] This application uses high-precision area loading of high-definition detail maps and lightweight textures for ordinary soil and rock, which significantly reduces video memory usage.

[0154] In this embodiment, this application provides a virtual model rendering method based on mine environment information. Its advantages are as follows: This application acquires mine environment information and constructs a semantic mesh model using this information; it uses a quadtree standard splitting algorithm to divide the semantic mesh model into multiple model leaf blocks, and determines the simplification probability value of each model leaf block based on its semantics; it divides the multiple model leaf blocks into a simplified mesh set and a precision mesh set according to the magnitude of the simplification probability value; using the boundary of the precision mesh set as a constraint, it uses the QEM quadratic error metric algorithm to simplify the simplified mesh set and samples the simplified region to obtain simplified region data; it performs visualization rendering based on the simplified region data and the precision mesh set to obtain a mine virtual model. This application can simplify some triangular faces, eliminating the need for visualization rendering of a large number of faces, thus shortening rendering time and improving rendering efficiency; moreover, this application uses high-precision mesh faces as constraints for simplification, which can reduce the probability of distorted meshes after simplification, thereby avoiding rendering distortion and improving the rendering effect.

[0155] This application also provides a virtual model rendering device based on mine environment information. See [link to documentation]. Figure 2 The diagram shows a schematic representation of a virtual model rendering device based on mine environment information according to an embodiment of this application.

[0156] As an example, the virtual model rendering device based on mine environment information may include: The construction module 201 is used to acquire mine environment information and use the mine environment information to construct a semantic mesh model, wherein the semantic mesh model is a three-dimensional model with added semantic labels to triangular facets; The determination module 202 is used to split the semantic grid model into multiple model leaf blocks using the quadtree standard splitting algorithm, and to determine the simplified probability value of each model leaf block based on the semantics of each model leaf block; The partitioning module 203 is used to divide the multiple model leaf blocks into a simplified grid set and a precision grid set according to the magnitude of the simplified probability value; The simplification module 204 is used to simplify the simplified grid set using the boundary of the precision grid set as a constraint condition, and to sample the simplified region to obtain simplified region data. The rendering module 205 is used to perform visualization rendering based on the simplified area data and the precision grid set to obtain a virtual model of the mine.

[0157] Optionally, the step of simplifying the simplified grid set using the boundary of the precision grid set as a constraint, and sampling the simplified region to obtain simplified region data, includes: Extract the boundary vertices of the precision mesh set, and determine the spacing between each triangular facet and the boundary vertex within the simplified mesh set; Assign a corresponding boundary weight value to each triangular facet within the simplified mesh set according to the interval distance; The QEM quadratic error metric algorithm is used to perform edge folding simplification on the facets of the simplified mesh set based on the boundary weight values, and the simplified region is sampled to obtain simplified region data.

[0158] Optionally, the step of using the QEM quadratic error metric algorithm to perform edge folding simplification on the patches of the simplified mesh set based on the boundary weight values, and sampling the simplified region to obtain simplified region data, includes: The simplified mesh set is sorted according to the boundary weight values ​​to obtain a patch queue; Based on the patch queue, traverse all triangular edges of the simplified mesh set and calculate the total distortion cost of each triangular edge after folding; Based on the total distortion cost, the simplified mesh set is folded edge by edge to obtain the simplified mesh region; After optimizing the simplified patch region to obtain the optimized patch region, the optimized patch region is sampled to obtain simplified region data. The optimization process includes repairing holes in the simplified patch region, removing overlapping patches in the simplified patch region, and verifying the topology of the simplified patch region.

[0159] Optionally, the step of performing visualization rendering based on the simplified regional data and the precision grid set to obtain a virtual model of the mine includes: Unify the spatial coordinates of the simplified region data and the precision grid set, and then stitch the simplified region data and the precision grid set together according to the spatial coordinates to obtain the model to be displayed; Based on the semantics of each grid in the model to be displayed, a visual texture is loaded, and the model to be displayed is rendered visually based on the visual texture to obtain a virtual model of the mine.

[0160] Optionally, the step of splitting the semantic grid model into multiple model leaf blocks using the standard quadtree splitting algorithm, and determining the simplified probability value of each model leaf block based on the semantics of each model leaf block, includes: The semantic grid model is split into multiple model leaf blocks using the standard quadtree splitting algorithm; Determine the multi-dimensional feature index parameters for each leaf block of the model, wherein the multi-dimensional feature index parameters include: semantic weight ratio, average curvature of the surface, patch spatial density, and fault boundary ratio; The multi-dimensional feature index parameters are normalized and weighted to obtain the simplified probability value of each model leaf block.

[0161] Optionally, the step of acquiring mine environment information and using the mine environment information to construct a semantic mesh model includes: Obtain the mining environment information uploaded by the construction personnel, wherein the mining environment information includes: spatial geometric data, engineering attribute data and semantically labeled raw data; The mining environment information is preprocessed to obtain mining processed data, wherein the preprocessing includes noise reduction, coordinate alignment and duplicate point removal; A mining grid model is constructed based on the mining data using the Poisson reconstruction algorithm, and a unique digital semantic code is bound to each triangular facet of the mining grid model to obtain a semantic grid model.

[0162] This application also provides a virtual model rendering system based on mine environment information; see [link to relevant documentation]. Figure 3 The diagram shows a schematic representation of a virtual model rendering system based on mine environment information provided in an embodiment of this application.

[0163] As an example, the virtual model rendering system based on mine environment information may include: a rendering processing platform and several visualization terminals. The rendering platform is communicatively connected to several visualization terminals, and the rendering platform is applicable to the virtual model rendering method based on mine environment information as described in the above embodiments.

[0164] The visualization terminal can be a smart terminal for construction workers, such as a tablet computer.

[0165] Reference Figure 4 The computer device shown in this application is a virtual model rendering method based on mine environment information, and may specifically include the following: The computer device 12 described above is in the form of a general-purpose computing device. The components of the computer device 12 may include, but are not limited to: one or more processors or processing units 16, system memory 28, and bus 18 connecting different system components (including system memory 28 and processing unit 16).

[0166] Bus 18 refers to one or more of several types of bus 18 architectures, including memory bus 18 or memory controller, peripheral bus 18, graphics acceleration port, processor, or local bus 18 using any of the various bus 18 architectures. For example, these architectures include, but are not limited to, Industry Standard Architecture (ISA) bus 18, Micro Channel Architecture (MAC) bus 18, Enhanced ISA bus 18, Audio / Video Electronics Standards Association (VESA) local bus 18, and Peripheral Component Interconnect (PCI) bus 18.

[0167] Computer device 12 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by computer device 12, including volatile and non-volatile media, removable and non-removable media.

[0168] System memory 28 may include computer system readable media in the form of volatile memory, such as random access memory (RAM) 30 and / or cache memory 32. Computer device 12 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, storage system 34 may be used to read and write non-removable, non-volatile magnetic media (commonly referred to as a "hard disk drive"). Although Figure 4 As not shown, a disk drive for reading and writing to a removable non-volatile disk (such as a "floppy disk") and an optical disk drive for reading and writing to a removable non-volatile optical disk (such as a CD-ROM, DVD-ROM, or other optical media) may be provided. In these cases, each drive may be connected to bus 18 via one or more data media interfaces. The memory may include at least one program product having a set (e.g., at least one) of program modules 42 configured to perform the functions of the embodiments of this application.

[0169] A program / utility 40 having a set (at least one) of program modules 42 may be stored, for example, in memory. Such program modules 42 include—but are not limited to—an operating system, one or more application programs, other program modules 42, and program data. Each or some combination of these examples may include an implementation of a network environment. Program modules 42 typically perform the functions and / or methods described in the embodiments of this application.

[0170] Computer device 12 can also communicate with one or more external devices 14 (e.g., keyboard, pointing device, display 24, camera, etc.), and with one or more devices that enable a user to interact with the computer device 12, and / or with any device that enables the computer device 12 to communicate with one or more other computing devices (e.g., network card, modem, etc.). This communication can be performed through input / output (I / O) interface 22. Furthermore, computer device 12 can also communicate with one or more networks (e.g., local area network (LAN)), wide area network (WAN), and / or public networks (e.g., the Internet) via network adapter 20. As shown, network adapter 20 communicates with other modules of computer device 12 via bus 18. It should be understood that, although... Figure 4 Not shown, it can be combined with computer device 12 to use other hardware and / or software modules, including but not limited to: microcode, device drivers, redundant processing unit 16, external disk drive array, RAID system, tape drive and data backup storage system 34, etc.

[0171] The processing unit 16 executes various functional applications and data processing by running programs stored in the system memory 28, such as implementing the virtual model rendering method based on mine environment information provided in the embodiments of this application.

[0172] That is, when the processing unit 16 executes the above program, it achieves the following: Obtain mine environment information and use the mine environment information to construct a semantic mesh model, wherein the semantic mesh model is a three-dimensional model with added semantic labels to triangular facets; The semantic grid model is split into multiple model leaf blocks using the standard quadtree splitting algorithm, and the simplified probability value of each model leaf block is determined based on the semantics of each model leaf block. Based on the magnitude of the simplified probability value, the multiple model leaf blocks are divided into a simplified grid set and a precision grid set; Using the boundary of the precision grid set as a constraint, the simplified grid set is simplified using the QEM quadratic error metric algorithm, and the simplified region is sampled to obtain simplified region data; A virtual model of the mine is obtained by visual rendering based on the simplified regional data and the precision grid set.

[0173] That is, to implement the following when the program is executed by the processor: Obtain mine environment information and use the mine environment information to construct a semantic mesh model, wherein the semantic mesh model is a three-dimensional model with added semantic labels to triangular facets; The semantic grid model is split into multiple model leaf blocks using the standard quadtree splitting algorithm, and the simplified probability value of each model leaf block is determined based on the semantics of each model leaf block. Based on the magnitude of the simplified probability value, the multiple model leaf blocks are divided into a simplified grid set and a precision grid set; Using the boundary of the precision grid set as a constraint, the simplified grid set is simplified using the QEM quadratic error metric algorithm, and the simplified region is sampled to obtain simplified region data; A virtual model of the mine is obtained by visual rendering based on the simplified regional data and the precision grid set.

[0174] Any combination of one or more computer-readable media may be used. A computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. A computer-readable storage medium can be, for example—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples (a non-exhaustive list) of computer-readable storage media include: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this document, a computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in connection with an instruction execution system, apparatus, or device.

[0175] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including—but not limited to—electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, capable of transmitting, propagating, or transmitting programs for use by or in connection with an instruction execution system, apparatus, or device.

[0176] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof. These programming languages ​​include object-oriented programming languages—such as Java, Smalltalk, and C++—and conventional procedural programming languages—such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider). The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably.

[0177] Although preferred embodiments of the present application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the embodiments of the present application.

[0178] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes said element.

[0179] The above provides a detailed description of the virtual model rendering method, system, device, and medium based on mine environment information provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A virtual model rendering method based on mine environment information, characterized in that, The method includes: Obtain mine environment information and use the mine environment information to construct a semantic mesh model, wherein the semantic mesh model is a three-dimensional model with added semantic labels to triangular facets; The semantic grid model is split into multiple model leaf blocks using the standard quadtree splitting algorithm, and the simplified probability value of each model leaf block is determined based on the semantics of each model leaf block. Based on the magnitude of the simplified probability value, the multiple model leaf blocks are divided into a simplified grid set and a precision grid set; Using the boundary of the precision grid set as a constraint, the simplified grid set is simplified using the QEM quadratic error metric algorithm, and the simplified region is sampled to obtain simplified region data; A virtual model of the mine is obtained by visual rendering based on the simplified regional data and the precision grid set.

2. The virtual model rendering method based on mine environment information according to claim 1, characterized in that, The process of simplifying the simplified grid set using the boundary of the precision grid set as a constraint, and sampling the simplified region to obtain simplified region data, includes: Extract the boundary vertices of the precision mesh set, and determine the spacing between each triangular facet and the boundary vertex within the simplified mesh set; Assign a corresponding boundary weight value to each triangular facet within the simplified mesh set according to the interval distance; The QEM quadratic error metric algorithm is used to perform edge folding simplification on the facets of the simplified mesh set based on the boundary weight values, and the simplified region is sampled to obtain simplified region data.

3. The virtual model rendering method based on mine environment information according to claim 2, characterized in that, The QEM quadratic error metric algorithm is used to perform edge folding simplification on the patches of the simplified mesh set based on the boundary weight values, and the simplified region is sampled to obtain simplified region data, including: The simplified mesh set is sorted according to the boundary weight values ​​to obtain a patch queue; Based on the patch queue, traverse all triangular edges of the simplified mesh set and calculate the total distortion cost of each triangular edge after folding; Based on the total distortion cost, the simplified mesh set is folded edge by edge to obtain the simplified mesh region; After optimizing the simplified patch region to obtain the optimized patch region, the optimized patch region is sampled to obtain simplified region data. The optimization process includes repairing holes in the simplified patch region, removing overlapping patches in the simplified patch region, and verifying the topology of the simplified patch region.

4. The virtual model rendering method based on mine environment information according to claim 1, characterized in that, The step of performing visualization rendering based on the simplified regional data and the precision grid set to obtain a virtual model of the mine includes: Unify the spatial coordinates of the simplified region data and the precision grid set, and then stitch the simplified region data and the precision grid set together according to the spatial coordinates to obtain the model to be displayed; Based on the semantics of each grid in the model to be displayed, a visual texture is loaded, and the model to be displayed is rendered visually based on the visual texture to obtain a virtual model of the mine.

5. The virtual model rendering method based on mine environment information according to claim 1, characterized in that, The step of splitting the semantic grid model into multiple model leaf blocks using the standard quadtree splitting algorithm, and determining the simplified probability value of each model leaf block based on the semantics of each model leaf block, includes: The semantic grid model is split into multiple model leaf blocks using the standard quadtree splitting algorithm; Determine the multi-dimensional feature index parameters for each leaf block of the model, wherein the multi-dimensional feature index parameters include: semantic weight ratio, average curvature of the surface, patch spatial density, and fault boundary ratio; The multi-dimensional feature index parameters are normalized and weighted to obtain the simplified probability value of each model leaf block.

6. The virtual model rendering method based on mine environment information according to any one of claims 1-5, characterized in that, The process of acquiring mine environment information and using that information to construct a semantic grid model includes: Obtain the mining environment information uploaded by the construction personnel, wherein the mining environment information includes: spatial geometric data, engineering attribute data and semantically labeled raw data; The mining environment information is preprocessed to obtain mining processed data, wherein the preprocessing includes noise reduction, coordinate alignment and duplicate point removal; A mining grid model is constructed based on the mining data using the Poisson reconstruction algorithm, and a unique digital semantic code is bound to each triangular facet of the mining grid model to obtain a semantic grid model.

7. A virtual model rendering device based on mine environment information, characterized in that, The device includes: A construction module is used to acquire mine environment information and use the mine environment information to construct a semantic mesh model, wherein the semantic mesh model is a three-dimensional model with added semantic labels to triangular facets; The determination module is used to split the semantic grid model into multiple model leaf blocks using the standard quadtree splitting algorithm, and to determine the simplified probability value of each model leaf block based on the semantics of each model leaf block; A partitioning module is used to divide multiple model leaf blocks into a simplified grid set and a precision grid set according to the magnitude of the simplified probability value; The simplification module is used to simplify the simplified grid set using the boundary of the precision grid set as a constraint condition, and to sample the simplified region to obtain simplified region data. The rendering module is used to perform visualization rendering based on the simplified area data and the precision grid set to obtain a virtual model of the mine.

8. A virtual model rendering system based on mine environment information, characterized in that, The system includes: a rendering processing platform and several visualization terminals; The rendering platform is communicatively connected to several visualization terminals, and the rendering platform is applicable to the virtual model rendering method based on mine environment information as described in any one of claims 1-6.

9. An electronic device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, when the processor executes the computer program, it implements the virtual model rendering method based on mining environment information as described in any one of claims 1-6.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer-executable program, which is used to cause a computer to execute the virtual model rendering method based on mining environment information as described in any one of claims 1-6.

Images