Augmented reality observation method and system for physical simulation test of rock burst in deep engineering

Abstract

This invention discloses an augmented reality observation method and system for physical simulation experiments of rockbursts in deep engineering, belonging to the interdisciplinary field of augmented reality and deep engineering. This invention proposes a novel observation paradigm that overcomes the aforementioned limitations. In physical simulation experiments of rockburst disasters in deep engineering, it achieves real-time fusion of multi-dimensional data on the rockburst disaster gestation process, enabling direct observation of the entire rockburst gestation process and multi-perspective phenomena. Ultimately, it achieves on-demand insight into all elements of the rockburst gestation process—"temporally continuous and spatially covered," "gestation-triggering-destruction," and "geology-environment-engineering-monitoring"—throughout all time and space. This invention aims to construct an immersive, spatially accurate, and interactively controllable augmented reality observation method for physical simulation experiments of deep engineering disasters, overcoming the triple bottlenecks of traditional methods in terms of process concealment, spatial inaccuracy, and passive analysis. Ultimately, it provides a scientific paradigm for in-depth exploration of the rockburst disaster gestation process through visualization observation.

Description

Technical Field

[0001] This invention belongs to the field of interdisciplinary technology of augmented reality and deep engineering, and in particular relates to an augmented reality observation method and system for rockburst physics simulation experiments in deep engineering. Background Technology

[0002] With the increasing depletion of shallow resources, the scale and depth of deep engineering projects, such as deep-buried tunnels and deep metal mines, are growing daily. Compared to shallow environments, deep earth environments present more complex geological conditions, including high stress, high permeability, and high temperature, as well as extremely active tectonic zones, becoming a bottleneck for deep engineering projects. To overcome this bottleneck, my country has vigorously carried out scientific research on physical simulation of disasters in deep engineering projects. The aim is to reproduce extreme conditions in deep environments through laboratory settings, realistically replicate the disaster gestation process in deep engineering projects, and further explore and reveal the gestation mechanisms and disaster patterns of major disasters in deep engineering projects.

[0003] Rockbursts, as a highly destructive and widespread extreme geological hazard in deep engineering, require careful study of their formation mechanism. Therefore, conducting physical simulation experiments of rockburst hazards in deep engineering is a key approach to exploring their formation process and underlying mechanisms. Typical physical simulation experiments of deep engineering hazards rely on large-scale experimental equipment, producing geological physical model samples that are enormous, often reaching meter-scale. Within such large and completely opaque physical models, researchers need to simultaneously apply high-intensity triaxial geostress and conduct complex engineering operations such as miniature excavation in real time. However, the instantaneous nature, high destructiveness, and complexity of the formation process of rockburst hazards make it difficult for researchers to clearly and accurately capture, observe, and record their dynamic formation process within opaque physical models, significantly increasing the difficulty of scientific research on the formation mechanism of rockburst hazards. Therefore, there is an urgent need for a physical simulation test observation method for rockburst disasters in deep engineering, so as to achieve dynamic and transparent observation of the rockburst incubation process, accurate reproduction of key rockburst cases, and in-depth mining of scientific information such as evolution laws and incubation mechanisms, thereby providing strong technical support for revealing the incubation mechanism of rockburst disasters.

[0004] However, current deep engineering observation methods are insufficient to meet the needs of the aforementioned physical simulation experiments, failing to achieve a fusion of virtual and real observations of engineering disturbances and internal surrounding rock responses. In deep engineering disaster physical simulation experiments, rockburst initiation is often the result of a complex coupling between engineering activities such as excavation and the evolution of surrounding rock stress and internal rock mass fracturing. Existing observation methods face severe observational fragmentation. Researchers can only capture the macroscopic deformation and eventual failure of physical model samples of geological bodies using equipment such as cameras, while information reflecting the internal initiation process, such as microseismic data and internal fracturing, can only be passively displayed as abstract two-dimensional charts or independent three-dimensional digital models detached from the physical entity. This fragmentation of observation methods prevents the intuitive integration of external macroscopic engineering activities and internal surrounding rock disaster responses under the same spatiotemporal reference. Furthermore, due to the lack of transparent observation methods for the interior of the sample, researchers cannot observe the internal microscopic evolution through the physical entity outside the physical model sample of the geological body. This makes it difficult for traditional observation methods to intuitively and dynamically capture and correlate the evolution and internal connections of the entire chain of rockburst "gestation-triggering-destruction" process on a spatiotemporal scale, greatly limiting the scientific exploration of the gestation mechanism of rockburst disasters in deep engineering. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides an augmented reality observation method and system for deep engineering rockburst physical simulation experiments. Through a closed-loop technology of "full-chain dynamic virtual effect generation - high-precision virtual-real fusion and overlay - intelligent interactive analysis," it aims to construct an immersive, spatially accurate, and interactively controllable augmented reality observation method for deep engineering disaster physical simulation experiments. This method overcomes the triple bottlenecks of traditional methods in terms of process concealment, spatial inaccuracy, and passive analysis, ultimately providing a scientific paradigm for in-depth exploration of the rockburst disaster gestation process through visualization.

[0006] The technical solution of this invention is as follows:

[0007] On the one hand, the present invention provides an augmented reality observation method for deep engineering rockburst physical simulation experiments, comprising the following steps:

[0008] Acquire multi-source data from physical simulation tests of rockburst disasters in deep engineering, and based on this, construct an augmented reality virtual content generation system for the rockburst incubation process, including constructing a lightweight voxel model of the physical model sample of the geological body in the physical simulation test of rockburst disasters in deep engineering, 3D crack effect of the rock mass, and rockburst damage effect;

[0009] Establish a spatial binding relationship between the augmented reality virtual content generation system for the rockburst incubation process and the internal structure of a physical model sample of a real geological body;

[0010] Based on the spatial binding relationship between the augmented reality virtual content generation system for the rockburst incubation process and the physical model specimen of the real geological body, the identification, registration and real-time tracking of the physical model specimen of the real geological body are realized, thereby superimposing the augmented reality virtual content generation system for the rockburst incubation process onto the physical model specimen of the real geological body.

[0011] Furthermore, the multi-source data of the deep engineering rockburst disaster physical simulation test includes geological structural data and rockburst monitoring data. The geological structural data includes lithology, fluids, and geological structures. The lithology includes sedimentary rocks, metamorphic rocks, and igneous rocks. The fluids include water, gas, and oil. The geological structures include folds, faults, hard structural planes, bedding, bands, dikes, and lithological interfaces. The rockburst monitoring data includes rock mass fracturing events during the disaster incubation process, as well as stress fields, strain fields, energy fields, and fracturing fields.

[0012] Furthermore, the acquisition of multi-source data from physical simulation experiments of rockburst disasters in deep engineering, and the construction of an augmented reality virtual content generation system for the rockburst incubation process based on this data, specifically includes the following steps:

[0013] A1: Obtain geological structure data from multiple geological structure map model files submitted by users, and construct a lightweight voxel model of the physical model sample of the geological body for the physical simulation test of rockburst disaster in deep engineering.

[0014] A2: Acquire rockburst monitoring data and combine it with geological structure data and lightweight voxel models of physical model specimens of geological bodies from the physical simulation test of rockburst disaster in deep engineering to simulate the crack propagation process and generate 3D crack virtual content. Then, construct a 3D crack model library and use the 3D crack model library combined with lightweight voxel models of physical model specimens of geological bodies from the physical simulation test of rockburst disaster in deep engineering to simulate the crack propagation process and generate 3D crack effects.

[0015] A3: Determine the rockburst-affected zone based on the crack propagation range of the generated 3D crack effect, and simulate the rockburst damage effect.

[0016] Furthermore, A1 specifically includes the following steps:

[0017] A1.1: Read multiple geological structure map model files submitted by the user, and extract and process geological structure data for each geological structure map model file to obtain processed geological structure data;

[0018] Specifically, the process involves: first, extracting geological structure data from the geological structure map model file; then, deduplicating, completing, and standardizing the format of the extracted geological structure data; and finally, organizing the geological data corresponding to each geological structure map model file into a structured JSON object to obtain the processed geological structure data.

[0019] A1.2: Based on the geological structure map model file, an initial three-dimensional geological mesh model is constructed through a three-dimensional rendering engine, and the processed geological structure data is maintained in the model database of the initial three-dimensional geological mesh model;

[0020] The model database includes the mapping relationship between each vertex in the three-dimensional geological network model and geological structure data. Each mapping relationship includes the unique ID of the vertex, lithological properties, fluid properties, geological structure properties, and spatial location coordinates of the vertex.

[0021] A1.3: For each geological structure map model file, the initial three-dimensional geological mesh model is queried from its respective model database to confirm its lithological properties. UV maps are then mapped and physical properties, including density and compressive strength, are bound to each initial three-dimensional geological network model according to the lithological properties.

[0022] A1.4: Using spatial transformation alignment technology, multiple three-dimensional geological mesh models that have been UV-mapped and bound with physical properties are seamlessly stitched together according to the spatial coordinate relationship between the geological structure map model files into a complete three-dimensional geological mesh model of the physical model sample of the geological body for physical simulation test of rockburst disaster in deep engineering.

[0023] A1.5: The sparse octree algorithm is used to recursively divide the three-dimensional geological mesh model of the physical model sample of the geological body in the physical simulation test of rockburst disaster in deep engineering into a sparse octree.

[0024] A1.6: Traverse all non-empty leaf nodes in the sparse octree, integrate and serialize their spatial coordinates, dimensions and bound physical properties, and export them in sparse voxel format to finally generate a lightweight voxel model of the physical model sample of the geological body for the physical simulation test of rockburst disaster in deep engineering.

[0025] Furthermore, A2 specifically includes the following steps:

[0026] A2.1: Obtain rockburst monitoring data of the rockburst disaster incubation process in deep engineering, and invert the mechanical causes and morphological characteristics of typical cracks based on the rockburst monitoring data to generate 3D crack virtual content, and then construct a 3D crack model library; the 3D crack model library includes 3D crack virtual content under various crack propagation modes.

[0027] A2.2: Receive the location of the microseismic event from the sensor, map the three-dimensional coordinates of the microseismic event location to the corresponding spatial position of the lightweight voxel model of the physical model sample of the geological body in the physical simulation test of rockburst disaster in deep engineering, and accurately locate the crack initiation point;

[0028] A2.3: Based on the direction of the local maximum principal stress at the crack initiation point and the fracture toughness of the rock, the theoretical propagation angle and length of the crack are calculated. On this basis, a random disturbance factor determined by the heterogeneity of the rock mass is introduced to simulate the uncertainty of the real crack path and obtain the corrected propagation angle and length.

[0029] The random perturbation factor determined by the heterogeneity of the rock mass is used to introduce random bias when calculating the crack propagation path, so as to simulate the irregular crack propagation phenomenon caused by the inhomogeneity of the internal structure of real rock.

[0030] A2.4: Based on the corrected extension angle and length, find the corresponding 3D crack virtual content from the 3D crack model library, and then dynamically embed the 3D crack virtual content into the corresponding position in the lightweight voxel model of the physical model sample of the geological body in the physical simulation test of rockburst disaster in deep engineering through mesh Boolean operation;

[0031] A2.5: The distance field fusion algorithm is used to smooth the geometric joint between the embedded 3D crack virtual content and the lightweight voxel model of the physical model of the geological body of the deep engineering rockburst disaster physical simulation test. At the same time, the environmental occlusion intensity and normal map parameters of the crack area are dynamically adjusted according to the three-dimensional embedding depth of the 3D crack virtual content to obtain the final 3D crack effect.

[0032] Specifically, the distance field of the lightweight voxel model of the geological body physical model specimen in the physical simulation test of rockburst disaster in deep engineering is set as follows: The distance field of the 3D crack virtual content is Then the fusion distance field Defined as:

[0033] (1);

[0034] In the formula, Let be the coordinate vector of any point in three-dimensional space. A positive parameter to control the smoothness of the blending process;

[0035] Subsequently, the fusion distance field was analyzed. The zero-level set is re-meshed to generate the final transition surface, visually achieving a natural transition between the 3D crack virtual content and the lightweight voxel model of the physical model specimen of the geological body in the physical simulation test of rockburst disaster in deep engineering.

[0036] Furthermore, A3 specifically includes the following steps:

[0037] A3.1: Calculate and delineate the rockburst influence area based on the expansion range and spatial distribution of the generated 3D crack effect;

[0038] The calculation and delineation of the rockburst impact zone are achieved by constructing a unified rockburst risk function, which is defined as follows:

[0039] (2);

[0040] In the formula, Due to the risk of rockburst, The coordinate vector is The distance of the crack at that location The coordinate vector is Stress intensity at the point, The critical stress threshold for rockburst. The attenuation coefficient is the effect of cracks.

[0041] The threshold for judging rockburst risk is set as follows: When satisfied When this is the case, the coordinate vector can be considered as... The location fell into the area affected by the rockburst.

[0042] A3.2: Using the implicit function surface trimming algorithm, target rock blocks located within the rockburst influence zone are separated from the lightweight voxel model of the physical model sample of the geological body in the physical simulation test of rockburst disaster in deep engineering; the target rock blocks refer to specific rock mass areas that will be damaged and need to be dynamically simulated in the physical simulation test of rockburst disaster in deep engineering.

[0043] Specifically, the implicit clipping function is defined as follows:

[0044] (3);

[0045] In the formula, For pruning implicit functions, This is an indicator function for the rockburst influence zone, within which... Set to 1, and set to 0 for the outside. The cutting factor;

[0046] When satisfied The cutting surface is formed, and the lightweight voxel model of the physical model sample of the geological body physical model test of the physical simulation test of rockburst disaster in deep engineering can be divided into two parts: the target rock block and the non-rockburst area by extracting the isosurface.

[0047] A3.3: Based on the energy released by the rockburst, a dynamic fragmentation algorithm is used to decompose the target rock block into fragments of random size, and each fragment inherits the physical properties of the original target rock block.

[0048] The dynamic fragmentation algorithm is based on the displacement and stress field responses of the target rock block, and utilizes an energy-driven implicit fracture function. Generate a fragmented distribution when the following conditions are met: When the region is greater than or equal to the fragmentation trigger threshold, it is marked as a fragmentable region. Subsequently, several fragmentation planes are generated by applying random perturbation seeds within the fragmentable region, and the fracture implicit function is used. The spatial gradient controls its size and orientation, ultimately forming several fragments of random size;

[0049] (4);

[0050] In the formula, For the implicit function of fracture, The instantaneous available energy of a rockburst The local stress of the target rock block. Fracture toughness;

[0051] A3.4: Simulate the flight trajectory and collision behavior of debris under the influence of gravity and air resistance, while reproducing the collapse and visual impact effects during the rockburst process;

[0052] The specific method is as follows: use the physics engine to set the mass, shape and initial velocity of each fragment, and calculate the displacement, rotation and collision bounce of the fragment in the scene in real time;

[0053] The specific method for reproducing the collapse and visual impact effects during a rockburst is as follows:

[0054] The emissivity, lifetime, and velocity distribution function of dust particles are defined, and the dust particles have different sizes. The starting point of the dust particles is the edge of the debris, and the transparency of the dust particles decreases over time to simulate the generation and dispersion of dust. The emissivity of the dust particles is the number of dust particles generated per unit time. The lifetime is the duration from the generation to the disappearance of each dust particle. The velocity distribution function is used to describe the law followed by the velocity of dust particles.

[0055] The shock wave generated by the energy released by rockburst is simplified as spherical expansion, and a shock wave front model is constructed. ,in The location of the shock wave front. For time, To measure the propagation speed of the shock wave front, brightness enhancement and transparency attenuation are applied at a set distance from the shock wave front, so that the fragments produce a brief brightening effect when approaching the shock wave and gradually attenuate as they move away from the wave front, thus realistically reproducing the collapse, dust, and impact scouring effects during the rockburst process.

[0056] Furthermore, the establishment of the spatial binding relationship between the augmented reality virtual content generation system for the rockburst incubation process and the internal structure of the physical model sample of the real geological body specifically includes the following steps:

[0057] B1: Construct a local spatial coordinate system;

[0058] The local spatial coordinate system takes the lower left rear vertex of the physical model specimen of the real geological body as its origin, and the X', Y', and Z' axes are respectively aligned with the main directions of the physical model specimen of the real geological body;

[0059] B2: Read the geological structure map model file, use edge detection and contour tracking algorithms to perform vectorized re-identification of key contours, realize semantic segmentation of the closed contour regions formed by the key contours, obtain several closed contour regions, and label each closed contour region with different lithologies according to the mapping relationship between each vertex of the 3D geological network model in the model database and the geological structure data. At the same time, calculate and store the geometric center coordinates of each closed contour region in the local spatial coordinate system; the key contours include strata boundaries and fault lines;

[0060] B3: Based on the generated key contours, the two-dimensional fault lines are meshed in three-dimensional space using a triangular sectioning algorithm in the local spatial coordinate system to construct a three-dimensional triangular mesh surface for characterizing the spatial distribution of geological structures, thus forming a three-dimensional topological structure.

[0061] The triangulation algorithm denotes the set of nodes on the key contour as:

[0062] (5);

[0063] In the formula, For the set of nodes on the key contour, For the three-dimensional nodes on the key contour, This refers to the index of the nodes on the key contour. This represents the total number of nodes on the key contour. These are the planar coordinates of the nodes in the local spatial coordinate system. The elevation of the node is obtained by interpolation from the elevation annotations in the geological structure map model file;

[0064] Node set on key contour Let the set of construction constraint edges, consisting of the input point set, be denoted as:

[0065] (6);

[0066] in, To construct the set of constraint edges, Indicates that by node With nodes The constructed constraint edges are defined as the set of all possible triangulations. And set a triangle The smallest interior angle is Therefore, the constrained Delaunay triangulation is represented as:

[0067] (7);

[0068] in, The optimal cut set;

[0069] The final 3D triangular mesh surface is obtained from the optimal cut set. All the triangles in the array are represented as follows:

[0070] (8);

[0071] in, It is a three-dimensional triangular mesh surface. This is the union operation for sets;

[0072] B4: Uniformly sample the surface of the three-dimensional geological grid model of the physical model sample of the geological body for the physical simulation test of rockburst disaster in deep engineering generated in A1.4 to generate a basic surface point cloud, and assign the lithological and physical properties mapped in step A1.3 to each point in the basic surface point cloud. Convert the voxel center point and its physical properties of the lightweight voxel model of the physical model sample of the physical simulation test of rockburst disaster in deep engineering generated in step A1.6 into an internal attribute point cloud. The basic surface point cloud and the internal attribute point cloud together constitute a three-dimensional attribute point cloud.

[0073] B5: The internal structural feature library is composed of a lightweight voxel model and a three-dimensional topological structure of the physical model sample of the geological body in the physical simulation test of rockburst disaster in deep engineering. The crack initiation point is located in the internal structural feature library and is used as the starting position of the 3D crack virtual content. At the same time, the expansion angle of the 3D crack virtual content is made along the normal direction of the joint surface to achieve spatial binding of the crack virtual effect.

[0074] B6: Based on the defined rockburst influence area and the magnitude of rockburst energy release, locate the origin coordinates of the explosion source in the local spatial coordinate system, and delineate the spherical influence range of the rockburst according to the magnitude of rockburst energy release. Lock the center coordinates of the rockburst explosion and visual impact effect during the rockburst process to the origin coordinates of the explosion source, thereby realizing the spatial binding of the rockburst explosion source area.

[0075] B7: Under a unified local spatial coordinate system, calculate the distance from the crack mesh vertex to the nearest joint surface. If it exceeds the threshold, project and correct it in the direction of the joint surface. At the same time, based on the spatial coordinates of the points contained in the 3D attribute point cloud and the compressive strength attribute in the physical attributes, establish a mapping function that associates the compressive strength value with the coordinates of the stress texture. Then, during rendering, the compressive strength value corresponding to the vertex position is sampled in real time. The coordinates that should be sampled on the stress texture are calculated through the mapping function, thereby replacing the traditional static UV mapping and realizing the dynamic adjustment of the stress texture and its coordinates, thus completing the spatial binding of the stress field texture.

[0076] The crack mesh vertices refer to the geometric vertices of the 3D crack virtual content dynamically embedded into the lightweight voxel model of the geological body physical model specimen of the deep engineering rockburst disaster physical simulation test through mesh Boolean operations;

[0077] B8: Integrate all spatial binding relationships generated by B5-B7 under a unified local spatial coordinate system, and output a structured table of spatial binding relationships between the augmented reality virtual content generation system for rockburst incubation process and the physical model specimen of the real geological body.

[0078] Furthermore, the spatial binding relationship between the augmented reality virtual content generation system based on the rockburst incubation process and the physical model specimen of the real geological body enables the identification, registration, and real-time tracking of the physical model specimen of the real geological body, thereby superimposing the augmented reality virtual content generation system of the rockburst incubation process onto the physical model specimen of the real geological body. This specifically includes the following steps:

[0079] C1: A two-level registration strategy from coarse to fine is adopted to solve for a 6-DOF fine registration pose transformation matrix, so as to achieve the initial matching and superposition of the augmented reality virtual content generation system of rockburst gestation process in complex environment with the physical model of real geological body;

[0080] Specifically, in the coarse registration stage, points located within homogeneous lithological regions and far from lithological boundaries within the generated 3D attribute point cloud are used as feature points. An approximate nearest neighbor search strategy based on lithological semantic constraints is employed to quickly match these feature points with the environmental point cloud acquired in real-time via mobile devices. That is, for each feature point... In environmental point clouds Find the nearest neighbor that satisfies lithological consistency. Establish a point-to-point correspondence between the augmented reality virtual content generator of the rockburst gestation process and the physical model of the real geological body:

[0081] (9);

[0082] in, For the set of feature points, Environmental point clouds The point in the middle; The feature point number, Environmental point clouds The numbering of points in the middle;

[0083] Subsequently, the RANSAC algorithm was used to match the nearest neighbors. After filtering, interior points are obtained, forming a set of matching interior points. The initial rigid body transformation matrix is ​​solved by minimizing the interior point error. :

[0084] (10);

[0085] in, Let be the initial rigid body transformation matrix. Indicates feature points Coordinates mapped to the environment point cloud coordinate system via rigid body transformation; The rigid body transformation matrix;

[0086] The obtained initial rigid body transformation matrix Least squares optimization is performed to obtain the optimized initial pose transformation matrix from the real-world coordinate system to the local space coordinate system. :

[0087] (11);

[0088] The fine registration stage uses the initial pose transformation matrix obtained from coarse registration. Based on this, a bidirectional iterative nearest-point registration method based on joint surface normal constraints is executed on the 3D attribute point cloud of key geological structural areas to solve for the 6-DOF fine registration pose transformation matrix. The key geological structural regions are faults, dominant joint sets, and lithological interfaces on the constructed three-dimensional triangular mesh surface.

[0089] Specifically, point clouds with normal vectors are generated based on the three-dimensional coordinates corresponding to key geological structural regions. , Let a point be represented by its corresponding normal vector. , For point clouds The index of the midpoint is obtained using the initial pose transformation matrix. Point cloud Projected onto a physical model sample of a real geological body, in an environmental point cloud Search for corresponding points in the middle to construct a scene point cloud. , In the point cloud of the scene and the point The corresponding points, based on point cloud pairs and normal vector The following optimization objective is constructed, and the fine-fit pose transformation matrix is ​​solved. :

[0090] (12);

[0091] in, Indicates the point Coordinates mapped to the environment point cloud coordinate system via rigid body transformation;

[0092] C2: Inject the 6-DOF fine-fit pose transformation matrix into the visual inertial odometry system, start real-time dynamic tracking, and realize real-time matching and superposition between the augmented reality virtual content generation system of rockburst incubation process and the physical model of real geological body;

[0093] Specifically, the front-end visual inertial odometry system acquires images of a real geological physical model and uses an ORB feature extractor to adaptively extract texture feature points on the surface of the real geological physical model sample. Simultaneously, it receives IMU data and obtains the pose changes between adjacent frames by pre-integrating the IMU data, compensating for motion blur caused by camera movement, and achieving preliminary inter-frame tracking. The texture feature points refer to the visual texture features on the surface of the geological physical model sample.

[0094] In a visual inertial odometry system, the back-end visual inertial odometry optimizes the tracking results of the front-end visual inertial odometry by constructing and continuously optimizing a factor graph. This factor graph simultaneously processes four constraints, including: IMU dynamic constraints, geological point cloud alignment constraints, visual reprojection constraints, and geological attitude correction constraints.

[0095] The IMU dynamic constraints are as follows:

[0096] (13);

[0097] (14);

[0098] (15);

[0099] In the formula, For a moment The position vector of the camera, Let the velocity vector of the camera be... The direction is represented by the rotation matrix. The acceleration measured by the IMU The angular velocity measured by the IMU. The vector of gravitational acceleration. For time intervals;

[0100] The visual reprojection constraint is:

[0101] (16);

[0102] In the formula, This is the reprojection error vector on the two-dimensional pixel plane. These are the pixel coordinates of the texture feature points observed in the current frame. The three-dimensional coordinates of the texture feature points. For camera projection function;

[0103] The alignment constraint for the geological point cloud is:

[0104] (17);

[0105] In the formula, This refers to the distance error between point clouds. For the 3D attribute point cloud under the current estimated pose, Environmental point clouds The point that minimizes the distance error between point clouds;

[0106] The geological occurrence correction constraint is:

[0107] (18);

[0108] In the formula, For the joint surface normal distance error, For the joint surface normal in the spatial binding relation table Let be a known point on the joint surface;

[0109] A triple-progressive fault-tolerance mechanism is adopted during real-time dynamic tracking:

[0110] First, an illumination-adaptive re-identification mechanism is adopted. When changes in ambient illumination cause the degradation of texture feature points to exceed a set level, the system automatically switches to a matching mode based on spectral features. It utilizes the stable reflection characteristics of different lithologies in specific bands to perform feature matching and tracking maintenance.

[0111] Secondly, topological reasoning under partial occlusion is adopted. When a part of the physical model sample of the real geological body is temporarily occluded, the fault topological reasoning engine is activated. The fault topological reasoning engine uses the generated three-dimensional triangular mesh surface and the spatial binding relationship table as the global prior knowledge base. When the occlusion of the region is detected, the spatial position of the geological structural features that can still be observed in the current frame image is used as real-time input. By solving the geometric constraints and interpolating the spatial path on the prior three-dimensional triangular mesh surface, the most likely orientation and geometric shape of the occluded region are inferred. Then, it is converted into geological point cloud alignment constraints and injected into the factor map for optimization.

[0112] In addition, an absolute pose recalibration technique is adopted. When the tracking quality is detected to have degraded beyond a set threshold or be completely lost, a repositioning process is triggered. The spatial position provided by the 6-DOF fine-fit pose transformation matrix is ​​used as the initial value for retrieval. A fast search and matching is performed in a pre-built keyframe database containing multi-view information. If the match is successful, the tracking pose is restored using the matched image. If it fails, the final absolute pose correction based on borehole coordinates is enabled. That is, the original borehole spatial coordinates with absolute accuracy recorded in the geological structure map model file are used as fixed references. The environmental point cloud and texture features currently perceived by the visual inertial odometry are forced to align with these fixed references to achieve reliable reset.

[0113] C3: Through spatial consistency verification, calculate in real time the Hausdorff distance between the environmental point cloud and extracted texture feature points perceived by the visual inertial odometry and the generated 3D attribute point cloud on key geological structures such as fault lines and rock strata interfaces. Automatically trigger local re-registration for areas where the deviation exceeds the threshold; otherwise, directly execute C4.

[0114] The local reregistration involves repeatedly executing the bidirectional iterative nearest-point registration method based on joint surface phase constraints described in step 3.1 for regions where the deviation exceeds the threshold, outputting a 6-DOF fine-registration pose matrix with dynamic health scoring. When the dynamic health score remains stable above the preset threshold, proceed to step C4.

[0115] C4: Sends tracking-ready commands and the final 6-DOF fine-tuned pose matrix to the augmented reality rendering engine. The augmented reality rendering engine calls the spatial binding relationship table to drive the augmented reality virtual content generation system of the rockburst gestation process to perform real-time rendering on the physical model of the real geological body.

[0116] On the other hand, the present invention also provides an augmented reality observation system for deep engineering rockburst physics simulation experiments, used to realize an augmented reality observation method for deep engineering rockburst physics simulation experiments, including:

[0117] Data layer: Used for storing and managing multi-source data from physical simulation experiments of rockburst hazards in deep engineering;

[0118] Model layer: Used for multi-source data from physical simulation experiments of rockburst disasters in deep engineering based on the data layer, to construct an augmented reality virtual content generation system for the rockburst incubation process and to store and maintain it;

[0119] Functional Layer: Used to establish the spatial binding relationship between the augmented reality virtual content generation system of the rockburst incubation process output by the data layer and the internal structure of the physical model specimen of the real geological body; based on the spatial binding relationship between the augmented reality virtual content generation system of the rockburst incubation process and the physical model specimen of the real geological body, the identification, registration and real-time tracking of the physical model specimen of the real geological body are realized, thereby superimposing the augmented reality virtual content generation system of the rockburst incubation process onto the physical model specimen of the real geological body.

[0120] Service Layer: Responsible for connecting the functional layer and the interaction layer. Based on the user input from the interaction layer, the service layer encapsulates the algorithm logic through a microservice architecture and provides standardized interfaces to support the entire chain of augmented reality observation business. Specifically, it includes spatial registration service and virtual-real fusion rendering service. The spatial registration service is the spatial binding relationship between the augmented reality virtual content generation system of rockburst gestation process and the physical model sample of real geological body.

[0121] Interactive Layer: Used for transparent observation by users through mobile devices; users point their mobile devices at the physical model sample of the real geological body, and the device screen will generate a semi-section, adjustable rockburst incubation process augmented reality virtual content generation system in real time, intuitively see the internal rockburst disaster incubation process, and provide a playback function.

[0122] Thirdly, this application proposes an electronic device comprising: one or more processors, and a memory for storing instructions that, when executed by the one or more processors, cause the one or more processors to perform the deep engineering rockburst physical simulation test augmented reality observation method.

[0123] Fourthly, this application proposes a computer-readable storage medium storing executable instructions that, when executed, cause a processor to perform the deep engineering rockburst physics simulation experiment augmented reality observation method.

[0124] Fifthly, this application proposes a computer program product, including a computer program or instructions that, when executed by a processor, implement the augmented reality observation method for deep engineering rockburst physical simulation experiments.

[0125] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0126] (1) This invention achieves "transparency" and "full-chain" dynamic observation of the entire process of rockburst disaster incubation. Addressing the challenge of the invisibility of internal disaster processes in deep engineering physical simulation experiments, this invention constructs a lightweight voxel model of the geological body physical model sample for deep engineering rockburst disaster physical simulation experiments. This transforms the originally abstract geological structural data and rockburst monitoring data into visualized 3D crack virtual content and rockburst damage effects in real time. This function connects the spatiotemporal evolution chain from micro-fracture initiation and crack expansion to the final rockburst, enabling researchers to directly see through the internal structure of the physical model and intuitively obtain a complete dynamic image of disaster incubation.

[0127] (2) This invention establishes a precise spatial mapping between virtual monitoring data and real physical models. Addressing the shortcomings of traditional augmented reality observations, such as easy drift within the virtual environment and lack of positioning basis, this invention establishes a spatial binding relationship between the augmented reality virtual content generation system for the rockburst incubation process and the physical model specimen of the real geological body. By strictly anchoring virtual elements (such as cracks and stress fields) onto geological structures (such as joint surfaces and lithological boundaries) in a local spatial coordinate system, an anatomical-level fusion of virtual information and physical structure is achieved. This high-precision virtual-real overlay function ensures that the observed disaster phenomena are spatially accurate and reliable, providing a rigorous spatial benchmark for verifying the rockburst mechanism.

[0128] (3) This invention ensures the continuity and anti-interference capability of the observation system under complex experimental environments. To address potential interference such as sudden changes in illumination or equipment obstruction at the physical simulation test site, this invention incorporates geological semantic information as a constraint during the visual-inertial odometry tracking process. The system possesses illumination-adaptive re-identification and fault topology reasoning functions, enabling it to maintain a stable tracking pose using geological features even when geological textures degrade or key parts are obstructed. This allows the invention to adapt to harsh experimental conditions, ensuring uninterrupted observation and no data loss.

[0129] (4) This invention constructs a proactive exploratory observation paradigm of "human-computer interaction". It changes the traditional passive data recording mode in experiments and provides interactive functions based on mobile devices. Researchers can perform semi-sectioning operations on the lightweight voxel model in real time to actively explore the stress field texture distribution at any depth inside the model; at the same time, they can use the time axis control function of the 3D crack virtual content to pause, rewind, and review the disaster evolution process. This fully free interactive capability greatly improves the efficiency of capturing key disaster-causing information and supports researchers in conducting deeper analysis and deconstruction of the rockburst incubation mechanism. Attached Figure Description

[0130] Figure 1 This is a flowchart of the augmented reality observation method for deep engineering rockburst physical simulation experiment in an embodiment of the present invention;

[0131] Figure 2 This is a flowchart illustrating the modeling of the spatial relationship between augmented reality virtual content and a real geological and physical model in an embodiment of the present invention;

[0132] Figure 3 This is a diagram illustrating the 3D recognition, registration, and real-time tracking alignment scheme between augmented reality virtual content and a real geological and physical model in an embodiment of the present invention.

[0133] Figure 4 This is a structural diagram of the augmented reality observation system for deep engineering rockburst physics simulation experiment in an embodiment of the present invention. Detailed Implementation

[0134] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0135] This invention proposes a novel observation paradigm that can overcome the aforementioned limitations. Its core objective is to achieve real-time fusion of multi-dimensional data on the rockburst disaster gestation process in deep engineering physical simulation experiments, enabling direct observation of the entire rockburst gestation process and multi-perspective phenomena, and ultimately achieving on-demand insight into all elements of "continuous time-space coverage," the entire chain of "gestation-triggering-destruction," and "geology-environment-engineering-monitoring."

[0136] Augmented reality (AR) technology demonstrates irreplaceable value in addressing the unique scenarios and needs of physical simulation experiments for rockburst hazards in deep engineering. This technology efficiently collects and integrates multi-source heterogeneous experimental data acquired in real-time during these experiments, and precisely overlays virtual visualizations generated from the experimental data onto real geological physical model samples. This allows researchers to directly observe, through the AR overlay, the entire chain of rockburst processes—from rock mass fracturing to rockburst destruction—with high precision and dynamic presentation within the real experimental setting. This constructs an interactive, dynamically visualized scene of the rockburst incubation process within a complex geological body, offering internal perspective. Therefore, developing an AR observation method for physical simulation experiments of rockburst hazards in deep engineering is of significant theoretical and practical value for clearly and intuitively presenting the entire process of hazard incubation, thereby deepening the scientific understanding and exploration of rockburst mechanisms.

[0137] Example 1:

[0138] This embodiment provides an augmented reality observation method for physical simulation experiments of rockbursts in deep engineering, such as... Figure 1 As shown, it includes the following steps:

[0139] Step 1: Obtain multi-source data from the physical simulation test of rockburst disaster in deep engineering, and based on this, construct an augmented reality virtual content generation system for the rockburst incubation process, including constructing a lightweight voxel model of the physical model sample of the geological body in the physical simulation test of rockburst disaster in deep engineering, the 3D crack effect of the rock mass, and the rockburst damage effect;

[0140] The multi-source data from the physical simulation experiment of rockburst disaster in deep engineering includes geological structural data and rockburst monitoring data. The geological structural data includes lithology, fluids, and geological structures. The lithology includes sedimentary rocks, metamorphic rocks, and igneous rocks, etc. The fluids include water, gas, oil, etc. The geological structures include folds, faults, hard structural planes, bedding, bands, dikes, and lithological interfaces, etc. The rockburst monitoring data includes rock mass fracturing events during the disaster incubation process, as well as multi-physical field monitoring data such as stress field, strain field, energy field, and fracture field.

[0141] Specifically, the steps include the following:

[0142] Step 1.1: Obtain geological structure data from multiple geological structure map model files submitted by users, and construct a lightweight voxel model of the physical model specimen of the geological body for the physical simulation test of rockburst disaster in deep engineering;

[0143] The geological structure map model file is a three-dimensional digital geological model submitted by the user, which contains five levels of information: geometric morphology (three-dimensional mesh, outline and boundary, etc.), geological structure (lithology, fluid, geological structure), physical and mechanical parameters (density, compressive strength, etc.), spatial reference (original borehole spatial coordinates, etc.), and appearance (texture, etc.).

[0144] Step 1.1.1: Read multiple geological structure map model files submitted by the user, and extract and process geological structure data for each geological structure map model file to obtain processed geological structure data;

[0145] Specifically, the process involves: first, extracting geological structural data, such as lithology, fluids, and geological structures, from the geological structural map model file; then, deduplicating, completing, and standardizing the format of the extracted geological structural data; and finally, organizing the geological data corresponding to each geological structural map model file into a structured JSON object to obtain the processed geological structural data.

[0146] In this embodiment, the geological structure map model file is in FBX or GITF format; the processed geological structure data is in JSON format.

[0147] Step 1.1.2: Based on the geological structure map model file, construct the initial three-dimensional geological mesh model through the three-dimensional rendering engine, and maintain the processed geological structure data in the model database of the initial three-dimensional geological mesh model;

[0148] The model database includes the mapping relationship between each vertex in the three-dimensional geological network model and geological structure data. Each mapping relationship includes the unique ID of the vertex, lithological properties, fluid properties, geological structure properties, and spatial location coordinates of the vertex.

[0149] Step 1.1.3: For each geological structure map model file, the initial three-dimensional geological mesh model is queried from its respective model database to confirm its lithological properties. UV maps (such as granite textures) are mapped to each initial three-dimensional geological network model according to the lithological properties, and physical properties, including density and compressive strength, are bound.

[0150] Step 1.1.4: Using spatial transformation alignment technology, multiple three-dimensional geological mesh models that have been UV-mapped and bound with physical properties are seamlessly stitched together according to the spatial coordinate relationship between the geological structure map model files into a complete three-dimensional geological mesh model of the physical model sample of the geological body for the physical simulation test of rockburst disaster in deep engineering, referred to as the complete three-dimensional geological mesh model.

[0151] The spatial transformation and alignment technique establishes a unified global coordinate system and uses three pre-calibrated global reference points as spatial benchmarks to calculate the rigid body transformation matrix from each local model coordinate system to the global coordinate system. This allows multiple independent 3D geological mesh models to be accurately transformed and seamlessly stitched into a complete 3D geological model within the unified global coordinate system. The unified global coordinate system has its origin set at the lower left rear vertex of the complete 3D geological mesh model, and its X, Y, and Z axes are defined to be parallel to the length, height, and width geometric principal axes of the complete 3D geological mesh model, respectively.

[0152] Step 1.1.5: The sparse octree algorithm is then used to recursively divide the complete 3D geological grid model to obtain a sparse octree, which aims to build a lightweight voxel model that supports multi-resolution query and rendering.

[0153] The core of the sparse octree algorithm is a recursive subdivision process, mathematically expressed as follows: For a given node... , Represents a three-dimensional bounding box. Representing the 3D geological mesh model within the 3D spatial bounding box, the first step is to determine whether the node is an empty leaf node, i.e. If not empty, then further by function By detecting the 3D bounding box of the node The function takes the lithological or material properties of all geological grid cells as an example. If all cell properties are identical, the node is considered a single-attribute leaf node, and the function returns True; otherwise, it returns False. If a node is neither an empty leaf node nor a single-attribute leaf node, it is a mixed lithology node, and its coordinate axis midpoint needs to be considered. Divided into eight sub-cubes Each sub-cube is treated as a child node. Let be the coordinates of the midpoint of the node's 3D bounding box along the X-axis. Let be the coordinates of the midpoint of the node's 3D space enclosed along the Y-axis. Let be the coordinates of the midpoint of the node's 3D space enclosed along the Z-axis. Number the child nodes; then recursively perform the subdivision process for each child node. , The function will recursively perform three steps: empty leaf node judgment, single attribute leaf node judgment, and child node splitting. This process continues until all nodes are classified as empty leaf nodes or single attribute leaf nodes, and finally generates a sparse tree structure that only stores non-empty nodes, namely a sparse octree.

[0154] Step 1.1.6: Traverse all non-empty leaf nodes (i.e. voxels representing homogeneous lithology) in the sparse octree, integrate and serialize their spatial coordinates, dimensions, and bound physical properties (such as density and compressive strength), and export them using an efficient sparse voxel format (such as OpenVDB) to finally generate a lightweight voxel model of the physical model sample of the geological body for the physical simulation test of rockburst disaster in deep engineering, for subsequent use;

[0155] Step 1.2: Obtain rockburst monitoring data, and combine it with geological structure data and lightweight voxel models of physical model specimens of geological bodies from the physical simulation test of rockburst disaster in deep engineering to simulate the crack propagation process and generate highly realistic 3D crack virtual content. Then, construct a 3D crack model library, and use the 3D crack model library combined with lightweight voxel models of physical model specimens of geological bodies from the physical simulation test of rockburst disaster in deep engineering to simulate the crack propagation process and generate 3D crack effects.

[0156] Step 1.2.1: Obtain rockburst monitoring data of the rockburst disaster incubation process in deep engineering, and invert the mechanical causes and morphological characteristics of typical cracks based on the rockburst monitoring data to generate 3D crack virtual content, and then construct a 3D crack model library; the 3D crack model library includes 3D crack virtual content under various crack propagation modes;

[0157] Step 1.2.2: Receive the location of the microseismic event from the sensor, map the three-dimensional coordinates of the microseismic event location to the corresponding spatial position of the lightweight voxel model of the physical model sample of the geological body in the physical simulation test of rockburst disaster in deep engineering, and accurately locate the crack initiation point.

[0158] Step 1.2.3: Based on the direction of the local maximum principal stress at the crack initiation point and parameters such as the fracture toughness of the rock, calculate the theoretical propagation angle and length of the crack. On this basis, introduce the random disturbance factor determined by the heterogeneity of the rock mass to simulate the uncertainty of the real crack path and obtain the corrected propagation angle and length.

[0159] The random perturbation factor determined by the heterogeneity of the rock mass is used to artificially introduce random deviations that conform to the geological laws of the real world when calculating the crack propagation path, so as to simulate the irregular crack propagation phenomenon caused by the inhomogeneity of the internal structure of real rocks.

[0160] Step 1.2.4: Based on the corrected extension angle and length, find the corresponding 3D crack virtual content from the 3D crack model library, and then dynamically embed the 3D crack virtual content into the corresponding position in the lightweight voxel model of the physical model sample of the geological body in the physical simulation test of rockburst disaster in deep engineering through mesh Boolean operation;

[0161] Step 1.2.5: The distance field fusion algorithm is used to smooth the geometric joint between the embedded 3D crack virtual content and the lightweight voxel model of the geological body physical model sample of the deep engineering rockburst disaster physical simulation test, so as to achieve visual seamless fusion. At the same time, according to the three-dimensional embedding depth of the 3D crack virtual content, the environmental occlusion intensity and normal map parameters of the crack area are dynamically adjusted to enhance the visual stereoscopic effect and obtain the final 3D crack effect.

[0162] The distance field fusion algorithm is based on the signed distance field representations of both the 3D crack virtual content and the lightweight voxel model of the geological body physical model specimen from the deep engineering rockburst disaster physical simulation test. By smoothly synthesizing the distance field functions of the two, a continuously differentiable fusion surface with a transition band is constructed. Specifically, the distance field of the lightweight voxel model of the geological body physical model specimen from the deep engineering rockburst disaster physical simulation test is set as follows: The distance field of the 3D crack virtual content is Then the fusion distance field Defined as:

[0163] (1);

[0164] In the formula, Let be the coordinate vector of any point in three-dimensional space, i.e. , The positive parameter controls the smoothness of the blending. The higher the value, the sharper the blending region; the lower the value, the smoother the transition. This method can avoid the sharp boundaries and discontinuities in normals that are easy to occur in geometric Boolean operations.

[0165] Subsequently, the fusion distance field was analyzed. The zero-level set is re-meshed to generate the final transition surface, visually achieving a natural transition between the 3D crack virtual content and the lightweight voxel model of the physical model specimen of the geological body in the physical simulation test of rockburst disaster in deep engineering.

[0166] Step 1.3: Determine the rockburst-affected zone based on the crack propagation range of the generated 3D crack effect, and simulate dynamic rockburst damage effects such as rockburst ejection and fragmentation;

[0167] Step 1.3.1: Based on the expansion range and spatial distribution of the generated 3D crack effect, calculate and delineate the rockburst influence area;

[0168] The calculation and delineation of the rockburst impact zone are achieved by constructing a unified rockburst risk function, which is defined as follows:

[0169] (2);

[0170] In the formula, Due to the risk of rockburst, The coordinate vector is The distance of the crack at that location The coordinate vector is Stress intensity at the point, This is the critical stress threshold for rockburst (the cumulative stress threshold of the rock mass). The attenuation coefficient is the effect of cracks.

[0171] The threshold for judging rockburst risk is set as follows: The rockburst risk assessment threshold is determined based on the construction site safety regulations or rock mass grade, and is used to determine the minimum risk level for rockburst occurrence. When the following conditions are met... When this is the case, the coordinate vector can be considered as... The location fell into the area affected by the rockburst.

[0172] Step 1.3.2: Using the implicit function surface trimming algorithm, the target rock block located within the rockburst influence zone is separated from the lightweight voxel model of the physical model sample of the geological body in the physical simulation test of rockburst disaster in deep engineering, providing a geometric basis for subsequent fragmentation simulation; the target rock block refers to a specific rock mass area that will be damaged and needs to be dynamically simulated in the physical simulation test of rockburst disaster in deep engineering.

[0173] The implicit function surface trimming algorithm is based on the signed distance field representation of a lightweight voxel model of a geological body physical model sample from a physical simulation experiment of rockburst disaster in deep engineering. It constructs a trimming implicit function that includes the rockburst-affected zone. And extract its zero level set to complete the geometric separation; specifically, define the clipping implicit function as:

[0174] (3);

[0175] In the formula, For pruning implicit functions, This is an indicator function for the rockburst influence zone, within which... Set to 1, and set to 0 for the outside. The cutting factor;

[0176] When satisfied The cutting surface is formed, and the lightweight voxel model of the physical model sample of the geological body physical model test of the physical simulation test of rockburst disaster in deep engineering can be divided into two parts: the target rock block and the non-rockburst area by extracting the isosurface.

[0177] Step 1.3.3: Based on the energy released by the rockburst, the target rock block is decomposed into fragments of random size using a dynamic fragmentation algorithm. Each fragment inherits the physical properties of the original target rock block, such as elastic modulus and Poisson's ratio.

[0178] The dynamic fragmentation algorithm is based on the displacement and stress field responses of the target rock block, and utilizes an energy-driven implicit fracture function. Generate a fragmented distribution when the following conditions are met: When the region is greater than or equal to the fragmentation trigger threshold, it is marked as a fragmentable region. Subsequently, several fragmentation planes are generated by applying random perturbation seeds within the fragmentable region, and the fracture implicit function is used. The spatial gradient controls its size and orientation, ultimately forming several fragments of random size;

[0179] (4);

[0180] In the formula, For the implicit function of fracture, The instantaneous available energy of a rockburst The local stress of the target rock block. Fracture toughness;

[0181] Step 1.3.4: Simulate the flight trajectory and collision behavior of the debris under the action of gravity and air resistance, and at the same time reproduce the collapse and visual impact effects of the rockburst process;

[0182] The specific method for simulating the flight trajectory and collision behavior of debris under the influence of gravity and air resistance is as follows:

[0183] The physics engine is used to set the mass, shape, and initial velocity of each fragment, and to calculate the displacement, rotation, and collision bounce of the fragments in the scene in real time.

[0184] The specific method for reproducing the collapse and visual impact effects during a rockburst is as follows:

[0185] The emissivity, lifetime, and velocity distribution function of dust particles are defined, and the dust particles have different sizes. The starting point of the dust particles is the edge of the debris, and the transparency of the dust particles decreases over time to simulate the generation and dispersion of dust. The emissivity of the dust particles is the number of dust particles generated per unit time. The lifetime is the duration from the generation to the disappearance of each dust particle. The velocity distribution function is used to describe the law followed by the velocity of dust particles.

[0186] The shock wave generated by the energy released by rockburst is simplified as spherical expansion, and a shock wave front model is constructed. ,in The location of the shock wave front. For time, To measure the propagation speed of the shock wave front, by applying brightness enhancement and transparency attenuation near the shock wave front, the fragments produce a visual effect of brief high brightness when approaching the shock wave and gradual attenuation as they move away from the front and back of the wave, thus realistically reproducing the collapse, dust, and impact scouring effects during the rockburst process.

[0187] Step 2: Establish a spatial binding relationship between the augmented reality virtual content generation system for the rockburst incubation process and the internal structure of the physical model sample of the real geological body, solving the fundamental problem of target positioning for virtual content overlay, such as... Figure 2 As shown, it includes the following steps:

[0188] Step 2.1: Construct a local spatial coordinate system;

[0189] The local spatial coordinate system takes the lower left rear vertex of the physical model specimen of the real geological body as its origin, and the X', Y', and Z' axes are respectively aligned with the main directions of the physical model specimen of the real geological body. This local spatial coordinate system will serve as a unified reference framework for all subsequent spatial data, including geological structures and virtual effects, to ensure that all spatial data are calculated and expressed under the same reference.

[0190] Step 2.2: Read the geological structure map model file, and use edge detection and contour tracking algorithms to perform vectorized re-identification of key contours. This enables semantic segmentation of the closed contour regions formed by the key contours, resulting in several closed contour regions. Based on the mapping relationship between each vertex of the 3D geological network model in the model database and the geological structure data in Step 1.1.2, different lithologies are labeled for each closed contour region. Simultaneously, the geometric center coordinates of each closed contour region are calculated and stored in the local spatial coordinate system. The key contours include strata boundaries and fault lines, etc.

[0191] Step 2.3: Based on the key contours generated in Step 2.2, the two-dimensional fault lines are meshed in three-dimensional space using a triangulation algorithm in the local spatial coordinate system to construct a three-dimensional triangular mesh surface for characterizing the spatial distribution of geological structures, forming a three-dimensional topological structure for subsequent spatial calculations.

[0192] The triangulation algorithm denotes the set of nodes on the key contour as:

[0193] (5);

[0194] In the formula, For the set of nodes on the key contour, For the three-dimensional nodes on the key contour, This refers to the index of the nodes on the key contour. This represents the total number of nodes on the key contour. These are the planar coordinates of the nodes in the local spatial coordinate system. The elevation of the node is obtained by interpolation from the elevation labels (such as contour lines and elevation points) in the geological structure map model file;

[0195] Node set on key contour Let the set of construction constraint edges, consisting of the input point set, be denoted as:

[0196] (6);

[0197] in, To construct the set of constraint edges, Indicates that by node With nodes The constructed constraint edges are used to define the set of all possible triangulations. And set a triangle The smallest interior angle is Therefore, the constrained Delaunay triangulation can be expressed as:

[0198] (7);

[0199] in, The optimal cut set;

[0200] The final 3D triangular mesh surface is obtained from the optimal cut set. All the triangles in the equation can be represented as:

[0201] (8);

[0202] in, It is a three-dimensional triangular mesh surface. This is the union operation for sets;

[0203] Step 2.4: Uniformly sample the surface of the three-dimensional geological mesh model of the physical model sample of the deep engineering rockburst disaster physical simulation test geological body generated in Step 1.1.4 to generate a basic surface point cloud. Assign each point in the basic surface point cloud the lithological and physical properties mapped in Step 1.1.3. Use the lightweight voxel model of the physical model sample of the deep engineering rockburst disaster physical simulation test geological body generated in Step 1.1.6 as the core data source of internal properties. Convert its voxel center points and physical properties into a high-density internal property point cloud to enhance the feature expression of the internal structure of the model. The basic surface point cloud and the internal property point cloud together constitute a unified three-dimensional property point cloud for spatial binding.

[0204] Step 2.5: The lightweight voxel model of the physical model sample of the geological body for the physical simulation test of rockburst disaster in deep engineering, output from Step 1.1.6, and the three-dimensional topology output from Step 2.3 together constitute the internal structural feature library. In the internal structural feature library, the crack initiation point determined in Step 1.2.2 is located, and this crack initiation point is used as the starting position of the 3D crack virtual content. At the same time, the expansion angle of the 3D crack virtual content is made to be along the normal direction of the joint surface, so as to realize the spatial binding of the crack virtual effect.

[0205] Step 2.6: Based on the rockburst influence area and the magnitude of rockburst energy released in Step 1.3.1, locate the origin coordinates of the explosion source in the local spatial coordinate system, and delineate the spherical influence range of the rockburst according to the magnitude of rockburst energy released. Lock the center coordinates of the rockburst explosion and visual impact effect during the rockburst process to the origin coordinates of the explosion source, thereby realizing the spatial binding of the rockburst explosion source area.

[0206] Step 2.7: Under a unified local spatial coordinate system, calculate the distance from the crack mesh vertex to the nearest joint surface. If it exceeds the threshold, project and correct it towards the joint surface. At the same time, based on the spatial coordinates of the points in the 3D attribute point cloud obtained in Step 2.4 and the compressive strength attribute (physical attribute), establish a mapping function that associates the compressive strength value with the coordinates of the stress texture (usually a hot color band, such as from blue low stress to red high stress). Then, during rendering, sample the compressive strength value corresponding to the position of each vertex in real time, and calculate the coordinates that should be sampled on the stress texture through the mapping function. This replaces the traditional static UV mapping, realizes dynamic adjustment of the stress texture and its coordinates, and completes the spatial binding of the stress field texture.

[0207] The crack mesh vertices refer to the geometric vertices of the 3D crack virtual content dynamically embedded into the lightweight voxel model of the geological body physical model sample of the deep engineering rockburst disaster physical simulation test through mesh Boolean operations in step 1.2.3.

[0208] Step 2.8: Integrate all spatial binding relationships generated in Steps 2.5-2.7 under a unified local spatial coordinate system, and output a structured spatial binding relationship table between the augmented reality virtual content generation system for rockburst incubation process and the physical model sample of the real geological body. This table is stored in a lightweight JSON format for subsequent identification, registration and alignment processes.

[0209] The spatial binding relationship table includes: local spatial coordinate system definition (coordinate system origin, three-axis direction, unit, etc.), crack virtual effect spatial binding (unique identifier, geological structure surface ID, key point three-dimensional coordinates of the crack main path, etc.), stress field texture spatial binding (UV-mechanical parameter comparison table, etc.), rockburst source area spatial binding (rockburst effect unique identifier, three-dimensional coordinates of the source center point, spherical influence radius, rockburst released energy, etc.), geological occurrence data (dip and dip angle of the main joint surfaces, and the spatial location of each structural surface, etc.), and global metadata (spatial binding relationship table version number, generation time, bound geological model ID, etc.).

[0210] Step 3: Based on the spatial binding relationship between the augmented reality virtual content generation system for the rockburst incubation process and the physical model specimen of the real geological body, the real geological body physical model specimen is quickly, accurately, and robustly identified, registered, and tracked in real time, thereby stably and accurately superimposing the augmented reality virtual content generation system for the rockburst incubation process onto the real geological body physical model specimen.

[0211] like Figure 3 As shown, the specific steps include the following:

[0212] Step 3.1: Using a two-level registration strategy from coarse to fine, the 6-DOF fine registration pose transformation matrix is ​​obtained to achieve the initial matching and superposition of the augmented reality virtual content generation system for rockburst gestation process in complex environments and the physical model of real geological bodies;

[0213] Specifically, in the coarse registration stage, points located within homogeneous lithological regions and far from lithological boundaries in the 3D attribute point cloud generated in step 2.4 are used as feature points. An approximate nearest neighbor search strategy based on lithological semantic constraints is employed to quickly match the feature points with the environmental point cloud collected in real time via mobile devices. That is, for each feature point... In environmental point clouds Find the nearest neighbor that satisfies lithological consistency. Quickly establish a point-to-point correspondence between the augmented reality virtual content generator of the rockburst formation process and the physical model of the real geological body:

[0214] (9);

[0215] in, For the set of feature points, Environmental point clouds The point in the middle; The feature point number, Environmental point clouds The numbering of points in the middle;

[0216] Subsequently, the RANSAC algorithm was used to match the nearest neighbors. After filtering, the interior points form a matching set of interior points. The initial rigid body transformation matrix is ​​solved by minimizing the interior point error. :

[0217] (10);

[0218] in, Let be the initial rigid body transformation matrix. Indicates feature points Coordinates mapped to the environment point cloud coordinate system via rigid body transformation; The rigid body transformation matrix;

[0219] The obtained initial rigid body transformation matrix Least squares optimization can further reduce interior point errors, resulting in the optimized initial pose transformation matrix from the real-world coordinate system to the local spatial coordinate system defined in step 2.1. :

[0220] (11);

[0221] The origin of the real-world coordinate system is usually located at the optical center of the camera at the moment of SLAM system initialization. The Z2 axis is parallel to the initial optical axis of the camera (pointing to the shooting direction), the Y2 axis is perpendicular to the ground and upward (determined by the direction of gravitational acceleration), and the X2 axis is determined by the right-hand rule (i.e., X2 = Y2 × Z2), forming a right-handed, orthogonal, globally fixed three-dimensional Cartesian coordinate system.

[0222] The fine registration stage uses the initial pose transformation matrix obtained from coarse registration. Based on this, a bidirectional iterative nearest-point registration method based on joint surface normal constraints is executed on the 3D attribute point cloud of key geological structural regions. In this process, the joint surface normal constraint is used as an iterative weight factor to prioritize the alignment accuracy of key geological structural regions, thereby solving for the 6-DOF fine registration pose transformation matrix. The key geological structure region refers to the faults, dominant joint groups, lithological interfaces, and other structures related to the rockburst incubation process on the three-dimensional triangular mesh surface constructed in step 2.3.

[0223] In this method, point clouds with normal vectors are generated based on the three-dimensional coordinates corresponding to key geological structural regions. , Let a point be represented by its corresponding normal vector. , For point clouds The index of the midpoint is obtained using the initial pose transformation matrix. Point cloud Projected onto a physical model sample of a real geological body, in an environmental point cloud Search for corresponding points in the middle to construct a scene point cloud. , In the point cloud of the scene and the point The corresponding points, based on point cloud pairs and normal vector The following optimization objective is constructed, and the fine-fit pose transformation matrix is ​​solved. :

[0224] (12);

[0225] in, Indicates the point Coordinates mapped to the environment point cloud coordinate system via rigid body transformation;

[0226] Step 3.2: Inject the 6-DOF fine-fit pose transformation matrix into the visual inertial odometry (VIO) system, start real-time dynamic tracking, and realize real-time matching and superposition between the augmented reality virtual content generation system of the rockburst incubation process and the physical model of the real geological body;

[0227] A real-time tracking thread based on a visual inertial odometry system is initiated, and the real-time tracking thread is continuously constrained and corrected through geological semantic information; the geological semantic information includes geological knowledge that can provide prior constraints for visual tracking, such as lithology type, geological structure occurrence, and spatial relationship of structural planes.

[0228] Specifically:

[0229] The front-end visual inertial odometry acquires images of a real geological physical model at a frame rate of 30fps, and uses an ORB feature extractor to adaptively extract texture feature points from the surface of the real geological physical model sample. Simultaneously, it receives 200Hz IMU (Inertial Measurement Unit) data. By pre-integrating the IMU data, it obtains the pose changes between adjacent frames, effectively compensating for motion blur caused by rapid camera movement, ensuring the continuity of inter-frame tracking, and achieving preliminary inter-frame tracking. The texture feature points refer to the visual texture features on the surface of the geological physical model sample, including natural textures formed by lithological changes and artificial textures.

[0230] The back-end visual inertial odometry optimizes the tracking results of the front-end visual inertial odometry by constructing and continuously optimizing a factor graph. This factor graph simultaneously processes four constraints, including: IMU dynamic constraints, geological point cloud alignment constraints, visual reprojection constraints, and geological attitude correction constraints.

[0231] The IMU dynamic constraints are based on the IMU's dynamic equations, providing continuous, high-frequency pose change constraints to ensure the physical rationality of the motion.

[0232] (13);

[0233] (14);

[0234] (15);

[0235] In the formula, For a moment The position vector of the camera, Let the velocity vector of the camera be... The direction is represented by the rotation matrix. The acceleration measured by the IMU The angular velocity measured by the IMU. The vector of gravitational acceleration. For time intervals;

[0236] The visual reprojection constraint backprojects the tracked texture feature points into three-dimensional space and calculates their reprojection error in subsequent frames to ensure the geometric accuracy of visual tracking.

[0237] (16);

[0238] In the formula, This is the reprojection error vector on the two-dimensional pixel plane. These are the pixel coordinates of the texture feature points observed in the current frame. The three-dimensional coordinates of the texture feature points. For camera projection function;

[0239] The geological point cloud alignment constraint, under the current 6-DOF fine-fit pose transformation matrix, performs local matching between the 3D attribute point cloud and the real-time perceived environmental point cloud, calculates the distance error between point clouds, and adds it as a univariate constraint to the factor map, forcing the tracking results to spatially conform to the known geological structure.

[0240] (17);

[0241] In the formula, This refers to the distance error between point clouds. For the 3D attribute point cloud under the current estimated pose, Environmental point clouds The point that minimizes the distance error between point clouds;

[0242] The geological attitude correction constraint identifies visible joint surfaces from the current real geological body physical model image every 10 frames, estimates their dip and dip angle (attitude), compares them with the geological attitude data in the spatial binding relationship table, and adds the generated error term as a constraint factor to the optimization, directly correcting pose drift using prior knowledge of geological structure:

[0243] (18);

[0244] In the formula, For the joint surface normal distance error, For the joint surface normal in the spatial binding relation table Let be a known point on the joint surface;

[0245] A triple progressive fault tolerance mechanism is adopted during real-time dynamic tracking to deal with problems such as sudden changes in illumination and local occlusion. The specific triple progressive fault tolerance mechanism is as follows:

[0246] First, an illumination-adaptive re-identification mechanism is adopted. When drastic changes in ambient illumination cause significant degradation of texture feature points, the matching mode is automatically switched to a matching mode based on spectral features (such as the near-infrared band). The stable reflection characteristics of different lithologies under specific bands are used for feature matching and tracking maintenance.

[0247] Secondly, topological inference under partial occlusion is adopted. When the key part of the physical model sample of the real geological body is temporarily occluded, the fault topological inference engine is activated. The core logic of the fault topological inference engine is based on the three-dimensional triangular mesh surface (three-dimensional topological structure) generated and persisted in step 2.3 and the geological occurrence data in the spatial binding relationship table output in step 2.8 as the global prior knowledge base. When the occlusion of the key area is detected, the engine takes the spatial location and occurrence information of geological structural features such as joint surfaces and lithological boundaries that can still be observed in the current frame image as real-time input. By solving the geometric constraints and interpolating the spatial path on the prior three-dimensional topological structure, the most likely orientation and geometric shape of the fault, joint and other structures in the occluded area are inferred. Then, it is converted into geological point cloud alignment constraints and injected into the factor map for optimization. Thus, during the period of visual information loss, the stability of the tracking pose is maintained by relying on the structural continuity of the geological model.

[0248] In addition, an absolute pose recalibration technique is adopted. When a significant decrease or complete loss of tracking quality is detected, a relocalization process is triggered. The spatial position provided by the 6-DOF fine-fit pose transformation matrix is ​​used as the initial value for retrieval. A fast search and matching is performed in a pre-built keyframe database containing multi-view information. If the matching is successful, the tracking pose is restored using the matched image. If the above method fails, the final absolute pose calibration based on borehole coordinates is enabled. That is, the original borehole spatial coordinates with absolute accuracy recorded in the geological structure map model file are used as fixed references. The environmental point cloud and texture features currently perceived by the visual inertial odometry (VIO) are forced to align with these fixed references to achieve reliable system-level reset.

[0249] The keyframe database stores keyframes with sufficient parallax and rich texture features selected during the tracking process, maintains the co-view relationship between keyframes and the mapping relationship between texture feature points and three-dimensional spatial coordinates; when tracking loss occurs, the features of the current frame are used to quickly retrieve and match in the database.

[0250] Step 3.3: Through spatial consistency verification, calculate in real time the Hausdorff distance between the environmental point cloud and the extracted texture feature points perceived by the visual inertial odometry (VIO) and the 3D attribute point cloud generated in step 2.4 on key geological structures such as fault lines and rock strata interfaces. Automatically trigger local re-registration for areas where the deviation exceeds the threshold; otherwise, directly execute step 3.4.

[0251] The local reregistration involves repeatedly executing the bidirectional iterative nearest-point registration method based on joint surface phase constraints described in step 3.1 for regions where the deviation exceeds the threshold, to recalculate the precise pose of the region. Finally, a 6-DOF fine-registration pose matrix with dynamic health scoring is output. When the dynamic health score remains stable above the preset threshold, proceed to step 3.4.

[0252] Step 3.4: Send the tracking-ready command and the final 6-DOF fine-tuned pose matrix to the augmented reality rendering engine. The augmented reality rendering engine calls the spatial binding relationship table output in step 2.8 to drive all virtual content to perform accurate, stable, and immersive real-time rendering on the physical model of the real geological body.

[0253] Example 2:

[0254] An augmented reality observation system for physical simulation experiments of rockburst hazards in deep engineering is provided to implement an augmented reality observation method for physical simulation experiments of rockburst hazards in deep engineering, such as... Figure 4 As shown, it includes:

[0255] Data Layer: This is the foundation of the entire system architecture, responsible for storing, managing, and maintaining all underlying data. It is used to store and manage multi-source data from the physical simulation test of rockburst disaster in deep engineering. Relying on the spatiotemporal alignment engine, it performs timestamp synchronization and spatial coordinate transformation on the multi-source data of the physical simulation test of rockburst disaster in deep engineering, and builds a multi-dimensional fusion data pool with spatiotemporal continuity.

[0256] The multi-source data from the physical simulation experiment of rockburst disaster in deep engineering includes geological structural data and rockburst monitoring data. The geological structural data includes lithology, fluids, and geological structures. The lithology includes sedimentary rocks, metamorphic rocks, and igneous rocks, etc. The fluids include water, gas, oil, etc. The geological structures include folds, faults, hard structural planes, bedding, bands, dikes, and lithological interfaces, etc. The rockburst monitoring data includes rock mass fracturing events during the disaster incubation process, as well as multi-physical field monitoring data such as stress field, strain field, energy field, and fracture field.

[0257] Model layer: Used for multi-source data from physical simulation experiments of rockburst disasters in deep engineering based on the data layer, to construct an augmented reality virtual content generation system for the rockburst incubation process and to store and maintain it;

[0258] Functional layer: Used to establish the spatial binding relationship between the augmented reality virtual content generation system of rockburst incubation process output by the data layer and the internal structure of the physical model sample of the real geological body; based on the spatial binding relationship between the augmented reality virtual content generation system of rockburst incubation process and the physical model sample of the real geological body, it realizes the rapid, accurate, and robust identification, registration and real-time tracking of the physical model sample of the real geological body, so as to stably and accurately superimpose the augmented reality virtual content generation system of rockburst incubation process onto the physical model sample of the real geological body;

[0259] Service Layer: Responsible for connecting the functional layer and the interaction layer. Based on the input of the user in the interaction layer, the service layer encapsulates the complex algorithm logic of the functional layer through a microservice architecture, and provides standardized interfaces to support the entire chain of augmented reality observation business. Specifically, it includes spatial registration services (the spatial binding relationship between the augmented reality virtual content generation system for rockburst gestation process and the physical model specimen of the real geological body) and virtual-real fusion rendering services, etc.

[0260] Interactive Layer: This layer allows users to conduct transparent observations via mobile devices such as tablets and smartphones. Users can point their mobile devices at a physical model sample of a real geological body, and the device screen will generate a semi-section, adjustable augmented reality virtual content generation system of the rockburst gestation process in real time. This provides an intuitive view of the internal rockburst disaster gestation process and gives users full freedom of control, enabling a deep mechanism exploration paradigm that allows for "sliding back through the disaster process, cutting through and viewing the internal structure, and marking key areas in real time".

[0261] Example 3:

[0262] This embodiment proposes an electronic device, including: one or more processors, and a memory, wherein the memory is used to store instructions, and when the instructions are executed by the one or more processors, the one or more processors execute the deep engineering rockburst physical simulation test augmented reality observation method.

[0263] The electronic device may be a mobile phone, computer, or tablet computer, etc., and includes a memory and a processor. The memory stores a computer program, which, when executed by the processor, implements the augmented reality observation method for deep engineering rockburst physics simulation experiments as described in the embodiments. It is understood that the electronic device may also include input / output (I / O) interfaces and communication components.

[0264] The processor is used to execute all or part of the steps in the augmented reality observation method for deep engineering rockburst physics simulation experiments as described in the above embodiments. The memory is used to store various types of data, which may include, for example, instructions for any application or method in an electronic device, as well as application-related data.

[0265] The processor can be implemented as an Application Specific Integrated Circuit (ASIC), Digital Signal Processor (DSP), Programmable Logic Device (PLD), Field Programmable Gate Array (FPGA), controller, microcontroller, microprocessor, or other electronic components, and is used to execute the deep engineering rockburst physical simulation test augmented reality observation method described in the above embodiments.

[0266] Example 4:

[0267] This embodiment proposes a computer-readable storage medium that stores executable instructions. When these instructions are executed, if they are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium.

[0268] The computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the deep engineering rockburst physical simulation test augmented reality observation method described in the various embodiments of this application.

[0269] The aforementioned storage media include: flash memory, hard disks, multimedia cards, card-type memory (e.g., SD (Secure Digital Memory Card) or DX (Memory Data Register, MDR) memory), random access memory (RAM), static random-access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic storage, disks, optical discs, servers, APP (Application) app stores, and other media capable of storing program verification codes. These media store computer programs, which, when executed by a processor, can implement the various steps of the deep engineering rockburst physical simulation experimental augmented reality observation method described above.

[0270] Example 5:

[0271] This embodiment proposes a computer program product, including a computer program or instructions, which, when executed by a processor, implements the augmented reality observation method for deep engineering rockburst physical simulation experiments.

[0272] Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or part of the technical solution, can be embodied in the form of a computer program product.

[0273] The various embodiments in this application are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0274] The scope of protection of this application is not limited to the embodiments described above. Obviously, those skilled in the art can make various modifications and variations to this disclosure without departing from the scope and spirit of this disclosure. If such modifications and variations fall within the scope of this disclosure and its equivalents, then the intent of this disclosure also includes these modifications and variations.

Claims

1. An augmented reality observation method for physical simulation experiments of rockburst in deep engineering, characterized in that, Includes the following steps: Acquire multi-source data from physical simulation tests of rockburst disasters in deep engineering, and based on this, construct an augmented reality virtual content generation system for the rockburst incubation process, including constructing a lightweight voxel model of the physical model sample of the geological body in the physical simulation test of rockburst disasters in deep engineering, 3D crack effect of the rock mass, and rockburst damage effect; Establish a spatial binding relationship between the augmented reality virtual content generation system for the rockburst incubation process and the internal structure of a physical model sample of a real geological body; Based on the spatial binding relationship between the augmented reality virtual content generation system for the rockburst incubation process and the physical model specimen of the real geological body, the identification, registration and real-time tracking of the physical model specimen of the real geological body are realized, thereby superimposing the augmented reality virtual content generation system for the rockburst incubation process onto the physical model specimen of the real geological body.

2. The augmented reality observation method for deep engineering rockburst physical simulation experiment according to claim 1, characterized in that, The multi-source data from the physical simulation test of rockburst disaster in deep engineering includes geological structural data and rockburst monitoring data. The geological structural data includes lithology, fluids, and geological structures. The lithology includes sedimentary rocks, metamorphic rocks, and igneous rocks. The fluids include water, gas, and oil. The geological structures include folds, faults, hard structural planes, bedding, banding, dikes, and lithological interfaces. The rockburst monitoring data includes rock mass fracturing events during the disaster incubation process, as well as stress fields, strain fields, energy fields, and fracturing fields.

3. The augmented reality observation method for deep engineering rockburst physical simulation experiment according to claim 1, characterized in that, The acquisition of multi-source data from physical simulation experiments of rockburst disasters in deep engineering, and the construction of an augmented reality virtual content generation system for the rockburst incubation process based on this data, specifically includes the following steps: A1: Obtain geological structure data from multiple geological structure map model files submitted by users, and construct a lightweight voxel model of the physical model sample of the geological body for the physical simulation test of rockburst disaster in deep engineering. A2: Acquire rockburst monitoring data and combine it with geological structure data and lightweight voxel models of physical model specimens of geological bodies from the physical simulation test of rockburst disaster in deep engineering to simulate the crack propagation process and generate 3D crack virtual content. Then, construct a 3D crack model library and use the 3D crack model library combined with lightweight voxel models of physical model specimens of geological bodies from the physical simulation test of rockburst disaster in deep engineering to simulate the crack propagation process and generate 3D crack effects. A3: Determine the rockburst-affected zone based on the crack propagation range of the generated 3D crack effect, and simulate the rockburst damage effect.

4. The augmented reality observation method for deep engineering rockburst physical simulation experiment according to claim 3, characterized in that, A1 specifically includes the following steps: A1.1: Read multiple geological structure map model files submitted by the user, and extract and process geological structure data for each geological structure map model file to obtain processed geological structure data; Specifically, the process involves: first, extracting geological structure data from the geological structure map model file; then, deduplicating, completing, and standardizing the format of the extracted geological structure data; and finally, organizing the geological data corresponding to each geological structure map model file into a structured JSON object to obtain the processed geological structure data. A1.2: Based on the geological structure map model file, an initial three-dimensional geological mesh model is constructed through a three-dimensional rendering engine, and the processed geological structure data is maintained in the model database of the initial three-dimensional geological mesh model; The model database includes the mapping relationship between each vertex in the three-dimensional geological network model and geological structure data. Each mapping relationship includes the unique ID of the vertex, lithological properties, fluid properties, geological structure properties, and spatial location coordinates of the vertex. A1.3: For each geological structure map model file, the initial three-dimensional geological mesh model is queried from its respective model database to confirm its lithological properties. UV maps are then mapped and physical properties, including density and compressive strength, are bound to each initial three-dimensional geological network model according to the lithological properties. A1.4: Using spatial transformation alignment technology, multiple three-dimensional geological mesh models that have been UV-mapped and bound with physical properties are seamlessly stitched together according to the spatial coordinate relationship between the geological structure map model files into a complete three-dimensional geological mesh model of the physical model sample of the geological body for physical simulation test of rockburst disaster in deep engineering. A1.5: The sparse octree algorithm is used to recursively divide the three-dimensional geological mesh model of the physical model sample of the geological body in the physical simulation test of rockburst disaster in deep engineering into a sparse octree. A1.6: Traverse all non-empty leaf nodes in the sparse octree, integrate and serialize their spatial coordinates, dimensions and bound physical properties, and export them in sparse voxel format to finally generate a lightweight voxel model of the physical model sample of the geological body for the physical simulation test of rockburst disaster in deep engineering.

5. The augmented reality observation method for deep engineering rockburst physical simulation experiment according to claim 3, characterized in that, A2 specifically includes the following steps: A2.1: Obtain rockburst monitoring data of the rockburst disaster incubation process in deep engineering, and invert the mechanical causes and morphological characteristics of typical cracks based on the rockburst monitoring data to generate 3D crack virtual content, and then construct a 3D crack model library; the 3D crack model library includes 3D crack virtual content under various crack propagation modes. A2.2: Receive the location of the microseismic event from the sensor, map the three-dimensional coordinates of the microseismic event location to the corresponding spatial position of the lightweight voxel model of the physical model sample of the geological body in the physical simulation test of rockburst disaster in deep engineering, and accurately locate the crack initiation point; A2.3: Based on the direction of the local maximum principal stress at the crack initiation point and the fracture toughness of the rock, the theoretical propagation angle and length of the crack are calculated. On this basis, a random disturbance factor determined by the heterogeneity of the rock mass is introduced to simulate the uncertainty of the real crack path and obtain the corrected propagation angle and length. The random perturbation factor determined by the heterogeneity of the rock mass is used to introduce random bias when calculating the crack propagation path, so as to simulate the irregular crack propagation phenomenon caused by the inhomogeneity of the internal structure of real rock. A2.4: Based on the corrected extension angle and length, find the corresponding 3D crack virtual content from the 3D crack model library, and then dynamically embed the 3D crack virtual content into the corresponding position in the lightweight voxel model of the physical model sample of the geological body in the physical simulation test of rockburst disaster in deep engineering through mesh Boolean operation; A2.5: The distance field fusion algorithm is used to smooth the geometric joint between the embedded 3D crack virtual content and the lightweight voxel model of the physical model of the geological body of the deep engineering rockburst disaster physical simulation test. At the same time, the environmental occlusion intensity and normal map parameters of the crack area are dynamically adjusted according to the three-dimensional embedding depth of the 3D crack virtual content to obtain the final 3D crack effect. Specifically, the distance field of the lightweight voxel model of the geological body physical model specimen in the physical simulation test of rockburst disaster in deep engineering is set as follows: The distance field of the 3D crack virtual content is Then the fusion distance field Defined as: （1）； In the formula, Let be the coordinate vector of any point in three-dimensional space. A positive parameter to control the smoothness of the blending process; Subsequently, the fusion distance field was analyzed. The zero-level set is re-meshed to generate the final transition surface, visually achieving a natural transition between the 3D crack virtual content and the lightweight voxel model of the physical model specimen of the geological body in the physical simulation test of rockburst disaster in deep engineering.

6. The augmented reality observation method for deep engineering rockburst physical simulation experiment according to claim 3, characterized in that, The A3 specifically includes the following steps: A3.1: Calculate and delineate the rockburst influence area based on the expansion range and spatial distribution of the generated 3D crack effect; The calculation and delineation of the rockburst impact zone are achieved by constructing a unified rockburst risk function, which is defined as follows: （2）； In the formula, Due to the risk of rockburst, The coordinate vector is The distance of the crack at that location The coordinate vector is Stress intensity at the point, The critical stress threshold for rockburst. The attenuation coefficient is the effect of cracks. The threshold for judging rockburst risk is set as follows: When satisfied When this is the case, the coordinate vector can be considered as... The location fell into the area affected by the rockburst. A3.2: Using the implicit function surface trimming algorithm, target rock blocks located within the rockburst influence zone are separated from the lightweight voxel model of the physical model sample of the geological body in the physical simulation test of rockburst disaster in deep engineering; the target rock blocks refer to specific rock mass areas that will be damaged and need to be dynamically simulated in the physical simulation test of rockburst disaster in deep engineering. Specifically, the implicit clipping function is defined as follows: （3）； In the formula, For pruning implicit functions, This is an indicator function for the rockburst influence zone, within which... Set to 1, and set to 0 for the outside. The cutting factor; When satisfied The cutting surface is formed, and the lightweight voxel model of the physical model sample of the geological body physical model test of the physical simulation test of rockburst disaster in deep engineering can be divided into two parts: the target rock block and the non-rockburst area by extracting the isosurface. A3.3: Based on the energy released by the rockburst, a dynamic fragmentation algorithm is used to decompose the target rock block into fragments of random size, and each fragment inherits the physical properties of the original target rock block. The dynamic fragmentation algorithm is based on the displacement and stress field responses of the target rock block, and utilizes an energy-driven implicit fracture function. Generate a fragmented distribution when the following conditions are met: When the region is greater than or equal to the fragmentation trigger threshold, it is marked as a fragmentable region. Subsequently, several fragmentation planes are generated by applying random perturbation seeds within the fragmentable region, and the fracture implicit function is used. The spatial gradient controls its size and orientation, ultimately forming several fragments of random size; （4）； In the formula, For the implicit function of fracture, The instantaneous available energy of a rockburst The local stress of the target rock block. Fracture toughness; A3.4: Simulate the flight trajectory and collision behavior of debris under the influence of gravity and air resistance, while reproducing the collapse and visual impact effects during the rockburst process; The specific method is as follows: use the physics engine to set the mass, shape and initial velocity of each fragment, and calculate the displacement, rotation and collision bounce of the fragment in the scene in real time; The specific method for reproducing the collapse and visual impact effects during a rockburst is as follows: The emissivity, lifetime, and velocity distribution function of dust particles are defined, and the dust particles have different sizes. The starting point of the dust particles is the edge of the debris, and the transparency of the dust particles decreases over time to simulate the generation and dispersion of dust. The emissivity of the dust particles is the number of dust particles generated per unit time. The lifetime is the duration from the generation to the disappearance of each dust particle. The velocity distribution function is used to describe the law followed by the velocity of dust particles. The shock wave generated by the energy released by rockburst is simplified as spherical expansion, and a shock wave front model is constructed. ,in The location of the shock wave front. For time, To measure the propagation speed of the shock wave front, brightness enhancement and transparency attenuation are applied at a set distance from the shock wave front, so that the fragments produce a brief brightening effect when approaching the shock wave and gradually attenuate as they move away from the wave front, thus realistically reproducing the collapse, dust, and impact scouring effects during the rockburst process.

7. The augmented reality observation method for deep engineering rockburst physical simulation experiment according to claim 1, characterized in that, The establishment of the spatial binding relationship between the augmented reality virtual content generation system for the rockburst incubation process and the internal structure of the physical model sample of a real geological body specifically includes the following steps: B1: Construct a local spatial coordinate system; The local spatial coordinate system takes the lower left rear vertex of the physical model specimen of the real geological body as its origin, and the X', Y', and Z' axes are respectively aligned with the main directions of the physical model specimen of the real geological body; B2: Read the geological structure map model file, use edge detection and contour tracking algorithms to perform vectorized re-identification of key contours, realize semantic segmentation of the closed contour regions formed by the key contours, obtain several closed contour regions, and label each closed contour region with different lithologies according to the mapping relationship between each vertex of the 3D geological network model in the model database and the geological structure data. At the same time, calculate and store the geometric center coordinates of each closed contour region in the local spatial coordinate system; the key contours include strata boundaries and fault lines; B3: Based on the generated key contours, the two-dimensional fault lines are meshed in three-dimensional space using a triangular sectioning algorithm in the local spatial coordinate system to construct a three-dimensional triangular mesh surface for characterizing the spatial distribution of geological structures, thus forming a three-dimensional topological structure. The triangulation algorithm denotes the set of nodes on the key contour as: （5）； In the formula, For the set of nodes on the key contour, For the three-dimensional nodes on the key contour, This refers to the index of the nodes on the key contour. This represents the total number of nodes on the key contour. These are the planar coordinates of the nodes in the local spatial coordinate system. The elevation of the node is obtained by interpolation from the elevation annotations in the geological structure map model file; Node set on key contour Let the set of construction constraint edges, consisting of the input point set, be denoted as: （6）； in, To construct the set of constraint edges, Indicates that by node With nodes The constructed constraint edges are defined as the set of all possible triangulations. And set a triangle The smallest interior angle is Therefore, the constrained Delaunay triangulation is represented as: （7）； in, The optimal cut set; The final 3D triangular mesh surface is obtained from the optimal cut set. All the triangles in the array are represented as follows: （8）； in, It is a three-dimensional triangular mesh surface. This is the union operation for sets; B4: Uniformly sample the surface of the three-dimensional geological grid model of the physical model sample of the geological body for the physical simulation test of rockburst disaster in deep engineering generated in A1.4 to generate a basic surface point cloud, and assign the lithological and physical properties mapped in step A1.3 to each point in the basic surface point cloud. Convert the voxel center point and its physical properties of the lightweight voxel model of the physical model sample of the physical simulation test of rockburst disaster in deep engineering generated in step A1.6 into an internal attribute point cloud. The basic surface point cloud and the internal attribute point cloud together constitute a three-dimensional attribute point cloud. B5: The internal structural feature library is composed of a lightweight voxel model and a three-dimensional topological structure of the physical model sample of the geological body in the physical simulation test of rockburst disaster in deep engineering. The crack initiation point is located in the internal structural feature library and is used as the starting position of the 3D crack virtual content. At the same time, the expansion angle of the 3D crack virtual content is made along the normal direction of the joint surface to achieve spatial binding of the crack virtual effect. B6: Based on the defined rockburst influence area and the magnitude of rockburst energy release, locate the origin coordinates of the explosion source in the local spatial coordinate system, and delineate the spherical influence range of the rockburst according to the magnitude of rockburst energy release. Lock the center coordinates of the rockburst explosion and visual impact effect during the rockburst process to the origin coordinates of the explosion source, thereby realizing the spatial binding of the rockburst explosion source area. B7: Under a unified local spatial coordinate system, calculate the distance from the crack mesh vertex to the nearest joint surface. If it exceeds the threshold, project and correct it in the direction of the joint surface. At the same time, based on the spatial coordinates of the points contained in the 3D attribute point cloud and the compressive strength attribute in the physical attributes, establish a mapping function that associates the compressive strength value with the coordinates of the stress texture. Then, during rendering, the compressive strength value corresponding to the vertex position is sampled in real time. The coordinates that should be sampled on the stress texture are calculated through the mapping function, thereby replacing the traditional static UV mapping and realizing the dynamic adjustment of the stress texture and its coordinates, thus completing the spatial binding of the stress field texture. The crack mesh vertices refer to the geometric vertices of the 3D crack virtual content dynamically embedded into the lightweight voxel model of the geological body physical model specimen of the deep engineering rockburst disaster physical simulation test through mesh Boolean operations; B8: Integrate all spatial binding relationships generated by B5-B7 under a unified local spatial coordinate system, and output a structured table of spatial binding relationships between the augmented reality virtual content generation system for rockburst incubation process and the physical model specimen of the real geological body.

8. The augmented reality observation method for deep engineering rockburst physical simulation experiment according to claim 1, characterized in that, The spatial binding relationship between the augmented reality virtual content generation system for the rockburst incubation process and the physical model specimen of the real geological body enables the identification, registration, and real-time tracking of the physical model specimen of the real geological body, thereby superimposing the augmented reality virtual content generation system for the rockburst incubation process onto the physical model specimen of the real geological body. Specifically, this includes the following steps: C1: A two-level registration strategy from coarse to fine is adopted to solve for a 6-DOF fine registration pose transformation matrix, so as to achieve the initial matching and superposition of the augmented reality virtual content generation system of rockburst gestation process in complex environment with the physical model of real geological body; Specifically, in the coarse registration stage, points located within homogeneous lithological regions and far from lithological boundaries within the generated 3D attribute point cloud are used as feature points. An approximate nearest neighbor search strategy based on lithological semantic constraints is employed to quickly match these feature points with the environmental point cloud acquired in real-time via mobile devices. That is, for each feature point... In environmental point clouds Find the nearest neighbor that satisfies lithological consistency. Establish a point-to-point correspondence between the augmented reality virtual content generator of the rockburst gestation process and the physical model of the real geological body: （9）； in, For the set of feature points, Environmental point clouds The point in the middle; The feature point number, Environmental point clouds The numbering of points in the middle; Subsequently, the RANSAC algorithm was used to match the nearest neighbors. After filtering, interior points are obtained, forming a set of matching interior points. The initial rigid body transformation matrix is ​​solved by minimizing the interior point error. : （10）； in, Let be the initial rigid body transformation matrix. Indicates feature points Coordinates mapped to the environment point cloud coordinate system via rigid body transformation; The rigid body transformation matrix; The obtained initial rigid body transformation matrix Least squares optimization is performed to obtain the optimized initial pose transformation matrix from the real-world coordinate system to the local space coordinate system. : （11）； The fine registration stage uses the initial pose transformation matrix obtained from coarse registration. Based on this, a bidirectional iterative nearest-point registration method based on joint surface normal constraints is executed on the 3D attribute point cloud of key geological structural areas to solve for the 6-DOF fine registration pose transformation matrix. The key geological structural regions are faults, dominant joint sets, and lithological interfaces on the constructed three-dimensional triangular mesh surface. Specifically, point clouds with normal vectors are generated based on the three-dimensional coordinates corresponding to key geological structural regions. , Let a point be represented by its corresponding normal vector. , For point clouds The index of the midpoint is obtained using the initial pose transformation matrix. Point cloud Projected onto a physical model sample of a real geological body, in an environmental point cloud Search for corresponding points in the middle to construct a scene point cloud. , In the point cloud of the scene and the point The corresponding points, based on point cloud pairs and normal vector The following optimization objective is constructed, and the fine-fit pose transformation matrix is ​​solved. : （12）； in, Indicates the point Coordinates mapped to the environment point cloud coordinate system via rigid body transformation; C2: Inject the 6-DOF fine-fit pose transformation matrix into the visual inertial odometry system, start real-time dynamic tracking, and realize real-time matching and superposition between the augmented reality virtual content generation system of rockburst incubation process and the physical model of real geological body; Specifically, the front-end visual inertial odometry system acquires images of a real geological physical model and uses an ORB feature extractor to adaptively extract texture feature points on the surface of the real geological physical model sample. Simultaneously, it receives IMU data and obtains the pose changes between adjacent frames by pre-integrating the IMU data, compensating for motion blur caused by camera movement, and achieving preliminary inter-frame tracking. The texture feature points refer to the visual texture features on the surface of the geological physical model sample. In a visual inertial odometry system, the back-end visual inertial odometry optimizes the tracking results of the front-end visual inertial odometry by constructing and continuously optimizing a factor graph. This factor graph simultaneously processes four constraints, including: IMU dynamic constraints, geological point cloud alignment constraints, visual reprojection constraints, and geological attitude correction constraints. The IMU dynamic constraints are as follows: （13）； （14）； （15）； In the formula, For a moment The position vector of the camera, Let the velocity vector of the camera be... The direction is represented by the rotation matrix. The acceleration measured by the IMU The angular velocity measured by the IMU. The vector of gravitational acceleration. For time intervals; The visual reprojection constraint is: （16）； In the formula, This is the reprojection error vector on the two-dimensional pixel plane. These are the pixel coordinates of the texture feature points observed in the current frame. The three-dimensional coordinates of the texture feature points. For camera projection function; The alignment constraint for the geological point cloud is: （17）； In the formula, This refers to the distance error between point clouds. For the 3D attribute point cloud under the current estimated pose, Environmental point clouds The point that minimizes the distance error between point clouds; The geological occurrence correction constraint is: （18）； In the formula, For the joint surface normal distance error, For the joint surface normal in the spatial binding relation table Let be a known point on the joint surface; A triple-progressive fault-tolerance mechanism is adopted during real-time dynamic tracking: First, an illumination-adaptive re-identification mechanism is adopted. When changes in ambient illumination cause the degradation of texture feature points to exceed a set level, the system automatically switches to a matching mode based on spectral features. It utilizes the stable reflection characteristics of different lithologies in specific bands to perform feature matching and tracking maintenance. Secondly, topological reasoning under partial occlusion is adopted. When a part of the physical model sample of the real geological body is temporarily occluded, the fault topological reasoning engine is activated. The fault topological reasoning engine uses the generated three-dimensional triangular mesh surface and the spatial binding relationship table as the global prior knowledge base. When the occlusion of the region is detected, the spatial position of the geological structural features that can still be observed in the current frame image is used as real-time input. By solving the geometric constraints and interpolating the spatial path on the prior three-dimensional triangular mesh surface, the most likely orientation and geometric shape of the occluded region are inferred. Then, it is converted into geological point cloud alignment constraints and injected into the factor map for optimization. In addition, an absolute pose recalibration technique is adopted. When the tracking quality is detected to have degraded beyond a set threshold or be completely lost, a repositioning process is triggered. The spatial position provided by the 6-DOF fine-fit pose transformation matrix is ​​used as the initial value for retrieval. A fast search and matching is performed in a pre-built keyframe database containing multi-view information. If the match is successful, the tracking pose is restored using the matched image. If it fails, the final absolute pose correction based on borehole coordinates is enabled. That is, the original borehole spatial coordinates with absolute accuracy recorded in the geological structure map model file are used as fixed references. The environmental point cloud and texture features currently perceived by the visual inertial odometry are forced to align with these fixed references to achieve reliable reset. C3: Through spatial consistency verification, calculate in real time the Hausdorff distance between the environmental point cloud and extracted texture feature points perceived by the visual inertial odometry and the generated 3D attribute point cloud on key geological structures such as fault lines and rock strata interfaces. Automatically trigger local re-registration for areas where the deviation exceeds the threshold; otherwise, directly execute C4. The local reregistration involves repeatedly executing the bidirectional iterative nearest-point registration method based on joint surface phase constraints described in step 3.1 for regions where the deviation exceeds the threshold, outputting a 6-DOF fine-registration pose matrix with dynamic health scoring. When the dynamic health score remains stable above the preset threshold, proceed to step C4. C4: Sends tracking-ready commands and the final 6-DOF fine-tuned pose matrix to the augmented reality rendering engine. The augmented reality rendering engine calls the spatial binding relationship table to drive the augmented reality virtual content generation system of the rockburst gestation process to perform real-time rendering on the physical model of the real geological body.

9. An augmented reality observation system for deep engineering rockburst physical simulation experiments, used to implement the augmented reality observation method for deep engineering rockburst physical simulation experiments as described in any one of claims 1-8, characterized in that, include: Data layer: Used for storing and managing multi-source data from physical simulation experiments of rockburst hazards in deep engineering; Model layer: Used for multi-source data from physical simulation experiments of rockburst disasters in deep engineering based on the data layer, to construct an augmented reality virtual content generation system for the rockburst incubation process and to store and maintain it; Functional layer: Used to establish the spatial binding relationship between the augmented reality virtual content generation system of rockburst incubation process output by the data layer and the internal structure of the physical model sample of the real geological body; Based on the spatial binding relationship between the augmented reality virtual content generation system for the rockburst incubation process and the physical model specimens of real geological bodies, the identification, registration, and real-time tracking of the physical model specimens of real geological bodies are achieved, thereby superimposing the augmented reality virtual content generation system for the rockburst incubation process onto the physical model specimens of real geological bodies. Service Layer: Responsible for connecting the functional layer and the interaction layer. Based on the user input from the interaction layer, the service layer encapsulates the algorithm logic through a microservice architecture and provides standardized interfaces to support the entire chain of augmented reality observation business. Specifically, it includes spatial registration service and virtual-real fusion rendering service. The spatial registration service is the spatial binding relationship between the augmented reality virtual content generation system of rockburst gestation process and the physical model sample of real geological body. Interaction layer: Used for transparent observation by users through mobile devices; Users point their mobile devices at a physical model sample of a real geological body, and the device screen will generate a semi-section, adjustable augmented reality virtual content generation system of rockburst incubation process in real time, providing an intuitive view of the internal rockburst disaster incubation process and offering a playback function.

10. A computer program product, characterized in that, Includes a computer program or instructions that, when executed by a processor, implement the augmented reality observation method for deep engineering rockburst physical simulation experiment as described in any one of claims 1-8.

Images