Point cloud tree biomass estimation method and related equipment

Abstract

The invention discloses a point cloud tree biomass estimation method, which comprises the following steps of: firstly, clustering different branches in preprocessed point cloud data, and then, carrying out fine-grained voxelization on the point cloud of each branch to generate a dense voxel grid; the outer boundary of each branch is approximated by using a convex hull, and the number of points located in the convex hull is obtained based on the convex hull and a voxel network, that is, whether each voxel is filled by point cloud is determined by occupying a modeling technology, so that the space in the branch is filled. And finally, calculating the volume of each tree according to the total amount of the occupied voxels, and converting the volume into biomass in combination with parameters such as wood density and the like, thereby realizing accurate estimation of the aboveground biomass of the tree. According to the method, the complex three-dimensional structure of the tree can be captured more accurately, and estimation errors caused by structure simplification are reduced.

Description

Technical Field

[0001] The invention belongs to the field of tree biomass calculation, and specifically relates to a point cloud tree biomass estimation method and related equipment. Background Art

[0002] Aboveground biomass (AGB) refers to the total amount of organic matter contained in trees and their appendages above the ground, including trunks, branches, leaves, flowers and fruits, but excluding roots and underground parts. Aboveground biomass of trees is an important indicator for measuring the carbon cycle, carbon storage and forest productivity of forest ecosystems. Accurately estimating tree biomass is not only helpful in assessing the carbon sequestration capacity of forests, but also has important significance for forest management, ecological protection and climate change research.

[0003] The traditional direct measurement method of biomass is to cut down and weigh all trees in a designated area, collect and dry the trunks, branches, leaves, flowers and fruits of each tree to obtain the biomass value of each tree. Then, by adding up the biomass values ​​of each tree, the total biomass of the entire survey area is obtained. This method is costly, time-consuming, and difficult to apply to large-scale forest biomass estimation.

[0004] The existing indirect biomass estimation method is to select a specific number of sample plots within the survey area, investigate variables such as tree height and diameter at breast height (DBH), and then measure each tree within the selected sample plots. The biomass is calculated using the allometric equation, and the total biomass of the survey area is extrapolated from the biomass estimated from the sample plots. The basic form of the allometric equation is:

[0005] Among them, B is biomass, D is DBH, H is tree height, and a, b, c are fitted regression coefficients. This method reduces the workload of field surveys and allows large-scale estimation of forest biomass. However, this method is usually based on data from specific regions or tree species, and its accuracy decreases when applied across regions or tree species. In addition, the measurement of tree height and DBH itself may have certain errors, especially in dense forest stands or complex terrain. The limitations of measurement tools and methods will further affect the accuracy of prediction factors.

[0006] Currently, the SimpleTree QSM method does not rely on single predictors such as tree height and diameter at breast height, but instead achieves more accurate biomass estimation by capturing the three-dimensional structural information of trees. This method uses high-density point cloud data obtained by laser scanning (LiDAR) to build tree geometric models, uses these geometric models to calculate the volume of trees, and converts them into biomass in combination with parameters such as wood density. However, the SimpleTree QSM method relies on simplified tree geometric models, which may ignore subtle changes in the crown and the specific distribution of branches and trunks, affecting the accuracy of the overall biomass estimation. Summary of the invention

[0007] The present invention provides a point cloud tree biomass estimation method and related equipment, which solves the problem that the existing biomass measurement method affects the accuracy of overall biomass estimation.

[0008] To achieve the above object, the present invention provides the following technical solutions: A point cloud tree biomass estimation method, comprising: De-noising the tree point cloud data, and down-sampling the de-noised point cloud data using a voxel grid to obtain down-sampled point cloud data; Perform branch clustering on the downsampled point cloud data to obtain branch clustering results; voxelize the point cloud of each branch according to the branch clustering result to obtain the voxel grid of each branch; Use the convex hull to approximate the outer boundary of each branch and obtain the number of voxel grids inside the convex hull of each branch; Tree aboveground biomass data were obtained based on the total volume of each tree, which was approximated based on the product of the individual voxel size and the number of voxels located within all convex hulls.

[0009] Preferably, the steps of performing denoising on the tree point cloud data and downsampling the denoised point cloud data using a voxel grid are specifically as follows: First, the point cloud to be processed is treated with Matlab's point cloud denoising function pcdenoise to remove noise points. Then, the voxel grid to which each point cloud belongs is calculated. For each voxel, the point closest to the voxel center is selected as the representative point to generate the downsampled point cloud.

[0010] Preferably, the specific method for computing the voxel grid to which each point cloud belongs is: First, the side length of the voxel grid is given. Each voxel represents a cube in three-dimensional space. According to the coordinates of the point and the size of the voxel, the three-dimensional index of the voxel to which each point belongs is determined, and the center coordinates of each voxel are calculated. Then, by calculating the distance from each point to the center of the voxel, the point closest to the center of each voxel is selected as the representative point. For all the generated voxels, only the representative point is retained to generate the downsampled point cloud.

[0011] Preferably, the step of performing branch clustering on the downsampled point cloud data to obtain the branch clustering result is specifically as follows: First, the k-nearest neighbor algorithm is used to establish edge connections for each point, and then the point cloud is Delaunay triangulated to generate a set of non-overlapping triangular meshes. According to the edges of the triangles, connections between points are established to supplement the topological relationships not covered in the k-nearest neighbor algorithm connection. The K-nearest neighbor connection and the Delaunay triangulation connection are merged to form a fully connected graph containing all important spatial relationships. The lowest point is selected from the point cloud as the root node, and the shortest path from each point to the root node in the fully connected graph is calculated from the root node. Then, the farthest leaf node that each node can reach is calculated through the backtracking algorithm. For each node, only those neighbors whose shortest path length to the root node is less than or equal to the current node are retained. The frequency of each node is corrected through multiple iterations, and eventually each node will point to a stable farthest leaf node. Finally, a label is assigned to each node to represent its farthest leaf node. Different labels represent different clusters, and each cluster represents a branch of a tree.

[0012] Preferably, the specific step of edge connection is to use the knnsearch function to find the k neighbors of each point, each point will be connected with its k nearest neighbors to form an edge, and two pruning conditions are used to filter the edges to remove unimportant edges in the graph, including: Add the median and standard deviation of the neighbor distance of each point to get a threshold. If the distance of an edge exceeds this threshold, it means that the edge is connected to a relatively distant neighbor, which is a noise or unimportant edge, and it will be pruned; According to the maximum distance between the neighbors of each point, the edges whose maximum distance is above the 90th percentile are selected for pruning to obtain the initial graph object.

[0013] Preferably, the specific steps of performing Delaunay triangulation on the point cloud are: using the delaunayTriangulation function provided by matlab, generating a Delaunay triangulation of the point cloud, generating a set of triangles, the edge of each triangle constitutes an edge connecting points, extracting all the edges of the triangulation, checking which Delaunay edges already exist in the k-nearest neighbor graph, removing duplicate edges, and then adding the Delaunay edges to the k-nearest neighbor graph, calculating the Euclidean distance of each edge, using it as the weight of the edge, and obtaining a weighted graph combining k-nearest neighbor and Delaunay triangulation.

[0014] Preferably, the outer boundary of each branch is approximated using the convex hull, and the steps of obtaining the number of points inside the convex hull based on the convex hull and the voxel network are specifically as follows: Use the convhull function in Matlab to calculate the convex hull of each branch point cloud. The convex hull is approximated to the outer boundary of the branch shape. Calculate the normal vector of the convex hull surface and normalize the obtained normal vector. If the dot product of the line connecting the point and a vertex on the convex hull surface and the normal vector of the convex hull surface is greater than or equal to zero, it means that the point is on the same side of the surface, or the point is on the surface. Perform such checks on all surfaces. If the point meets the conditions for all surfaces, the point is inside the convex hull; otherwise, the point is outside the convex hull.

[0015] A point cloud tree biomass estimation system, comprising: Preprocessing module: used to perform noise reduction on tree point cloud data, and downsample the noise-reduced point cloud data using voxel grids to obtain downsampled point cloud data; Branch clustering module: used to perform branch clustering on the downsampled point cloud data to obtain branch clustering results; Voxelization module: used to voxelize the point cloud of each branch according to the branch clustering results to obtain a voxel grid; Computation module: used to approximate the outer boundary of each branch using the convex hull and obtain the number of voxel grids inside the convex hull of each branch; Estimation module: used to approximate the total volume of each tree based on the product of the size of a single voxel and the number of voxels located within all convex hulls, and obtain the tree aboveground biomass data based on the total volume of each tree.

[0016] A computer device comprises a memory, a processor and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of a point cloud tree biomass estimation method are implemented.

[0017] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of a point cloud tree biomass estimation method are implemented.

[0018] Compared with the prior art, the present invention has the following beneficial effects: The present invention provides a point cloud tree biomass estimation method, which first clusters different branches in the preprocessed point cloud data, and then fine-grained voxelizes the point cloud of each branch to generate a dense voxel grid. By using the convex hull to approximate the outer boundary of each branch, the number of points inside the convex hull is obtained based on the convex hull and voxel network, that is, the occupancy modeling technology, to determine whether each voxel is filled by the point cloud, thereby filling the space inside the branch. Finally, the volume of each tree is calculated based on the total amount of occupied voxels, and the volume is converted into biomass in combination with parameters such as wood density, thereby achieving an accurate estimation of the aboveground biomass of trees. This method can more accurately capture the complex three-dimensional structure of trees and reduce estimation errors caused by structural simplification. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 A flow chart of a point cloud tree biomass estimation method of the present invention; Figure 2 This is a diagram of the branch clustering part of the present invention; Figure 3 In the embodiment of the present invention, the tree point cloud branch clustering result is visualized, and different colors represent different branches.

[0020] Figure 4 The present invention occupies part of the modeling architecture diagram; Figure 5 This is a block diagram of a point cloud tree biomass estimation system of the present invention. DETAILED DESCRIPTION

[0021] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0022] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0023] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0024] In order to enable those skilled in the art to better understand the technical solution of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings.

[0025] like Figure 1 As shown, the present invention provides a point cloud tree biomass estimation method, comprising: De-noising the tree point cloud data, and down-sampling the de-noised point cloud data using a voxel grid to obtain down-sampled point cloud data; Perform branch clustering on the downsampled point cloud data to obtain branch clustering results; The point cloud of each branch is voxelized according to the branch clustering results to obtain a voxel grid; Use the convex hull to approximate the outer boundary of each branch and obtain the number of voxel grids inside the convex hull of each branch; Tree aboveground biomass data were obtained based on the total volume of each tree, which was approximated based on the product of the individual voxel size and the number of voxels located within all convex hulls.

[0026] The detailed steps are: S1: De-noising the tree point cloud data, and down-sampling the de-noised point cloud data using a voxel grid to obtain down-sampled point cloud data.

[0027] Specifically, we first use the point cloud denoising function pcdenoise in Matlab to remove noise points from the point cloud, and then calculate the voxel grid for each point cloud. For each voxel, we select the point closest to the voxel center as the representative point to generate the downsampled point cloud, which can reduce the number of points while retaining the overall topological structure of the point cloud.

[0028] The specific method for computing the voxel grid to which each point cloud belongs is as follows: first, the side length of the voxel grid is given as 0.06, and each voxel represents a cube in three-dimensional space. According to the coordinates of the point and the size of the voxel, the three-dimensional index of the voxel to which each point belongs is determined, the center coordinates of each voxel are calculated, and then the distance from each point to the center of the voxel is calculated, and the point closest to the center of each voxel is selected as the representative point. Only the representative point is retained for all the generated voxels, so that a downsampled point cloud is generated, which can reduce the number of points while retaining the overall topological structure of the point cloud.

[0029] S2: See Figure 2, perform branch clustering on the downsampled point cloud data to obtain branch clustering results, ensuring that different point clouds can be distinguished to which branch of the tree they belong.

[0030] Specifically, first use the k-nearest neighbor algorithm to establish edge connections for each point. Then perform Delaunay triangulation on the point cloud to generate a set of non-overlapping triangular meshes. According to the edges of the triangles, connections between points are established to supplement the topological relationships not covered by the k-nearest neighbor algorithm connection. K-nearest neighbor connections and Delaunay triangulation connections are merged to form a fully connected graph containing all important spatial relationships. Next, the lowest point is selected from the point cloud as the root node, and the shortest path from each point in the fully connected graph to the root node is calculated from the root node. This step can accurately reflect the growth direction and branch topology of the tree. Then, the farthest leaf node that each node can reach is calculated by the backtracking algorithm. For each node, only those neighbors whose shortest path length to the root node is less than or equal to the current node are retained. The frequency of each node is corrected through multiple iterations, and eventually each node will point to a stable farthest leaf node. Finally, a label is assigned to each node to represent its farthest leaf node. Different labels represent different clusters, and each cluster represents a branch of the tree. See Figure 3 In one embodiment of the present invention, the tree point cloud branch clustering results are visualized, and different colors represent different branches.

[0031] The specific steps of edge connection are to find the k neighbors of each point using the knnsearch function. knnsearch is a MATLAB function used to perform K nearest neighbor search, and k is specified as 10. Each point will be connected to its k nearest neighbors to form an edge. In order to remove unimportant edges in the graph, two pruning conditions are used to filter the edges: 1. The median and standard deviation of the neighbor distance of each point are added to obtain a threshold. If the distance of an edge exceeds this threshold, it means that the edge is connected to a relatively distant neighbor, which may be a noise or an unimportant edge, so it can be considered to be pruned (deleted); 2. According to the maximum distance between the neighbors of each point, select those edges whose maximum distance is above the 90th percentile for pruning. In this way, the initial graph object is obtained.

[0032] The specific steps of performing Delaunay triangulation on the point cloud to obtain the edges of the graph are to use the delaunayTriangulation function provided by matlab to generate a Delaunay triangulation of the point cloud. This will produce a set of triangles, and the edges of each triangle constitute an edge connecting points. All the edges of the triangulation are extracted, and the Delaunay edges that already exist in the above-mentioned k-nearest neighbor graph are checked, and these duplicate edges are removed. Then the Delaunay edges are added to the K-nearest neighbor graph, and the Euclidean distance of each edge is calculated and used as the weight of the edge. Finally, a weighted graph combining K-nearest neighbor and Delaunay triangulation is returned.

[0033] S3: voxelize the point cloud of each branch according to the branch clustering results to obtain a voxel grid.

[0034] Specifically, for each branch point cloud, its minimum and maximum ranges are calculated, and then a three-dimensional grid within the range is created using the meshgrid function in MATLAB. The size of each grid is defined as 0.01. S4: See Figure 4 , use the convex hull to approximate the outer boundary of each branch, and get the number of voxel grids inside the convex hull of each branch.

[0035] Specifically, the convhull function in Matlab is used to calculate the convex hull of each branch point cloud. The convex hull approximates the outer boundary of the branch shape. The normal vector of the convex hull surface (simplex) is calculated and the normal vector is normalized. If the dot product of the line connecting the point and a vertex on the convex hull surface and the normal vector of the convex hull surface is greater than or equal to zero, it means that the point is on the same side of the surface, or the point is on the surface. This kind of check needs to be performed on all surfaces. If the point meets the conditions for all surfaces, the point is inside the convex hull; otherwise, the point is outside the convex hull.

[0036] S5: Approximate the total volume of each tree based on the product of the size of a single voxel and the number of voxels located within all convex hulls. Combining the volume of each tree (Volume) and the wood density (WD), the volume is converted to biomass (Biomass). The calculation formula is as follows:

[0037] In summary, the present invention relates to a tree biomass estimation technology, which clusters point clouds belonging to different branches, then performs fine-grained voxelization on the point clouds of each branch, fills the space inside each branch with a dense voxel grid, and estimates the volume of the tree based on the occupied voxel volume. This avoids the problem of ignoring subtle changes in the crown and the specific distribution of branches caused by tree geometric modeling.

[0038] By clustering the point clouds belonging to different branches and performing occupancy modeling for each cluster, it is possible to estimate tree biomass directly from the tree point cloud. This can minimize estimation deviations and avoid the problems of traditional destructive felling that leads to high costs and time, and overly simplified geometric modeling that leads to low biomass accuracy.

[0039] In one embodiment of the present invention, the biomass of 61 trees was estimated by point cloud data and compared with reference data. The evaluation indicators include bias, mean absolute percentage error (MAPE), root mean square error (RMSE) and relative root mean square error (Relative RMSE).

[0040] Among them, see Table 1, the estimation results of the patent method and the traditional tree geometric modeling QSM method on the point cloud data of 61 trees are compared. The deviation of the patent method is -0.0081, which is close to zero, indicating that there is almost no systematic error between the predicted value and the true value; the deviation of the QSM method is 0.0964, which is a certain overestimation. The average absolute percentage error of the patent method is 0.2342, indicating that its prediction is more accurate in terms of relative error; the average absolute percentage error of the QSM method is 0.2400, and the relative error of the prediction is slightly larger. The RMSE of the patent method is small (3.5471), indicating that its prediction error is relatively small; while the RMSE of the QSM method is large (5.8524), indicating that its prediction error is large and the performance is poor. The Relative RMSE of the patent method is low (0.5199), indicating that its prediction error is small relative to the true value and the performance is good; the Relative RMSE of the QSM method is high (0.8578), which means that its prediction relative error is large and the performance is poor.

[0041] In summary, this patented method outperforms the QSM method in all indicators. It has smaller deviation, smaller relative error, smaller root mean square error, and is more accurate and reliable overall.

[0042] Table 1 Comparison of results

[0043] like Figure 5 As shown, the present invention also provides a point cloud tree biomass estimation system, comprising: Preprocessing module: used to perform noise reduction on tree point cloud data, and downsample the noise-reduced point cloud data using voxel grids to obtain downsampled point cloud data; Branch clustering module: used to perform branch clustering on the downsampled point cloud data to obtain branch clustering results; Voxelization module: used to voxelize the point cloud of each branch according to the branch clustering results to obtain a voxel grid; Computation module: used to approximate the outer boundary of each branch using the convex hull and obtain the number of voxel grids inside the convex hull of each branch; Estimation module: used to approximate the total volume of each tree based on the product of the size of a single voxel and the number of voxels located within all convex hulls, and obtain the tree aboveground biomass data based on the total volume of each tree.

[0044] A terminal device is provided in one embodiment of the present invention. The terminal device of this embodiment includes: a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps in the above-mentioned method embodiments are implemented. Alternatively, when the processor executes the computer program, the functions of the modules / units in the above-mentioned device embodiments are implemented.

[0045] The computer program may be divided into one or more modules / units, and the one or more modules / units are stored in the memory and executed by the processor to accomplish the present invention.

[0046] The terminal device may be a computing device such as a desktop computer, a notebook, a PDA, a cloud server, etc. The terminal device may include, but is not limited to, a processor and a memory.

[0047] The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.

[0048] The memory may be used to store the computer program and / or module, and the processor implements various functions of the terminal device by running or executing the computer program and / or module stored in the memory and calling the data stored in the memory.

[0049] If the module / unit integrated in the terminal device is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by the processor, the steps of the above-mentioned various method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.

[0050] Although the embodiments of the present invention are described above in conjunction with the accompanying drawings, the present invention is not limited to the above-mentioned specific embodiments and application fields, and the above-mentioned specific embodiments are only illustrative and instructive, rather than restrictive. Under the guidance of the specification, a person skilled in the art can also make many forms without departing from the scope of protection of the claims of the present invention, all of which belong to the scope of protection of the present invention.

Claims

1. A point cloud tree biomass estimation method, characterized in that: include: De-noising the tree point cloud data, and down-sampling the de-noised point cloud data using a voxel grid to obtain down-sampled point cloud data; Perform branch clustering on the downsampled point cloud data to obtain branch clustering results; voxelize the point cloud of each branch according to the branch clustering result to obtain the voxel grid of each branch; Use the convex hull to approximate the outer boundary of each branch and obtain the number of voxel grids inside the convex hull of each branch; Tree aboveground biomass data were obtained based on the total volume of each tree, which was approximated based on the product of the individual voxel size and the number of voxels located within all convex hulls.

2. The point cloud tree biomass estimation method according to claim 1, characterized in that: The specific steps of denoising the tree point cloud data and downsampling the denoised point cloud data using a voxel grid are as follows: First, the point cloud to be processed is treated with Matlab's point cloud denoising function pcdenoise to remove noise points. Then, the voxel grid to which each point cloud belongs is calculated. For each voxel, the point closest to the voxel center is selected as the representative point to generate the downsampled point cloud.

3. The point cloud tree biomass estimation method according to claim 2, characterized in that: The specific method for computing the voxel grid to which each point cloud belongs is: First, the side length of the voxel grid is given. Each voxel represents a cube in three-dimensional space. According to the coordinates of the point and the size of the voxel, the three-dimensional index of the voxel to which each point belongs is determined, and the center coordinates of each voxel are calculated. Then, by calculating the distance from each point to the center of the voxel, the point closest to the center of each voxel is selected as the representative point. For all the generated voxels, only the representative point is retained to generate the downsampled point cloud.

4. The point cloud tree biomass estimation method according to claim 1, characterized in that: The steps of performing branch clustering on the downsampled point cloud data and obtaining the branch clustering results are as follows: First, the k-nearest neighbor algorithm is used to establish edge connections for each point, and then the point cloud is Delaunay triangulated to generate a set of non-overlapping triangular meshes. According to the edges of the triangles, connections between points are established to supplement the topological relationships not covered in the k-nearest neighbor algorithm connection. The K-nearest neighbor connection and the Delaunay triangulation connection are merged to form a fully connected graph containing all important spatial relationships. The lowest point is selected from the point cloud as the root node, and the shortest path from each point to the root node in the fully connected graph is calculated from the root node. Then, the farthest leaf node that each node can reach is calculated through the backtracking algorithm. For each node, only those neighbors whose shortest path length to the root node is less than or equal to the current node are retained. The frequency of each node is corrected through multiple iterations, and eventually each node will point to a stable farthest leaf node. Finally, a label is assigned to each node to represent its farthest leaf node. Different labels represent different clusters, and each cluster represents a branch of a tree.

5. The point cloud tree biomass estimation method according to claim 4, characterized in that: The specific steps of edge connection are to use the knnsearch function to find the k neighbors of each point, each point will be connected to its k nearest neighbors to form an edge, and two pruning conditions are used to filter the edges and remove unimportant edges in the graph, including: Add the median and standard deviation of the neighbor distance of each point to get a threshold. If the distance of an edge exceeds this threshold, it means that the edge is connected to a relatively distant neighbor, which is a noise or unimportant edge, and it will be pruned; According to the maximum distance between the neighbors of each point, the edges whose maximum distance is above the 90th percentile are selected for pruning to obtain the initial graph object.

6. The point cloud tree biomass estimation method according to claim 4, characterized in that: The specific steps of Delaunay triangulation of point cloud are: use the delaunayTriangulation function provided by matlab to generate a Delaunay triangulation of point cloud, produce a set of triangles, the edge of each triangle constitutes an edge connecting points, extract all the edges of triangulation, check which Delaunay edges already exist in the k nearest neighbor graph, remove duplicate edges, and then add Delaunay edges to the k nearest neighbor graph, calculate the Euclidean distance of each edge, use it as the weight of the edge, and obtain a weighted graph combining k nearest neighbor and Delaunay triangulation.

7. The point cloud tree biomass estimation method according to claim 1, characterized in that: The convex hull is used to approximate the outer boundary of each branch. The steps to obtain the number of points inside the convex hull based on the convex hull and voxel network are as follows: Use the convhull function in Matlab to calculate the convex hull of each branch point cloud. The convex hull is approximated to the outer boundary of the branch shape. Calculate the normal vector of the convex hull surface and normalize the obtained normal vector. If the dot product of the line connecting the point and a vertex on the convex hull surface and the normal vector of the convex hull surface is greater than or equal to zero, it means that the point is on the same side of the surface, or the point is on the surface. Perform such checks on all surfaces. If the point meets the conditions for all surfaces, the point is inside the convex hull; otherwise, the point is outside the convex hull.

8. A point cloud tree biomass estimation system, characterized in that: include: Preprocessing module: used to perform noise reduction on tree point cloud data, and downsample the noise-reduced point cloud data using voxel grids to obtain downsampled point cloud data; Branch clustering module: used to perform branch clustering on the downsampled point cloud data to obtain branch clustering results; Voxelization module: used to voxelize the point cloud of each branch according to the branch clustering results to obtain a voxel grid; Computation module: used to approximate the outer boundary of each branch using the convex hull and obtain the number of voxel grids inside the convex hull of each branch; Estimation module: used to approximate the total volume of each tree based on the product of the size of a single voxel and the number of voxels located within all convex hulls, and obtain the tree aboveground biomass data based on the total volume of each tree.

9. 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, the steps of the point cloud tree biomass estimation method as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the point cloud tree biomass estimation method according to any one of claims 1 to 7 are implemented.

Images