Unmanned aerial vehicle autonomous coal inventory method based on laser radar

Abstract

The invention relates to an unmanned aerial vehicle autonomous coal inventory method based on a laser radar. The method comprises the following steps: constructing an offline point cloud map of a coal pile environment; controlling the unmanned aerial vehicle to rise to a preset height from a preset range near the original point of the off-line map, collecting an initial point cloud map of the local environment, and registering the initial point cloud map with the off-line map to obtain an optimized initial pose; laser radar point cloud data are continuously collected in real time in the flight process of the unmanned aerial vehicle, and the pose of the unmanned aerial vehicle is estimated based on an offline map so as to update a three-dimensional point cloud map of the environment; the local point cloud density of the three-dimensional point cloud map is monitored in real time, a neighborhood radius parameter of a DBSCAN algorithm is dynamically adjusted, and meanwhile, a point cloud segmentation strategy is optimized in combination with an IMU motion state, so that coal pile point cloud is obtained; incremental mapping of the point cloud of the coal pile is realized through the dynamic point cloud management capability of the ikd-tree, and real-time volume calculation of the coal pile is realized based on the point cloud of the coal pile. Compared with the prior art, the system has the advantages of autonomous coal pile measurement, high measurement accuracy and efficiency and the like.

Description

Technical Field

[0001] The present invention relates to the technical field of point cloud processing, and particularly to an autonomous coal inventory method for unmanned aerial vehicles based on lidar. Background Art

[0002] With the development of unmanned aerial vehicle technology, unmanned aerial vehicles have gradually been applied to the exploration and measurement fields of coal mines. Unmanned aerial vehicles can quickly and accurately obtain the terrain and environmental data of coal mine areas, greatly improving the inventory efficiency.

[0003] However, the existing coal mine inventory technologies based on unmanned aerial vehicles mainly rely on visual imaging technologies, such as ordinary cameras or optical cameras. These imaging devices are greatly affected by changes in environmental light. Especially in the case of insufficient light or strong light reflection, it is difficult to guarantee the image quality, resulting in a reduction in the accuracy and reliability of measurement results. In addition, it is difficult for optical imaging technologies to directly obtain accurate depth information, which limits the precise modeling and three-dimensional reconstruction of coal seams.

[0004] As a mature three-dimensional imaging technology, lidar (Light Detection and Ranging) can provide high-precision depth information in complex environments and has been widely used in fields such as autonomous driving and topographic mapping. Compared with traditional visual imaging technologies, lidar is not affected by light and can work stably under various environmental conditions such as day and night, cloudy days, etc., and can accurately obtain large-scale, high-precision three-dimensional point cloud data. These point cloud data provide more accurate information for the modeling and spatial analysis of coal seams.

[0005] Although lidar technology shows great advantages in many fields, applying it to the autonomous coal inventory task of unmanned aerial vehicles still faces certain challenges. In the existing technology, unmanned aerial vehicles based on lidar usually rely on static and manually set flight paths for data collection, lacking the ability of autonomous environmental perception and path planning. In addition, how to efficiently process and utilize the large-scale point cloud data obtained by lidar to achieve fast and accurate coal seam inventory and measurement while ensuring data accuracy is still an urgent problem to be solved. Summary of the Invention

[0006] The purpose of the present invention is to provide an autonomous coal inventory method for unmanned aerial vehicles based on lidar, which can give full play to the advantages of lidar, combine the flight characteristics of unmanned aerial vehicles, and realize the automated inventory of coal mine areas through autonomous navigation and path planning technologies, improving the efficiency and accuracy of coal seam measurement, and having important application value and market prospects.

[0007] The purpose of the present invention can be achieved through the following technical solutions:

[0008] An autonomous coal inventory method for unmanned aerial vehicles based on lidar, comprising the following steps:

[0009] S1. Construct an offline point cloud map of the coal pile environment;

[0010] S2. Control the drone to rise from a preset range near the origin of the offline map to a preset height, collect the initial point cloud map of the local environment, register the initial point cloud map with the offline map to obtain an optimized initial pose;

[0011] S3. Continuously collect lidar point cloud data in real time during the flight of the drone, and estimate the pose of the drone in real time based on the offline map and the initial pose to update the three-dimensional point cloud map of the environment;

[0012] S4. Dynamically adjust the neighborhood radius parameter of the DBSCAN algorithm by real-time monitoring of the local point cloud density of the three-dimensional point cloud map, and at the same time optimize the point cloud segmentation strategy in combination with the IMU motion state to obtain the coal pile point cloud;

[0013] S5. Realize incremental mapping of the coal pile point cloud through the dynamic point cloud management ability of ikd-tree, and realize real-time coal pile volume calculation based on the coal pile point cloud.

[0014] In step S2, the iterative closest point algorithm is used for accurate registration of the initial point cloud map and the offline map to optimize the initial pose:

[0015]

[0016] where p i is a point in the offline map, q i is the corresponding point in the current point cloud, T is the transformation matrix from the current point cloud to the map point cloud, and finally a pose transformation is solved to minimize the error, obtaining the initial pose of the drone relative to the offline map.

[0017] In step S2, to avoid false matching, a matching success rate threshold is set. If the error of the matching result between the initial point cloud map and the offline map is greater than the set matching success rate threshold, the starting position of the drone is readjusted or the collection of point cloud data is restarted.

[0018] In step S3, the SLAM algorithm is used for real-time positioning and mapping, including the following steps:

[0019] S31. Initialize the initial pose in the SLAM algorithm;

[0020] S32. Obtain the acceleration and angular velocity data provided by the IMU to estimate the relative motion of the drone: By integrating the IMU data, the speed and displacement of the drone in a short period of time are obtained;

[0021]

[0022] P(t) = P(t - 1) + V(t)Δt

[0023] where a(t) is the acceleration provided by the IMU, V(t) is the velocity, and P(t) is the displacement;

[0024] S33. Use the pose estimated by the IMU as the initial value, and use the ICP algorithm to match the point cloud collected in real time with the point cloud in the offline map to obtain a more accurate pose estimate;

[0025] S34. Based on the pose information, continuously update the 3D point cloud map of the environment during the flight of the UAV. Each time new point cloud data is obtained, it is transformed into the point cloud position in the global coordinate system through the pose information for map update.

[0026] The specific method for dynamically adjusting the neighborhood radius parameter of the DBSCAN algorithm is as follows: In the dense area, perform small-range high-precision scanning, and set the neighborhood radius parameter within the first preset range. In the sparse area, switch to large-range search and set the neighborhood radius parameter within the second preset range. The first preset range is smaller than the second preset range.

[0027] The specific method for optimizing the point cloud segmentation strategy by combining the IMU motion state is as follows: Judge the UAV state according to the IMU motion state. When the UAV hovers, reduce the minimum cluster point number threshold to capture the characteristics of small coal piles; when the UAV cruises, increase the cluster point number to resist the interference of motion blur.

[0028] In step S4, to further improve the accuracy of point cloud segmentation, perform secondary screening of the geometric features of the preliminarily extracted coal pile point cloud: Calculate the local curvature based on principal component analysis, and remove the discrete noise points with curvature greater than the preset threshold.

[0029] In step S5, the incremental mapping of the coal pile point cloud is specifically as follows: Downsample each frame of the lidar point cloud, use the radius search function of the ikd-tree to remove the redundant points that are too close to the existing point cloud, and retain the remaining new points. Accumulate the new point clouds of the preset number of frames to form a sub-map; Perform DBSCAN segmentation on each sub-map to obtain the coal pile point cloud; Use the range search function of the ikd-tree to expand the boundary of the sub-map in the direction of the historical point cloud according to the preset distance; Perform Delaunay triangulation on the expanded sub-map with the goal of maximizing the radius of the inscribed circle of the triangle, and transform the extracted point cloud into a series of interconnected triangles to form the surface of the coal pile.

[0030] The specific implementation of real-time coal pile volume calculation based on coal pile point cloud is as follows: Select a reference plane from the coal pile point cloud, and regard each triangle as a cylinder. The height of the cylinder is the vertical distance from the vertex of the triangle to the reference plane. The volume of the cylinder is obtained by multiplying the area of the triangle by the corresponding height. Add up the volumes of all the cylinders to get the volume of the coal pile in the sub-map. Add up the volumes of all the coal piles in the sub-maps to get the total volume of the coal pile.

[0031] The offline map is constructed using the SLAM algorithm.

[0032] Compared with the prior art, the present invention has the following beneficial effects:

[0033] 1. The present invention first constructs an offline map, uses the Iterative Closest Point (ICP) algorithm for precise matching to obtain an accurate initial pose relative to the offline map, and then realizes a SLAM algorithm for positioning and mapping based on the offline point cloud map by tightly coupling lidar and IMU data. By using the offline map, the amount of point cloud data processing is reduced, and the processing efficiency is improved.

[0034] 2. The present invention uses the DBSCAN algorithm adaptive to the dynamic environment to segment the point cloud and extract the coal pile-related area. By dynamically adjusting the neighborhood radius parameter of the DBSCAN algorithm and secondary screening of geometric features, the problems of low efficiency and poor accuracy of traditional segmentation algorithms caused by the irregularity of the coal pile surface and environmental interference are solved, and high-precision segmentation of the coal pile point cloud is realized.

[0035] 3. Aiming at the problems of delay and resource bottleneck caused by traditional volume calculation relying on the full amount of point cloud, the present invention proposes a real-time volume calculation method based on ikd-tree (incremental dynamic k-d tree). The point cloud is transformed into triangles forming the coal pile surface through Delaunay triangulation. By selecting a reference plane and regarding each triangle as a cylinder, the volume of the cylinder is calculated, and the volumes of all the cylinders are added up to obtain the total volume of the coal pile, realizing simple and efficient independent coal inventory. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 It is a flowchart of the method of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0037] The present invention will be described in detail below with reference to the drawings and specific embodiments. This embodiment is implemented on the premise of the technical solution of the present invention, and the detailed implementation manner and specific operation process are given, but the protection scope of the present invention is not limited to the following embodiments.

[0038] Traditional methods for measuring the volume of coal piles mainly rely on manual measurement and simple geometric calculations, and these methods have several significant problems. First, the measurement accuracy is often insufficient. Especially when dealing with irregular coal piles, it is difficult to accurately estimate their actual volume. Second, the operation process is complex and time-consuming. Manual measurement requires multiple round trips to different points for data collection, increasing the possibility of human errors. Moreover, measuring in some dangerous environments poses safety hazards, especially in places where landslides or collapses may occur. In addition, the data update of traditional methods lags behind, making it difficult to reflect the changes of coal piles in real time, affecting the timeliness of resource management and the effectiveness of decision-making. Therefore, it is urgent to introduce new technologies to improve the accuracy and efficiency of measurement and ensure the safety and economy of coal management.

[0039] This embodiment provides a method for the drone to autonomously measure coal piles based on lidar, enabling the drone to achieve stable positioning in the offline map and simultaneously establish a point cloud map of the coal pile environment, so as to achieve autonomous coal pile measurement. As Figure 1 shown, this method includes the following steps:

[0040] S1, construct an offline point cloud map of the coal pile environment.

[0041] The offline map provides a high-precision environmental model for the drone. By comparing the data collected by the drone at the initial position with the offline map data, the initial pose of the drone in the offline map can be obtained. In addition, through the offline map, the geometric parameters of the coal pile environment can be obtained, which is convenient for managers to design the flight trajectory of the drone. Based on this, an offline point cloud map of the coal pile environment needs to be obtained first before coal pile measurement. In order to construct the offline point cloud map of the coal pile environment, this embodiment is based on the existing SLAM algorithm, coupling the data of the lidar and the IMU to estimate the position of the drone in real time and establish a point cloud map.

[0042] In this embodiment, the lidar uses the mid360 of the Liovx series, which comes with an IMU sensor. The lidar is placed obliquely on the top of the drone around the X-axis (for example, 60 degrees), which enables the drone to achieve autonomous positioning of the drone and the construction of the coal pile point cloud map using only one lidar, saving computing resources.

[0043] The drone synchronously collects environmental data using LiDAR (Light Detection and Ranging) and IMU (Inertial Measurement Unit). The LiDAR obtains point cloud data of the environment, while the IMU provides acceleration and angular velocity information for estimating the motion state of the drone. During data collection, existing SLAM (Simultaneous Localization and Mapping) algorithms are used to estimate the pose of the drone in real time and construct a point cloud map simultaneously, finally obtaining an offline point cloud map. The offline point cloud map needs to be further processed by manually removing the parts of the environment that may change, such as the coal pile part, and retaining the point cloud of the static part. This processing can reduce the impact of environmental changes on the positioning accuracy of the drone.

[0044] S2, control the drone to rise from a preset range near the origin of the offline map to a preset height, collect the initial point cloud map of the local environment, and register the initial point cloud map with the offline map to obtain an optimized initial pose.

[0045] To obtain the initial pose of the drone in the environment, first, the drone rises to a certain height at the origin of the offline map near the starting position to ensure that comprehensive local environmental point cloud data can be collected. This process will help form an initial point cloud map representing the current local environment. During this process, the starting position of the drone should be as close as possible to the origin of the offline map to improve the success rate and accuracy of matching. After taking off and rising to a certain height, the drone will rotate and scan the surrounding environment, thereby obtaining a more complete initial point cloud of the local environment.

[0046] Then, use the Iterative Closest Point (ICP) algorithm to perform precise registration of the initial point cloud map and the offline map to optimize the initial pose:

[0047]

[0048] where p i is a point in the offline map, q i is the corresponding point in the current point cloud, T is the transformation matrix from the current point cloud to the map point cloud, and finally a pose transformation is solved to minimize the error, obtaining the initial pose of the drone relative to the offline map.

[0049] In addition, to avoid incorrect matching, set a matching success rate threshold. If the error between the matching results of the initial point cloud map and the offline map is greater than the set matching success rate threshold, readjust the starting position of the drone or re-collect the point cloud data to ensure good registration effect and avoid errors caused by sensor noise or environmental changes.

[0050] Through the above methods, the drone can accurately obtain the pose transformation in the environment, providing reliable data support for subsequent autonomous navigation and mission execution.

[0051] S3. During the flight of the drone, continuously collect lidar point cloud data in real time, and estimate the pose of the drone in real time based on the offline map and the initial pose to update the three-dimensional point cloud map of the environment.

[0052] In order for the drone to obtain its own position information in the offline map, a SLAM algorithm based on the offline point cloud map for positioning and mapping needs to be designed. The positioning module of the algorithm relies on the tight-coupling fusion of lidar and IMU information. In the case of an offline map, the system first estimates a preliminary pose prediction through the IMU data at the current moment, and then matches the feature points in the offline map with the point cloud at the current moment to obtain an optimized pose estimate. The goal of the mapping module is to construct a three-dimensional map of the environment through continuously collected lidar data. The pose information of each frame of point cloud can be obtained through the positioning module, and the global point cloud map in the global coordinate system can be obtained by combining the pose information.

[0053] Specifically, it includes the following steps:

[0054] S31. Initialize the initial pose in the SLAM algorithm;

[0055] S32. Obtain the acceleration and angular velocity data provided by the IMU to estimate the relative motion of the drone: By integrating the IMU data, the velocity and displacement of the drone in a short period of time are obtained:

[0056]

[0057] P(t) = P(t - 1) + V(t)Δt

[0058] where a(t) is the acceleration provided by the IMU, V(t) is the velocity, and P(t) is the displacement;

[0059] S33. Use the pose estimated by the IMU as the initial value, and use the ICP algorithm to match the real-time collected point cloud with the point cloud in the offline map to obtain a more accurate pose estimate;

[0060] S34. Based on the pose information, continuously update the three-dimensional point cloud map of the environment during the flight of the drone. Each time new point cloud data is obtained, it is transformed into the point cloud position in the global coordinate system through the pose information for map update.

[0061] By continuously collecting point cloud data and performing positioning and map updates, the drone can construct a three-dimensional map of the coal pile environment with high precision in real time.

[0062] S4, by real-time monitoring of the local point cloud density of the three-dimensional point cloud map, dynamically adjusting the neighborhood radius parameter of the DBSCAN algorithm, and combining the IMU motion state to optimize the point cloud segmentation strategy, the coal pile point cloud is obtained.

[0063] In view of the low efficiency and poor accuracy of traditional segmentation algorithms caused by the irregular surface of coal piles and environmental interference, this embodiment proposes a dynamic segmentation method that integrates density perception and geometric feature analysis. By real-time monitoring of the local point cloud density of the three-dimensional point cloud map, the neighborhood radius parameter of the DBSCAN (Density-Based Spatial Clustering of Applications with Noise) algorithm is dynamically adjusted: a small-range high-precision scan is used in dense areas, and the neighborhood radius parameter is set within the first preset range. In sparse areas, it switches to a large-range search and sets the neighborhood radius parameter within the second preset range. The first preset range is smaller than the second preset range. At the same time, the point cloud segmentation strategy is optimized in combination with the IMU motion state: the drone state is judged according to the IMU motion state. When the drone is hovering, the minimum clustering point threshold is lowered to capture the characteristics of tiny coal piles; when the drone is cruising, the clustering points are increased to resist motion blur interference.

[0064] In order to further improve the accuracy of point cloud segmentation, the geometric features of the initially extracted coal pile point cloud are secondary screened: the local curvature is calculated based on principal component analysis (PCA), and discrete noise points (such as floating dust) with curvature greater than a preset threshold are eliminated.

[0065] This step significantly reduces the mis-segmentation rate and improves the ability to retain edge details through dynamic parameter adjustment and multi-feature fusion.

[0066] S5, through the dynamic point cloud management capability of ikd-tree, realizes incremental mapping of coal pile point cloud and realizes real-time coal pile volume calculation based on coal pile point cloud.

[0067] In order to solve the delay and resource bottleneck problems caused by the reliance on full point cloud in traditional volume calculation, this embodiment proposes a real-time volume calculation architecture based on ikd-tree (incremental dynamic kd tree). Through the dynamic point cloud management capability of ikd-tree, incremental mapping and volume fusion of coal pile point cloud are realized.

[0068] This method first downsamples each frame of lidar point cloud, and uses the radius search function of ikd-tree (radius 0.1m) to remove redundant points that are too close to the existing point cloud, retaining the remaining new points. The new point clouds of ten frames are accumulated to form a sub-map; DBSCAN segmentation is performed on each sub-map to obtain the point cloud of the coal pile part. In this embodiment, the KdTree module of the PCL library is used to calculate the local density in real time, the search radius is set to 0.5m, and 4 threads are started for parallel acceleration. The neighborhood radius parameter of the DBSCAN algorithm is dynamically adjusted according to the local point cloud density: in the dense area (≥50 points / m 2 ) a small-range high-precision scan of 0.3m is used to distinguish the detailed structure, and in the sparse area (<50 points / m 2 ) it is switched to a large-range search of 0.5m to avoid missed detection. At the same time, the segmentation strategy is optimized by combining the IMU motion state: when the drone hovers, the minimum clustering point threshold is reduced to 50 points to capture the characteristics of small coal piles; when cruising, it is increased to 100 points to resist motion blur interference. To further improve the segmentation accuracy, geometric feature secondary screening is performed on the preliminarily extracted coal pile point cloud: based on principal component analysis (PCA), the local curvature (λ3 / (λ1+λ2+λ3)) (representing the three eigenvalues of the covariance matrix) is calculated, and the discrete noise points (such as floating dust) with a curvature greater than 0.15 are removed.

[0069] Using the range search function of ikd-tree, the sub-map boundary is extended 0.2m in the direction of the historical point cloud; Delaunay triangulation is performed on the extended sub-map with the goal of maximizing the radius of the inscribed circle of the triangle to ensure a good fit of the point cloud data. The extracted point cloud is converted into a series of interconnected triangles to form the surface of the coal pile.

[0070] Next, to calculate the volume, a reference plane (usually the ground of the coal pile) is selected, and each triangle is regarded as a cylinder. The height of the cylinder is the vertical distance from the vertex of the triangle to the reference plane, and the volume is obtained by multiplying the area of the triangle by the height:

[0071] V i =A i ·h i

[0072] where A i is the area of the triangle, and h i is the vertical height from the vertex of the triangle to the reference plane.

[0073] The volumes of all cylinders are added up to obtain the volume of the coal pile in the sub-map:

[0074] V total =∑V i

[0075] Add up the volumes of all sub - maps to obtain the total volume of the coal pile.

[0076] The preferred specific embodiments of the present invention have been described in detail above. It should be understood that those of ordinary skill in the art can make many modifications and variations based on the concept of the present invention without creative efforts. Therefore, all technical solutions that can be obtained by those skilled in the art in the technical field based on the concept of the present invention through logical analysis, reasoning, or limited experiments on the basis of the prior art should fall within the protection scope determined by the claims.

Claims

1. An autonomous coal yard inventory method for drones based on lidar, characterized in that, It includes the following steps: S1. Construct an offline point cloud map of the coal pile environment; S2. Control the drone to rise from a preset range near the origin of the offline map to a preset height, collect the initial point cloud map of the local environment, register the initial point cloud map and the offline map to obtain an optimized initial pose; S3. Continuously collect lidar point cloud data in real time during the flight of the drone, and estimate the pose of the drone in real time based on the offline map and the initial pose to update the three-dimensional point cloud map of the environment; S4. Dynamically adjust the neighborhood radius parameter of the DBSCAN algorithm by real-time monitoring of the local point cloud density of the three-dimensional point cloud map, and at the same time optimize the point cloud segmentation strategy by combining the IMU motion state to obtain the coal pile point cloud; S5. Through the dynamic point cloud management ability of the ikd-tree, realize incremental mapping of the coal pile point cloud, and calculate the real-time coal pile volume based on the coal pile point cloud.

2. The method for the unmanned aerial vehicle to automatically measure the coal yard based on lidar according to claim 1, wherein In step S2, the iterative closest point algorithm is used to accurately register the initial point cloud map and the offline map to optimize the initial pose: where p i is a point in the offline map, q i is the corresponding point in the current point cloud, T is the transformation matrix from the current point cloud to the map point cloud. Finally, a pose transformation is solved to minimize the error, and the initial pose of the UAV relative to the offline map is obtained.

3. The method for the autonomous coal yard inventory of an unmanned aerial vehicle based on lidar according to claim 1, wherein In step S2, to avoid false matching, a matching success rate threshold is set. If the matching result error between the initial point cloud map and the offline map is greater than the set matching success rate threshold, the starting position of the drone is readjusted or the point cloud data is collected again.

4. The method for autonomous coal yard inventory using a lidar-based drone according to claim 1, wherein Step S3 uses the SLAM algorithm for real-time positioning and mapping, including the following steps: S31. Initialize the initial pose in the SLAM algorithm; S32. Obtain the acceleration and angular velocity data provided by the IMU to estimate the relative motion of the drone: By integrating the IMU data, the speed and displacement of the drone in a short time are obtained: P(t) = P(t - 1) + V(t)Δt where a(t) is the acceleration provided by the IMU, V(t) is the speed, and P(t) is the displacement; S33. Take the pose estimated by the IMU as the initial value, and use the ICP algorithm to match the point cloud collected in real time with the point cloud in the offline map to obtain a more accurate pose estimate; S34. Based on the pose information, continuously update the three-dimensional point cloud map of the environment during the flight of the drone. Each time new point cloud data is obtained, it is transformed into the point cloud position in the global coordinate system through the pose information for map update.

5. A method for autonomous coal yard inventory of an unmanned aerial vehicle based on lidar according to claim 1, characterized in that, The specific dynamic adjustment of the neighborhood radius parameter of the DBSCAN algorithm is as follows: In the dense area, small-range high-precision scanning is adopted, and the neighborhood radius parameter within the first preset range is set. In the sparse area, it switches to large-range search, and the neighborhood radius parameter within the second preset range is set. The first preset range is smaller than the second preset range.

6. The method for the autonomous coal yard inventory of an unmanned aerial vehicle based on lidar according to claim 1, wherein, The specific optimization of the point cloud segmentation strategy by combining the IMU motion state is as follows: Judge the state of the drone according to the IMU motion state. When the drone hovers, reduce the minimum clustering point number threshold to capture small coal pile features; when the drone cruises, increase the clustering point number to resist motion blur interference.

7. A method for autonomous coal yard inventory of an unmanned aerial vehicle based on lidar according to claim 1, characterized in that, In step S4, to further improve the accuracy of point cloud segmentation, geometric feature secondary screening is performed on the preliminarily extracted coal pile point cloud: Calculate the local curvature based on the principal component analysis, and remove the discrete noise points with curvature greater than the preset threshold.

8. A method for autonomous coal yard inventory of an unmanned aerial vehicle based on lidar according to claim 1, characterized in that, In the step S5, the incremental mapping of the coal pile point cloud is specifically as follows: downsample each frame of the lidar point cloud, use the radius search function of the ikd-tree to remove redundant points that are too close to the existing point cloud, retain the remaining new points, and accumulate the new point clouds of a preset number of frames to form a sub-map; perform DBSCAN segmentation on each sub-map to obtain the coal pile point cloud; use the range search function of the ikd-tree to expand the boundary of the sub-map in the direction of the historical point cloud according to a preset distance; perform Delaunay triangulation on the expanded sub-map with the goal of maximizing the radius of the inscribed circle of the triangle, and convert the extracted point cloud into a series of interconnected triangles to form the surface of the coal pile.

9. A method for autonomous coal yard mapping of an unmanned aerial vehicle based on lidar according to claim 8, characterized in that, The realization of real-time coal pile volume calculation based on the coal pile point cloud is specifically as follows: select a reference plane from the coal pile point cloud, and regard each triangle as a cylinder. The height of the cylinder is the vertical distance from the vertex of the triangle to the reference plane, and the volume of the cylinder is obtained by multiplying the area of the triangle by the corresponding height; add the volumes of all the cylinders to obtain the volume of the coal pile in the sub-map; add the volumes of the coal piles in all the sub-maps to obtain the total coal pile volume.

10. The method for autonomous coal yard inventory using a lidar-based drone according to claim 1, characterized in that, The offline map is constructed using the SLAM algorithm.

Images