Three-dimensional measurement system for geological structures based on inertial navigation

Abstract

The present application relates to geological exploration and three-dimensional modeling technical field, disclose a three-dimensional measurement system of geological structure based on inertial navigation. The system includes through the processing of inertial measurement data stream by solving module, the instantaneous spatial attitude and relative displacement vector of the measurement carrier are obtained, and the basic three-dimensional geological structure model is constructed. The confidence of the solving result is calculated in real time by the quantization module, and the space-time uncertainty distribution field is generated; the model correction module uses the distribution field to carry out error propagation analysis, and the low confidence area in the model is iteratively optimized in geometric form. The system imports the lithology, stratum and structure interpretation data from the outside, and associates it to the model as an attribute. The optimized model, associated attribute and uncertainty distribution field are packaged into a composite three-dimensional geological measurement result file by the result packaging module. The present application realizes the dynamic quantification and visualization of measurement uncertainty, and can independently optimize the model accordingly, improves the reliability and precision of inertial navigation three-dimensional geological modeling.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of geological exploration and three-dimensional modeling, in particular to a three-dimensional geological structure measurement system based on inertial navigation. BACKGROUND

[0002] Currently, the three-dimensional geological structure measurement based on inertial navigation technology generally adopts a flow data processing mode. The system collects motion data through an inertial measurement component, obtains the attitude and displacement of the carrier through calculation, and then constructs a three-dimensional geological structure framework composed of a triangular facet network. After completing the geometric model construction, the attribute information such as lithology and stratum is usually extracted from an independent geological database and associated with the corresponding area of the model to form the final geological interpretation result. This technical path realizes the conversion from physical measurement to digital model.

[0003] The existing technical solution has obvious defects. The inertial navigation calculation itself has cumulative error, and the error continuously diverges with the increase of time and motion distance. The conventional system only outputs the position and attitude of the optimal calculation value, and lacks quantitative evaluation of the real-time reliability of the calculation result itself. The three-dimensional geometric model constructed hides uneven precision distribution, and the user cannot know which areas in the model have high confidence and which areas have been distorted due to error accumulation. At the same time, the post-correction of the model often relies on manual experience or introduces external absolute positioning data for global correction, and cannot automatically and finely optimize the local area according to the internal quality index of the model.

[0004] The technical problem to be solved is how to dynamically quantify and visualize the spatial and temporal variation of the precision uncertainty in the inertial navigation three-dimensional measurement process, and how to use this quantified uncertainty information to drive the three-dimensional geological model to perform self-check and iterative optimization, and output reliable model results with precision specifications from the system level. SUMMARY

[0005] The present application provides a three-dimensional geological structure measurement system based on inertial navigation to solve the problems raised in the background.

[0006] To achieve the above-mentioned purpose, the present application provides a three-dimensional geological structure measurement system based on inertial navigation, which comprises:

[0007] A calculation module receives a continuous motion data stream from an inertial measurement component and sequentially calculates the instantaneous spatial attitude and relative displacement vector of the measurement carrier;

[0008] A three-dimensional model construction module receives the instantaneous spatial attitude and relative displacement vector and constructs a basic three-dimensional geological structure model expressed by a triangular facet network;

[0009] a geological attribute attachment device, which imports lithology, stratigraphic and structural interpretation data corresponding to a measurement area from an external geological database and associates and maps the lithology, stratigraphic and structural interpretation data as attribute fields to corresponding triangular facets in the basic three-dimensional geological structure model;

[0010] a quantification module, which calculates and records the solving confidence of each of the instantaneous spatial attitude and relative displacement vector in real time during the solving process of the solving module, and generates a spatio-temporal uncertainty distribution field synchronized with the basic three-dimensional geological structure model;

[0011] a model correction module, which performs error propagation analysis on the basic three-dimensional geological structure model by using the spatio-temporal uncertainty distribution field, and iteratively optimizes and adjusts the geometric shape of a low-confidence area in the basic three-dimensional geological structure model according to the analysis result;

[0012] a result packaging module, which standardizes and packages the iteratively optimized and adjusted geometric shape, the associated and mapped attribute fields and the spatio-temporal uncertainty distribution field, and outputs a composite three-dimensional geological measurement result file that can be independently analyzed.

[0013] Preferably, the receiving the instantaneous spatial attitude and relative displacement vector and constructing a basic three-dimensional geological structure model expressed by a triangular facet network comprises:

[0014] receiving the instantaneous spatial attitude and relative displacement vector, and performing spline interpolation on continuous discrete solving points according to a time stamp sequence to form a continuous and smooth three-dimensional spatial trajectory line;

[0015] performing hierarchical voxel division on a spatial region passed through by the continuous and smooth three-dimensional spatial trajectory line according to a preset spatial grid resolution to generate a multi-level resolution voxelized spatial container;

[0016] generating a spatial discrete point set representing a geological surface shape in the multi-level resolution voxelized spatial container according to a set point density distribution rule based on the continuous and smooth three-dimensional spatial trajectory line;

[0017] performing local neighborhood normal estimation and curvature analysis on the spatial discrete point set, and identifying and extracting a spatial plane equation of a geological structure surface based on normal clustering and curvature characteristics;

[0018] fusing the spatial discrete point set and the spatial plane equation of the geological structure surface to construct a basic three-dimensional geological structure model expressed by a triangular facet network.

[0019] Preferably, the spline interpolation on continuous discrete solving points according to a time stamp sequence to form a continuous and smooth three-dimensional spatial trajectory line comprises:

[0020] acquiring a sequence of discrete solution points arranged in time sequence from the solution module, each discrete solution point containing three-dimensional coordinates and attitude quaternion;

[0021] using a non-uniform rational B-spline curve fitting algorithm to generate an initial spatial curve passing through all the control points, with the sequence of discrete solution points as the control points;

[0022] performing fairness evaluation on the initial spatial curve, and calculating the curvature variation rate of the curve on each segment;

[0023] inserting additional interpolation control points in the curve segment where the curvature variation rate exceeds the preset threshold, and using a piecewise cubic Hermite interpolation method to locally re-fit the initial spatial curve;

[0024] performing parameterization processing on the re-fitted overall curve to generate a parameter equation with time as the only parameter, and the curve expressed by the parameter equation is the continuous and smooth three-dimensional spatial trajectory line.

[0025] Preferably, the hierarchical voxel division according to the preset spatial grid resolution generates a multi-resolution voxelized spatial container, including:

[0026] setting a basic space bounding box that completely contains the continuous and smooth three-dimensional spatial trajectory line;

[0027] equally dividing the basic space bounding box at the highest resolution to form a voxel grid of the finest level;

[0028] gradually generating voxel grids of coarser resolutions by merging adjacent multiple finest level voxels into one large voxel, thereby forming a multi-resolution voxel pyramid structure from fine to coarse;

[0029] establishing a spatial index for each voxel in the voxel pyramid structure, and marking the voxels passed through by the continuous and smooth three-dimensional spatial trajectory line as active voxels;

[0030] creating a dynamic storage container for storing and managing all active voxels and their spatial indexes at different levels, thereby forming the multi-resolution voxelized spatial container;

[0031] According to the continuous and smooth three-dimensional spatial trajectory line, a set of spatial discrete points representing the morphology of the geological surface is generated in the multi-resolution voxelized spatial container according to a set point density distribution rule, including:

[0032] selecting a specific resolution level in the multi-resolution voxelized spatial container that matches the current detail performance requirement;

[0033] The dense sampling points are generated at equal arc length intervals on the continuous smooth three-dimensional space trajectory line;

[0034] In the normal plane of each sampling point, a specified number of spatial points are randomly generated within the active voxel range of the specific resolution level according to a preset point cloud lateral diffusion radius and point spacing;

[0035] All generated spatial points are subjected to a de-duplication process, and spatial points falling within the marked non-geological obstacle voxels are removed;

[0036] The three-dimensional coordinates of all finally retained spatial points are collected to form the spatial discrete point set representing the geological surface morphology.

[0037] Preferably, the spatial plane equation of the geological structure surface is identified and extracted according to the normal clustering and curvature characteristics, comprising:

[0038] S1: For each point in the spatial discrete point set, calculate its K-neighbor point set, and estimate the local normal vector and curvature value of the point using principal component analysis method;

[0039] S2: Perform clustering analysis according to the local normal vector direction of all points, and group points with similar normal vector directions into the same candidate point set;

[0040] S3: For each candidate point set, estimate the optimal spatial plane model using the iterative random sample consensus algorithm, and calculate the distance residual of the inliers in the candidate point set to the optimal spatial plane model;

[0041] S4: Select the optimal spatial plane model with a higher inlier proportion than a set threshold and the smallest distance residual sum of squares, and output its mathematical equation parameters as a geological structure surface;

[0042] Repeat S1 to S4 until all significant candidate point sets are processed, thereby extracting multiple spatial plane equations of the geological structure surfaces.

[0043] Preferably, the spatial discrete point set and the spatial plane equation of the geological structure surface are fused to construct a basic three-dimensional geological structure model expressed by a triangular facet network, comprising:

[0044] Using the spatial discrete point set as constraint points, an initial triangular mesh surface is generated in three-dimensional space using the Delaunay triangulation algorithm;

[0045] Each spatial plane equation of the geological structure surface is used as a geometric constraint condition to cut the initial triangular mesh surface, so that the mesh is split at the structure surface boundary;

[0046] Splitting the triangular mesh, eliminating the abnormal triangles and ensuring the mesh edges near the geological structure surface align with the structure surface boundary;

[0047] Storing the vertex coordinates, triangular patch indexes and normal vector information of the optimized triangular mesh, and constructing the basic three-dimensional geological structure model expressed by the triangular patch network.

[0048] Preferably, the mapping of the lithology, stratigraphic and structural interpretation data as attribute fields to the corresponding triangular patches in the basic three-dimensional geological structure model comprises:

[0049] Analyzing the lithology, stratigraphic and structural interpretation data imported from the external geological database and converting it into a standardized data format containing attribute key-value pairs;

[0050] In the basic three-dimensional geological structure model, calculating the geometric center point coordinates of each triangular patch;

[0051] Matching the geometric center point coordinates of each triangular patch with the spatialized attribute map provided by the external geological database to determine the lithology code, stratigraphic unit code and structural attribution of the triangular patch;

[0052] In the data structure of the basic three-dimensional geological structure model, creating additional attribute fields for each triangular patch and writing the matched lithology code, stratigraphic unit code and structural attribution into the corresponding additional attribute fields to complete the attribute association mapping.

[0053] Preferably, the real-time calculation and recording of the calculation confidence of each instantaneous spatial pose and relative displacement vector generates a spatiotemporal uncertainty distribution field synchronized with the basic three-dimensional geological structure model, comprising:

[0054] At each calculation of the calculation module, synchronously obtaining the original reading noise level of the inertial sensor, the calculation iteration residual and the consistency measure with the previous calculation result;

[0055] According to the pre-built error propagation model, the original reading noise level, the calculation iteration residual and the consistency measure are integrated to calculate the joint uncertainty estimate of the instantaneous spatial pose and relative displacement vector obtained by this calculation, which is the reciprocal of the calculation confidence at this time;

[0056] Recording the timestamp, spatial position and corresponding joint uncertainty estimate of each calculation time as a spatiotemporal data point;

[0057] Spatially interpolating all spatiotemporal data points to generate a scalar field that covers the entire measurement spatiotemporal domain and continuously changes in space and time dimensions, which is the spatiotemporal uncertainty distribution field.

[0058] Preferably, the error propagation analysis of the basic three-dimensional geological structure model is performed using the spatiotemporal uncertainty distribution field, and the geometric shape of the low-confidence area in the basic three-dimensional geological structure model is iteratively optimized and adjusted according to the analysis result, comprising:

[0059] The spatiotemporal uncertainty distribution field is spatially superimposed on the basic three-dimensional geological structure model, and the joint uncertainty estimate corresponding to the geometric center of each triangular facet is read;

[0060] Triangular facets with joint uncertainty estimates exceeding the allowed upper limit are marked as low-confidence areas;

[0061] For triangular facets in the low-confidence area, the optimized coordinates of the vertices in this area are recalculated using the moving least squares method surface fitting technique according to the geometric trend of the adjacent area and the spatial plane equation constraint of the geological structure surface;

[0062] The optimized vertex coordinates are used to update the basic three-dimensional geological structure model, and the local curvature and structure surface fitting degree of the updated model are recalculated;

[0063] The region marking and vertex optimization process is iteratively performed with the structure surface fitting degree as the convergence criterion until the geometric shape of all areas meets the preset structure surface fitting degree requirement, and the iterative optimization and adjustment is completed.

[0064] Preferably, the result packaging module performs standardized packaging to output a composite three-dimensional geological survey result file, comprising:

[0065] Defining a hierarchical data structure containing a geometric grid layer, an attribute data layer, and a metadata layer;

[0066] Converting the geometric shape data after iterative optimization and adjustment into a standardized triangular mesh data format and storing it in the geometric grid layer;

[0067] Encoding the associated attribute fields and the data of the spatiotemporal uncertainty distribution field in key-value pair format and storing them in the attribute data layer;

[0068] Writing the measurement task identifier, coordinate system information, data version, and packaging time information into the metadata layer;

[0069] Merging and compressing the data of the geometric grid layer, attribute data layer, and metadata layer into a binary stream;

[0070] Adding a custom file header to the front of the compressed data stream, which contains the starting offset, length, and decompression identifier of each data layer, and finally outputting the composite three-dimensional geological survey result file that can be independently parsed.

[0071] Compared with the prior art, the present application has the following advantages:

[0072] By synchronously calculating the solving confidence of each instantaneous attitude and displacement vector in the data solving process, and generating a three-dimensional model space-time synchronized uncertainty distribution field, the dynamic point-by-point quantification of measurement accuracy is realized. The distribution field records the propagation path and size of the error, so that each triangular facet of the model corresponds to a clear confidence index. The traditionally opaque model output is transformed into a spatial data volume with high transparency and complete metadata reliability. Users can intuitively identify the high reliability area and potential distortion area of the model.

[0073] Using the generated space-time uncertainty distribution field for error propagation analysis, and based on this, the geometric form of the low confidence area in the model is iteratively optimized. The accuracy evaluation data is transformed from a post-reference information into an active control parameter in the model generation process. The system can automatically identify the model sections with serious error accumulation, and adjust their spatial form through algorithm. Without relying on external absolute reference, it suppresses the diffusion and amplification of inertial navigation error in the three-dimensional model, and improves the geometric fidelity and internal consistency of the model as a whole. BRIEF DESCRIPTION OF DRAWINGS

[0074] Figure 1 The timing diagram of the three-dimensional geological structure measurement system based on inertial navigation described in the present application;

[0075] Figure 2 The flowchart for constructing the basic three-dimensional geological structure model;

[0076] Figure 3 The flowchart for voxelization and spatial discrete point generation;

[0077] Figure 4 The solving performance analysis diagram for the inertial solving and trajectory generation stage;

[0078] Figure 5 The quality evaluation diagram of the three-dimensional geological structure measurement results. DETAILED DESCRIPTION

[0079] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0080] Please refer to Figure 1The present application provides a three-dimensional geological structure measurement system based on inertial navigation, which comprises: when the system starts to work, a calculation module continuously receives a continuous motion data stream from an inertial measurement component and sequentially calculates an instantaneous spatial attitude and a relative displacement vector of a measurement carrier. A three-dimensional model construction module receives the instantaneous spatial attitude and the relative displacement vector, and the core task of the three-dimensional model construction module is to construct a basic three-dimensional geological structure model expressed by a triangular facet network. A geological attribute attachment device imports lithology, stratum and structure interpretation data corresponding to a measurement area from an external geological database, and maps the data as attribute fields to corresponding triangular facets in the constructed basic three-dimensional geological structure model. While the calculation module is running, a quantification module calculates and records a calculation confidence degree of each instantaneous spatial attitude and relative displacement vector in real time, thereby generating a time-space uncertainty distribution field synchronized with the basic three-dimensional geological structure model in time and space. A model correction module then uses the time-space uncertainty distribution field to perform error propagation analysis on the basic three-dimensional geological structure model, and iteratively optimizes and adjusts the geometric morphology in a low confidence degree area in the model according to the analysis result. Finally, an achievement packaging module standardizes and packages the iteratively optimized and adjusted geometric morphology, the mapped attribute fields and the time-space uncertainty distribution field, and outputs a composite three-dimensional geological measurement achievement file which can be independently analyzed, thereby completing the whole measurement and modeling process.

[0081] In one embodiment of the present application, reference is made to Figure 2, the three-dimensional model construction module receives the instantaneous spatial pose and relative displacement vector from the solving module, and the module performs spline interpolation on the continuous discrete solving points according to the timestamp sequence to form a continuous and smooth three-dimensional space trajectory. A sequence of discrete solving points arranged in time sequence is obtained from the solving module, and each discrete solving point in the sequence includes three-dimensional coordinates and attitude quaternion. An initial space curve passing through all control points is generated by using a non-uniform rational B-spline curve fitting algorithm with the discrete solving point sequence as the control points. Smoothness evaluation is performed on the initial space curve, and the curvature change rate of the curve on each segment is calculated. Additional interpolation control points are inserted into the curve segment whose curvature change rate exceeds the preset threshold, and the initial space curve is locally refitted by using a piecewise cubic Hermite interpolation method. The overall curve after refitting is parameterized to generate a parameter equation with time as the only parameter. The curve expressed by the parameter equation is the continuous and smooth three-dimensional space trajectory. The space region passed through by the continuous and smooth three-dimensional space trajectory is hierarchically voxelized according to a preset spatial grid resolution to generate a multi-resolution voxelized space container. According to the continuous and smooth three-dimensional space trajectory, a set of spatial discrete points representing the morphology of the geological surface is generated in the multi-resolution voxelized space container according to a set point density distribution rule. The set of spatial discrete points is subjected to local neighborhood normal estimation and curvature analysis, and the spatial plane equation of the geological structure surface is identified and extracted according to the normal clustering and curvature characteristics. The set of spatial discrete points and the spatial plane equation of the geological structure surface are fused to construct a basic three-dimensional geological structure model expressed by a triangular facet network.

[0082] In specific implementations, the three-dimensional model construction module receives the instantaneous spatial pose and relative displacement vector from the solving module, and the module performs spline interpolation on the continuous discrete solving points according to the timestamp sequence to form a continuous and smooth three-dimensional space trajectory. A sequence of discrete solving points arranged in time sequence is obtained from the solving module, and each discrete solving point in the sequence includes three-dimensional coordinates and attitude quaternion. An initial space curve passing through all control points is generated by using a non-uniform rational B-spline curve fitting algorithm with the discrete solving point sequence as the control points. Smoothness evaluation is performed on the initial space curve, and the curvature change rate of the curve on each segment is calculated. Additional interpolation control points are inserted into the curve segment whose curvature change rate exceeds the preset threshold, and the initial space curve is locally refitted by using a piecewise cubic Hermite interpolation method. The overall curve after refitting is parameterized to generate a parameter equation with time as the only parameter. The curve expressed by the parameter equation is the continuous and smooth three-dimensional space trajectory. In some embodiments, the parameter equation can be expressed as:

[0083]

[0084] wherein: represents the parameter a three-dimensional spatial coordinate of the point in the three-dimensional space, is a normalized parameter linearly related to the time stamp, represents the non-uniform rational B-spline basis function, is a three-dimensional coordinate control point of the point in the sequence of discrete solution points, is the total number of control points.

[0085] It can be understood that the three-dimensional model construction module divides the spatial region passed through by the continuous smooth three-dimensional spatial trajectory line into hierarchical voxels according to a preset spatial grid resolution, to generate a multi-resolution voxelized spatial container. In a specific implementation, a basic space bounding box completely containing the continuous smooth three-dimensional spatial trajectory line is set, and the basic space bounding box is equally divided to form a voxel grid of the finest level at the highest resolution. A voxel grid of a coarser resolution is generated at each level by merging a plurality of finest level voxels adjacent to each other into one large voxel, thereby constituting a multi-resolution voxel pyramid structure from fine to coarse. A spatial index is established for each voxel in the voxel pyramid structure, and a voxel passed through by the continuous smooth three-dimensional spatial trajectory line is marked as an active voxel, to create a dynamic storage container. The dynamic storage container stores and manages all active voxels and their spatial indexes hierarchically, to constitute a multi-resolution voxelized spatial container.

[0086] A set of spatial discrete points representing the geological surface morphology is generated according to a set point density distribution rule of the continuous smooth three-dimensional spatial trajectory line in the multi-resolution voxelized spatial container. In some embodiments, a specific resolution level matching the current detail performance requirement is selected in the multi-resolution voxelized spatial container, dense sampling points are generated at equal arc length intervals on the continuous smooth three-dimensional spatial trajectory line, and a specified number of spatial points are randomly generated in the range of active voxels of the specific resolution level in the normal plane of each sampling point according to a preset point cloud transverse diffusion radius and point spacing. All generated spatial points are subjected to a deduplication process, which eliminates spatial points falling within voxels marked as non-geological obstacles, and the three-dimensional coordinates of all spatial points finally retained are collected to form a set of spatial discrete points representing the geological surface morphology.

[0087] ​Optionally, local neighborhood normal estimation and curvature analysis are performed on the set of spatial discrete points, and the spatial plane equation of the geological structure surface is identified and extracted according to the normal clustering and curvature characteristics. In a specific implementation, the K-neighbor point set of each point in the set of spatial discrete points is calculated, the local normal vector and curvature value of the point are estimated by using the principal component analysis method, clustering analysis is performed according to the local normal vector direction of all points, and points with similar normal vector directions are classified into the same candidate point set. The random sample consensus algorithm is used to iteratively estimate the optimal spatial plane model for each candidate point set, and the distance residual of the inliers in the candidate point set to the optimal spatial plane model is calculated, and the optimal spatial plane model with a higher inlier proportion than a set threshold and the smallest distance residual sum of squares is selected, and the mathematical equation parameters thereof are output as a geological structure surface. The above steps are repeatedly executed until all significant candidate point sets are processed, and thus the spatial plane equations of multiple geological structure surfaces are extracted.

[0088] It can be understood that the set of spatial discrete points is fused with the spatial plane equation of the geological structure surface to construct a basic three-dimensional geological structure model expressed by a triangular facet network. In a specific implementation, the set of spatial discrete points is used as a constraint point, and a Delaunay triangulation algorithm is used to generate an initial triangular mesh surface in a three-dimensional space. The spatial plane equation of each geological structure surface is used as a geometric constraint condition to cut the initial triangular mesh surface, the geometric constraint condition causes the mesh to split at the boundary of the structure surface, local topological optimization is performed on the split triangular mesh to eliminate abnormal triangles and ensure that the mesh edges near the geological structure surface are aligned with the structure surface boundary. The vertex coordinates, triangular facet indexes, and normal vector information of the optimized triangular mesh are stored to construct a basic three-dimensional geological structure model expressed by a triangular facet network.

[0089] In an embodiment of the present application, reference is made to Figure 3, a base space bounding box completely containing the continuously smooth 3D space trajectory is set, and the base space bounding box is equally divided at the highest resolution to form a voxel grid at the finest level. Voxel grids at coarser resolutions are generated by merging multiple adjacent finest level voxels into one large voxel, thereby forming a multi-resolution voxel pyramid structure from fine to coarse. A spatial index is established for each voxel in the voxel pyramid structure, and voxels traversed by the continuously smooth 3D space trajectory are marked as active voxels. A dynamic storage container is created for storing and managing all active voxels and their spatial indexes at different levels, thereby forming a multi-level resolution voxelized space container. A specific resolution level matching the current detail performance requirement is selected in the multi-level resolution voxelized space container, and dense sampling points are generated at equal arc length intervals on the continuously smooth 3D space trajectory. A specified number of spatial points are randomly generated within the range of active voxels at the specific resolution level in the normal plane of each sampling point according to a preset point cloud lateral diffusion radius and point spacing. All generated spatial points are de-duplicated, and spatial points falling within voxels marked as non-geological obstacle voxels are removed. The three-dimensional coordinates of all the finally retained spatial points are collected to form a spatial discrete point set representing the geological surface morphology.

[0090] In a specific implementation, a base space bounding box completely containing the continuously smooth 3D space trajectory is set, and the base space bounding box is equally divided at the highest resolution to form a voxel grid at the finest level. The highest resolution determines the minimum size of the voxels. Voxel grids at coarser resolutions are generated by merging multiple adjacent finest level voxels into one large voxel, thereby forming a multi-resolution voxel pyramid structure from fine to coarse. In some embodiments, the edge length of a voxel at a certain level of the voxel pyramid structure is related to the edge length of the finest level voxel as follows:

[0091]

[0092] wherein: represents the current level index, represents the index of the finest level, is the edge length of the finest level voxel, is the edge length of the voxel at the Edge length of the hierarchical voxel. A spatial index is established for each voxel in the voxel pyramid structure, which is implemented in the form of a Morton code or a three-dimensional array coordinate. The voxels that are crossed by the continuously smooth three-dimensional space trajectory are marked as active voxels, and a dynamic storage container is created to store and manage all active voxels and their spatial indexes in a hierarchical manner, forming a multi-resolution voxelized space container.

[0093] In the multi-resolution voxelized space container, a specific resolution level that matches the current detail performance requirement is selected, and the selection of the specific resolution level is determined according to the preset rendering accuracy or computing resource constraint. Dense sampling points are generated on the continuously smooth three-dimensional space trajectory according to equal arc length intervals, which ensure that the sampling points are uniformly distributed on the trajectory. A specified number of spatial points are randomly generated in the range of active voxels at the specific resolution level in the normal plane of each sampling point according to a preset point cloud lateral diffusion radius and point spacing, and the point cloud lateral diffusion radius defines the lateral spatial range of the generated point cloud around the sampling point. In some embodiments, the process of generating spatial points is performed in a two-dimensional disc region defined in the normal plane using a Poisson disc sampling algorithm, which ensures that the generated spatial points meet the minimum point spacing constraint. All generated spatial points are subjected to a de-duplication process, which accurately compares the three-dimensional coordinates of the spatial points and removes points with duplicate coordinates. Spatial points falling within non-geological obstacle voxels defined by pre-imported geological models or obstacle annotation information are removed.

[0094] It can be understood that the generation of the set of spatial discrete points is completely performed within the framework of the multi-resolution voxelized space container, and the marking of active voxels limits the spatial range of point cloud generation, thereby improving the computing efficiency. Optionally, the point density distribution rule can be dynamically adjusted, for example, a smaller equal arc length interval and a larger point cloud lateral diffusion radius are used in a section of the trajectory with high curvature to generate more dense spatial points to depict complex geological surface morphology. In specific implementations, the process of randomly generating spatial points uses a pseudo-random number generator, which uses the spatial coordinates of the sampling points as seeds to ensure the repeatability of the generated results.

[0095] Optionally, the implementation of the dynamic storage container in memory can be an octree or a sparse voxel grid to adapt to the data management needs of large-scale geological scenes. In specific implementations, the generation process of the set of spatial discrete points is closely coupled with the voxel pyramid structure, and the generated spatial points are recorded in the active voxels of the specific resolution level to which they belong, and this association facilitates spatial queries and fast switching between multi-resolution detail levels.

[0096] In one embodiment of the present application, the K-Nearest Neighbor set of each point in the spatial discrete point set is calculated, and the local normal vector and curvature value of the point is estimated by Principal Component Analysis method. Clustering analysis is performed according to the local normal vector direction of all points, and points with similar normal vector direction are grouped into the same candidate point set. For each candidate point set, the optimal spatial plane model is iteratively estimated by Random Sample Consensus algorithm, and the distance residual of inliers in the candidate point set to the optimal spatial plane model is calculated. The optimal spatial plane model with inlier ratio higher than a set threshold and minimum distance residual sum of squares is selected, and the mathematical equation parameters of the optimal spatial plane model are output as a geological structure plane. The above steps are repeated until all significant candidate point sets are processed, and multiple spatial plane equations of geological structure planes are extracted. The spatial discrete point set is used as constraint points, and the Delaunay triangulation algorithm is used to generate an initial triangular mesh surface in three-dimensional space. The spatial plane equation of each geological structure plane is used as a geometric constraint condition to cut the initial triangular mesh surface, so that the mesh is split at the boundary of the structure plane. The split triangular mesh is locally topologically optimized to eliminate abnormal triangles and ensure that the mesh edges near the geological structure plane are aligned with the structure plane boundary. The vertex coordinates, triangular patch index, and normal vector information of the optimized triangular mesh are stored to construct a basic three-dimensional geological structure model represented by a triangular patch network.

[0097] In specific implementations, the K-Nearest Neighbor set of each point in the spatial discrete point set is calculated by constructing a KD-tree of the spatial discrete point set and performing nearest neighbor search. The local normal vector and plane feature coefficient of the point are estimated by Principal Component Analysis method, which constructs a covariance matrix of the K-Nearest Neighbor set of the point and performs eigenvalue decomposition. In some embodiments, the plane feature coefficient of a point in the spatial discrete point set can be calculated by the eigenvalue of the covariance matrix of its neighboring points, and the formula is as follows:

[0098]

[0099] wherein: , , is the eigenvalue of the covariance matrix, is the plane feature coefficient of the point , and a smaller value indicates that the point is located in an approximately planar region.

[0100] According to the local normal vector direction of all points, a clustering analysis is performed using a region growing algorithm or a clustering algorithm based on a normal vector angle threshold to classify points with similar normal vector directions into the same candidate point set. For each candidate point set, an optimal spatial plane model is iteratively estimated using a random sample consensus algorithm, which randomly selects three points from the candidate point set to calculate an initial plane model in each iteration, and calculates the distance residuals of all inliers in the candidate point set to the initial plane model. An optimal spatial plane model with a higher proportion of inliers and a minimum sum of squared distance residuals is selected, and the proportion of inliers refers to the proportion of the number of points with a distance less than a preset tolerance threshold to the plane model to the total number of points. The mathematical equation parameters of the optimal spatial plane model are output as a geological structure plane. The above steps are repeatedly performed until all significant candidate point sets are processed, thereby extracting multiple spatial plane equations of geological structure planes.

[0101] It can be understood that an initial triangular mesh surface is generated in a three-dimensional space using a Delaunay triangulation algorithm with a set of spatial discrete points as constraint points. The Delaunay triangulation algorithm ensures that the generated triangular mesh satisfies the empty circle property. The spatial plane equation of each geological structure plane is used as a geometric constraint condition to cut the initial triangular mesh surface, which means that the triangular mesh needs to be divided by the infinite plane represented by the spatial plane equation of the geological structure plane. In some embodiments, the cutting operation is implemented by calculating the intersection points of each edge of the triangular mesh and the spatial plane equation of the geological structure plane, and inserting new vertices at the intersection points, and then splitting the original triangle into multiple new triangles. The geometric constraint condition causes the mesh to split at the boundary of the structure plane.

[0102] The split triangular mesh is subjected to local topological optimization, which eliminates deformed triangles through edge flipping, vertex deletion and re-triangulation operations, and ensures that the mesh edges near the geological structure plane are aligned with the structure plane boundary. Alternatively, ensuring that the mesh edges near the geological structure plane are aligned with the structure plane boundary can be achieved by constraining the triangulation algorithm, which adds the edges formed by the intersection of the geological structure plane and the mesh as fixed constraint edges to the triangulation process. The vertex coordinates, triangle patch indices and normal vector information of the optimized triangular mesh are stored to construct a basic three-dimensional geological structure model represented by a triangular patch network. In specific implementations, the storage process saves the vertex coordinates in the form of a floating-point number array, the triangle patch indices in the form of an integer array, and the normal vector information of each vertex or patch in the form of a floating-point number array.

[0103] Optionally, the storage of the triangular facet index adopts a vertex list plus a facet list format, and each facet is composed of three indexes pointing to the vertex list. It can be understood that the basic three-dimensional geological structure model expressed by the triangular facet network is the geometric basis for subsequent geological attribute attachment and model correction operations. In specific implementation, the data structure of the basic three-dimensional geological structure model also maintains the adjacency relationship information between the vertices and the facets, which provides convenience for subsequent grid analysis and processing. In specific implementation, the data structure of the basic three-dimensional geological structure model further maintains the adjacency relationship information between the vertices and the facets on the basis of storing the vertex coordinates, the triangular facet index and the normal vector information, and the adjacency relationship information is realized by recording the index list of all triangular facets to which each vertex belongs, and simultaneously recording the index list of the adjacent triangular facets of the shared edge of each triangular facet.

[0104] In an embodiment of the present application, the geological attribute attachment device parses the lithology, stratigraphic and structural interpretation data imported from the external geological database, and converts it into a standardized data format containing attribute key-value pairs. The geometric center point coordinates of each triangular facet are calculated in the basic three-dimensional geological structure model, the geometric center point coordinates of each triangular facet are positionally matched with the spatialized attribute chart provided in the external geological database, and the lithology code, stratigraphic unit code and structural attribution of the triangular facet are determined. In the data structure of the basic three-dimensional geological structure model, an additional attribute field is created for each triangular facet, and the matched determined lithology code, stratigraphic unit code and structural attribution are written into the corresponding additional attribute field, completing the attribute correlation mapping. The quantization module synchronously obtains the original reading noise level of the inertial sensor, the iterative residual of the calculation module and the consistency measure with the previous calculation result each time the calculation module calculates. According to the pre-built error propagation model, the original reading noise level, the iterative residual of the calculation and the consistency measure are comprehensively calculated to obtain the joint uncertainty estimate of the instantaneous spatial attitude and the relative displacement vector obtained by this calculation, which is the inverse of the calculation confidence at this time. The timestamp, spatial position and corresponding joint uncertainty estimate of each calculation time are recorded as a spatiotemporal data point. Spatial interpolation is performed on all spatiotemporal data points to generate a scalar field that covers the entire measurement spatiotemporal domain and continuously changes in space and time dimensions. The scalar field is the spatiotemporal uncertainty distribution field.

[0105] In a specific implementation, the geological attribute attachment device parses lithology, stratigraphic and structural interpretation data imported from an external geological database, which can be a data source in the form of a geographic information system layer or a special geological database. The imported lithology, stratigraphic and structural interpretation data are converted into a standardized data format containing attribute key-value pairs, which enables the system to uniformly process structured attribute data from different sources. The geometric center point coordinates of each triangular facet in the basic three-dimensional geological structure model are calculated by taking the arithmetic mean of the coordinates of the three vertices of the triangular facet. The geometric center point coordinates of each triangular facet are matched with the spatialized attribute map provided by the external geological database, which is a vector or raster data layer with geographic coordinate reference, to determine the lithology code, stratigraphic unit code and structural attribution of the triangular facet. Additional attribute fields are created for each triangular facet in the data structure of the basic three-dimensional geological structure model, and the matched lithology code, stratigraphic unit code and structural attribution are written into the corresponding additional attribute fields to complete the attribute association mapping. See Table 1.

[0106] Table 1: Geological attribute matching table

[0107]

[0108] It can be understood that the quantization module synchronously obtains the raw reading noise level of the inertial sensor, the solution iteration residual, and the consistency measure with the previous solution result each time the solution module is solved. The raw reading noise level is obtained from the factory calibration parameters of the inertial sensor or real-time online estimation, the solution iteration residual is the residual norm after the filter or optimization algorithm converges, and the consistency measure is calculated by comparing the displacement difference and attitude change difference between adjacent solution points. According to the pre-built error propagation model, the raw reading noise level, the solution iteration residual, and the consistency measure are integrated to calculate the joint uncertainty estimate of the instantaneous spatial attitude and relative displacement vector obtained by this solution, and the joint uncertainty estimate is taken as the inverse of the solution confidence at this moment. In some embodiments, the calculation of the joint uncertainty estimate can use a linear combination model:

[0109]

[0110] wherein: represents the quantized value of the raw reading noise level at time , represents the quantized value of the solution iteration residual at time , represents the quantized value of the consistency measure at time , , , is a weight coefficient determined according to sensor characteristics and solving algorithm, is the joint uncertainty estimate at time .

[0111] The timestamp, spatial position and corresponding joint uncertainty estimate at each solving time are recorded as a spatiotemporal data point, and the spatiotemporal data points constitute a set of discrete spatiotemporal samples with uncertainty labels. Spatial interpolation is performed on all spatiotemporal data points, and the spatial interpolation method can use Kriging interpolation or radial basis function interpolation to generate a scalar field that covers the entire measurement spatiotemporal domain and varies continuously in space and time dimensions. This scalar field is the spatiotemporal uncertainty distribution field. Optionally, the spatiotemporal uncertainty distribution field is represented as a three-dimensional grid data structure in computer memory, and each grid node of the three-dimensional grid data structure stores a scalar representing the joint uncertainty estimate at that location.

[0112] In specific implementations, the process of generating the spatiotemporal uncertainty distribution field requires handling the time dimension as an implicit parameter. The time dimension is integrated into the spatial interpolation through the correspondence between the timestamps of the solving times and the spatial trajectories. In some embodiments, spatial interpolation can be completed in two steps: first, interpolation in the time series along the continuously smooth three-dimensional trajectory line, and then radial interpolation in the trajectory line plane direction, thereby constructing the complete spatiotemporal uncertainty distribution field.

[0113] Referring to Figure 4 , this is a solving performance analysis chart for the inertial solving and trajectory generation stage. As the position / attitude error increases, the solving confidence decreases simultaneously (e.g., at 100 seconds, the error reaches a peak value, and the confidence decreases to 0.6), which conforms to the logic that "the larger the error, the lower the solving reliability". When the curvature change rate (representing the complexity of the trajectory) increases, the solving error also increases, which reflects the interference of complex trajectories on the accuracy of inertial solving. In the later stage (after 60 seconds), both the error and the curvature show an upward trend, indicating that the motion state of the measurement carrier is gradually becoming complex, and the solving difficulty is increasing. Such charts are used to evaluate the solving reliability of inertial navigation and are the core index support for "motion data acquisition → trajectory generation" in three-dimensional geological structure measurement. High-precision and high-confidence solving results can provide accurate spatial trajectory basis for subsequent three-dimensional geological model construction.

[0114] In one embodiment of the present application, the model revision module spatially superimposes the spatiotemporal uncertainty distribution field with the base three-dimensional geological structure model, reads the joint uncertainty estimate corresponding to the geometric center of each triangular facet. Marking out the triangular facets whose joint uncertainty estimate exceeds the allowed upper limit, it demarcates them as low-confidence regions, and for the triangular facets within the low-confidence regions, according to the geometric trend of their adjacent regions and the spatial plane equation constraint of the geological structure surface, it recalculates the optimized coordinates of the vertices in the region using the moving least squares method surface fitting technique. Using the optimized vertex coordinates to update the base three-dimensional geological structure model and recalculating the local curvature and structure surface fitting degree of the updated model, it iteratively performs the above region marking and vertex optimization process until the geometric morphology of all regions meets the preset structure surface fitting degree requirement to complete the iterative optimization adjustment. The achievement packaging module defines a layered data structure including a geometric grid layer, an attribute data layer, and a metadata layer, converts the geometric morphology data after iterative optimization adjustment into a standardized triangular mesh data format and stores it in the geometric grid layer. The attribute fields associated with the mapping and the data of the spatiotemporal uncertainty distribution field are encoded in key-value pair format and stored in the attribute data layer, and the measurement task identifier, coordinate system information, data version, and packaging time information are written into the metadata layer. The data of the geometric grid layer, the attribute data layer, and the metadata layer are merged and compressed into a binary stream, and a custom file header is added to the front of the compressed data stream, which contains the starting offset, length, and decompression identifier of each data layer. Finally, it outputs a composite three-dimensional geological survey achievement file that can be independently parsed.

[0115] In specific implementation, the model revision module spatially superimposes the spatiotemporal uncertainty distribution field with the base three-dimensional geological structure model, reads the joint uncertainty estimate corresponding to the geometric center of each triangular facet, which is obtained by querying the interpolation of the spatiotemporal uncertainty distribution field at the center point coordinates. Marking out the triangular facets whose joint uncertainty estimate exceeds the allowed upper limit, it demarcates them as low-confidence regions, and for the triangular facets within the low-confidence regions, according to the geometric trend of their adjacent regions and the spatial plane equation constraint of the geological structure surface, it recalculates the optimized coordinates of the vertices in the region using the moving least squares method surface fitting technique. The moving least squares method surface fitting technique defines a local reference domain for each vertex to be optimized, and fits a local surface function within the reference domain using weighted least squares. In some embodiments, the weight function used in the moving least squares method surface fitting can be taken as:

[0116]

[0117] wherein: ​It is the first one to be optimized. The coordinates of the vertices, It is the first in its neighborhood The coordinates of the points participating in the fitting. It is a smoothing parameter that controls the range of influence of the weights. Point With point The square of the Euclidean distance between them Point Point The impact weight of the optimization process.

[0118] The optimized vertex coordinates are used to update the basic 3D geological structure model, and the local curvature and structural surface fit of the updated model are recalculated. The structural surface fit measures the degree of closeness between the updated model surface and the spatial plane equation of the known geological structural surface. Using the structural surface fit as the convergence criterion, the structural surface fit is quantified by calculating the average distance from the triangular facets in the low-confidence region to the corresponding geological structural surface. The above region marking and vertex optimization process is iteratively executed until the geometry of all regions meets the preset structural surface fit requirements, completing the iterative optimization adjustment.

[0119] It is understandable that the results encapsulation module defines a hierarchical data structure, which includes a geometric mesh layer, an attribute data layer, and a metadata layer. The iteratively optimized geometric data is converted into a standardized triangular mesh data format, such as STL or PLY, and stored in the geometric mesh layer. The associated mapping attribute fields and the spatiotemporal uncertainty distribution field data are encoded in a key-value pair format, which allows for flexible storage and retrieval of different attribute types, and stored in the attribute data layer. In some embodiments, the spatiotemporal uncertainty distribution field data can be stored as a three-dimensional scalar array in the attribute data layer and associated with the vertices or faces of the geometric mesh layer through indexes. The measurement task identifier, coordinate system information, data version, and encapsulation time information are written into the metadata layer, and the data in the metadata layer is organized using an Extensible Markup Language (Extensible Markup Language) format.

[0120] The data of the geometric grid layer, the attribute data layer and the metadata layer are merged and compressed in binary stream. The binary stream merging ensures compact storage of the data, and the compression adopts a lossless compression algorithm to reduce the file size. A custom file header is added in the front of the compressed data stream, which contains the starting offset, length and decompression identifier of each data layer. The starting offset is used to locate the starting position of each layer data. Finally, the output is a composite 3D geological survey result file that can be independently parsed. In specific implementation, the composite 3D geological survey result file has a custom file extension, and when parsing, the file header needs to be read first, and then the information in the file header is used to decompress and separate each layer data for processing. Optionally, the key-value pair data in the attribute data layer can be serialized in JSON or similar format and then stored in compressed form.

[0121] Referring to Figure 5 This is a quality evaluation chart of a 3D geological structure survey result. All index scores are higher than 90%, indicating that the core dimensions of the geological survey result, such as geometric accuracy, attribute integrity and model fitting degree, all reach a high level; the file parsability score is the highest (≈97%), reflecting the high standardization degree of the result packaging; the geometric accuracy and uncertainty quantification scores are consistent (≈95%), indicating that the model accuracy and error quantification are highly reliable. This type of chart is used for comprehensive evaluation of the quality of the geological survey result, and is the core acceptance basis for the "result output" link of the 3D geological structure survey system. The high-score result can provide reliable 3D data support for subsequent geological analysis and engineering application.

[0122] Although embodiments of the present application have been shown and described, it is to be understood that various modifications, substitutions, replacements and changes can be made to these embodiments without departing from the principles and spirit of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. A three-dimensional geologic structure surveying system based on inertial navigation, characterized by, The system comprises: a solving module, receiving a continuous motion data stream from an inertial measurement component, and successively solving an instantaneous spatial attitude and a relative displacement vector of a measurement carrier; a three-dimensional model construction module, receiving the instantaneous spatial attitude and the relative displacement vector, and constructing a basic three-dimensional geological structure model expressed by a triangular facet network; a geological attribute attachment device, importing lithology, stratum and structure interpretation data corresponding to a measurement area from an external geological database, and associating and mapping the lithology, stratum and structure interpretation data to corresponding triangular facets in the basic three-dimensional geological structure model as attribute fields; a quantization module, calculating and recording a solving confidence degree of each of the instantaneous spatial attitude and the relative displacement vector in real time during a solving process of the solving module, and generating a spatio-temporal uncertainty distribution field synchronized with the basic three-dimensional geological structure model; a model correction module, performing error propagation analysis on the basic three-dimensional geological structure model by using the spatio-temporal uncertainty distribution field, and iteratively optimizing and adjusting a geometric shape of a low-confidence-degree area in the basic three-dimensional geological structure model according to an analysis result; an achievement packaging module, standardizing and packaging the iteratively optimized and adjusted geometric shape, the associated and mapped attribute fields and the spatio-temporal uncertainty distribution field, and outputting a composite three-dimensional geological measurement achievement file which can be independently analyzed; the receiving the instantaneous spatial attitude and the relative displacement vector, and constructing the basic three-dimensional geological structure model expressed by the triangular facet network, comprises: receiving the instantaneous spatial attitude and the relative displacement vector, and performing spline interpolation on continuous discrete solving points according to a time stamp sequence to form a continuous and smooth three-dimensional spatial trajectory line; performing hierarchical voxel division on a space region passed through by the continuous and smooth three-dimensional spatial trajectory line according to a preset spatial grid resolution to generate a multi-level resolution voxelized space container; generating a spatial discrete point set representing a geological surface shape in the multi-level resolution voxelized space container according to a set point density distribution rule according to the continuous and smooth three-dimensional spatial trajectory line; performing local neighborhood normal estimation and curvature analysis on the spatial discrete point set, identifying and extracting a spatial plane equation of a geological structure surface according to normal clustering and curvature characteristics; and fusing the spatial discrete point set and the spatial plane equation of the geological structure surface to construct the basic three-dimensional geological structure model expressed by the triangular facet network; the real-time calculation and recording of the solving confidence degree of each of the instantaneous spatial attitude and the relative displacement vector, and the generation of the spatio-temporal uncertainty distribution field synchronized with the basic three-dimensional geological structure model, comprises: synchronously acquiring an original reading noise level of an inertial sensor, a solving iterative residual error and a consistency measurement with a previous solving result each time the solving module is solved; and calculating a joint uncertainty estimate value of an instantaneous spatial attitude and a relative displacement vector obtained by this time solving according to a pre-built error propagation model, the original reading noise level, the solving iterative residual error and the consistency measurement, as an inverse of a solving confidence degree at this time. Combined uncertainty estimation The calculation employs a linear combination model: wherein: represents the quantized value of the original read noise level at time , represents the quantized value of the resolved iteration residual at time , represents the quantized value of the consistency measure at time , , , is a weight coefficient determined from sensor characteristics and resolving algorithm, is the joint uncertainty estimate at time ; The time stamp, spatial position and corresponding joint uncertainty estimation of each calculation time are recorded as a space-time data point; all space-time data points are spatially interpolated to generate a scalar field covering the entire measurement space-time domain and continuously changing in space and time dimensions, which is the space-time uncertainty distribution field; the space-time uncertainty distribution field is represented as a three-dimensional grid data structure in computer memory, and each grid node of the three-dimensional grid data structure stores a scalar representing the joint uncertainty estimation at the node.

2. The inertial navigation-based three-dimensional geologic structure surveying system according to claim 1, wherein, The continuous and smooth three-dimensional space trajectory is formed by spline interpolation of the continuous discrete calculation points according to the time stamp sequence, including: obtaining a sequence of discrete calculation points arranged in time sequence from the calculation module, each discrete calculation point containing three-dimensional coordinates and attitude quaternion; using a non-uniform rational B-spline curve fitting algorithm, taking the sequence of discrete calculation points as control points, generating an initial space curve passing through all control points; evaluating the fairness of the initial space curve, calculating the curvature change rate of the curve on each segment; inserting additional interpolation control points in the curve segment where the curvature change rate exceeds the preset threshold, and locally re-fitting the initial space curve using piecewise cubic Hermite interpolation method; performing parameterization processing on the overall curve after re-fitting to generate a parameter equation with time as the only parameter, and the curve expressed by the parameter equation is the continuous and smooth three-dimensional space trajectory.

3. The inertial navigation-based three-dimensional geologic structure surveying system according to claim 2, wherein, The hierarchical voxel division is performed according to the preset spatial grid resolution to generate a multi-resolution voxelized space container, including: setting a basic space bounding box that completely contains the continuous and smooth three-dimensional space trajectory; equally dividing the basic space bounding box at the highest resolution to form a voxel grid of the finest level; generating voxel grids of coarser resolutions by merging adjacent multiple finest level voxels into one large voxel, thereby forming a multi-resolution voxel pyramid structure from fine to coarse; establishing a spatial index for each voxel in the voxel pyramid structure, and marking the voxels passed through by the continuous and smooth three-dimensional space trajectory as active voxels; creating a dynamic storage container for storing and managing all active voxels and their spatial indexes at different levels, thereby forming the multi-resolution voxelized space container; According to the continuous and smooth three-dimensional space trajectory, a set of spatial discrete points representing the morphology of the geological surface is generated in the multi-resolution voxelized space container according to a set point density distribution rule, including: selecting a specific resolution level in the multi-resolution voxelized space container that matches the current detail representation requirement; generating dense sampling points on the continuous and smooth three-dimensional space trajectory according to equal arc length intervals; in the law plane of each sampling point, a specified number of spatial points are randomly generated within the range of active voxels of the specific resolution level according to the preset point cloud lateral diffusion radius and point spacing; de-duplicate all generated spatial points, and eliminate spatial points falling within voxels marked as non-geologic obstacles; collect three-dimensional coordinates of all finally retained spatial points to form a set of spatial discrete points representing the geological surface morphology.

4. The inertial navigation-based three-dimensional geologic structure surveying system according to claim 3, wherein, The method for identifying and extracting spatial plane equations of geological structure surfaces according to normal clustering and curvature characteristics comprises: S1: for each point in the set of spatial discrete points, calculate a K-neighbor point set of the point, and estimate a local normal vector and a curvature value of the point by using a principal component analysis method; S2: perform clustering analysis according to the local normal vector directions of all points, and classify points with similar normal vector directions into a same candidate point set; S3: for each candidate point set, iteratively estimate an optimal spatial plane model by using a random sample consensus algorithm, and calculate distance residuals of inliers in the candidate point set to the optimal spatial plane model; S4: screen out an optimal spatial plane model with a higher inlier proportion than a set threshold and a minimum distance residual sum of squares, and output mathematical equation parameters of the optimal spatial plane model as a geological structure surface; repeat S1 to S4 until all significant candidate point sets are processed, so as to extract spatial plane equations of multiple geological structure surfaces.

5. The inertial navigation-based three-dimensional geologic structure surveying system according to claim 4, wherein, The method for fusing the set of spatial discrete points and the spatial plane equations of the geological structure surfaces to construct a basic three-dimensional geological structure model expressed by a triangular facet network comprises: generate an initial triangular mesh surface in a three-dimensional space by using a Delaunay triangulation algorithm with the set of spatial discrete points as constraint points; cut the initial triangular mesh surface by taking the spatial plane equation of each geological structure surface as a geometric constraint condition, so that the mesh is split at a structure surface boundary; perform local topological optimization on the split triangular mesh to eliminate abnormal triangles and ensure that mesh edges near the geological structure surface are aligned with the structure surface boundary; store vertex coordinates, triangular facet indexes, and normal vector information of the optimized triangular mesh to construct the basic three-dimensional geological structure model expressed by the triangular facet network.

6. The inertial navigation-based three-dimensional geologic structure surveying system according to claim 1, wherein, The method for associating and mapping the lithology, stratigraphic, and structural interpretation data to corresponding triangular facets in the basic three-dimensional geological structure model as attribute fields comprises: analyze the lithology, stratigraphic, and structural interpretation data imported from an external geological database, and convert the data into a standardized data format containing attribute key-value pairs; calculate a geometric center point coordinate of each triangular facet in the basic three-dimensional geological structure model; perform position matching of the geometric center point coordinate of each triangular facet with a spatialized attribute chart provided by the external geological database to determine a lithology code, a stratigraphic unit code, and a structure attribution of the triangular facet; create additional attribute fields for each triangular facet in a data structure of the basic three-dimensional geological structure model, and write the matched and determined lithology code, stratigraphic unit code, and structure attribution into corresponding additional attribute fields to complete attribute association and mapping.

7. The inertial navigation-based three-dimensional geologic structure surveying system according to claim 6, wherein, The error propagation analysis on the basic three-dimensional geological structure model is performed by using the spatiotemporal uncertainty distribution field, and the geometric shape of a low-confidence region in the basic three-dimensional geological structure model is iteratively optimized and adjusted according to an analysis result, and the error propagation analysis on the basic three-dimensional geological structure model by using the spatiotemporal uncertainty distribution field comprises the following steps: The spatiotemporal uncertainty distribution field is spatially superimposed on the basic three-dimensional geological structure model, and a joint uncertainty estimate corresponding to a geometric center of each triangular facet is read; Triangular facets with a joint uncertainty estimate exceeding an allowed upper limit are marked, and are defined as a low-confidence region; For triangular facets in the low-confidence region, optimized coordinates of vertices in the region are recalculated by using a moving least square method surface fitting technique according to geometric trends of adjacent regions and a spatial plane equation constraint of the geological structure surface; The basic three-dimensional geological structure model is updated by using the optimized vertex coordinates, and local curvatures and structure surface fitting degrees of the updated model are recalculated; The region marking and vertex optimization process are iteratively performed until geometric shapes of all regions meet preset structure surface fitting degree requirements, and the iterative optimization and adjustment are completed, with the structure surface fitting degree as a convergence criterion.

8. The inertial navigation-based three-dimensional geologic structure surveying system according to claim 1, wherein, The achievement packaging module performs standardized packaging and outputs a composite three-dimensional geological survey achievement file, and the achievement packaging module comprises the following steps: A layered data structure is defined, and the structure comprises a geometric grid layer, an attribute data layer and a metadata layer; Geometric shape data after the iterative optimization and adjustment are converted into a standardized triangular mesh data format, and are stored in the geometric grid layer; Attribute fields and data of the spatiotemporal uncertainty distribution field are encoded in a key-value pair format, and are stored in the attribute data layer; Survey task identification, coordinate system information, data version and packaging time information are written into the metadata layer; Data in the geometric grid layer, the attribute data layer and the metadata layer are merged and compressed into a binary stream; A custom file header is added in front of the compressed data stream, and the file header comprises a starting offset, a length and a decompression identifier of each data layer, and finally outputs the composite three-dimensional geological survey achievement file which can be independently analyzed.

Images