A single-tree volume non-destructive measurement method based on ground-based laser radar

Abstract

This invention provides a non-destructive measurement method for single-tree volume based on ground-based lidar, belonging to the field of single-tree volume measurement. It focuses on extracting structural features of tree trunks with different morphologies and calculating trunk volume. The method uses Masson pine as the research object, with 52 individual trees as samples. Real values ​​of different heights and diameters were collected and the single-tree volume was calculated. Simultaneously, a ground-based lidar system was used to scan a high-density three-dimensional lidar point cloud. First, the original point cloud data was sampled and processed. The growth of seed points was constrained by normal vectors and angles to accurately extract the tree trunk point cloud. Then, point cloud slicing was used to simulate the cutting of a real disk, and an improved convex hull algorithm was used to calculate the area of ​​a single layer of point cloud. In occluded areas, a fitted circle was constructed using local point cloud data to calculate the area of ​​that layer. Finally, the single-tree volume was obtained using the method of infinitesimal elements. The correlation between the calculated volume and the actual calculated volume is R. 2 The value reached 0.995, and the RMSE was 0.017m. 3 The nRMSE is 2.38%, indicating that the trunk can be accurately extracted and the volume of a single piece of wood can be calculated without contact.

Description

Technical Field

[0001] This invention relates to the field of tree trunk algorithms, and in particular to a non-destructive method for measuring the volume of a single log based on ground-based lidar. Background Technology

[0002] With the increasing demand for refined management of forest resources, the accurate and efficient extraction of individual tree trunk information has become a key link in forestry digitalization, especially in areas dominated by economic forests. The accurate extraction of trunk parameters (such as diameter at breast height (DBH) and volume) is crucial for timber quality assessment, biomass estimation, logging strategy formulation, and forest health monitoring. Traditionally, obtaining trunk parameters has relied primarily on manual measurement, a highly labor-intensive and inefficient process. Some key structural attributes, such as the variation of trunk diameter along height, trunk curvature, and cross-sectional profile, often require segmental measurement or destructive sampling methods, which are not only time-consuming and labor-intensive but also cause irreversible damage to the plot structure. In tall trees or dense canopies, issues such as partial trunk obstruction and poor accessibility also limit the feasibility and accuracy of measurements.

[0003] LiDAR technology, capable of acquiring high-precision point clouds using echoes for 3D reconstruction of target objects, is widely used in tree parameter extraction. Ground-based lidar systems (TLS), due to their high spatial resolution, are increasingly applied to forest trunk modeling and parameter estimation. However, the complex forest understory environment, the irregular spatial distribution of trees, the overlapping and occlusion of trunks, and noise in point cloud data all pose significant challenges to the automatic extraction of individual tree trunks. Furthermore, the lack of more refined trunk surface reconstruction methods and the inability to accurately calculate the volume of individual trees using cylindrical fitting of trunk growth methods are also issues.

[0004] This invention provides a non-destructive measurement method for single-tree volume based on ground-based lidar, focusing on extracting and calculating the volume of tree trunks from structural features of different morphologies. Using Masson pine as the research object and 52 individual trees as samples, the method collects the true values ​​of diameters at different heights and calculates the volume of each individual tree. Simultaneously, a ground-based lidar system scans a high-density three-dimensional lidar point cloud. First, the original point cloud data is sampled and processed, and the growth of seed points is constrained by normal vectors and angles to accurately extract the tree trunk point cloud. Next, point cloud slicing is used to simulate the cutting of a real disk, and an improved convex hull algorithm is used to calculate the area of ​​a single layer of point cloud. In occluded areas, a fitted circle is constructed using local point cloud data to calculate the area of ​​that layer. Finally, the volume of each individual tree is obtained using the method of infinitesimal elements. Summary of the Invention

[0005] This patent provides a non-destructive measurement method for the volume of a single tree trunk based on ground-based lidar. Building upon the two-dimensional convex hull algorithm, this patent proposes a double-constraint convex hull algorithm, optimizing the convex hull boundary to ensure a close fit between the boundary and the tree trunk point cloud. Simultaneously, a distance threshold is added, improving computational efficiency while maintaining accuracy, thereby enhancing the model's accurate description of the tree trunk shape.

[0006] The present invention provides a non-destructive measurement method for the volume of a single timber based on ground-based lidar, comprising the following steps: S101: Data Acquisition The main tree species studied was Masson pine. The diameter at breast height (DBH), tree height, diameter at different relative positions, height below the branch, and crown width of 153 Masson pine trees were collected and measured. The bark volume of each sample tree was calculated based on the diameter at different relative positions. At the same time, the FARO Focus3D terrestrial 3D laser scanner was used to scan the Masson pine trees to obtain high-precision 3D point clouds. The point clouds of multiple sites were stitched together based on a common target sphere. S102: Data Preprocessing The average density of the point cloud data collected by the ground-based lidar is as high as 5000 p / m 2 To avoid the reduction in computational efficiency caused by the large amount of data from ground-based lidar, voxel downsampling is used to process the original ground-based lidar. This method calculates the centroid of each voxel unit and uses the coordinates of the centroid to replace all points in the unit, thus completing point cloud downsampling while ensuring that the distribution characteristics of the point cloud remain unchanged. Based on the distribution of points, this patent sets the voxel side length r to 0.1m; Through morphological comparison and screening, 52 trees with typical different morphologies were finally selected as experimental subjects. S103: Tree Trunk Point Cloud Extraction To achieve non-destructive measurement of the volume of individual trees, it is still necessary to accurately identify the point cloud of the individual tree trunk from the scattered and disordered laser point cloud. This patent proposes a trunk extraction algorithm based on normal vector constraints. The specific steps of the algorithm are as follows: Step 1: Initialize the complete point set Points, seed point set SeedPoints, adjacent point set AdjacentPoints, trunk point set TrunkPoints, and temporary point set TempPoints; Step 2: Read the preprocessed point cloud data of a single tree and establish a normal vector for the point cloud data; for each scan point p in the point cloud, search for its N nearest neighbor points, and then calculate the local plane P of these points in the least squares sense. This plane can be represented as: (1) In the formula, Let d be the normal vector of plane P, and d be the distance from P to the origin. The normal vector of the plane fitted by the N nearest neighbors can be considered as the normal vector of the current scan point; the normal vector of plane P can be obtained by principal component analysis (PCA), and the calculation shows that P passes through the centroid of its N neighborhood points. And normal vector satisfy =1, first perform eigenvalue decomposition on the covariance matrix M in equation (2) to obtain each eigenvalue of M, and the eigenvector corresponding to the smallest eigenvalue of M is the normal vector of P. (2) The point cloud normal vector represents the direction of the plane formed by its neighboring points. This patent establishes a kd-tree for the entire point cloud and uses a search radius of 1m to calculate the point cloud normal vector using the 10 nearest neighbor points. Step 3: Based on the establishment of the normal vector, initialize the Points point set, assign a normal vector attribute value to each point, and set the category of all points to unclassified; Step 4: When hour, It can be by and The dot product is obtained, This indicates the degree of perpendicularity of the point cloud normal vector at that location. When the value approaches 0, it indicates that the normal vector is almost perpendicular to the Z-axis; therefore, we choose... Regions with a value greater than 0.05 and a point cloud height greater than 0.05m and less than 0.2m are used as initial seed points for subsequent growth, and points that meet the conditions are added to SeedPoints; Step 5: To further improve algorithm efficiency and reduce the amount of data traversed in loops, the algorithm adopts a segmented growth method, adding the seed point of the previous segment as the ignition seed point of the new segment. In this patent, a 0.3m interval and a 0.3m layer thickness are used to layer the single tree point cloud, and the layered point cloud is added to the TempPoints collection. The pseudocode for point cloud slicing and layering is as follows: Step Six: For each temporary point in each layer, add each seed point in a loop, and calculate the 10 nearest neighbors of the seed point by Euclidean distance. Determine whether the angle between each point and the normal vector of the seed point is less than the angle threshold. Change the category of points that meet the conditions to trunk points, and also compare them with seed points. The difference is less than The point is recorded as the new seed point; Step 7: Loop through each slice of point cloud and gradually add the points that meet the conditions to the trunk point cloud set to generate the final trunk point cloud; S104: Double Constraint Convex Hull Algorithm Based on the Graham convex hull algorithm, a dual-constraint convex hull algorithm is proposed. This algorithm mainly improves upon the Graham convex hull algorithm. It slices the tree trunk point cloud in the same way as the slicing algorithm in S103 and inputs them one by one into the optimization algorithm to complete the accurate extraction of boundary points and calculate the area of ​​a single layer. The algorithm mainly uses the angle and CHL convex hull side length threshold to gradually approximate the real situation of the point cloud distribution, making the convex hull more closely fit the tree trunk surface. Then, the discretized Green's formula is used to calculate the area of ​​each layer of convex hull. Sort the generated convex hulls counterclockwise. At this point, for each convex set... vertex The area calculation formula is (Equation (3)): (3) In the formula, This is the convex hull area calculated using the optimized convex hull algorithm. The coordinates of the convex hull vertices arranged in counter-clockwise order This represents the number of points in the convex hull set; typically, the initial point is repeatedly added to the end of the point set to close the polygon. S105: Hough Circle Fitting Algorithm Due to the scanning method of ground-based lidar data, areas with an average height of over 15m are difficult to capture, which easily leads to missing point cloud data and incomplete tree trunk point clouds. The double-constraint convex hull algorithm struggles to complete the point cloud when data is missing, resulting in an underestimation of the calculated single-layer area. To address this, Hough circle fitting is used to fit the shape of the missing point cloud locations at higher elevations and calculate the area, thus compensating for the inaccurate calculation of upper-layer areas. The specific principle of the Hough circle fitting algorithm is as follows: Iterate through each edge point (x) i , y i ): Calculate the possible centers of the circle for all possible radii (a k , b k ), and at the same time, the corresponding accumulator A[a k ][b k ][r k Add 1; find local maxima in the accumulator that exceed the threshold. These maxima are the parameters (a*, b*, r*) of the detected circle.

[0007] S106: Calculation of Single Timber Volume The volume calculation of a single piece of wood adopts the idea of ​​the infinitesimal method, assuming that the trunk of each slice layer is a number of frustum-like cones, and using the frustum calculation formula to calculate the volume of the trunk of each layer (formula (4)): (4) Because the improved convex hull algorithm DB is sensitive to missing point cloud data, when the point cloud data is incomplete, the area is calculated by circle fitting. In order to further realize the automated calculation of volume, the algorithm is designed with an area ratio threshold ARatio to automatically determine whether the area ratio of DB algorithm and circle fitting is greater than ARatio. When it is greater than the threshold, it indicates that DB algorithm is greatly affected by missing data, and the area of ​​this layer is the result of circle fitting. To further correct the area results of each layer and prevent the existence of large or small fitted values, the algorithm uses a standard deviation threshold to assist in constraining the area of ​​each slice layer; the reliability of the area of ​​each layer is verified by comparing the area of ​​the current layer with the area of ​​the previous layers; a decrease in area usually indicates that the normal stem is gradually thinning, while a large abrupt change indicates that it is misfitted; in the case of detecting an area abrupt change, the algorithm will discard the data of this layer to prevent erroneous calculation (Formula (5)): (5) In the formula , represents the difference between the area of ​​the current layer and the average area of ​​its previous 10 layers; K represents the control factor of the threshold; S represents the standard deviation of Diff; Although ground-based lidar systems can collect high-density point clouds, their ability to collect information on objects above 15m is weak, and they cannot capture relatively complete point cloud information, resulting in the inability to accurately identify tree trunks at high altitudes. The algorithm uses conical fitting to calculate the volume of the tree trunk at high altitudes (Formula (6)): (6) In the formula, S n The area of ​​the fitted point cloud at the highest detected point is represented by H; H represents the height of a single tree; slice represents the number of slice layers. Indicates the slice layer height; The required volume of a single piece of timber is the sum of the volumes of the lower layers and the volume of the fitted cone of the top layer, i.e. (7).

[0008] The beneficial effects of this invention are: This application proposes a non-destructive measurement method for single-tree volume based on ground-based lidar, using TLS (Terrain LiDAR Systems) data. The method focuses on extracting structural features of tree trunks with different morphologies and calculating trunk volume. Using Masson pine as the research object and 52 individual trees as samples, the method destructively collects the true values ​​of different heights and diameters and calculates the volume of each tree. Simultaneously, a ground-based lidar system is used to scan a high-density three-dimensional lidar point cloud. First, the original point cloud data is sampled and processed. The growth of seed points is constrained by normal vectors and angles to accurately extract the tree trunk point cloud. Then, the point cloud is sliced ​​to simulate the cutting of a real disk, and an improved convex hull algorithm is used to calculate the area of ​​a single layer of point cloud. In occluded areas, a fitted circle is constructed using local point cloud data to calculate the area of ​​that layer. Finally, the volume of each tree is obtained using the method of infinitesimal elements. The correlation between the volume calculated in this invention and the actual calculated volume is R. 2 The value reached 0.995, and the RMSE was 0.017m. 3 The nRMSE was 2.38%, indicating that the invention can accurately extract tree trunks and calculate the volume of individual timbers without contact, and the results are reliable. Attached Figure Description

[0009] To more clearly illustrate the technical solutions in this disclosure, the accompanying drawings used in some embodiments of this disclosure will be briefly described below. Obviously, the drawings described below are only drawings of some embodiments of this disclosure, and those skilled in the art can obtain other drawings based on these drawings. In addition, the drawings described below can be regarded as schematic diagrams and are not intended to limit the actual size of the product, the actual flow of the method, the actual timing of the signals, etc. involved in the embodiments of this disclosure.

[0010] Figure 1 The flowchart of the proposed algorithm is shown below; (a) data preprocessing; (b) trunk extraction; (c) fitting of diameter and area of ​​single-layer slices; (d) calculation of single-wood volume.

[0011] Figure 2 This indicates the result of accurately identifying tree trunk point clouds using this algorithm, with red dots representing tree trunks.

[0012] Figure 3 To ensure the accuracy of volume calculations using this algorithm, the black dashed line represents a 1:1 line. Detailed Implementation

[0013] To enable those skilled in the art to better understand the technical solutions in this application, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this application.

[0014] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0015] The present invention will be further described in detail below through embodiments.

[0016] An example of a non-destructive measurement method for the volume of a single piece of timber based on ground-based lidar includes the following steps: Data collection The main tree species studied was Masson pine. Data on diameter at breast height (DBH), tree height, diameter at different relative positions, branch height, and crown width were collected and measured from 153 Masson pine trees. The bark-inclusive volume of each sample tree was calculated based on the diameter at different relative positions. Simultaneously, a FARO Focus3D terrestrial 3D laser scanner was used to scan the Masson pine trees and obtain high-precision 3D point clouds. Multiple point clouds were stitched together based on a common target sphere, achieving an average point cloud density of 4410 p / m². 2 .

[0017] Data preprocessing The average density of the point cloud data collected by the ground-based lidar is as high as 5000 p / m 2 To avoid the reduction in computational efficiency caused by the large amount of data from ground-based lidar, voxel downsampling is used to process the original ground-based lidar data. This method calculates the centroid of each voxel unit and uses the coordinates of the centroid to replace all points within the unit, thus achieving point cloud downsampling while ensuring that the distribution characteristics of the point cloud remain unchanged.

[0018] In this patent, based on the distribution of points, the voxel side length r is set to 0.1m.

[0019] Through morphological comparison and screening, 52 trees with typical different morphologies were finally selected as experimental subjects.

[0020] Tree trunk point cloud extraction To achieve non-destructive measurement of individual tree volume, it is still necessary to accurately identify the individual tree trunk point cloud from the scattered and disordered laser point cloud. This patent proposes a trunk extraction algorithm based on normal vector constraints. The specific steps of the algorithm are as follows: Step 1: Initialize the complete point set Points, seed point set SeedPoints, adjacent point set AdjacentPoints, trunk point set TrunkPoints, and temporary point set TempPoints; Step 2: Read the preprocessed point cloud data of a single tree and establish normal vectors for the point cloud data. For each scanned point p in the point cloud, search for its N nearest neighbor points, and then calculate the local plane P of these points in the least squares sense. This plane can be represented as: (1) In the formula, Let d be the normal vector of plane P, and d be the distance from P to the origin. The normal vector of the plane fitted by the N nearest neighbors can be considered as the normal vector of the current scan point; the normal vector of plane P can be obtained by principal component analysis (PCA), and the calculation shows that P passes through the centroid of its N neighborhood points. And normal vector satisfy =1, first perform eigenvalue decomposition on the covariance matrix M in equation (2) to obtain each eigenvalue of M, and the eigenvector corresponding to the smallest eigenvalue of M is the normal vector of P.

[0021] (2) The point cloud normal vector represents the direction of the plane formed by its neighboring points. This patent establishes a kd-tree for all point clouds and uses a search radius of 1m to calculate the point cloud normal vector using 10 nearest neighbor points.

[0022] Step 3: Based on the establishment of the normal vector, initialize the Points point set, assign a normal vector attribute value to each point, and set the category of all points to unclassified; Step 4: When hour, It can be by and The dot product is obtained, This indicates the degree of perpendicularity of the point cloud normal vector at that location. When the value approaches 0, it indicates that the normal vector is almost perpendicular to the Z-axis. Therefore, we choose... Regions with a value greater than 0.05 and a point cloud height greater than 0.05m and less than 0.2m are used as initial seed points for subsequent growth, and points that meet the conditions are added to SeedPoints.

[0023] Step 5: To further improve algorithm efficiency and reduce the amount of data traversed in loops, the algorithm adopts a segmented growth method, adding the seed point of the previous segment as the ignition seed point of the new segment. In this patent, a 0.3m interval and a 0.3m layer thickness are used to layer the single-tree point cloud, and the layered point cloud is added to the TempPoints collection. The pseudocode for point cloud slicing and layering is as follows: Step Six: For each temporary point in each layer, add each seed point in a loop, and calculate the 10 nearest neighbors of the seed point by Euclidean distance. Determine whether the angle between each point and the normal vector of the seed point is less than the angle threshold. Change the category of points that meet the conditions to trunk points, and also compare them with seed points. The difference is less than The point is recorded as the new seed point.

[0024]

[0025] Step 7: Loop through each slice of point cloud and gradually add the points that meet the conditions to the trunk point cloud set to generate the final trunk point cloud.

[0026] Double-constraint convex hull algorithm Based on the Graham convex hull algorithm, a dual-constraint convex hull algorithm is proposed, which mainly improves upon the Graham convex hull algorithm. Using the same method as the slicing algorithm in S103, the tree trunk point cloud is sliced ​​and input into the optimization algorithm one by one to accurately extract boundary points and calculate the area of ​​each layer. This algorithm mainly uses the angle and CHL convex hull side length threshold to gradually approximate the true distribution of the point cloud, making the convex hull more closely fit the tree trunk surface. Then, the discretized Green's theorem is used to calculate the area of ​​each layer of the convex hull.

[0027] Sort the generated convex hulls counterclockwise. At this point, for each convex set... vertex The area calculation formula is (Equation (3)): (3) In the formula, This is the convex hull area calculated using the optimized convex hull algorithm. The coordinates of the convex hull vertices arranged in counter-clockwise order The number of points in the convex hull set; typically, the initial point is repeatedly added to the end of the point set to close the polygon.

[0028] Hough circle fitting algorithm Due to the scanning method of ground-based lidar, areas with an average height of over 15m are difficult to capture, easily leading to missing point cloud data and incomplete tree trunk point clouds. The double-constraint convex hull algorithm struggles to complete the point cloud when data is missing, resulting in an underestimation of the calculated single-layer area. This patent employs Hough circle fitting to fit the shape of the missing point cloud locations at higher elevations and calculate the area, thus overcoming the inaccurate calculation of upper-layer areas. The specific principle is as follows: Iterate through each edge point (x) i , y i ): Calculate the possible centers of the circle for all possible radii (a k , b k ), and at the same time, the corresponding accumulator A[a k ][b k ][r k Add 1; find local maxima in the accumulator that exceed the threshold. These maxima are the parameters (a*, b*, r*) of the detected circle.

[0029] Single timber volume calculation The volume calculation of a single piece of wood adopts the idea of ​​the infinitesimal method, assuming that the trunk of each slice layer is a number of frustum-like cones, and using the frustum calculation formula to calculate the volume of the trunk of each layer (formula (4)): (4) Because the improved convex hull algorithm DB is sensitive to missing point cloud data, when the point cloud data is incomplete, the area is calculated by circle fitting. In order to further realize the automated calculation of volume, the algorithm is designed with an area ratio threshold ARatio to automatically determine whether the area ratio of the DB algorithm and the circle fitting is greater than ARatio. When it is greater than the threshold, it indicates that the DB algorithm is greatly affected by missing data, and the area of ​​the layer is the result of circle fitting.

[0030] To further correct the area results of each layer and prevent the existence of excessively large or small fitted values, the algorithm uses a standard deviation threshold to assist in constraining the area of ​​each slice layer. The reliability of the area of ​​each layer is verified by comparing the area of ​​the current layer with the areas of the previous layers. A decrease in area usually indicates that the normal stem is gradually thinning, while a large abrupt change indicates that it has been misfitted. In the case of a detected abrupt change in area, the algorithm discards the data of the current layer to prevent erroneous calculation (Equation (5)): (5) In the formula , represents the difference between the area of ​​the current layer and the average area of ​​the previous 10 layers; K represents the control factor of the threshold; S represents the standard deviation of Diff.

[0031] Although ground-based lidar systems can collect high-density point clouds, their ability to collect information on objects above 15m is weak, and they cannot capture relatively complete point cloud information, resulting in the inability to accurately identify tree trunks at high altitudes. The algorithm uses conical fitting to calculate the volume of the tree trunk at high altitudes (Formula (6)): (6) In the formula, S n The area of ​​the fitted point cloud at the highest detected point is represented by H; H represents the height of a single tree; slice represents the number of slice layers. Indicates the height of the slice layer.

[0032] The required volume of a single piece of timber is the sum of the volumes of the lower layers and the volume of the fitted cone of the top layer, i.e. (7) Algorithm performance evaluation Evaluation of the accuracy of volume calculation The volume of 52 individual trees surveyed in the sample plot was used as the actual value and compared with the volume calculated by the algorithm proposed in this invention. This patent uses a coefficient of determination (COP). ), root mean square error ( The results are evaluated (Equation (8-10)).

[0033] (8) (9) (10) in, This represents the actual value from the field survey; This represents the volume calculated by the algorithm proposed in this invention; Indicates the number of trees; and , respectively The average, maximum, and minimum values ​​are represented. .

[0034] Volume calculation results Compared with the true value, the extracted volume ( Figure 3 ), The value was 0.995, and the RMSE was 0.017m. 3 The nRMSE is 2.38%. Based on the evaluation indicators, it can be seen that the error between the volume calculated by the algorithm proposed in this invention and the actual volume of a single timber measured by destructive methods is extremely small, demonstrating very high accuracy in calculating the volume of a single timber. The low nRMSE value fully demonstrates the high stability and reliability of this algorithm, and the calculated volume can meet the actual needs of high-precision forestry resource surveys and monitoring.

[0035] The foregoing has described the basic principles, main features, and advantages of the present invention. It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered illustrative and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the scope of the invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

[0036] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This narrative style is merely for clarity. Those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

Claims

1. A non-destructive measurement method for the volume of a single timber based on ground-based lidar, characterized in that: Includes the following steps: S101: Data Acquisition The tree species studied was Masson pine. The diameter at breast height (DBH), tree height, diameter at different relative positions, height below the branch, and crown width of 153 Masson pine trees were collected and measured. The bark volume of each sample tree was calculated based on the diameter at different relative positions. At the same time, the FAROFocus3D terrestrial 3D laser scanner was used to scan the Masson pine trees to obtain high-precision 3D point clouds. The point clouds of multiple sites were stitched together based on a common target sphere. S102: Data Preprocessing The method of voxel downsampling is used to process the original ground-based lidar. This method calculates the centroid of each voxel unit and uses the coordinates of the centroid to replace all points in the unit, thereby completing point cloud downsampling while ensuring that the distribution characteristics of the point cloud remain unchanged. Based on the distribution of points, the voxel side length r is set to 0.1m; Through morphological comparison and screening, 52 trees with typical different morphologies were finally selected as experimental subjects. S103: Tree Trunk Point Cloud Extraction A tree trunk extraction algorithm based on normal vector constraints is proposed. The specific steps of the algorithm are as follows: Step 1: Initialize the complete point set Points, seed point set SeedPoints, adjacent point set AdjacentPoints, trunk point set TrunkPoints, and temporary point set TempPoints; Step 2: Read the preprocessed point cloud data of a single tree and establish a normal vector for the point cloud data; for each scan point p in the point cloud, search for its N nearest neighbor points, and then calculate the local plane P of these points in the least squares sense. This plane is represented as: 　　　（1） In the formula, Let d be the normal vector of plane P, and d be the distance from P to the origin. The normal vector of the plane fitted by the N nearest neighbors is considered to be the normal vector of the current scan point; the normal vector of plane P is obtained by principal component analysis (PCA), and the calculation shows that P passes through the centroid of its N neighborhood points. And normal vector satisfy =1, first perform eigenvalue decomposition on the covariance matrix M in equation (2) to obtain each eigenvalue of M, and the eigenvector corresponding to the smallest eigenvalue of M is the normal vector of P. 　　　（2） In the formula, p i For each point cloud, N represents the number of nearest neighbors. The point cloud normal vector represents the direction of the plane formed by its neighboring points. After building a kd-tree for all point clouds, a search radius of 1m is used, and the point cloud normal vector is calculated using 10 nearest neighbors. Step 3: Based on the establishment of the normal vector, initialize the Points point set, assign a normal vector attribute value to each point, and set the category of all points to unclassified; Step 4: When hour, Depend on and The dot product is obtained, This indicates the degree of perpendicularity of the point cloud normal vector at that location. When the value approaches 0, it indicates that the normal vector is almost perpendicular to the Z-axis; therefore, we choose... Regions with a value greater than 0.05 and a point cloud height greater than 0.05m and less than 0.2m are used as initial seed points for subsequent growth, and points that meet the conditions are added to SeedPoints; Step 5: The algorithm adopts a segmented growth method and adds the seed point of the previous segment as the ignition seed point of the new segment; the single tree point cloud is layered with an interval of 0.3m and a layer thickness of 0.3m, and the layered point cloud is added to the TempPoints set. Step Six: For each temporary point in each layer, add each seed point in a loop, and calculate the 10 nearest neighbors of the seed point by Euclidean distance. Determine whether the angle between each point and the normal vector of the seed point is less than the angle threshold. Change the category of points that meet the conditions to trunk points, and also compare them with seed points. The difference is less than The point is recorded as the new seed point; Step 7: Loop through each slice of point cloud and gradually add the points that meet the conditions to the trunk point cloud set to generate the final trunk point cloud; S104: Double Constraint Convex Hull Algorithm The Graham convex hull algorithm is an improvement on the double-constraint convex hull algorithm. It slices the tree trunk point cloud using the same method as the slicing algorithm in S103 and inputs each slice into the optimization algorithm to accurately extract boundary points and calculate the area of ​​a single layer. The algorithm uses angle and CHL convex hull side length thresholds to gradually approximate the true distribution of the point cloud, making the convex hull more closely fit the tree trunk surface. Then, it uses a discretized Green's formula to calculate the area of ​​each convex hull layer. Sort the generated convex hulls counterclockwise. At this point, for each convex set... vertex The area calculation formula is (Equation (3)): 　　　（3） In the formula, This is the convex hull area calculated using the optimized convex hull algorithm. The coordinates of the convex hull vertices arranged in counter-clockwise order This represents the number of points in the convex hull set; typically, the initial point is repeatedly added to the end of the point set to close the polygon. S105: Hough Circle Fitting Algorithm Due to the scanning method of ground-based lidar, areas with an average height of over 15m are difficult to capture, which easily leads to missing point cloud data and incomplete tree trunk point clouds. The double-constraint convex hull algorithm struggles to complete the point cloud when data is missing, resulting in an underestimation of the calculated single-layer area. To address this, Hough circle fitting is used to fit the shape of the missing point cloud locations at higher elevations and calculate the area, thus compensating for the inaccurate calculation of upper-layer areas. The specific principle of the Hough circle fitting algorithm is as follows: Iterate through each edge point (x) i , y i ): Calculate the possible centers of the circle for all possible radii (a k , b k ), and at the same time, the corresponding accumulator A[a k ][b k ][r k Add 1; find local maxima in the accumulator that exceed the threshold. These maxima are the parameters (a*, b*, r*) of the detected circle. S106: Calculation of Single Timber Volume The volume calculation of a single piece of wood adopts the idea of ​​the infinitesimal method, assuming that the trunk of each slice layer is a number of frustum-like cones, and using the frustum calculation formula to calculate the volume of the trunk of each layer (formula (4)): 　　　（4） Because the improved convex hull algorithm DB is sensitive to missing point cloud data, when the point cloud data is incomplete, the area is calculated by circle fitting. In order to further realize the automated calculation of volume, the algorithm is designed with an area ratio threshold ARatio to automatically determine whether the area ratio of DB algorithm and circle fitting is greater than ARatio. When it is greater than the threshold, it indicates that DB algorithm is greatly affected by missing data, and the area of ​​this layer is the result of circle fitting. To further refine the area results of each layer and prevent the existence of excessively large or small fitted values, the algorithm uses a standard deviation threshold to assist in constraining the area of ​​each slice layer. The reliability of the area of ​​each layer is verified by comparing the area of ​​the current layer with the areas of the previous layers. A decrease in area usually indicates that the normal stem is gradually thinning, while a large abrupt change indicates that it has been misfitted. Although ground-based lidar systems can collect high-density point clouds, their ability to collect information on objects above 15m is weak, and they cannot capture relatively complete point cloud information, resulting in the inability to accurately identify tree trunks at high altitudes. The algorithm uses conical fitting to calculate the volume of the tree trunk at high altitudes (Formula (6)): 　　　（6） In the formula, S n The area of ​​the fitted point cloud at the highest detected point is represented by H; H represents the height of a single tree; slice represents the number of slice layers. Indicates the slice layer height; The required volume of a single piece of timber is the sum of the volumes of the lower layers and the volume of the fitted cone of the top layer, i.e. （7） 。

Images