Coal pile volume measurement method, system, device and medium based on laser point cloud

Abstract

The present application belongs to the field of coal yard management of thermal power plants, and discloses a coal pile volume measurement method, system, device and medium based on laser point cloud, which comprises the following steps: spatial interpolation of laser point cloud data of the coal pile area by using a rectangular grid, solving of a characteristic elevation array of each grid unit and construction of an input tensor, inputting of a pre-trained point cloud data optimization neural network model, cutting of ground point data in the input tensor and repair of defective point data to obtain an output tensor, planar projection of the output tensor into a rectangular grid to obtain a projected rectangular grid, and iteration of each grid unit in the projected rectangular grid to generate a voxel block of the current grid unit according to the characteristic elevation array of the current grid unit. The volume of the voxel block of each grid unit is calculated and added to obtain the volume of the coal pile. The ground area rejection is not complete due to manual framing and setting of a fixed horizontal plane elevation in the traditional modeling process, and the high requirement for point cloud quality in the construction of triangular faces in the traditional modeling process is reduced.

Description

Technical Field

[0001] This invention belongs to the field of coal yard management in thermal power plants, and relates to a method, system, equipment and medium for measuring coal pile volume based on laser point cloud. Background Technology

[0002] Coal inventory is a crucial aspect of coal yard management in thermal power plants. With the increasing intelligence and automation levels in these plants, traditional manual coal inventory methods are gradually being replaced by laser coal inventory technology. LiDAR can be mounted on the cantilever of a bucket wheel excavator, on the rails of a closed coal yard, or even carried by drones for scanning operations in open coal yards. For laser coal inventory mounted on bucket wheel excavators and rails, the origin of the point cloud coordinates is relatively fixed, and the ground position can be accurately determined through calibration. However, for drone-based laser coal inventory, ground area identification is often required, typically using manual selection, which is complex and lacks precision.

[0003] In addition, traditional laser point cloud reconstruction often uses a scheme to first construct a triangular mesh surface and then calculate the mesh projection volume to achieve volume measurement. The algorithm is complex and has high requirements for point cloud quality. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method, system, device and medium for measuring coal pile volume based on laser point cloud.

[0005] To achieve the above objectives, the present invention employs the following technical solution:

[0006] In a first aspect, the present invention provides a method for measuring the volume of a coal pile based on laser point clouds, comprising:

[0007] Acquire laser point cloud data of the coal pile area;

[0008] A rectangular grid is used to spatially interpolate the laser point cloud data of the coal pile area, and the characteristic elevation array of each grid cell is solved. An input tensor is constructed based on the characteristic elevation array of each grid cell. The input tensor is used to store the characteristic elevation array of each grid cell.

[0009] The input tensor is used to optimize the neural network model by inputting pre-trained point cloud data. The ground point data in the input tensor is cropped and defective point data is repaired to obtain the output tensor.

[0010] The output tensor plane is projected onto a rectangular grid to obtain a projected rectangular grid. Each grid cell in the projected rectangular grid is traversed, and a voxel block of the current grid cell is generated based on the characteristic elevation array of the current grid cell.

[0011] The volume of the voxel blocks in each grid cell is calculated and summed to obtain the volume of the coal pile.

[0012] Optionally, acquiring the laser point cloud data of the coal pile area includes:

[0013] Acquire laser point cloud data collected from the coal yard and construct a rectangular bounding box for the coal pile area;

[0014] The laser point cloud data collected from the coal yard is filtered by using a rectangular frame of the coal pile area to remove the laser point cloud data outside the rectangular frame of the coal pile area, thus obtaining the laser point cloud data of the coal pile area.

[0015] Before performing spatial interpolation of the laser point cloud data of the coal pile area using a rectangular grid, the method further includes:

[0016] Traverse all laser points in the laser point cloud data of the coal pile area, calculate the minimum distance between each laser point and its nearest neighbor laser points, and identify the current laser point as an outlier when the distance is ≥1.0 to 2.0 times the average laser point distance; remove all outliers from the laser point cloud data of the coal pile area.

[0017] Optionally, the spatial interpolation of the laser point cloud data of the coal pile area using a rectangular grid includes:

[0018] A rectangular grid covering the laser point cloud data of the coal pile area is established on the xy plane; wherein the grid cell length of the rectangular grid is 0.2 to 0.8 times the average laser point distance of the laser point cloud data of the coal pile area projected in the plane;

[0019] Iterate through each grid cell of the rectangular grid, replacing the planar coordinates of the laser point cloud data in the current grid cell with the center coordinates of the current grid cell, while keeping the elevation values ​​unchanged.

[0020] Optionally, the process of solving for the characteristic elevation array of each grid cell includes:

[0021] Traverse each grid cell of the rectangular grid by performing the following steps:

[0022] S11: Sort the elevation values ​​of all laser point cloud data in the current grid cell from largest to smallest to obtain the elevation array of the current grid cell;

[0023] S12: If the elevation array of the current grid cell contains only one elevation value, mark that elevation value as the characteristic elevation of the current grid cell and jump to the next grid cell; if the elevation array of the current grid cell is empty, then the characteristic elevation array of the current grid cell is empty and jump to the next grid cell.

[0024] If there are at least two elevation values ​​in the elevation array of the current grid cell, then the first elevation value in the elevation array of the current grid cell is taken as the feature elevation and S13 is performed.

[0025] S13: Calculate the numerical difference between any two adjacent elevation values ​​in the elevation array of the current grid cell to obtain the first numerical difference. If the first numerical difference is greater than or equal to the first set threshold, mark both elevation values ​​as characteristic elevations.

[0026] S14: If the last elevation value of the current grid cell is not marked as a feature elevation, calculate the numerical difference between the elevation value and all feature elevations to obtain several second numerical differences. If all second numerical differences are ≥ a second set threshold, mark the last elevation value as a feature elevation.

[0027] S15: Sort all feature elevations of the current grid cell from largest to smallest to obtain the feature elevation array of the current grid cell.

[0028] Optionally, constructing the input tensor based on the feature elevation array of each grid cell includes:

[0029] Obtain the maximum value N of the number of feature elevations in the feature elevation array of all grid cells, and create an input tensor of size W×H×N. The input tensor stores n feature elevations from the feature elevation array of each grid cell in the rectangular grid, and fills the empty positions with empty elements. Here, W is the number of grid cells on the long side of the rectangular grid, and H is the number of grid cells on the short side of the rectangular grid.

[0030] Optionally, the point cloud data optimized neural network model includes an encoding layer and a decoding layer connected in sequence;

[0031] The encoding layer consists of a padding layer, a convolutional layer, and a max pooling layer connected in sequence. The padding layer is used to pad the input tensor when the number of grid cells on the long side of the rectangular grid is not equal to the number of grid cells on the short side of the rectangular grid, so that the number of grid cells on the short side of the rectangular grid is equal to the number of grid cells on the short side of the rectangular grid. The padding region is zero-padding.

[0032] The decoding layer consists of a max-unpooling layer, a convolutional layer, and a pruning layer connected in sequence; the pruning layer is used to remove the padding regions in the encoding layer.

[0033] Optionally, the step of traversing each grid cell in the projected rectangular grid and generating the voxel block of the current grid cell based on the feature elevation array of the current grid cell includes:

[0034] Traverse each grid cell in the projected rectangular grid and obtain the number n of characteristic elevations in the characteristic elevation array of the current grid cell;

[0035] When n is odd, generate (n-1) / 2 hexahedral voxels. The length and width of each voxel are equal to the length and width of the current mesh cell. The top and bottom elevations of the i-th voxel are the 2*i-1 and 2*i-th feature elevations in the feature elevation array, respectively. When i = (n-1) / 2, the bottom elevation of the voxel is 0.

[0036] When n is even, n / 2 hexahedral voxels are generated. The length and width of each voxel are equal to the length and width of the current mesh cell. The top and bottom elevations of the i-th voxel are the 2*i-1 and 2*i-th feature elevations of the feature elevation array, respectively.

[0037] No voxel blocks are generated when the feature elevation array is empty.

[0038] In a second aspect, the present invention provides a coal pile volume measurement system based on laser point clouds, comprising:

[0039] The data acquisition module is used to acquire laser point cloud data of the coal pile area;

[0040] The feature processing module is used to perform spatial interpolation on the laser point cloud data of the coal pile area using a rectangular grid, solve for the feature elevation array of each grid cell, and construct an input tensor based on the feature elevation array of each grid cell; wherein, the input tensor is used to store the feature elevation array of each grid cell.

[0041] The ground point segmentation module is used to input the input tensor into the pre-trained point cloud data to optimize the neural network model, and to crop and repair the ground point data in the input tensor to obtain the output tensor.

[0042] The voxel block generation module is used to project the output tensor plane onto a rectangular mesh to obtain a projected rectangular mesh, and traverse each mesh cell in the projected rectangular mesh to generate a voxel block for the current mesh cell based on the characteristic elevation array of the current mesh cell.

[0043] The volume calculation module is used to calculate the volume of the voxel blocks of each grid cell and add them together to obtain the volume of the coal pile.

[0044] In a third aspect, the present invention provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described method for measuring coal pile volume based on laser point clouds.

[0045] In a fourth aspect, the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the above-described method for measuring the volume of a coal pile based on laser point clouds.

[0046] Compared with the prior art, the present invention has the following beneficial effects:

[0047] This invention presents a method for measuring the volume of a coal pile based on laser point clouds. First, spatial interpolation is performed on the laser point cloud data of the coal pile area using a rectangular grid, and the characteristic elevation array of each grid cell is solved. Then, an input tensor is constructed based on the characteristic elevation array of each grid cell, and this input tensor is fed into a pre-trained point cloud data to optimize a neural network model. Ground point data in the input tensor is cropped and defective data is repaired to obtain an output tensor. By using a neural network to identify and remove point cloud data in the ground area and repair defective data, the incomplete removal of ground areas caused by manual selection and setting a fixed horizontal plane elevation in traditional modeling is eliminated, and point cloud voids caused by the quality of the original point cloud are avoided. Next, the output tensor is projected onto a rectangular grid to obtain a projected rectangular grid. Each grid cell in the projected rectangular grid is traversed, and a voxel block of the current grid cell is generated based on the characteristic elevation array of the current grid cell. Finally, the volume of the voxel blocks of each grid cell is calculated and added together to obtain the coal pile volume. By using characteristic elevation to capture key point clouds and achieve voxelized modeling, the high requirements for point cloud quality in the traditional modeling process of constructing triangular faces are reduced. Attached Figure Description

[0048] Figure 1 This is a flowchart of a coal pile volume measurement method based on laser point cloud according to an embodiment of the present invention.

[0049] Figure 2 This is a schematic diagram of rectangular frame filtering according to an embodiment of the present invention.

[0050] Figure 3 This is a schematic diagram illustrating the spatial interpolation principle of an embodiment of the present invention.

[0051] Figure 4 This is a diagram showing the spatial interpolation effect of an embodiment of the present invention.

[0052] Figure 5 This is a schematic diagram of the input tensor construction in an embodiment of the present invention.

[0053] Figure 6 This is a schematic diagram of the neural network model structure for point cloud data optimization in an embodiment of the present invention.

[0054] Figure 7 This is a schematic diagram of defect points caused by unreasonable mesh size in an embodiment of the present invention.

[0055] Figure 8 This is a schematic diagram of defect points caused by holes in the original point cloud, according to an embodiment of the present invention.

[0056] Figure 9This is a schematic diagram of the construction of a voxel block with a feature elevation of 1 according to an embodiment of the present invention.

[0057] Figure 10 This is a schematic diagram of the construction of a voxel block with two feature elevations according to an embodiment of the present invention.

[0058] Figure 11 This is a schematic diagram of the construction of a voxel block with 3 feature elevations according to an embodiment of the present invention. Detailed Implementation

[0059] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0060] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0061] The present invention will now be described in further detail with reference to the accompanying drawings:

[0062] See Figure 1 In one embodiment of the present invention, a method for measuring the volume of a coal pile based on laser point clouds is provided, comprising the following steps:

[0063] S1: Acquire laser point cloud data of the coal pile area.

[0064] S2: Spatial interpolation of the laser point cloud data of the coal pile area is performed using a rectangular grid, and the characteristic elevation array of each grid cell is solved. An input tensor is constructed based on the characteristic elevation array of each grid cell. The input tensor is used to store the characteristic elevation array of each grid cell.

[0065] S3: Input the input tensor into the pre-trained point cloud data to optimize the neural network model, and crop and repair the ground point data in the input tensor to obtain the output tensor.

[0066] S4: Project the output tensor plane onto the rectangular grid to obtain the projected rectangular grid, and traverse each grid cell in the projected rectangular grid. Generate the voxel block of the current grid cell based on the characteristic elevation array of the current grid cell.

[0067] S5: Calculate the volume of the voxel blocks in each grid cell and add them together to obtain the volume of the coal pile.

[0068] To address the shortcomings of traditional laser point cloud reconstruction methods, which require constructing a triangular mesh surface and then calculating the mesh projection volume for volume measurement, resulting in complex algorithms and high requirements for point cloud quality, this invention proposes a coal pile volume measurement method based on laser point clouds. First, a rectangular mesh is used to spatially interpolate the laser point cloud data of the coal pile area, and the characteristic elevation array of each mesh cell is solved. Then, an input tensor is constructed based on the characteristic elevation array of each mesh cell, and this input tensor is input into a pre-trained point cloud data to optimize the neural network model. The ground point data in the input tensor is then cropped and defective points are repaired to obtain the output tensor. By using a neural network to identify and remove point cloud data in the ground area and repair defective points, the method eliminates the incomplete removal of ground areas caused by manual selection and setting a fixed horizontal elevation in traditional modeling processes, and avoids point cloud voids caused by the quality of the original point cloud. Next, the output tensor plane is projected onto a rectangular grid to obtain a projected rectangular grid. Then, each grid cell in the projected rectangular grid is traversed. The voxel block of the current grid cell is generated based on the feature elevation array of the current grid cell. Finally, the volume of the voxel blocks of each grid cell is calculated and added together to obtain the volume of the coal pile. By using feature elevation to capture key point clouds and realize voxel modeling, the high requirements for the quality of point clouds when constructing triangular faces in the traditional modeling process are reduced.

[0069] In one possible implementation, acquiring laser point cloud data of the coal pile area includes: acquiring laser point cloud data collected from the coal yard and constructing a rectangular frame of the coal pile area; filtering the laser point cloud data collected from the coal yard using the rectangular frame of the coal pile area, removing laser point cloud data outside the rectangular frame of the coal pile area, and obtaining the laser point cloud data of the coal pile area.

[0070] Specifically, common point cloud file formats, such as LAS, TXT, and PCD, are parsed to obtain laser point cloud data [xi, yi, zi]. It should be noted that point cloud files store the three-dimensional coordinates of each point, i.e., a two-dimensional array in the form of [[x1, y2, z1], [x2, y2, z3], ..., [xn, yn, zn]]. The storage position of each point in the array is independent of its actual spatial position. Therefore, KDTree or octrees need to be introduced during post-processing to establish the spatial relationships between points, thereby improving the search efficiency of the point cloud. This is one of the drawbacks of traditional point cloud storage methods.

[0071] Then, the point cloud data is filtered using a rectangular bounding box around the coal pile area, removing all laser point cloud data located outside the bounding box (xi, yi). The plane containing the bounding box of the coal pile area is parallel to the xy-plane of the original point cloud coordinate system. Figure 2 The left frame is a rectangular frame of the coal pile area, the right frame is a schematic diagram of the filtered laser point cloud distribution, and both sides are top views of the xy plane.

[0072] In one possible implementation, before spatially interpolating the laser point cloud data of the coal pile area using a rectangular grid, the method further includes: traversing each laser point in the laser point cloud data of the coal pile area, calculating the minimum distance between each laser point and several nearest neighbor laser points, and identifying the current laser point as an outlier when the distance is ≥ 1.0 to 2.0 times the average laser point distance; and removing all outliers from the laser point cloud data of the coal pile area.

[0073] Specifically, during the acquisition of the original laser point cloud, factors such as reflectivity and equipment errors can cause the generated laser point cloud to drift instead of being generated at the object's surface. These drifting points are essentially noise, interfering with the processing of the normal laser point cloud and subsequent modeling and volume calculations, and therefore need to be removed. Typically, the nearest neighbor distance for each laser point is calculated, and the average nearest neighbor distance for all laser points is statistically analyzed. A laser point is considered an outlier when its nearest neighbor distance exceeds X times the average (X≥1).

[0074] In one possible implementation, spatial interpolation of the laser point cloud data of the coal pile area using a rectangular grid includes: establishing a rectangular grid on the xy plane that covers the laser point cloud data of the coal pile area; wherein the grid cell length of the rectangular grid is 0.2 to 0.8 times the average laser point distance of the laser point cloud data of the coal pile area projected in the plane; traversing each grid cell of the rectangular grid, replacing the planar coordinates of the laser point cloud data in the current grid cell with the center coordinates of the current grid cell, while keeping the elevation value unchanged.

[0075] Specifically, spatial interpolation is performed on the laser point cloud data to ensure that all laser point clouds are uniformly distributed in the xy plane; the specific steps are as follows:

[0076] 1. Establish a rectangular grid on the xy plane that covers all laser point cloud data. The minimum grid cell length is 0.2 to 0.8 times the average point distance of the laser point cloud data projected in the xy plane.

[0077] The specified grid size is W×H, where W is the number of grid cells on the longer side of the rectangular grid and H is the number of grid cells on the shorter side of the rectangular grid.

[0078] 2. Within any smallest grid cell, replace the point cloud data (xi,yi) with the center coordinates (Xj,Yj) of the smallest grid cell, while keeping the point cloud data zi unchanged.

[0079] See Figure 3 The diagram illustrates the spatial interpolation process, where triangular points represent the original laser point cloud data, and circular points represent the processed laser point cloud data. (See also...) Figure 4 This demonstrates the effect of spatial interpolation of laser point cloud data of a coal pile area using a rectangular grid.

[0080] In one possible implementation, solving the characteristic elevation array of each grid cell includes:

[0081] Traverse each grid cell of the rectangular grid by performing the following steps:

[0082] S11: Sort the elevation values ​​of all laser point cloud data in the current grid cell from largest to smallest to obtain the elevation array of the current grid cell.

[0083] S12: If the elevation array of the current grid cell contains only one elevation value, mark that elevation value as the characteristic elevation of the current grid cell and jump to the next grid cell; if the elevation array of the current grid cell is empty, then the characteristic elevation array of the current grid cell is empty and jump to the next grid cell.

[0084] If the elevation array of the current grid cell contains at least two elevation values, then the first elevation value in the elevation array of the current grid cell is taken as the feature elevation and S13 is performed.

[0085] S13: Calculate the numerical difference between any two adjacent elevation values ​​in the elevation array of the current grid cell to obtain the first numerical difference. If the first numerical difference is greater than or equal to the first set threshold, mark both elevation values ​​as characteristic elevations.

[0086] S14: If the last elevation value of the current grid cell is not marked as a feature elevation, calculate the numerical difference between the elevation value and all feature elevations to obtain several second numerical differences. If all second numerical differences are ≥ the second set threshold, mark the last elevation value as a feature elevation.

[0087] S15: Sort all feature elevations of the current grid cell from largest to smallest to obtain the feature elevation array of the current grid cell.

[0088] Specifically, the setting standards for the first and second threshold values ​​are generally calibrated during debugging based on the actual engineering conditions.

[0089] In one possible implementation, constructing the input tensor based on the feature elevation array of each grid cell includes: obtaining the maximum value N of the number of feature elevations in the feature elevation array of all grid cells, and creating an input tensor of size W×H×N. The input tensor stores n feature elevations from the feature elevation array of each grid cell in the rectangular grid, and fills the empty positions with empty elements; where W is the number of grid cells on the long side of the rectangular grid, and H is the number of grid cells on the short side of the rectangular grid.

[0090] For details, see Figure 5 Get the maximum number of elements in the feature elevation array of all grid cells, create a W×H×N input tensor to store n (0≤n≤Nmax) feature elevations of each grid cell in the W×H rectangular grid, and fill the remaining positions in the three-dimensional array with empty elements.

[0091] In one possible implementation, the point cloud data optimization neural network model includes an encoding layer and a decoding layer connected in sequence; the encoding layer includes a padding layer, a convolutional layer, and a max pooling layer connected in sequence; the padding layer is used to pad the input tensor when the number of grid cells on the long side of the rectangular grid is not equal to the number of grid cells on the short side of the rectangular grid, so that the number of grid cells on the short side of the rectangular grid is equal to the number of grid cells on the short side of the rectangular grid, and the padding region is zero-padding; the decoding layer includes a max unpooling layer, a convolutional layer, and a pruning layer connected in sequence; the pruning layer is used to remove the padding region filled in the encoding layer.

[0092] See Figure 6A point cloud data optimization neural network model is constructed to encode and decode the input tensor obtained in the first step. The encoding stage sequentially passes through a padding layer, a convolutional layer, and a max pooling layer. The padding layer is used to fill the input tensor when W and H are not equal, so that the grid size in the W and H dimensions is equal. The filled regions are zero-padding. The convolutional layer and the max pooling layer are repeated several times. The decoding stage sequentially passes through a max unpooling layer, a convolutional layer, and a pruning layer. The max unpooling layer and the convolutional layer are repeated several times. The pruning layer is used to remove the filled regions in the encoding layer. The final output tensor is completely consistent with the input tensor in terms of dimensions.

[0093] The difference between the output tensor and the input tensor is that the output tensor prunes the data identified as ground points. The significance of this ground point segmentation step is that in ordinary point cloud modeling, the area to be modeled needs to be manually selected, i.e., ground point regions need to be manually removed. Simply setting ground elevation for simple segmentation can easily lead to incomplete removal due to ground undulations. See also... Figures 7-8 This paper shows defective points caused by unreasonable grid size and defects caused by holes in the original point cloud. Solid points are normal data, while hollow points are defective data, generally due to data loss. By repairing local point cloud holes caused by unreasonable grid size selection, defects in the original point cloud scanning, and other factors, the paper avoids volume omissions caused by local point cloud holes.

[0094] Before using point cloud data to optimize neural network models, they need to be trained with labeled data. The dataset processing and training process is as follows:

[0095] 1) Collect several laser point cloud files after laser scanning of coal yards, and preprocess them in the manner described above to obtain the corresponding input tensors; wherein, the rectangular frame of the coal pile area is randomly generated by the system.

[0096] 2) The ground area in the original laser point cloud file is marked manually, and the defect points in the input tensor are marked. The system then generates the corresponding output tensor.

[0097] 3) Input the processed data into the point cloud data optimization neural network model for training.

[0098] In one possible implementation, the step of traversing each grid cell in the projected rectangular grid and generating a voxel block for the current grid cell based on the feature elevation array of the current grid cell includes: traversing each grid cell in the projected rectangular grid and obtaining the number n of feature elevations in the feature elevation array of the current grid cell; when n is odd, generating (n-1) / 2 hexahedral voxel blocks, where the length and width of each voxel block are equal to the length and width of the current grid cell, and the upper and lower base elevations of the i-th voxel block are respectively... The elevation values ​​are the (2*i-1)th and (2*i)th characteristic elevations in the characteristic elevation array. When i = (n-1) / 2, the bottom elevation of the voxel is 0. When n is even, n / 2 hexahedral voxels are generated, and the length and width of each voxel are equal to the length and width of the current mesh cell. The top and bottom elevations of the i-th voxel are the (2*i-1)th and (2*i)th characteristic elevations in the characteristic elevation array, respectively. When the characteristic elevation array is empty, no voxels are generated.

[0099] For details, see Figures 9 to 11 Based on the output tensor, voxel modeling is performed. The rectangular grid corresponding to the projection of the output tensor onto the xy plane is traversed. The number n of the feature elevation array stored in the output tensor within the current grid cell is determined. Then, voxel blocks are generated based on the number n of the feature elevations.

[0100] Finally, based on the voxel blocks generated in the previous step, the volume of the object within the target area is calculated. The method is as follows: traverse all voxel blocks and calculate their volumes, then add the volumes of all voxel blocks together to obtain the volume of the object within the target area.

[0101] The following are embodiments of the apparatus of the present invention, which can be used to execute embodiments of the method of the present invention. For details not disclosed in the apparatus embodiments, please refer to the embodiments of the method of the present invention.

[0102] In another embodiment of the present invention, a coal pile volume measurement system based on laser point cloud is provided, which can be used to implement the above-mentioned coal pile volume measurement method based on laser point cloud. Specifically, the coal pile volume measurement system based on laser point cloud includes a data acquisition module, a feature processing module, a ground point segmentation module, a voxel block generation module, and a volume calculation module.

[0103] The system comprises the following modules: a data acquisition module for acquiring laser point cloud data of the coal pile area; a feature processing module for spatial interpolation of the laser point cloud data of the coal pile area using a rectangular grid, solving for the feature elevation array of each grid cell, and constructing an input tensor based on the feature elevation array of each grid cell; the input tensor for storing the feature elevation array of each grid cell; a ground point segmentation module for inputting the input tensor into a pre-trained point cloud data to optimize a neural network model, cropping and repairing defective point data in the input tensor to obtain an output tensor; a voxel block generation module for projecting the output tensor plane onto a rectangular grid to obtain a projected rectangular grid, traversing each grid cell in the projected rectangular grid, and generating a voxel block for the current grid cell based on the feature elevation array of the current grid cell; and a volume calculation module for calculating and summing the volumes of the voxel blocks of each grid cell to obtain the volume of the coal pile.

[0104] All relevant content of each step involved in the aforementioned embodiments of the coal pile volume measurement method based on laser point clouds can be referenced to the functional description of the corresponding functional module of the coal pile volume measurement system based on laser point clouds in the embodiments of the present invention, and will not be repeated here.

[0105] The module division in this embodiment of the invention is illustrative and represents only one logical functional division. In actual implementation, other division methods may be used. Furthermore, the functional modules in the various embodiments of the invention can be integrated into a single processor, exist as separate physical entities, or be integrated into a single module. The integrated modules described above can be implemented in hardware or as software functional modules.

[0106] In another embodiment of the present invention, a computer device is provided, comprising a processor and a memory. The memory stores a computer program, which includes program instructions. The processor executes the program instructions stored in the computer storage medium. The processor may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing and control core of the terminal, suitable for implementing one or more instructions, specifically suitable for loading and executing one or more instructions in the computer storage medium to achieve a corresponding method flow or corresponding function. The processor described in this embodiment of the present invention can be used for the operation of a coal pile volume measurement method based on laser point clouds.

[0107] In another embodiment of the present invention, a storage medium is provided, specifically a computer-readable storage medium (Memory), which is a memory device in a computer device used to store programs and data. It is understood that the computer-readable storage medium here can include both the built-in storage medium in the computer device and extended storage media supported by the computer device. The computer-readable storage medium provides storage space that stores the terminal's operating system. Furthermore, the storage space also stores one or more instructions suitable for loading and execution by a processor. These instructions can be one or more computer programs (including program code). It should be noted that the computer-readable storage medium here can be a high-speed RAM memory or a non-volatile memory, such as at least one disk storage device. The processor can load and execute one or more instructions stored in the computer-readable storage medium to implement the corresponding steps of the coal pile volume measurement method based on laser point clouds in the above embodiments.

[0108] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0109] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0110] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0111] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0112] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for measuring the volume of a coal pile based on laser point clouds, characterized in that, include: Acquire laser point cloud data of the coal pile area; A rectangular grid is used to spatially interpolate the laser point cloud data of the coal pile area, and the characteristic elevation array of each grid cell is solved. An input tensor is constructed based on the characteristic elevation array of each grid cell. The input tensor is used to store the characteristic elevation array of each grid cell. The input tensor is used to optimize the neural network model by inputting pre-trained point cloud data. The ground point data in the input tensor is cropped and defective point data is repaired to obtain the output tensor. The output tensor plane is projected onto a rectangular grid to obtain a projected rectangular grid. Each grid cell in the projected rectangular grid is traversed, and a voxel block of the current grid cell is generated based on the characteristic elevation array of the current grid cell. Calculate the volume of each voxel block in the grid cell and add them together to obtain the volume of the coal pile; The spatial interpolation of laser point cloud data of the coal pile area using a rectangular grid includes: A rectangular grid covering the laser point cloud data of the coal pile area is established on the xy plane; wherein the grid cell length of the rectangular grid is 0.2 to 0.8 times the average laser point distance of the laser point cloud data of the coal pile area projected in the plane; Iterate through each grid cell of the rectangular grid, and replace the planar coordinates of the laser point cloud data in the current grid cell with the center coordinates of the current grid cell, while keeping the elevation values ​​unchanged. The process of traversing each grid cell in the projected rectangular grid and generating the voxel block of the current grid cell based on the feature elevation array of the current grid cell includes: Traverse each grid cell in the projected rectangular grid and obtain the number n of characteristic elevations in the characteristic elevation array of the current grid cell; When n is odd, generate (n-1) / 2 hexahedral voxels. The length and width of each voxel are equal to the length and width of the current mesh cell. The top and bottom elevations of the i-th voxel are the 2*i-1 and 2*i-th feature elevations in the feature elevation array, respectively. When i = (n-1) / 2, the bottom elevation of the voxel is 0. When n is even, n / 2 hexahedral voxels are generated. The length and width of each voxel are equal to the length and width of the current mesh cell. The top and bottom elevations of the i-th voxel are the 2*i-1 and 2*i-th feature elevations of the feature elevation array, respectively. No voxel blocks are generated when the feature elevation array is empty.

2. The method for measuring coal pile volume based on laser point clouds according to claim 1, characterized in that, The acquisition of laser point cloud data of the coal pile area includes: Acquire laser point cloud data collected from the coal yard and construct a rectangular bounding box for the coal pile area; The laser point cloud data collected from the coal yard is filtered by using a rectangular frame of the coal pile area to remove the laser point cloud data outside the rectangular frame of the coal pile area, thus obtaining the laser point cloud data of the coal pile area. Before performing spatial interpolation of the laser point cloud data of the coal pile area using a rectangular grid, the method further includes: Traverse all laser points in the laser point cloud data of the coal pile area, calculate the minimum distance between each laser point and its nearest neighbor laser points, and identify the current laser point as an outlier when the distance is ≥1.0 to 2.0 times the average laser point distance; remove all outliers from the laser point cloud data of the coal pile area.

3. The method for measuring coal pile volume based on laser point clouds according to claim 1, characterized in that, The characteristic elevation array for solving each grid cell includes: Traverse each grid cell of the rectangular grid by performing the following steps: S11: Sort the elevation values ​​of all laser point cloud data in the current grid cell from largest to smallest to obtain the elevation array of the current grid cell; S12: If the elevation array of the current grid cell contains only one elevation value, mark that elevation value as the characteristic elevation of the current grid cell and jump to the next grid cell; if the elevation array of the current grid cell is empty, then the characteristic elevation array of the current grid cell is empty and jump to the next grid cell. If there are at least two elevation values ​​in the elevation array of the current grid cell, then the first elevation value in the elevation array of the current grid cell is taken as the feature elevation and S13 is performed. S13: Calculate the numerical difference between any two adjacent elevation values ​​in the elevation array of the current grid cell to obtain the first numerical difference. If the first numerical difference is greater than or equal to the first set threshold, mark both elevation values ​​as characteristic elevations. S14: If the last elevation value of the current grid cell is not marked as a feature elevation, calculate the numerical difference between the elevation value and all feature elevations to obtain several second numerical differences. If all second numerical differences are ≥ a second set threshold, mark the last elevation value as a feature elevation. S15: Sort all feature elevations of the current grid cell from largest to smallest to obtain the feature elevation array of the current grid cell.

4. The method for measuring coal pile volume based on laser point clouds according to claim 1, characterized in that, The process of constructing the input tensor based on the feature elevation array of each grid cell includes: Obtain the maximum value N of the number of feature elevations in the feature elevation array of all grid cells, and create an input tensor of size W×H×N. The input tensor stores n feature elevations from the feature elevation array of each grid cell in the rectangular grid, and fills the empty positions with empty elements. Here, W is the number of grid cells on the long side of the rectangular grid, and H is the number of grid cells on the short side of the rectangular grid.

5. The method for measuring coal pile volume based on laser point clouds according to claim 1, characterized in that, The point cloud data optimized neural network model includes an encoding layer and a decoding layer connected in sequence; The encoding layer consists of a padding layer, a convolutional layer, and a max pooling layer connected in sequence. The padding layer is used to pad the input tensor when the number of grid cells on the long side of the rectangular grid is not equal to the number of grid cells on the short side of the rectangular grid, so that the number of grid cells on the short side of the rectangular grid is equal to the number of grid cells on the short side of the rectangular grid. The padding region is zero-padding. The decoding layer consists of a max-unpooling layer, a convolutional layer, and a pruning layer connected in sequence; the pruning layer is used to remove the padding regions in the encoding layer.

6. A coal pile volume measurement system based on laser point clouds, characterized in that, include: The data acquisition module is used to acquire laser point cloud data of the coal pile area; The feature processing module is used to perform spatial interpolation on the laser point cloud data of the coal pile area using a rectangular grid, solve for the feature elevation array of each grid cell, and construct an input tensor based on the feature elevation array of each grid cell; wherein, the input tensor is used to store the feature elevation array of each grid cell. The ground point segmentation module is used to input the input tensor into the pre-trained point cloud data to optimize the neural network model, and to crop and repair the ground point data in the input tensor to obtain the output tensor. The voxel block generation module is used to project the output tensor plane onto a rectangular mesh to obtain a projected rectangular mesh, and traverse each mesh cell in the projected rectangular mesh to generate a voxel block for the current mesh cell based on the characteristic elevation array of the current mesh cell. The volume calculation module is used to calculate the volume of the voxel blocks of each grid cell and add them together to obtain the volume of the coal pile. The spatial interpolation of laser point cloud data of the coal pile area using a rectangular grid includes: A rectangular grid covering the laser point cloud data of the coal pile area is established on the xy plane; wherein the grid cell length of the rectangular grid is 0.2 to 0.8 times the average laser point distance of the laser point cloud data of the coal pile area projected in the plane; Iterate through each grid cell of the rectangular grid, and replace the planar coordinates of the laser point cloud data in the current grid cell with the center coordinates of the current grid cell, while keeping the elevation values ​​unchanged. The process of traversing each grid cell in the projected rectangular grid and generating the voxel block of the current grid cell based on the feature elevation array of the current grid cell includes: Traverse each grid cell in the projected rectangular grid and obtain the number n of characteristic elevations in the characteristic elevation array of the current grid cell; When n is odd, generate (n-1) / 2 hexahedral voxels. The length and width of each voxel are equal to the length and width of the current mesh cell. The top and bottom elevations of the i-th voxel are the 2*i-1 and 2*i-th feature elevations in the feature elevation array, respectively. When i = (n-1) / 2, the bottom elevation of the voxel is 0. When n is even, n / 2 hexahedral voxels are generated. The length and width of each voxel are equal to the length and width of the current mesh cell. The top and bottom elevations of the i-th voxel are the 2*i-1 and 2*i-th feature elevations of the feature elevation array, respectively. No voxel blocks are generated when the feature elevation array is empty.

7. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the coal pile volume measurement method based on laser point cloud as described in any one of claims 1 to 5.

8. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the coal pile volume measurement method based on laser point cloud as described in any one of claims 1 to 5.

Images