Mine laneway laser radar map management method, device and equipment and storage medium

Abstract

The application discloses a kind of mine laneway's laser radar map management method, device, equipment and storage medium, it is related to mine laneway map management technical field, the method includes: to the original point cloud submap is preprocessed, by the strategy of registration and coherent point drift non-rigid registration combination of random sampling consistency algorithm, the point cloud submap after pre-processing is aligned with historical global map;When meeting preset map deletion trigger condition, through difference calculation, contour intersection extraction and peak value determination identification backfilling roadway area and mark;Subsequently, the point cloud data of marked area is deleted, generates three-dimensional global map, and carries out multidimensional quality assessment to it, when assessment is qualified, corresponding two-dimensional map is generated based on three-dimensional global map, and three-dimensional and two-dimensional map is output to preset terminal.The application can be in the underground environment of no GPS and sparse feature, realize the high-precision geometry of global map is consistent, and dynamically adapts the change of roadway backfilling.

Description

Technical Field

[0001] This application relates to the field of mine roadway map management technology, and in particular to methods, devices, equipment and storage media for managing LiDAR maps of underground mine roadways. Background Technology

[0002] With the development trend of intelligent and unmanned mining in metal mines, the construction and management of high-precision maps of underground mine roadways has become a key infrastructure supporting autonomous operating systems such as unmanned mining trucks and inspection robots. The underground environment is characterized by the absence of GPS signals, poor lighting conditions, severe dust interference, and repetitive roadway structures with sparse texture features, making traditional mapping methods relying on vision or GNSS assistance unsuitable. Therefore, Simultaneous Localization and Mapping (SLAM) technology based on lidar has become the mainstream solution for generating high-precision underground maps. Its core requirement is to maintain the geometric consistency of the global map during long-term operation and to dynamically respond to drastic topological and geometric changes in roadways caused by mining progress or backfilling operations.

[0003] Currently, for downhole lidar map construction, the industry generally adopts graph-optimized SLAM frameworks (such as LIO-SAM, LeGO-LOAM, etc.), which improve trajectory accuracy through loop closure detection and pose graph optimization, and store the map using voxel grids or point cloud databases. Some systems introduce submap mechanisms to support incremental updates, while some research attempts to combine semantic information or manual labels to assist localization. For map management, static snapshots or simple timestamp version control strategies are usually adopted, with local fusion or overwriting performed when new data arrives.

[0004] However, existing practices still have significant problems in practical underground applications: On the one hand, due to the highly similar and lack of salient features in tunnel structures, closed-loop detection is prone to failure, leading to cumulative errors and even geometric distortion in the map after long-term operation; on the other hand, current map management systems mostly assume unidirectional environmental expansion, failing to effectively handle the mine-specific phenomenon of "tunnel backfilling"—that is, the original passable area being filled and closed, resulting in a large amount of invalid or erroneous geometric information remaining in the map, which not only occupies storage resources but also misleads subsequent positioning and path planning. Furthermore, the lack of a quantitative evaluation mechanism for map quality makes it difficult to determine whether submaps are reliable, need to be redrawn, or removed. Therefore, how to achieve high-precision geometric consistency of the global map and dynamically adapt to changes in tunnel backfilling in underground environments without GPS and with sparse features has become an urgent problem to be solved. Summary of the Invention

[0005] The purpose of this application is to provide a method, device, equipment and storage medium for managing lidar maps of underground mine roadways, aiming to solve the technical problem of how to achieve high-precision geometric consistency of the global map and dynamically adapt to changes in roadway backfilling in an underground environment without GPS and with sparse features.

[0006] To achieve the above objectives, this application proposes a method for managing lidar maps of underground mine roadways, the method comprising: Obtain the original point cloud sub-map and preprocess the original point cloud sub-map to obtain the preprocessed point cloud sub-map. By combining random sampling consensus algorithm registration with coherent point drift non-rigid registration, the preprocessed point cloud sub-map is aligned with the historical global map to obtain the aligned point cloud map. When the preset map deletion trigger condition is met, the aligned point cloud map and the historical global map are subjected to differential calculation, contour intersection extraction and intersection peak determination, the backfilled alley area is identified and marked, and the marked point cloud map is obtained. Delete the point cloud data corresponding to the backfilled tunnel area from the marked point cloud map to obtain a three-dimensional global map; A comprehensive quality assessment of the three-dimensional global map is performed based on multi-dimensional quality assessment indicators to obtain the quality assessment results. When the quality assessment result is qualified, a corresponding two-dimensional map is generated based on the three-dimensional global map, and the three-dimensional global map and the two-dimensional map are output to a preset terminal.

[0007] In one embodiment, the step of aligning the preprocessed point cloud sub-map with the historical global map using a strategy combining random sampling consensus registration and coherent point drift non-rigid registration to obtain the aligned point cloud map includes: The preprocessed point cloud sub-map is coarsely registered with the historical global map using a random sampling consistency algorithm to obtain a coarsely registered point cloud sub-map. Based on preset deformation rigidity constraint coefficients and preset spatial smoothness constraint coefficients, the coarsely registered point cloud sub-map is finely registered with the historical global map using a coherent point drift non-rigid registration model to obtain a finely registered point cloud sub-map. Calculate the registration error between the finely registered point cloud sub-map and the historical global map; When the registration error is greater than or equal to the preset registration error threshold, return to the step of performing coarse registration between the preprocessed point cloud sub-map and the historical global map using the random sampling consistency algorithm to obtain the coarsely registered point cloud sub-map. When the registration error is less than the preset registration error threshold, the finely registered point cloud sub-map is used as the aligned point cloud map.

[0008] In one embodiment, the step of performing differential calculation, contour intersection extraction, and intersection peak determination on the aligned point cloud map and the historical global map when the preset map deletion trigger condition is met, and identifying and marking the backfilled alleyway area to obtain the marked point cloud map includes: When the preset map deletion trigger condition is met, the spatial difference set between the aligned point cloud map and the historical global map is calculated by the global map difference operator; The Alpha Shape Algorithm is used to extract the tunnel contour boundary from the spatial difference set, and a concave hull boundary is constructed based on the tunnel contour boundary. Calculate the contact line between the concave hull boundary and the historical global map, and smooth the contact line to obtain a smoothed contact line. Based on the curvature or height variation characteristics of the smoothed contact intersection line, identify the significant peaks of the smoothed contact intersection line and count the number of the significant peaks; When the number of significant peaks is less than a preset peak determination threshold, the area in the spatial difference set corresponding to the smoothed contact line is marked as a backfill tunnel area, and a marked point cloud map is obtained.

[0009] In one embodiment, the step of performing a comprehensive quality assessment of the three-dimensional global map based on multi-dimensional quality assessment indicators to obtain the quality assessment result includes: The three-dimensional global map is divided into a preset number of local regions, and the covariance matrix of the local regions is calculated. The average map entropy is then calculated based on the covariance matrix. Calculate the distance from each coordinate point in the marked point cloud map to the local fitting plane corresponding to each coordinate point, and calculate the mean plane variance based on the distance; The extreme values ​​of the marked point cloud map are compared with the preset work area coordinate range constraints to obtain the geographical range compliance result. The coordinate fluctuation range of the marked point cloud map is calculated, and the coordinate fluctuation range is compared with a preset coordinate fluctuation range threshold to obtain the stability evaluation result; Calculate the chamfer distance between the marked point cloud map and the preset baseline global map, and obtain the geometric similarity evaluation result based on the chamfer distance; The quality assessment result is obtained by combining the average map entropy, the average planar variance, the geographical scope compliance result, the stability assessment result, and the geometric similarity assessment result.

[0010] In one embodiment, the step of obtaining the original point cloud sub-map and preprocessing the original point cloud sub-map to obtain the preprocessed point cloud sub-map includes: A scanning command is sent to the vehicle-mounted lidar so that the vehicle-mounted lidar can scan the underground roadway environment, collect and return the original point cloud sub-map; The original point cloud sub-maps are subjected to initial quality screening and spatial validity checks to obtain a set of valid sub-maps; The effective sub-map set is subjected to noise filtering and voxel uniform downsampling to obtain the preprocessed point cloud sub-map.

[0011] In one embodiment, the step of performing initial quality screening and spatial validity checks on the original point cloud sub-map to obtain a set of valid sub-maps includes: Ground point extraction is performed on the original point cloud sub-map to obtain a ground point set; Calculate the area of ​​the convex hull region of the ground point set in the horizontal projection plane. The area of ​​the convex hull region is obtained based on the convex hull operator and the polygon area calculation operator. When the area of ​​the convex hull region is within a preset convex hull area threshold range and all three-dimensional coordinate points in the original point cloud sub-map are within a preset mine operation area, the original point cloud sub-map is considered a valid sub-map. Based on the multiple valid subgraphs, a set of valid subgraphs is obtained.

[0012] In one embodiment, the step of performing noise filtering and voxel uniform downsampling on the effective sub-map set to obtain the preprocessed point cloud sub-map includes: The point cloud data in the effective sub-graph set is subjected to noise suppression according to the discrete point filtering operator to obtain the noise-suppressed point cloud data. The discrete-point filtering operator is represented as follows: in, Point The local neighborhood, This indicates deviation from the metric function. The noise threshold is used to determine the noise level. This represents the point cloud data after noise suppression. This represents the discrete-point filtering operator. This represents the point cloud data in the set of valid subgraphs. express Any three-dimensional point cloud data point; Based on the preset voxel side length, a voxel mesh is constructed with the minimum coordinate in the noise-suppressed point cloud data as the reference starting point. The noise-suppressed point cloud data is divided into several voxel subsets according to the voxel index using the voxel grid. Based on the preset maximum number of retained points, a uniform sampling operation is performed in each voxel subset to obtain downsampled point cloud data. All the downsampled point cloud data are integrated to obtain a preprocessed point cloud sub-map.

[0013] Furthermore, to achieve the above objectives, this application also proposes a lidar map management device for underground mine roadways, the device comprising: The preprocessing module is used to acquire the original point cloud sub-map and preprocess the original point cloud sub-map to obtain the preprocessed point cloud sub-map. The map stitching module is used to align the preprocessed point cloud sub-map with the historical global map using a strategy that combines random sampling consistency algorithm registration with coherent point drift non-rigid registration, so as to obtain an aligned point cloud map. The area marking module is used to perform differential calculation, contour intersection extraction and intersection peak determination on the aligned point cloud map and the historical global map when the preset map deletion trigger condition is met, identify and mark the backfilled alley area, and obtain the marked point cloud map. The map deletion module is used to delete the point cloud data corresponding to the backfilled alleyway area from the marked point cloud map to obtain a three-dimensional global map. The quality assessment module is used to perform a comprehensive quality assessment of the three-dimensional global map based on multi-dimensional quality assessment indicators, and obtain the quality assessment results. The map update module is used to generate a corresponding two-dimensional map based on the three-dimensional global map when the quality assessment result is qualified, and output the three-dimensional global map and the two-dimensional map to a preset terminal.

[0014] In addition, to achieve the above objectives, this application also proposes a lidar map management device for underground mine roadways, the device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the lidar map management method for underground mine roadways as described above.

[0015] In addition, to achieve the above objectives, this application also proposes a storage medium, which is a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements the steps of the above-described method for managing lidar maps of underground mine roadways.

[0016] In addition, to achieve the above objectives, this application also provides a computer program product, which includes a computer program that, when executed by a processor, implements the steps of the above-described method for managing lidar maps of underground mine roadways.

[0017] One or more technical solutions proposed in this application have at least the following technical effects: First, the original point cloud sub-map is acquired and preprocessed to effectively remove invalid or low-quality data, improving the stability and efficiency of subsequent processing. Next, a strategy combining random sampling consistency registration and coherent point drift non-rigid registration is employed to accurately integrate the preprocessed point cloud sub-map into the historical global map. When preset map deletion trigger conditions are met, the system performs differential calculations on the aligned point cloud map and the historical global map, combining contour intersection extraction and peak value determination to accurately identify and mark backfilled tunnel areas. Subsequently, the corresponding point cloud data is deleted from the marked point cloud map, generating a 3D global map containing only the current actual tunnel structure, avoiding the retention of invalid information. Based on this, the map is comprehensively scored using multi-dimensional quality evaluation indicators to ensure that the updated map meets accuracy and consistency requirements. When the evaluation result is satisfactory, the system synchronously generates the corresponding 2D map and outputs both the 3D and 2D maps to a preset terminal. This application enables high-precision geometric consistency of the global map in downhole environments without GPS and with sparse features, and dynamically adapts to changes in tunnel backfilling. Attached Figure Description

[0018] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0019] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 A flowchart illustrating the first embodiment of the lidar map management method for underground mine roadways in this application; Figure 2 This is a schematic diagram of two-dimensional map generation provided in Embodiment 1 of the method for managing lidar maps of underground mine roadways in this application. Figure 3 A flowchart illustrating Embodiment 2 of the method for managing lidar maps of underground mine roadways in this application; Figure 4A simplified flowchart illustrating the lidar map management method for underground mine roadways provided in Embodiment 2 of this application; Figure 5 This is a schematic diagram of the map management system architecture for the lidar map management method for underground mine roadways provided in Embodiment 2 of this application; Figure 6 This is a schematic diagram of the module structure of a lidar map management device for underground mine roadways according to an embodiment of this application. Figure 7 This is a schematic diagram of the equipment structure of the hardware operating environment involved in the lidar map management method for underground mine roadways in this application embodiment.

[0021] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0022] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of this application and are not intended to limit this application.

[0023] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.

[0024] It should be noted that the executing entity of this application embodiment can be a computing service device with data processing, network communication, and program execution functions, such as a tablet computer, personal computer, or mobile phone, or an electronic device or map management system capable of realizing the above functions. The following uses a map management system as an example to describe this embodiment and the subsequent embodiments.

[0025] The user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0026] Based on this, the embodiments of this application provide a method for managing lidar maps of underground mine roadways, referring to... Figure 1 , Figure 1 This is a flowchart illustrating the first embodiment of the laser radar map management method for underground mine roadways according to this application.

[0027] In this embodiment, the laser radar map management method for underground mine roadways includes steps S10 to S60: Step S10: Obtain the original point cloud sub-map and preprocess the original point cloud sub-map to obtain the preprocessed point cloud sub-map. It should be noted that the original point cloud submap refers to a set of three-dimensional point cloud data collected and initially stitched together by lidar in a local time period or local space in the mine. It usually corresponds to a submap unit in the SLAM system and has not yet undergone filtering, downsampling or geometric optimization.

[0028] As an example, the steps of acquiring the original point cloud sub-map and preprocessing the original point cloud sub-map to obtain the preprocessed point cloud sub-map include: sending a scanning command to the vehicle-mounted lidar to enable the vehicle-mounted lidar to scan the underground roadway environment, collect and return the original point cloud sub-map; performing initial quality screening and spatial validity verification on the original point cloud sub-map to obtain a set of valid sub-maps; and performing noise filtering and voxel uniform downsampling processing on the set of valid sub-maps to obtain the preprocessed point cloud sub-map.

[0029] The effective submap set refers to the subset of submaps selected from the original point cloud submaps after initial quality screening and spatial validity verification, which meets the preset geometric integrity, positioning reliability, and spatial coverage rationality. These submaps are considered to have the basic credibility to participate in subsequent map construction and fusion, excluding low-quality or invalid data caused by sensor failure, severe motion distortion, or extreme degradation of environmental features (e.g., data collected in completely open or repetitive structure areas).

[0030] As an example, the step of performing initial quality screening and spatial validity verification on the original point cloud sub-map to obtain a set of valid sub-maps includes: extracting ground data from the original point cloud sub-map to obtain a set of ground points; calculating the area of ​​the convex hull region of the ground point set in a horizontal projection plane, wherein the area of ​​the convex hull region is obtained based on the convex hull operator and the polygon area calculation operator; when the area of ​​the convex hull region is within a preset convex hull area threshold range and all three-dimensional coordinate points in the original point cloud sub-map are within a preset mine operation area, the original point cloud sub-map is considered a valid sub-map; and a set of valid sub-maps is obtained based on multiple valid sub-maps.

[0031] A ground point set refers to the set of three-dimensional points belonging to the bottom surface of a lane extracted from the original point cloud submap using ground segmentation algorithms (e.g., methods based on normal vectors, height thresholds, or plane fitting). It is used to characterize the geometry of the vehicle driving area. A horizontal projection plane is a planar coordinate system formed by projecting points in three-dimensional space (such as the ground point set) perpendicularly along the direction of gravity (usually the Z-axis) onto a two-dimensional plane parallel to the ground. It is often used to simplify the geometric analysis of lane layouts. The convex hull area refers to the area of ​​the smallest convex polygon enclosed by the convex hull of the ground point set within the horizontal projection plane. It is used to quantify the lateral extent of the lanes covered by the submap. A convex hull operator is a geometric algorithm (e.g., Graham's scan algorithm or Andrew's monotonic chain algorithm) used to calculate the convex hull of a set of two-dimensional points. Its output is a sequence of vertices of the smallest convex polygon enclosing all input points. The polygon area calculation operator is a mathematical operation module that calculates the area enclosed by a convex polygon on a horizontal projection plane using numerical methods (e.g., the shoelace formula) based on the vertex order output by the convex hull operator. The preset convex hull area threshold range refers to a reasonable area range (e.g., 20 square meters to 200 square meters) set based on prior statistical characteristics such as typical underground roadway width, effective LiDAR scanning distance, and point cloud density. This range is used to eliminate invalid sub-maps that are too small or too large due to incomplete scanning, stationary vehicles, or narrow intersections. The preset mine operation area refers to the legal geographic space area predefined in the mine's digital twin model or engineering drawings that allows unmanned driving equipment to operate. It is usually stored as a three-dimensional boundary or polygonal region and is used to exclude invalid point cloud data located in collapsed areas, unexcavated areas, or restricted areas due to positioning drift or sensor mis-triggers.

[0032] As an example, the steps of performing noise filtering and voxel uniform downsampling on the effective sub-map set to obtain a preprocessed point cloud sub-map include: performing noise suppression on the point cloud data in the effective sub-map set according to a discrete point filtering operator to obtain noise-suppressed point cloud data; constructing a voxel grid based on the minimum coordinate in the noise-suppressed point cloud data as the reference starting point according to a preset voxel side length; dividing the noise-suppressed point cloud data into several voxel subsets according to the voxel index through the voxel grid; performing a uniform sampling operation in each voxel subset according to a preset maximum number of retained points to obtain downsampled point cloud data; and integrating all the downsampled point cloud data to obtain the preprocessed point cloud sub-map.

[0033] Discrete point filtering operators are algorithmic modules used to identify and remove isolated noise points or outliers in a point cloud (e.g., methods based on Statistical Outlier Removal (SOR)). They determine whether a point is noise by analyzing the distribution characteristics of each point and its neighbors (such as average distance and standard deviation). Preset voxel side length refers to the uniform side length (e.g., 0.1 meters) of each cubic voxel unit in the X, Y, and Z directions, manually set when constructing a 3D voxel mesh. This parameter is adjusted according to the original point cloud density and the need for detail preservation in subsequent processing, directly affecting the granularity of the downsampled data. A voxel mesh is a discretized spatial index structure composed of regularly arranged cubic units (i.e., voxels) that divide 3D space. Each voxel is uniquely identified by its integer coordinate index, used for efficiently organizing and managing the spatial distribution of point cloud data. A voxel subset is a local set of points from the original point cloud that fall within the same voxel unit. Each voxel subset corresponds to a specific voxel in the voxel mesh and is the basic unit for performing local uniform sampling. The preset maximum number of points to retain refers to the maximum number of points allowed to be retained in each voxel subset (e.g., 5 points). This value is set based on the trade-off between the need to retain the details of the tunnel structure and computational efficiency. It is used to control the density of the point cloud after downsampling, ensuring that the representativeness of geometric features is maintained while reducing the amount of data.

[0034] The specific steps of preprocessing are as follows: (1) After the vehicle-mounted mapping system is started, it scans the underground mine roadway environment using lidar and collects the original point cloud sub-map, which is represented as: in, This represents the first point in the original point cloud sub-map. i A 3D point cloud data point, express Three-dimensional spatial coordinates, Representing three-dimensional real space, N This represents the total number of points in the 3D point cloud data.

[0035] (2) To avoid the cumulative error caused by low-quality point clouds directly participating in global mapping, the collected point cloud sub-maps are... The convex hull area is used for quality assessment and subgraph range determination.

[0036] ① Sub-image quality assessment: First, ground point extraction is performed on the point cloud sub-image to obtain the ground point set. And calculate the area of ​​its convex hull region in the horizontal projection plane: in, Represents the convex hull operator. This represents the polygon area calculation operator. A subgraph quality discrimination function is constructed by performing interval determination on the convex hull area: when If a subgraph is deemed complete in terms of spatial coverage and point cloud distribution, it is considered suitable as a valid subgraph for subsequent stitching; otherwise, it is judged as a subgraph of subquality and is removed, forming a set of subgraphs. Among them, the polygon area threshold range The settings are based on the scale of underground roadways and the statistical characteristics of point cloud density, in order to take into account spatial variations under different operating scenarios.

[0037] ② After the sub-graph passes the geometric quality assessment, spatial constraints on the working area are further introduced. The predefined mine working area is set as follows: in, The boundary of the predefined mine operation area is defined if and only if all points in the subgraph are located within the predefined underground operation area. If the subgraph passes the spatial validity check, it is allowed to proceed to the subsequent point cloud processing and map building process; otherwise, it is considered an invalid subgraph that has exceeded the boundary and is removed, resulting in a set of valid subgraphs. .

[0038] (3) After the subgraph passes the above judgment, the system performs point cloud noise filtering based on discrete point filtering, voxel uniform downsampling, and point cloud non-rigid registration based on RANSAC-CPD.

[0039] ① First, a discrete point filtering operator is used to suppress noise in the point cloud. The discrete point filtering operator is expressed as follows: in, Point The local neighborhood, This indicates deviation from the metric function. The noise threshold is used to determine the noise level. This represents the point cloud data after noise suppression. This represents the discrete-point filtering operator. This represents the point cloud data in the set of valid subgraphs. express Any 3D point cloud data point; the deviation metric function is a mathematical function used to quantify the degree of spatial geometric inconsistency between the current point cloud sub-map and the historical global map. It is usually calculated by measuring the point-to-point distance, point-to-area distance or chamfer distance in the overlapping area of ​​the two maps, and outputs a scalar value that reflects the size of the registration deviation or structural drift. This value serves as the basis for determining whether to trigger re-registration, backfill detection or map update.

[0040] ②On this basis, for Perform uniform voxel downsampling. Let the voxel side length be... A voxel mesh is constructed with reference to the minimum coordinates of the point cloud, and its voxel index is defined as: in, To find the minimum point cloud location, divide the point cloud into several voxel subsets based on voxel indices. Uniform sampling is performed within each voxel, and the downsampling result is expressed as follows: in, This represents the maximum number of points allowed to be retained within a single element, thereby achieving point cloud density uniformity while maintaining spatial structure consistency, resulting in downsampled point cloud data. .

[0041] Step S20: The preprocessed point cloud sub-map is aligned with the historical global map by combining random sampling consistency algorithm registration with coherent point drift non-rigid registration strategy to obtain the aligned point cloud map. It should be noted that Random Sample Consensus (RANSAC) registration refers to the process of using the RANSAC algorithm to perform initial pose estimation on the preprocessed point cloud sub-map and the historical global map. This process involves repeatedly selecting a small number of matching point pairs, fitting a rigid transformation model (including rotation and translation), and evaluating the model quality based on the number of interior points. The final output is the optimal rigid transformation that aligns the most point pairs, achieving a coarse but robust global alignment. Coherent Point Drift (CPD) non-rigid registration refers to the process of using the CPD algorithm to further refine the local deformation or accumulated errors that may exist between the point cloud sub-map and the historical global map, based on the initial alignment provided by RANSAC. This process models the registration problem as a probability density estimation problem, solving a non-rigid (such as affine or thin-plate spline) transformation by maintaining the coherence of the overall motion of the point set, thereby improving the continuity and consistency of the map in geometric details.

[0042] Understandably, the map management system first extracts matching point pairs from the preprocessed point cloud sub-map and the historical global map (e.g., through nearest neighbor search), and calls the random sampling consensus algorithm registration module. It repeatedly and randomly selects the minimum point set from a large number of candidate transformations to solve the rigid transformation model. By statistically analyzing the number of interior points, it filters out the optimal rotation and translation parameters, thus achieving a coarse but robust global alignment. This step effectively overcomes the mismatch problem caused by sparse features or repetitive structures in the downhole environment. Then, the map management system uses this rigid transformation result as the initial pose and starts the coherent point drift non-rigid registration module. It treats the preprocessed point cloud sub-map as a "moving point set" and the corresponding area in the historical global map as a "fixed point set." Through iterative optimization of the probabilistic correspondence and non-rigid motion fields (such as affine or low-frequency deformation), it gradually corrects subtle misalignments in the sub-map caused by long-term SLAM accumulated errors or local geometric distortions. Finally, the map management system will precisely embed the point cloud sub-map after non-rigid correction into the historical global map according to the optimized transformation parameters, and output a geometrically continuous and structurally consistent aligned point cloud map, thereby providing a high-precision spatial reference for subsequent map fusion and updates.

[0043] Step S30: When the preset map deletion trigger condition is met, perform differential calculation, contour intersection extraction and intersection peak determination on the aligned point cloud map and the historical global map, identify and mark the backfilled alley area, and obtain the marked point cloud map. It should be noted that the preset map deletion trigger conditions refer to a set of trigger rules pre-set by the map management system to determine whether a roadway backfilling event may occur. These typically include one or more of the following: time dimension (e.g., more than a set period since the last update), spatial dimension (e.g., the newly acquired sub-map and historical map do not overlap in a specific area or the coverage is below a preset threshold), and task status (e.g., receiving an instruction from the mine scheduling system that an area has been closed or backfilled). A backfilled roadway area refers to a section of the original passageway that has been manually closed with gravel, concrete, or other filling materials during mine operations due to the completion of the work plan or safety requirements. Its physical space is no longer passable, and its geometric structure has undergone irreversible changes. In lidar scanning, this manifests as a long-term lack of effective point cloud returns in the area or significant holes and abrupt boundary changes between the old and new maps at that location.

[0044] As an example, the step of performing differential calculation, contour intersection extraction, and intersection peak determination on the aligned point cloud map and the historical global map when the preset map deletion trigger condition is met, and identifying and marking the backfill tunnel area to obtain the marked point cloud map includes: when the preset map deletion trigger condition is met, calculating the spatial difference set between the aligned point cloud map and the historical global map using the global map differential operator; extracting the tunnel contour boundary on the spatial difference set using the Alpha Shape Algorithm, and constructing a humped boundary based on the tunnel contour boundary; calculating the contact intersection line between the humped boundary and the historical global map, and smoothing the contact intersection line to obtain a smoothed contact intersection line; identifying significant peaks of the smoothed contact intersection line based on the curvature or height change characteristics of the smoothed contact intersection line, and counting the number of significant peaks; when the number of significant peaks is less than a preset peak determination threshold, marking the area in the spatial difference set corresponding to the smoothed contact intersection line as the backfill tunnel area, and obtaining the marked point cloud map.

[0045] The global map difference operator is an algorithm module used to calculate the geometric differences between the aligned point cloud map and the historical global map. It compares the two maps at the voxel or point level in a unified coordinate system to identify areas that exist only in the historical map but are missing from the current map (i.e., potential disappearance areas), outputting a set of points or voxels representing spatial changes. The spatial difference set, generated by the global map difference operator, is a set of 3D point data representing inconsistencies between the historical global map and the current aligned point cloud map. It mainly includes historical structures that no longer appear in the new map due to backfilling of tunnels, serving as the basic input for subsequent contour extraction and backfill determination. The Alpha Shapes algorithm is a computational geometry method (Alpha Shapes algorithm) that constructs adjustable-fineness boundary contours based on point sets. By setting a shape control parameter, it adaptively generates closed or open polygonal boundaries between convex and concave hulls while preserving the original point set topology, making it suitable for contour reconstruction of complex tunnel cross-sections. The tunnel outline boundary refers to a two-dimensional closed or multi-segment curve extracted from the spatial difference set on a horizontal projection plane using the Alpha Shape algorithm. This curve reflects the outer edge shape of the potential backfill area and is used to characterize the spatial extent and geometric shape of the area. The convex hull-like boundary is an approximate concave hull boundary generated based on the tunnel outline boundary by adjusting the parameters of the Alpha Shape algorithm or post-processing strategies (such as local concavity preservation). This boundary more closely resembles the actual tunnel orientation and reflects the internal concave structure, providing a more accurate description of the shape of irregular tunnel areas compared to a convex hull. The contact intersection line is a three-dimensional spatial curve formed by the intersection of the convex hull-like boundary, projected back into three-dimensional space, and the corresponding tunnel surface model (or point cloud) in the historical global map. This curve is located at the boundary between the suspected backfill area and the still-existing tunnel structure, serving as a key geometric clue to determine whether backfilling has actually occurred. Height variation characteristics refer to the coordinate variation pattern of the sampling point sequence along the contact intersection line in the vertical direction (Z-axis), including abrupt elevation changes, slope transitions, or local undulations. This helps identify whether the intersection line is located at a real roadway end face (usually exhibiting a continuous, gentle transition) or an artificial backfill boundary (often accompanied by steep cutoffs or plateau-like terminations). Significant peaks refer to local maxima where the curvature value or height gradient exceeds a certain multiple of the local neighborhood mean when performing curvature or height variation analysis on the smoothed contact intersection line. These points typically correspond to sharp corners, inflection points, or abrupt changes in the boundary profile, and their number and distribution reflect the complexity of the region's geometry. The preset peak value threshold is an empirical threshold (e.g., 3) statistically set based on the geometric characteristics of typical unbackfilled roadway ends (such as the number of peaks generated by natural structures like turns and intersections). This threshold distinguishes between normal structural changes and simple closed boundaries caused by backfilling. When the number of significant peaks is below this threshold, it indicates that the boundary is too smooth or regular, conforming to the characteristics of backfill closure, thus triggering backfill area marking.

[0046] If the deletion trigger condition is met at the current moment (in this embodiment, it is within a preset time window and no map deletion operation has been performed on that day), the map deletion module is entered; otherwise, this step is skipped and the map update module is entered directly. The map deletion module sequentially performs new-old map difference calculation, contour intersection extraction, and intersection peak value determination.

[0047] Let the historical global map be The currently constructed global map is First, the spatial difference set between the two is calculated using a global map difference operator. The spatial difference set is represented as follows: in, The point cloud spatial consistency threshold, set This includes areas of structural difference caused by backfilled roadways or newly added uncollected roadways. express Any three-dimensional point in the array, express Any three-dimensional point in the array.

[0048] Subsequently, in the difference set Extract the tunnel contour boundary and construct a concave hull-like boundary using the Alpha Shapes algorithm. And further calculate its contact line with the old map. , is represented as: in, These are shape control parameters used to balance the fineness and stability of the contour.

[0049] For the obtained intersection line Perform curve smoothing and feature analysis to construct its curvature or height variation function. Based on this, the peak value discrimination rule is defined as follows: in, This indicates the number of significant peaks on the intersection curve. A preset peak value threshold is set. When the intersection line shows multiple significant peak values, the area is judged to have a complex shape and is more likely to originate from areas with missing or uncovered point cloud data, thus avoiding misjudging it as a backfilled alley and deleting it. Conversely, it is judged as a relatively flat backfilled area and is used as a candidate area for map deletion, thereby reducing the probability of accidental deletion while ensuring map accuracy.

[0050] Step S40: Delete the point cloud data corresponding to the backfill tunnel area from the marked point cloud map to obtain a three-dimensional global map; It should be noted that the three-dimensional global map refers to a complete three-dimensional point cloud map that reflects the actual passable roadway space structure in the current mine, formed by completely removing the point cloud data corresponding to the backfill roadway area from the marked point cloud map and integrating the remaining effective areas after the backfill roadway area has been identified and marked by the map management system. This map has geometric consistency, timeliness and topological accuracy.

[0051] Understandably, the map management system first determines whether each point in the marked point cloud map is located within the backfilled tunnel area, based on the marked information (including its 3D spatial boundaries or corresponding voxel index set). Then, all point cloud data identified as belonging to the backfilled tunnel area is removed from the map data structure, while the remaining unmarked point cloud data is retained to maintain the integrity of the effective tunnel structure. This generates an accurate 3D global map reflecting the current actual navigable environment underground, avoiding misleading subsequent positioning and path planning due to the retention of false geometric information from closed areas.

[0052] Step S50: Perform a comprehensive quality assessment on the three-dimensional global map based on multi-dimensional quality assessment indicators to obtain the quality assessment results; It should be noted that multi-dimensional quality assessment indicators refer to multiple complementary metrics used by the map management system to quantify and evaluate the overall reliability of the 3D global map. These include geometric dimensions (e.g., registration error between sub-maps calculated based on Chamfer distance), structural dimensions (e.g., reflecting the smoothness of roadway walls through plane fitting variance), information dimensions (e.g., measuring feature richness using the information entropy of point cloud distribution), and coverage dimensions (e.g., the coverage rate of effective point clouds within the preset mine operation area). These indicators together constitute a comprehensive criterion for judging map accuracy, completeness, and usability. The quality assessment result refers to the structured evaluation conclusion output by the map management system after fusion analysis based on the calculated values ​​of the above multi-dimensional quality assessment indicators and in combination with preset thresholds or weighting rules. It is usually presented in the form of numerical scores, grade labels (e.g., "high / medium / low quality"), or specific problem diagnoses (e.g., "local registration misalignment" or "severe feature loss"), and is used to guide whether to accept the map update, trigger the redrawing mechanism, or issue an alarm for manual intervention.

[0053] As an example, the steps for comprehensively evaluating the quality of the 3D global map based on multi-dimensional quality assessment indicators to obtain the quality assessment result include: dividing the 3D global map into a preset number of local regions and calculating the covariance matrix of the local regions, and calculating the average map entropy based on the covariance matrix; calculating the distance from each coordinate point in the marked point cloud map to the local fitting plane corresponding to each coordinate point, and calculating the average plane variance based on the distance; comparing the coordinate extreme values ​​of the marked point cloud map with the preset work area coordinate range constraints to obtain the geographical scope compliance result; calculating the coordinate fluctuation amplitude of the marked point cloud map and comparing the coordinate fluctuation amplitude with the preset coordinate fluctuation amplitude threshold to obtain the stability assessment result; calculating the chamfer distance between the marked point cloud map and the preset benchmark global map, and obtaining the geometric similarity assessment result based on the chamfer distance; and combining the average map entropy, the average plane variance, the geographical scope compliance result, the stability assessment result, and the geometric similarity assessment result to obtain the quality assessment result.

[0054] The preset quantity refers to the preset number of local regions used to achieve refined evaluation by segmentation. A local region refers to a subset of point clouds within several sub-spaces obtained by uniformly or adaptively dividing the 3D global map according to the preset quantity. Each local region covers a continuous segment of the tunnel and serves as the basic unit for independently calculating quality indicators. The covariance matrix is ​​a 3×3 symmetric matrix calculated using the centroid of all point cloud coordinates within a local region as its origin. It describes the distribution direction and dispersion of point clouds within that region in 3D space. The average map entropy is the arithmetic mean of the information entropy (reflecting the disorder or structural complexity of point cloud distribution) calculated based on the eigenvalues ​​of the covariance matrices of each local region. A lower value indicates a more regular overall geometric structure and clearer features. The local fitting plane is the best approximation plane fitted to the point cloud data within each local region using methods such as least squares. It characterizes the main orientation of the tunnel wall or ground surface in that segment. Mean plane variance refers to the squared mean of the vertical distances from each point in a 3D global map to the local fitted plane corresponding to its local area. It measures the deviation of the point cloud from the ideal plane; a smaller value indicates a smoother local geometry and a more stable map. Coordinate extreme values ​​refer to the maximum and minimum values ​​of all points in the 3D global map along the X, Y, and Z coordinate axes. They define the spatial boundaries of the map in the global coordinate system. Preset work area coordinate range constraints refer to the spatial boundary restrictions (usually defined by the minimum / maximum values ​​of X / Y / Z) of legally passable roadways in the global coordinate system, provided by the mine digital twin system or scheduling platform. These constraints serve as the basis for determining whether the map has exceeded its boundaries.

[0055] The geographical scope compliance result refers to the judgment conclusion obtained by comparing the coordinate extreme values ​​of the 3D global map with the preset work area coordinate range constraints. If all extreme values ​​are within the allowable range, it is judged as compliant; otherwise, it is considered out of bounds. Coordinate fluctuation amplitude refers to the maximum value or standard deviation of the positional offset of the overall point cloud centroid or key control points in the X, Y, and Z directions of the 3D global map over multiple consecutive update cycles, used to measure the stability of the map over time. The preset coordinate fluctuation amplitude threshold is the maximum allowable fluctuation range pre-set according to the accuracy of the downhole positioning system and the long-term consistency requirements of the map. Exceeding this threshold is considered an abnormal map drift. The stability assessment result is the judgment result output after comparing the calculated coordinate fluctuation amplitude with the preset coordinate fluctuation amplitude threshold, used to indicate whether the current map is in a stable state. The preset baseline global map refers to a historical reference map maintained by the system, which has undergone multiple rounds of optimization and verification and is considered highly reliable. It is usually the most recently fused 3D global map that has passed quality assessment and is used as the comparison benchmark for the geometric consistency of the current map. Chamfer distance refers to the weighted average of the bidirectional nearest point distances between the 3D global map and the preset benchmark global map. It is used to quantify the overall degree of geometric difference between the two point cloud maps. The geometric similarity evaluation result is an evaluation conclusion generated based on the magnitude of the chamfer distance and a preset similarity threshold. The smaller the chamfer distance, the higher the geometric similarity, indicating that the current map is more consistent with the benchmark map in structure.

[0056] First, a comprehensive quality assessment is conducted on the stitched and abridged 3D global map. If it passes the assessment, the result is updated, a 2D map is generated, and the map is stored in the database. Otherwise, it waits for the next sub-map upload.

[0057] Let the current stitched and subtracted 3D global map be... First, a comprehensive quality assessment is conducted, which consists of multiple quality indicators, including mean map entropy (MME), mean plane variance (MPV), prior geographic extent constraints, coordinate fluctuation amplitude, and Chamfer distance. Through multi-dimensional joint judgment, a closed-loop quality control mechanism is formed from submap input to global map output. The judgment function for the comprehensive quality assessment is expressed as follows: in, This represents the spatial coordinates of a 3D point in a 3D global map. CD represents the range of coordinate fluctuation of a 3D point in a 3D global map. It is an abbreviation for Chamfer Distance. The preset map structure complexity threshold, To preset the map flatness threshold, This indicates the permissible spatial constraints of the work area. This represents the set of constraints on coordinate fluctuation amplitude. As a baseline global map, This is a preset quality threshold for the baseline global map. Based on the analysis of benchmark datasets such as the static mine road dataset WHU-TLS, the high-precision single-object dataset Stanford Bunny, and the dynamic road dataset KITTI, a threshold is set to meet the registration robustness requirement of over 90% in mine scenarios. , Average map entropy (MME) is calculated by dividing the global map... Statistical calculations are performed on a local region, which is defined as follows: The average map entropy is expressed as follows: in, express The division obtained k A local area, express The local entropy value, For the first The covariance matrix of a local point cloud is used to quantify the structural complexity and disorder of the map. The lower the entropy value, the more regular the point cloud structure.

[0058] The mean plane variance is expressed as follows: in, This represents the distance from a point to its corresponding local fitted plane. This index is used to evaluate the accuracy of plane reconstruction. The lower the value, the more stable the plane structure. express middle Local plane variance; express Any three-dimensional point cloud data point in the dataset.

[0059] By searching the historical database map, a preset work area threshold is established. Prior geographic range constraints are imposed by comparing the extreme values ​​of global map coordinates with preset values. The constraint form is as follows: Preset coordinate fluctuation range working area threshold Used to evaluate the overall stability of a map, it is defined as: Chamfer distance is used to quantify the geometric similarity between an updated map and a baseline map. Chamfer distance is expressed as follows: in, and Representing sets and A three-dimensional point vector; Indicates the number of points in the set. This represents the Euclidean distance between two points.

[0060] Step S60: When the quality assessment result is qualified, a corresponding two-dimensional map is generated based on the three-dimensional global map, and the three-dimensional global map and the two-dimensional map are output to a preset terminal.

[0061] It should be noted that a 2D map refers to a planar raster map or vector map generated based on a 3D global map through vertical projection (usually projecting the point cloud onto the XY plane along the direction of gravity) combined with tunnel contour extraction, topology simplification, and rasterization processing. It retains key navigation information such as tunnel centerlines, intersections, turns, and passable areas, and is suitable for higher-level applications such as path planning and task scheduling. A pre-configured terminal refers to a target device or software platform pre-configured in the intelligent mining system to receive and use map data, including but not limited to the onboard computing unit of an unmanned mining truck, the visualization system of a ground dispatch and monitoring center, the task management module of an inspection robot, or the data interface of a digital twin platform.

[0062] Understandably, when the overall quality assessment result is "qualified", the global map will be updated; otherwise, the update will be abandoned and the system will wait for the next submap upload.

[0063] After the quality assessment is passed, based on the qualified global 3D map Two-dimensional maps are generated based on calculation formulas for hull-like boundaries and contact intersections. Its projection and boundary optimization process is defined as follows: in, This represents the projection operator from 3D to 2D. This represents a boundary smoothing operator based on edge detection and curve fitting, used to extract and optimize the lane contour, forming a continuous and clearly structured two-dimensional lane representation. The two-dimensional map accurately reflects the lane orientation and intersection topology, combining the accuracy advantages of three-dimensional data with the visualization effects of two-dimensional maps, providing clear and practical structured data for interface display.

[0064] Subsequently, the updated 3D global map global_map.pcd and the generated 2D map global_map.json were uploaded to the MySQL database for storage.

[0065] Finally, the 3D and 2D map results are output separately. The 3D point cloud map is stored in a MySQL map database, supporting quick access to real-time positioning; the 2D map is transmitted to the graphical display unit of the user interface, enabling dynamic visualization of the entire underground panorama and local areas. The system triggers the map update process in real time based on the frequency of sub-map uploads, ensuring that the output map is updated synchronously with the real-time environment. This forms a complete closed loop from point cloud processing, map construction, dynamic updates, data storage to multi-terminal applications, providing reliable data support for underground safety monitoring and location services.

[0066] Please refer to Figure 2 , Figure 2 This diagram illustrates the generation of a two-dimensional map provided in Embodiment 1 of the LiDAR map management method for underground mine roadways according to this application. The diagram shows the entire process of generating and updating a LiDAR map for underground mine roadways from an original point cloud sub-map to a two-dimensional map. The upper left side of the diagram shows the source global map (blue), i.e., the historical global map, representing the currently constructed historical three-dimensional point cloud map; the lower side shows the sub-map (red), i.e., the preprocessed point cloud sub-map, representing newly acquired local point cloud data to be fused. Both are input into the map stitching module, and point cloud registration precisely aligns the sub-map to the coordinate system of the source global map, forming the fused "global map". Figure 1 (Red and blue mixed), and map downsampling was performed to optimize the data volume, resulting in a more compact "global map". Figure 2 Subsequently, the system performs differential calculations between the old and new maps, identifies areas of geometric change caused by backfilling, and removes these invalid areas through the map deletion module, generating a "new global map" (green), which retains only the currently existing passable lane structures. Finally, based on the "new global map," the "2D map generation" module extracts key topological information to generate a "new 2D map" (light green) for vehicle routing planning, fully presenting the dynamic evolution process from 3D point cloud to 2D navigation map.

[0067] This embodiment provides a method for managing lidar maps in underground mine roadways. First, an original point cloud sub-map is acquired and preprocessed to effectively remove invalid or low-quality data, improving the stability and efficiency of subsequent processing. Next, a strategy combining random sampling consistency algorithm registration and coherent point drift non-rigid registration is employed to accurately integrate the preprocessed point cloud sub-map into the historical global map. When a preset map deletion trigger condition is met, the system performs differential calculations on the aligned point cloud map and the historical global map, combining contour intersection extraction and peak value determination to accurately identify and mark backfilled roadway areas. Subsequently, the corresponding point cloud data is deleted from the marked point cloud map, generating a 3D global map containing only the current actual roadway structure, avoiding the retention of invalid information. Based on this, the 3D global map is comprehensively scored using multi-dimensional quality evaluation indicators to ensure that the updated map meets accuracy and consistency requirements. When the evaluation result is satisfactory, the system synchronously generates a corresponding 2D map and outputs both the 3D and 2D maps to a preset terminal. This embodiment can achieve high-precision geometric consistency of the global map in underground environments without GPS and with sparse features, and dynamically adapt to changes in roadway backfilling.

[0068] Based on the first embodiment of this application, in the second embodiment of this application, the content that is the same as or similar to that in Embodiment 1 above can be referred to the above description, and will not be repeated hereafter. Based on this, please refer to... Figure 3 , Figure 3 This is a flowchart illustrating the second embodiment of the lidar map management method for underground mine roadways according to this application. Step S20 of the lidar map management method for underground mine roadways includes steps S21 to S25: Step S21: The preprocessed point cloud sub-map is coarsely registered with the historical global map using a random sampling consistency algorithm to obtain a coarsely registered point cloud sub-map. Step S22: Based on the preset deformation rigidity constraint coefficient and the preset spatial smoothness constraint coefficient, the coarsely registered point cloud sub-map is finely registered with the historical global map through the coherent point drift non-rigid registration model to obtain the finely registered point cloud sub-map. Step S23: Calculate the registration error between the finely registered point cloud sub-map and the historical global map; Step S24: When the registration error is greater than or equal to the preset registration error threshold, return to the step of performing coarse registration between the preprocessed point cloud sub-map and the historical global map using the random sampling consistency algorithm to obtain the coarsely registered point cloud sub-map. Step S25: When the registration error is less than the preset registration error threshold, the finely registered point cloud sub-map is used as the aligned point cloud map.

[0069] It should be noted that the historical global map refers to a complete 3D point cloud map maintained by the map management system, which integrates all previously valid point cloud sub-maps and has undergone consistency optimization. It represents the best global understanding of the current accessible underground mine roadway environment and serves as a benchmark reference for new sub-map registration and updates. The preset deformation rigidity constraint coefficient is a weighting parameter used in the coherent point drift non-rigid registration model to control the degree to which non-rigid transformation deviates from rigid transformation. This coefficient is set based on the typical small deformation characteristics of underground roadways caused by mining disturbances or accumulated sensor errors; a higher value indicates a greater tendency to maintain a rigid structure. The preset spatial smoothness constraint coefficient is a regularization parameter used in the coherent point drift non-rigid registration model to adjust the smoothness of the motion field of adjacent points. This coefficient is set according to the geometric continuity requirements of the point cloud to suppress excessive local distortion and ensure a natural and continuous roadway outline after registration. The finely registered point cloud sub-map refers to the optimized result obtained after fine geometric correction of the coarsely registered point cloud sub-map after processing by the coherent point drift non-rigid registration model; it is the alignment result that has not yet passed the final error verification.

[0070] The Coherent Point Drift Non-rigid Registration Model (CPD) is a non-rigid alignment algorithm that models the point set registration problem as a probability density estimation. It solves for a global non-rigid transformation that satisfies motion coherence and smoothness constraints by maximizing the likelihood of the moving point set on the fixed point set using a Gaussian mixture model. This model is suitable for correcting local geometric distortions caused by accumulated errors in SLAM. Registration error refers to a statistical measure (e.g., root mean square error) of the point-to-point or point-to-area distance between the aligned point cloud sub-map and the historical global map within the overlapping area. It quantifies the degree of geometric consistency between the two. The preset registration error threshold is an upper limit for judging registration quality (e.g., 0.15 meters) pre-set based on lidar accuracy, downhole positioning requirements, and map fusion fault tolerance. If the actual registration error is lower than this threshold, the alignment result is considered to meet the accuracy requirements and is acceptable as the final aligned point cloud map.

[0071] Understandably, this is based on the input point cloud to be registered. Taking the preprocessed point cloud sub-map as input, RANSAC coarse registration and CPD non-rigid fine registration are performed sequentially. RANSAC coarse registration estimates the initial rigid transformation through feature matching. The initial rigid transformation solution formula of the random sample consensus algorithm is expressed as follows: in, The reference point cloud for registration (i.e., the historical global map point cloud). This represents the process of generating the set of corresponding points based on the consistency of random sampling. The initial pose transformation obtained from coarse registration. This represents the rigid transformation matrix to be solved; Indicates from The generated set of corresponding interior points is a reliable set of corresponding points retained by the RANSAC algorithm after verification through random sampling; Indicates an index; Represents the point after rigid transformation .

[0072] Based on the coarse registration results, the CPD non-rigid registration model is further used to perform fine alignment (fine registration) of the point cloud. The optimization objective function of the coherent point drift non-rigid registration model is expressed as follows: in, The deformation stiffness constraint coefficient is . The spatial smoothing constraint coefficients are used to obtain the non-rigid transformation through iterative optimization. This enables high-precision alignment between the submap and the global map in complex downhole environments, providing a reliable foundation for subsequent map stitching and updates. Denotes the non-rigid transformation matrix to be solved. It is the gradient of the non-rigid transformation matrix; Represents the regularization term of a non-rigid transformation matrix, constraining the complexity of the transformation; This represents the gradient regularization term of a non-rigid transformation matrix.

[0073] This embodiment first uses a random sampling consensus algorithm to perform feature matching between the preprocessed point cloud sub-map and the historical global map to obtain an initial rigid registration result, effectively eliminating mismatch interference. Then, based on preset deformation rigidity constraint coefficients and preset spatial smoothness constraint coefficients, a coherent point drift non-rigid registration model is used to perform fine non-rigid alignment on the registered point cloud sub-map to correct local geometric distortions. Next, the registration error between the aligned point cloud sub-map and the historical global map is calculated, and it is determined whether it is less than a preset registration error threshold. If it does not meet the threshold, the initial feature matching step is returned to re-execute the registration; if it does meet the threshold, the current result is used as the final aligned point cloud map. This embodiment ensures registration accuracy and improves the geometric consistency of the fusion of the sub-map and the global map through iterative optimization.

[0074] For example, to help understand the implementation process of the lidar map management method for underground mine roadways obtained by combining this embodiment with the above-described embodiment one, please refer to... Figure 4 , Figure 4 A simplified flowchart illustrating a method for managing lidar maps in underground mine roadways is provided, specifically: First, point cloud map acquisition is performed, obtaining the original point cloud sub-map generated by the vehicle-mounted LiDAR scanning in the underground environment. Then, sub-map quality assessment is conducted, initially screening the acquired point cloud data. If the quality is unsatisfactory, the acquisition is repeated; otherwise, a signal processing task is triggered. The task is passed to the map stitching module, which is responsible for fusing the newly acquired sub-map with the historical global map. During stitching, the system determines whether a deletion signal has been triggered. If not, it directly proceeds to the subsequent map quality assessment and update process; if triggered, it enters the map deletion module to perform the identification and removal of backfill tunnels. The deletion module contains a series of sub-steps: first, noise filtering and voxel uniform downsampling based on discrete point filtering are applied to the input sub-map; then, a non-rigid point cloud registration method based on RANSAC-CPD is used to achieve accurate alignment; next, spatial difference regions are identified through differential calculation of the old and new maps, and contour intersections are extracted; then, the peak value of the intersection is used to determine whether it is a backfill tunnel; finally, the backfill tunnel deletion operation is performed, and the updated global map is output. After deletion is completed or skipped, the system performs a map quality assessment on the final map. If the assessment fails, it returns to the map stitching module for reprocessing; if it passes, it enters the map update stage, writing the new map into the database. Subsequently, a corresponding 2D map is generated and compared with the 3D global map. Figure 1 The data is then stored in a MySQL storage system. The entire process achieves closed-loop management from point cloud acquisition, preprocessing, stitching, dynamic deletion to quality assessment and storage.

[0075] Please refer to Figure 5 , Figure 5This is a schematic diagram of the map management system architecture for the LiDAR map management method for underground mine roadways provided in Embodiment 2 of this application. The system uses LiDAR as the input source to collect raw point cloud data of the underground environment. The data first enters the input processing section, which includes two sub-functions: sub-map quality assessment and sub-map range determination. These sub-functions are used to initially screen each frame or segment of point cloud sub-map for noise and structural integrity, and to determine whether it is located within a preset mine operation area, ensuring the validity and legality of the input data. The processed valid sub-maps are transmitted to the core signal processing task triggering module through the server communication channel. This module determines whether to execute map stitching or map deletion tasks based on the task scheduling strategy (such as timed triggering or detection of backfilling events). In the map stitching process, the system sequentially executes three steps: noise filtering, point cloud registration, and downsampling, accurately fusing the new sub-map into the historical global map. In the map deletion process, the system identifies changed areas through map difference calculation, extracts contour intersections, and determines the peak value of the intersections, thereby locating and clearing backfilled roadway areas. After stitching or deletion is completed, the system enters the map update phase, where the final map undergoes a quality assessment. If it passes, the map update operation is executed, and a corresponding 2D map is generated. Finally, the 3D and 2D map data are stored in a MySQL database. The entire processing result is output through the output processing section, including providing high-precision map support for vehicle positioning and displaying visual information to mine dispatchers through a user interface. This architecture achieves complete closed-loop management from data input, intelligent processing, dynamic maintenance to multi-terminal output.

[0076] It should be noted that the above examples are only for understanding this application and do not constitute a limitation on the lidar map management method for underground mine roadways of this application. Any simple modifications based on this technical concept are within the protection scope of this application.

[0077] This application also provides a lidar map management device for underground mine roadways; please refer to... Figure 6 The laser radar map management device for the underground mine roadways includes: The preprocessing module 10 is used to acquire the original point cloud sub-map and preprocess the original point cloud sub-map to obtain the preprocessed point cloud sub-map. The map stitching module 20 is used to align the preprocessed point cloud sub-map with the historical global map using a strategy that combines random sampling consistency algorithm registration with coherent point drift non-rigid registration, so as to obtain an aligned point cloud map. The area marking module 30 is used to perform differential calculation, contour intersection extraction and intersection peak determination on the aligned point cloud map and the historical global map when the preset map deletion trigger condition is met, identify and mark the backfilled alley area, and obtain the marked point cloud map. The map deletion module 40 is used to delete the point cloud data corresponding to the backfilled alley area from the marked point cloud map to obtain a three-dimensional global map. The quality assessment module 50 is used to perform a comprehensive quality assessment of the three-dimensional global map based on multi-dimensional quality assessment indicators, and obtain the quality assessment results. The map update module 60 is used to generate a corresponding two-dimensional map based on the three-dimensional global map when the quality assessment result is qualified, and output the three-dimensional global map and the two-dimensional map to a preset terminal.

[0078] The LiDAR map management device for underground mine roadways provided in this application adopts the LiDAR map management method for underground mine roadways in the above embodiments. It can solve the technical problem of how to achieve high-precision geometric consistency of the global map and dynamically adapt to changes in roadway backfilling in an underground environment without GPS and with sparse features. Compared with the prior art, the beneficial effects of the LiDAR map management device for underground mine roadways provided in this application are the same as those of the LiDAR map management method for underground mine roadways provided in the above embodiments. Moreover, other technical features in the LiDAR map management device for underground mine roadways are the same as those disclosed in the methods of the above embodiments, and will not be repeated here.

[0079] This application provides a lidar map management device for underground mine roadways. The lidar map management device for underground mine roadways includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the lidar map management method for underground mine roadways in the above embodiment 1.

[0080] The following is for reference. Figure 7 This document illustrates a structural schematic diagram of a lidar map management device suitable for implementing embodiments of this application in underground mine roadways. The lidar map management device for underground mine roadways in this application embodiment may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital radio receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Description), PMPs (Portable Media Players), vehicle terminals (e.g., vehicle navigation terminals), and fixed terminals such as digital TVs and desktop computers. Figure 7 The lidar map management device for underground mine roadways shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this application.

[0081] like Figure 7 As shown, the LiDAR map management device for underground mine roadways may include a processing unit 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to programs stored in ROM (Read Only Memory) 1002 or programs loaded from storage device 1003 into RAM (Random Access Memory) 1004. RAM 1004 also stores various programs and data required for the operation of the LiDAR map management device for underground mine roadways. The processing unit 1001, ROM 1002, and RAM 1004 are interconnected via bus 1005. Input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to I / O interface 1006: input devices 1007 including, for example, touchscreens, touchpads, keyboards, mice, image sensors, microphones, accelerometers, gyroscopes, etc.; output devices 1008 including, for example, LCDs (Liquid Crystal Displays), speakers, vibrators, etc.; storage devices 1003 including, for example, magnetic tapes, hard disks, etc.; and communication devices 1009. Communication device 1009 allows the LiDAR map management equipment in the mine roadway to exchange data wirelessly or via wired communication with other devices. Although the figure shows a LiDAR map management equipment in a mine roadway with various systems, it should be understood that it is not required to implement or possess all the systems shown. More or fewer systems can be implemented alternatively.

[0082] Specifically, according to the embodiments disclosed in this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from storage device 1003, or installed from ROM 1002. When the computer program is executed by processing device 1001, it performs the functions defined in the methods of the embodiments disclosed in this application.

[0083] The LiDAR map management device for underground mine roadways provided in this application, employing the LiDAR map management method for underground mine roadways described in the above embodiments, can solve the technical problem of achieving high-precision geometric consistency of the global map and dynamically adapting to changes in roadway backfilling in underground environments without GPS and with sparse features. Compared with the prior art, the beneficial effects of the LiDAR map management device for underground mine roadways provided in this application are the same as those of the LiDAR map management method for underground mine roadways provided in the above embodiments, and other technical features in this LiDAR map management device for underground mine roadways are the same as those disclosed in the previous embodiment method, and will not be repeated here.

[0084] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0085] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0086] This application provides a computer-readable storage medium having computer-readable program instructions (i.e., a computer program) stored thereon, which are used to execute the laser radar map management method for underground mine roadways in the above embodiments.

[0087] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, RAM (Random Access Memory), ROM (Read Only Memory), EPROM (Erasable Programmable Read Only Memory or Flash Memory), optical fibers, CD-ROM (CD-Read Only Memory), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, system, or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0088] The aforementioned computer-readable storage medium may be included in the lidar map management equipment for underground mine roadways; or it may exist independently and not be installed in the lidar map management equipment for underground mine roadways.

[0089] The aforementioned computer-readable storage medium carries one or more programs. When these programs are executed by a LiDAR map management device in a mine roadway, the LiDAR map management device in the mine roadway causes the following actions: It acquires an original point cloud sub-map and preprocesses the original point cloud sub-map to obtain a preprocessed point cloud sub-map; it aligns the preprocessed point cloud sub-map with a historical global map using a strategy combining random sampling consistency algorithm registration and coherent point drift non-rigid registration to obtain an aligned point cloud map; and when a preset map deletion trigger condition is met, it performs the following actions: The aligned point cloud map is compared with the historical global map by differential calculation, contour intersection extraction, and intersection peak determination to identify and mark the backfill tunnel area, thus obtaining a marked point cloud map. The point cloud data corresponding to the backfill tunnel area is deleted from the marked point cloud map to obtain a three-dimensional global map. The three-dimensional global map is comprehensively evaluated based on multi-dimensional quality evaluation indicators to obtain a quality evaluation result. When the quality evaluation result is qualified, a corresponding two-dimensional map is generated based on the three-dimensional global map, and the three-dimensional global map and the two-dimensional map are output to a preset terminal.

[0090] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including LAN (Local Area Network) or WAN (Wide Area Network)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0091] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0092] The modules described in the embodiments of this application can be implemented in software or hardware. The names of the modules do not necessarily limit the functionality of the unit itself.

[0093] The readable storage medium provided in this application is a computer-readable storage medium that stores computer-readable program instructions (i.e., a computer program) for executing the aforementioned method for managing lidar maps of underground mine roadways. This method solves the technical problem of achieving high-precision geometric consistency of a global map and dynamically adapting to changes in roadway backfilling in underground environments without GPS and with sparse features. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as those of the lidar map management method for underground mine roadways provided in the above embodiments, and will not be elaborated upon here.

[0094] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the above-described method for managing lidar maps of underground mine roadways.

[0095] The computer program product provided in this application can solve the technical problem of achieving high-precision geometric consistency of the global map and dynamically adapting to changes in roadway backfilling in underground environments without GPS and with sparse features. Compared with the prior art, the beneficial effects of the computer program product provided in this application are the same as those of the lidar map management method for underground mine roadways provided in the above embodiments, and will not be repeated here.

[0096] The above description is only a part of the embodiments of this application and does not limit the patent scope of this application. All equivalent structural transformations made under the technical concept of this application and using the contents of the specification and drawings of this application, or direct / indirect applications in other related technical fields, are included in the patent protection scope of this application.

Claims

1. A method of managing a laser radar map of a mine roadway, characterized by, The method includes: Obtain the original point cloud sub-map and preprocess the original point cloud sub-map to obtain the preprocessed point cloud sub-map. By combining random sampling consensus algorithm registration with coherent point drift non-rigid registration, the preprocessed point cloud sub-map is aligned with the historical global map to obtain the aligned point cloud map. When the preset map deletion trigger condition is met, the aligned point cloud map and the historical global map are subjected to differential calculation, contour intersection extraction and intersection peak determination, the backfilled alley area is identified and marked, and the marked point cloud map is obtained. Delete the point cloud data corresponding to the backfilled tunnel area from the marked point cloud map to obtain a three-dimensional global map; A comprehensive quality assessment of the three-dimensional global map is performed based on multi-dimensional quality assessment indicators to obtain the quality assessment results. When the quality assessment result is qualified, a corresponding two-dimensional map is generated based on the three-dimensional global map, and the three-dimensional global map and the two-dimensional map are output to a preset terminal. The steps of performing differential calculation, contour intersection extraction, and intersection peak determination on the aligned point cloud map and the historical global map when the preset map deletion trigger condition is met, and identifying and marking the backfilled alley area to obtain the marked point cloud map include: When the preset map deletion trigger condition is met, the spatial difference set between the aligned point cloud map and the historical global map is calculated by the global map difference operator; The Alpha Shape Algorithm is used to extract the tunnel contour boundary from the spatial difference set, and a concave hull boundary is constructed based on the tunnel contour boundary. Calculate the contact line between the concave hull boundary and the historical global map, and smooth the contact line to obtain a smoothed contact line. Based on the curvature or height variation characteristics of the smoothed contact intersection line, identify the significant peaks of the smoothed contact intersection line and count the number of the significant peaks; When the number of significant peaks is less than a preset peak determination threshold, the area in the spatial difference set corresponding to the smoothed contact line is marked as a backfill tunnel area, and a marked point cloud map is obtained.

2. The method of claim 1, wherein, The strategy of combining random sampling consensus algorithm registration with coherent point drift non-rigid registration to align the preprocessed point cloud sub-map with the historical global map to obtain the aligned point cloud map includes the following steps: The preprocessed point cloud sub-map is coarsely registered with the historical global map using a random sampling consistency algorithm to obtain a coarsely registered point cloud sub-map. Based on preset deformation rigidity constraint coefficients and preset spatial smoothness constraint coefficients, the coarsely registered point cloud sub-map is finely registered with the historical global map using a coherent point drift non-rigid registration model to obtain a finely registered point cloud sub-map. Calculate the registration error between the finely registered point cloud sub-map and the historical global map; When the registration error is greater than or equal to the preset registration error threshold, return to the step of performing coarse registration between the preprocessed point cloud sub-map and the historical global map using the random sampling consistency algorithm to obtain the coarsely registered point cloud sub-map. When the registration error is less than the preset registration error threshold, the finely registered point cloud sub-map is used as the aligned point cloud map.

3. The method as described in claim 1, characterized in that, The steps for conducting a comprehensive quality assessment of the 3D global map based on multi-dimensional quality assessment indicators to obtain the quality assessment results include: The three-dimensional global map is divided into a preset number of local regions, and the covariance matrix of the local regions is calculated. The average map entropy is then calculated based on the covariance matrix. Calculate the distance from each coordinate point in the marked point cloud map to the local fitting plane corresponding to each coordinate point, and calculate the mean plane variance based on the distance; The extreme values ​​of the marked point cloud map are compared with the preset work area coordinate range constraints to obtain the geographical range compliance result. The coordinate fluctuation range of the marked point cloud map is calculated, and the coordinate fluctuation range is compared with a preset coordinate fluctuation range threshold to obtain the stability evaluation result; Calculate the chamfer distance between the marked point cloud map and the preset baseline global map, and obtain the geometric similarity evaluation result based on the chamfer distance; The quality assessment result is obtained by combining the average map entropy, the average planar variance, the geographical scope compliance result, the stability assessment result, and the geometric similarity assessment result.

4. The method as described in claim 1, characterized in that, The steps of obtaining the original point cloud sub-map and preprocessing the original point cloud sub-map to obtain the preprocessed point cloud sub-map include: A scanning command is sent to the vehicle-mounted lidar so that the vehicle-mounted lidar can scan the underground roadway environment, collect and return the original point cloud sub-map; The original point cloud sub-maps are subjected to initial quality screening and spatial validity checks to obtain a set of valid sub-maps; The effective sub-map set is subjected to noise filtering and voxel uniform downsampling to obtain the preprocessed point cloud sub-map.

5. The method as described in claim 4, characterized in that, The steps of performing initial quality screening and spatial validity checks on the original point cloud sub-maps to obtain a set of valid sub-maps include: Ground point extraction is performed on the original point cloud sub-map to obtain a ground point set; Calculate the area of ​​the convex hull region of the ground point set in the horizontal projection plane. The area of ​​the convex hull region is obtained based on the convex hull operator and the polygon area calculation operator. When the area of ​​the convex hull region is within a preset convex hull area threshold range and all three-dimensional coordinate points in the original point cloud sub-map are within a preset mine operation area, the original point cloud sub-map is considered a valid sub-map. Based on the multiple valid subgraphs, a set of valid subgraphs is obtained.

6. The method as described in claim 4, characterized in that, The step of performing noise filtering and voxel uniform downsampling on the effective sub-map set to obtain the preprocessed point cloud sub-map includes: The point cloud data in the effective sub-graph set is subjected to noise suppression according to the discrete point filtering operator to obtain the noise-suppressed point cloud data. The discrete-point filtering operator is represented as follows: in, Point The local neighborhood, This indicates deviation from the metric function. The noise threshold is used to determine the noise level. This represents the point cloud data after noise suppression. This represents the discrete-point filtering operator. This represents the point cloud data in the set of valid subgraphs. express Any three-dimensional point cloud data point; Based on the preset voxel side length, a voxel mesh is constructed with the minimum coordinate in the noise-suppressed point cloud data as the reference starting point. The noise-suppressed point cloud data is divided into several voxel subsets according to the voxel index using the voxel grid. Based on the preset maximum number of retained points, a uniform sampling operation is performed in each voxel subset to obtain downsampled point cloud data. All the downsampled point cloud data are integrated to obtain a preprocessed point cloud sub-map.

7. A lidar map management device for underground mine roadways, characterized in that, The device employs the laser radar map management method for underground mine roadways as described in any one of claims 1 to 6, and the device comprises: The preprocessing module is used to acquire the original point cloud sub-map and preprocess the original point cloud sub-map to obtain the preprocessed point cloud sub-map. The map stitching module is used to align the preprocessed point cloud sub-map with the historical global map using a strategy that combines random sampling consistency algorithm registration with coherent point drift non-rigid registration, so as to obtain an aligned point cloud map. The area marking module is used to perform differential calculation, contour intersection extraction and intersection peak determination on the aligned point cloud map and the historical global map when the preset map deletion trigger condition is met, identify and mark the backfilled alley area, and obtain the marked point cloud map. The map deletion module is used to delete the point cloud data corresponding to the backfilled alleyway area from the marked point cloud map to obtain a three-dimensional global map. The quality assessment module is used to perform a comprehensive quality assessment of the three-dimensional global map based on multi-dimensional quality assessment indicators, and obtain the quality assessment results. The map update module is used to generate a corresponding two-dimensional map based on the three-dimensional global map when the quality assessment result is qualified, and output the three-dimensional global map and the two-dimensional map to a preset terminal.

8. A lidar map management device for underground mine roadways, characterized in that, The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the laser radar map management method for underground mine roadways as claimed in any one of claims 1 to 6.

9. A storage medium, characterized in that, The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, it implements the steps of the laser radar map management method for underground mine roadways as described in any one of claims 1 to 6.

Images