A method and system for underground space mapping based on laser SLAM

Abstract

The present invention relates to the field of positioning and map construction technology, and in particular to a method and system for underground space mapping based on laser SLAM. The method of the present invention comprises: obtaining point cloud data and IMU data of a target space containing artificial markers, wherein the artificial markers are used to enhance the environmental characteristics of the target space; obtaining a laser odometry factor using the A‑LOAM algorithm based on the point cloud data, and obtaining an IMU pre-integration factor based on the IMU data; constructing a factor graph based on the laser odometry factor and the IMU pre-integration factor; optimizing the factor graph using a nonlinear optimization method, and constructing a map based on the optimization results. The method of the present invention can improve the accuracy of underground space mapping.

Description

Technical Field

[0001] The present invention relates to the field of positioning and map construction technology. More specifically, the present invention relates to a method and system for underground space mapping based on laser SLAM. Background Art

[0002] With the continuous development of underground engineering construction, the demand for mapping narrow and long underground spaces, such as tunnels, underground parking lots, and mines, is growing. In narrow and long underground environments, robots face difficulties in accurately building maps using GNSS (Global Navigation Satellite System) signals due to signal obstruction and multipath effects. Current approaches to address this issue utilize SLAM (Simultaneous Localization and Mapping) algorithms that do not rely on positioning information for mapping. Specifically, when a mobile robot is in an unknown environment, it estimates its position and builds a map of the environment based on environmental information such as distance and images collected by its onboard sensors. SLAM is categorized into laser SLAM and visual SLAM, depending on the sensors used. Visual SLAM algorithms are effective in capturing feature information in texture-rich environments, but they are computationally intensive and significantly affected by lighting. Compared to visual SLAM algorithms, laser SLAM algorithms are less susceptible to lighting and require less computation, making them more effective for mapping underground spaces.

[0003] However, underground spaces usually have highly similar structures, such as tunnels and corridors, which have relatively regular shapes and layouts and lack obvious unique features. This makes it easy for laser SLAM to misjudge similar points at different locations as the same feature point when extracting feature points, resulting in feature matching errors.

[0004] Therefore, how to solve the problem of map drift caused by the scarcity of environmental features is a technical problem that needs to be solved urgently. Summary of the Invention

[0005] In order to solve the technical problem of low accuracy in underground space mapping caused by the scarcity of environmental features, the present invention provides solutions in the following aspects.

[0006] In a first aspect, the present invention provides a method for mapping an underground space based on laser SLAM, comprising: obtaining point cloud data and IMU data of a target space including artificial markers, wherein the artificial markers are used to enhance the environmental characteristics of the target space; obtaining a laser odometry factor based on the point cloud data using an A-LOAM algorithm, and obtaining an IMU pre-integration factor based on the IMU data; constructing a factor graph based on the laser odometry factor and the IMU pre-integration factor; optimizing the factor graph using a nonlinear optimization method, and constructing a map based on the optimization results.

[0007] Furthermore, the A-LOAM algorithm is used to obtain a laser odometry factor, including: obtaining edge points and plane points based on the point cloud data; multiplying an adaptive loss function by a first weight to obtain a first distance, and multiplying a second perpendicular distance from the plane point to the corresponding plane by a second weight to obtain a second distance, wherein the independent variable of the adaptive loss function is the first perpendicular distance from the edge point to the corresponding straight line, the first weight is positively correlated with the difference between the curvature of the edge point and the average curvature, and the second weight is negatively correlated with the second perpendicular distance; based on the first distance and the second distance, solving the optimal posture transformation matrix at adjacent moments; and using the posture transformation matrix as the laser odometry factor.

[0008] Furthermore, the calculation expression of the first weight is:

[0009] ;

[0010] Where, The current point cloud frame The first weight of the edge point, The current point cloud frame The curvature of the edge points, is the average curvature of all edge points in the current point cloud frame, is the gain coefficient, is the standard deviation of the curvature of the current point cloud frame.

[0011] Furthermore, the calculation expression of the first distance is:

[0012] ;

[0013] Where, The current point cloud frame The first distance from an edge point to the corresponding line, The current point cloud frame The first weight of the edge point, The current point cloud frame The first perpendicular distance from an edge point to the corresponding line, is the threshold value, which is positively correlated with the reference distance. is the Huber loss function.

[0014] Furthermore, the calculation expression of the second weight is:

[0015] ;

[0016] Where, The current point cloud frame The second weight of the plane point is the current point cloud frame The second perpendicular distance from a plane point to the corresponding plane, is the scale factor, which is determined according to the actual application scenario.

[0017] Furthermore, it also includes: in response to the first vertical distance or the second vertical distance exceeding a preset value in a continuous preset number of iterations, temporarily eliminating the corresponding edge points or plane points when solving the optimal posture transformation matrix.

[0018] Furthermore, before obtaining the laser odometry factor, the method further includes: adjusting the maximum number of iterations in the A-LOAM algorithm multiple times and performing corresponding mapping; and using the maximum number of iterations corresponding to the minimum trajectory error value as the final maximum number of iterations of the A-LOAM algorithm.

[0019] Furthermore, when acquiring point cloud data, it includes: scanning a target space including artificial marking points spaced apart along a channel direction by a laser radar to obtain the point cloud data, wherein the artificial marking points are reflective films.

[0020] Furthermore, when constructing and optimizing the factor graph, it includes: constructing the factor graph through a preset sliding window, and optimizing the factor graph based on the Gauss-Newton algorithm.

[0021] In a second aspect, the present invention provides an underground space mapping system based on laser SLAM, comprising a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, an underground space mapping method based on laser SLAM according to the first aspect is implemented.

[0022] The beneficial effects of the present invention are: by selecting the maximum number of iterations with the smallest trajectory error as the final maximum number of iterations of the A-LOAM algorithm, the present invention can reduce the cumulative error caused by long-term operation in a narrow and long underground space, thereby improving the accuracy of laser SLAM algorithm in underground space mapping; by adding artificial marking points at regular intervals in the underground space to enhance the characteristics of the underground space, the problem of drift when mapping the environment due to the scarcity of feature points in the underground space is avoided, thereby improving the accuracy of mapping; by fusing the IMU pre-integration factor and the laser odometry factor, underground space mapping can be achieved without relying on GNSS signals, thereby avoiding the problem of GNSS signals being easily lost in the underground space, resulting in confusing mapping and inaccurate positioning, thereby improving the accuracy of underground space mapping; by dynamically allocating weights to the first vertical distance and the second vertical distance, the interference of abnormal points in the optimization process is suppressed, thereby improving the accuracy of pose estimation, thereby improving the accuracy of subsequent mapping. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 is a flow chart schematically illustrating a method for mapping underground space based on laser SLAM according to an embodiment of the present invention;

[0024] Figure 2 is a pictorial diagram schematically illustrating a mobile platform according to an embodiment of the present invention;

[0025] Figure 3 is a flowchart schematically illustrating construction of a factor graph according to an embodiment of the present invention;

[0026] Figure 4 FIG. 4 is a block diagram schematically illustrating a structure of an underground space mapping system based on laser SLAM according to an embodiment of the present invention. DETAILED DESCRIPTION

[0027] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work shall fall within the scope of protection of the present invention.

[0028] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0029] Figure 1 FIG. 4 is a flow chart schematically illustrating a method for mapping underground space based on laser SLAM according to an embodiment of the present invention.

[0030] In a first aspect, the present invention provides an underground space mapping method based on laser SLAM, such as Figure 1As shown, the method of the present invention includes:

[0031] S1. Obtain point cloud data and IMU data.

[0032] Before data collection, multiple manual marking points are set at intervals along the direction of the passage, i.e. the direction of the equipment’s movement, in the target space (e.g., a narrow underground space such as a tunnel or pipe gallery). The equipment referred to can be a robot or a Figure 2 The mobile platform shown. Specifically, each artificial marker (in this embodiment, a reflective film) is placed on the wall between 0.5 and 1 meter above the ground to ensure that the reflective film can be scanned by the LiDAR. Furthermore, the reflective film can be arranged in a diamond pattern (with a diagonal length of 15 cm) to avoid being identical or similar to natural structures within the tunnel, thereby enhancing the uniqueness of the feature points. Furthermore, the reflective film has a retroreflectivity greater than or equal to 500 cd / lx / m².

[0033] After adding manual marking points, a mobile platform equipped with a mechanical lidar travels at a constant speed in the target space containing the manual marking points, while scanning the target space with the lidar to obtain point cloud data; and collecting IMU (Inertial Measurement Unit) data through the IMU, where the IMU data includes acceleration and angular velocity.

[0034] In this embodiment, a Leishen intelligent C16 laser radar is used to collect point cloud data of the target space. The measurement range of the laser radar is 100m and the measurement accuracy is ±3cm.

[0035] By adding artificial marking points in the target space, the environmental characteristics of the underground space are enhanced, and the problems of incorrect positioning and mapping confusion caused by the scarcity of underground space feature points are solved, thereby improving the accuracy of underground space mapping.

[0036] S2. Determine the laser odometry factor based on the point cloud data and the IMU pre-integration factor based on the IMU data.

[0037] In one embodiment, the laser odometry factor is obtained using the A-LOAM algorithm based on the point cloud data.

[0038] It should be noted that before determining the laser odometry factor, the maximum number of iterations for the A-LOAM algorithm was also adjusted. Specifically, the maximum number of iterations for the ceres library in the A-LOAM algorithm was adjusted multiple times, and corresponding mapping trajectory experiments were conducted. The trajectory error results corresponding to different maximum iteration numbers were compared, and the maximum number of iterations corresponding to the minimum trajectory error was used as the final maximum number of iterations for the ceres library.

[0039] By selecting the maximum number of iterations corresponding to the minimum trajectory error value as the final maximum number of iterations, the cumulative error and obvious drift problem of the traditional SLAM algorithm during long-term operation are avoided, thereby improving the accuracy of underground space mapping.

[0040] Specifically, the method for determining the laser odometry factor includes:

[0041] S201. Acquire edge points and plane points based on point cloud data.

[0042] Specifically, the curvature of each point in the original point cloud data is calculated, and edge points and plane points are selected as key features based on the magnitude of the curvature. Among them, edge points represent points with larger curvature in the point cloud, and plane points represent points with smaller curvature in the point cloud.

[0043] S202: Determine a first distance from an edge point to a corresponding straight line and a second distance from a plane point to a corresponding plane.

[0044] Specifically, when determining the first distance, for any edge point in the current point cloud frame, the two points closest to the edge point in the previous point cloud frame are determined, and a straight line is constructed using these two points. Then, the first distance from the edge point to the corresponding straight line is calculated based on the adaptive loss function and the first weight, wherein the first weight is positively correlated with the difference between the curvature of the edge point and the average curvature. Specifically, the calculation expression of the first weight is:

[0045] ;

[0046] Where, The current point cloud frame The first weight of the edge point, The current point cloud frame The curvature of the edge points, is the average value of the curvature of all edge points in the current point cloud frame (i.e., the average curvature), is the gain coefficient, ranging from 0.5 to 3, is the standard deviation of the curvature of all edge points in the current point cloud frame.

[0047] It is understandable that the greater the difference in curvature between an edge point and all edge points, the better the determined edge point can characterize the environmental characteristics (i.e., these points contain more environmental structural information). Therefore, by assigning higher weights to edge points with relatively large curvatures, these key feature points can play a greater role in the optimization process when the distance from the point to the line is subsequently calculated, thereby improving the accuracy of pose estimation, helping the algorithm to converge faster, reducing the number of iterations, and thus improving computational efficiency.

[0048] In one embodiment, the adaptive loss function may be a Huber loss function. Specifically, the calculation expression of the first distance is:

[0049] ;

[0050] Where, For the The first distance from an edge point to the corresponding line, The current point cloud frame The first weight of the edge point, The current point cloud frame The first perpendicular distance from an edge point to the corresponding line, is the threshold value, which is positively correlated with the reference distance. is the Huber loss function.

[0051] Specifically, The calculation expression is:

[0052] ;

[0053] In this embodiment, the reference distance is the median of the first vertical distances in history. Specifically, The threshold is 1.345 times the reference distance. By setting this threshold to be related to historical data, the loss function is more flexible in handling outliers, reducing estimation bias caused by improper threshold selection, and thus making the predicted value output by the model closer to the true value. In an optional embodiment, the reference distance can also be the mean of the historical first vertical distances.

[0054] When the absolute value of the first perpendicular distance from the edge point to the corresponding straight line is less than or equal to the threshold When , square loss is used, which has high convergence efficiency and is sensitive to normal points; but when the absolute value of the first perpendicular distance from the edge point to the corresponding straight line is greater than the threshold When , it indicates that the edge point is an outlier. At this time, the loss function turns to linear growth, and the gradient of the outlier point is limited to a fixed value. , rather than increasing infinitely with the error, thereby suppressing the gradient weight of the outliers and making the optimization direction more dominated by normal points, thereby avoiding the interference of outliers on pose estimation, and thus improving the accuracy and efficiency of determining the optimal pose transformation matrix.

[0055] At the same time, for any plane point in the current point cloud frame, determine the three points closest to the plane point in the previous point cloud frame, construct a plane with these three points, and then calculate the second perpendicular distance from the plane point to this plane. Based on the second weight and the second perpendicular distance, determine the second distance from the plane point to the corresponding plane, that is, multiply the second perpendicular distance by the second weight to obtain the second distance. Among them, the second weight is negatively correlated with the second perpendicular distance. Specifically, the calculation expression of the second weight is:

[0056] ;

[0057] Where, The current point cloud frame The second weight of the plane point, The current point cloud frame The second perpendicular distance from a plane point to the corresponding plane, is a scale factor used to control the decay rate of the second weight with the second distance, and is determined according to the actual application scenario.

[0058] It is understandable that the larger the scale factor, the slower the weight decreases, allowing more points with larger distances ( ) retains weights, which is suitable for dynamic or noisy scenes; the smaller the scale factor, the faster the weight decreases, and only points with very close distances are allowed to participate in the optimization process, which is suitable for high-precision and low-noise scenes.

[0059] Since dynamic objects or noise points are generally far away from the corresponding plane, by assigning smaller weights to points that are far away from the plane, these plane points can play a lower role in the optimization process. That is, the interference of dynamic objects or noise points in solving the optimal pose transformation matrix is ​​suppressed, making the optimization process more dependent on static plane points (such as the ground or wall), thereby improving the accuracy and robustness of point cloud registration, and further improving the accuracy of pose estimation and the accuracy of subsequent mapping.

[0060] S203: Based on the first distance and the second distance, solve the optimal posture transformation matrix at adjacent moments, and use the optimal posture transformation matrix as a laser odometry factor.

[0061] Based on the above process, the first distances corresponding to all edge points and the second distances corresponding to all plane points are calculated, then normalized, and all first and second distances are jointly minimized using the LM algorithm to solve the optimal LiDAR pose transformation matrix. In optional embodiments, other nonlinear optimization methods, such as the Gauss-Newton algorithm, can be used.

[0062] During the optimization process, when the residual of a point (edge ​​point or plane point) for three consecutive iterations (the residual corresponding to the edge point is the first vertical distance, and the residual corresponding to the plane point is the second vertical distance) exceeds the set value (in this embodiment, ), indicating that the point is a dynamic interference or severe noise point, its constraints are temporarily removed until the subsequent point cloud frames restore stable observation. That is, when the observation residuals of these points in a subsequent frame return to the normal range, they will re-participate in the subsequent optimization iterations.

[0063] By setting greater than Only then are the constraints on that point removed, ensuring that removal occurs within a reasonable error range, reducing misjudgments and ensuring inclusiveness for normal static points. By only removing points that exceed the set value multiple times in a row during the optimization process, accidental noise interference in single-frame data is eliminated, and only persistently abnormal dynamic targets are isolated to avoid misjudgments. By allowing temporarily removed points to re-participate in optimization after the residuals return to normal in subsequent frames, the continued interference of dynamic objects on pose estimation is avoided while preserving the integrity of the environment map. In summary, by temporarily removing outliers, the reliability and accuracy of determining the optimal pose transformation matrix are improved, thereby improving the accuracy of map construction.

[0064] In summary, by dynamically assigning weights to the first vertical distance and the second vertical distance, the interference of outliers on pose estimation is suppressed, so that the optimization direction is more dominated by normal points, which improves the accuracy of pose estimation and thus improves the accuracy of mapping.

[0065] In one embodiment, obtaining an IMU pre-integration factor based on IMU data includes: pre-integrating the acceleration and angular velocity measured by the IMU to obtain changes in relative posture, velocity, and rotation between adjacent moments, thereby obtaining the IMU pre-integration factor.

[0066] S3. Construct and optimize the factor graph based on the laser odometry factor and the IMU pre-integration factor, and build a map based on the optimization results.

[0067] The laser odometry factor and the IMU pre-integration factor are fused to construct a factor graph containing pose nodes and constraints. When constructing the factor graph, a sliding window approach can be used to limit the scale of the optimization calculations. Specifically, the size and parameters of the sliding window are set according to actual needs. The factor graph is then constructed using this sliding window. When a new node is added to the window, the oldest node in the window is removed. This sliding window-based factor graph construction not only ensures the accuracy of the factor graph but also avoids excessive computational overhead that could affect mapping speed, thereby improving both efficiency and accuracy.

[0068] During the optimization phase, the factor graph is optimized using a nonlinear least squares optimization method. In this embodiment, the Gauss-Newton algorithm is used to optimize the factor graph. Specifically, the initial pose and factor graph parameters are set, and then the Jacobian matrix and residual vector (corresponding to the factor graph optimization error) are calculated, where the Jacobian matrix describes the influence of each factor on the pose, and the residual vector represents the difference between the current pose estimate and the observed value. In each iterative optimization, the current pose estimate is updated based on the Jacobian matrix and the residual vector. Furthermore, after each iteration, it is determined whether the objective function (the sum of squared errors of all factors) is less than the set convergence threshold. If so, the optimization process is terminated; if not, the optimization continues. Through multiple iterative optimizations, the pose estimate finally obtained can effectively reduce the drift caused by the accumulated error, thereby improving the accuracy and stability of map construction.

[0069] Furthermore, a map is constructed based on the obtained pose estimates. Specifically, the positions of feature points in the map are determined based on the optimal estimated values ​​of the feature point nodes obtained by factor graph optimization. The obtained pose estimates are applied to the corresponding point cloud data. The point cloud data from different times and perspectives are fused into the same coordinate system. The point cloud is then converted from the sensor coordinate system to the global map coordinate system, completing the map construction.

[0070] Among them, the optimization error of the factor graph is defined as:

[0071] ;

[0072] Where, is the total number of IMU pre-integration constraints within the sliding window, is the number of edge points extracted from the current point cloud frame, is the number of plane points extracted from the current point cloud frame, The first The posture at a moment, The first The posture at a moment, The first IMU measurement value at a moment, is the motion model function (IMU pre-integration), indicating that the IMU is composed of From moment to moment The predicted pose at the moment; The current point cloud frame The first weight of the edge point, The current point cloud frame The first perpendicular distance from an edge point to the corresponding line, is the threshold, is the Huber loss function, The current point cloud frame The second weight of the plane point, The current point cloud frame The second perpendicular distance from a plane point to the corresponding plane, is the threshold.

[0073] By fusing IMU pre-integration factors and laser odometry factors to construct a factor graph, the team leveraged their complementary nature to reduce cumulative errors, thereby improving map accuracy. This approach also avoids the issues of GNSS signal loss in underground spaces, which can lead to chaotic mapping and inaccurate positioning. Optimization using a sliding window method and the Gauss-Newton algorithm reduced the computational effort during map construction, accelerated optimization, and improved map accuracy.

[0074] Figure 4 FIG1 is a block diagram schematically showing a structure of an underground space mapping system based on laser SLAM according to this embodiment.

[0075] In a second aspect, the present invention also provides an underground space mapping system based on laser SLAM. Figure 4 As shown, the underground space mapping system includes a processor and a memory, and the memory stores computer program instructions. When the computer program instructions are executed by the processor, an underground space mapping method based on laser SLAM according to the first aspect of the present invention is implemented.

[0076] The underground space mapping system also includes other components well known to those skilled in the art, such as a communication interface. The configuration and functions of these components are known in the art and will not be described in detail here.

[0077] In the present invention, the aforementioned memory can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. For example, the computer-readable storage medium can be any suitable magnetic or magneto-optical storage medium, such as Resistive Random Access Memory (RRAM), Dynamic Random Access Memory (DRAM), Static Random Access Memory (SRAM), Enhanced Dynamic Random Access Memory (EDRAM), High-Bandwidth Memory (HBM), Hybrid Memory Cube (HMC), etc., or any other medium that can be used to store the required information and can be accessed by an application, module, or both. Any such computer storage medium can be part of, accessible to, or connectable to a device. Any application or module described in the present invention can be implemented using computer-readable / executable instructions that can be stored or otherwise maintained by such a computer-readable medium.

[0078] In the description of this specification, “a plurality of” means at least two, for example, two, three or more, etc., unless otherwise clearly defined.

[0079] While several embodiments of the present invention have been shown and described herein, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Numerous modifications, variations, and alternatives will occur to those skilled in the art without departing from the concept and spirit of the present invention. It should be understood that various alternatives to the embodiments of the present invention described herein may be employed in practicing the present invention.

Claims

1. A method for underground space mapping based on laser SLAM, characterized in that: include: Acquiring point cloud data and IMU data of a target space including artificial markers, wherein the artificial markers are used to enhance environmental features of the target space; Based on the point cloud data, a laser odometry factor is obtained using the A-LOAM algorithm, and an IMU pre-integration factor is obtained based on the IMU data; wherein, obtaining the laser odometry factor using the A-LOAM algorithm includes: obtaining edge points and plane points based on the point cloud data; multiplying an adaptive loss function by a first weight to obtain a first distance, and multiplying a second perpendicular distance from the plane point to the corresponding plane by a second weight to obtain a second distance, wherein the independent variable of the adaptive loss function is the first perpendicular distance from the edge point to the corresponding straight line, the first weight is positively correlated with the difference between the curvature of the edge point and the average curvature, and the second weight is negatively correlated with the second perpendicular distance; based on the first distance and the second distance, solving the optimal posture transformation matrix at adjacent moments; and using the posture transformation matrix as the laser odometry factor; Constructing a factor graph based on the laser odometry factor and the IMU pre-integration factor; The factor graph is optimized using a nonlinear optimization method, and a map is constructed based on the optimization results.

2. The underground space mapping method based on laser SLAM according to claim 1, characterized in that: The calculation expression of the first weight is: Where, The current point cloud frame The first weight of the edge point, The current point cloud frame The curvature of the edge points, is the average curvature of all edge points in the current point cloud frame, is the gain coefficient, is the standard deviation of the curvature of the current point cloud frame.

3. The underground space mapping method based on laser SLAM according to claim 2, characterized in that: The calculation expression of the first distance is: Where, The current point cloud frame The first distance from an edge point to the corresponding line, The current point cloud frame The first weight of the edge point, The current point cloud frame The first perpendicular distance from an edge point to the corresponding line, is the threshold value, which is positively correlated with the reference distance. is the Huber loss function.

4. The underground space mapping method based on laser SLAM according to claim 1, characterized in that: The calculation expression of the second weight is: Where, The current point cloud frame The second weight of the plane point, The current point cloud frame The second perpendicular distance from a plane point to the corresponding plane, is the scale factor, which is determined according to the actual application scenario.

5. The underground space mapping method based on laser SLAM according to claim 1, characterized in that: Also includes: In response to the first vertical distance or the second vertical distance exceeding a preset value in a continuous preset number of iterations, the corresponding edge points or plane points are temporarily eliminated when solving the posture transformation matrix.

6. The underground space mapping method based on laser SLAM according to claim 1, characterized in that: Before obtaining the laser odometry factor, the following steps are also included: Adjusting the maximum number of iterations in the A-LOAM algorithm multiple times and performing corresponding mapping; The maximum number of iterations corresponding to the minimum trajectory error value is taken as the final maximum number of iterations of the A-LOAM algorithm.

7. The underground space mapping method based on laser SLAM according to claim 1, characterized in that: When acquiring point cloud data, it includes: scanning a target space including artificial marking points spaced apart along a channel direction by a laser radar to obtain the point cloud data, wherein the artificial marking points are reflective films.

8. The underground space mapping method based on laser SLAM according to claim 1, characterized in that: When constructing and optimizing the factor graph, it includes: constructing the factor graph through a preset sliding window, and optimizing the factor graph based on the Gauss-Newton algorithm.

9. An underground space mapping system based on laser SLAM, characterized in that: The invention comprises a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, an underground space mapping method based on laser SLAM according to any one of claims 1 to 8 is implemented.

Images