A method for extracting the crown volume of a single tree from ground-based laser point clouds based on spherical coordinate integration.

Abstract

This invention discloses a method for extracting the crown volume of a single tree from ground-based laser point clouds based on spherical coordinate integration, belonging to the field of forestry engineering technology. First, a point cloud segmentation method using a visual optimal threshold selection strategy is employed to extract the crown point cloud, improving the extraction and operability. Then, the crown points are projected from the 3D point cloud space to the 3D spherical coordinate space, innovatively proposing a new method for finely describing crown morphology. Finally, the crown is cut into several quadrangular pyramidal micro-elements and integrated to achieve crown volume extraction, improving the accuracy and universality of the extraction. This method provides a complete technical workflow for extracting the crown volume of a single tree from ground-based laser point clouds, contributing to improving the automation and intelligence level of forestry production and management. It has significant theoretical and engineering application reference value for forestry scientific research and forestry production, such as 3D tree morphology simulation, forest resource surveys, and aboveground biomass estimation.

Description

Technical Field

[0001] This invention belongs to the field of forestry engineering technology, specifically relating to a method for extracting the crown volume of a single tree from ground-based laser point clouds based on spherical coordinate integration. Background Technology

[0002] Real-time and accurate extraction of forest structure parameters is crucial for understanding terrestrial carbon storage and diagnosing ecosystem health. Canopy volume, representing the geometric volume of space occupied by the tree canopy, comprehensively reflects the degree of competition among tree groups and the intensity of photosynthesis. It is an important factor in surface ecological monitoring research such as three-dimensional tree morphology simulation, forest resource surveys, and surface biomass estimation. Traditional methods for measuring individual tree canopy volume mainly use measuring tools such as steel tape measures and altimeters to manually obtain canopy width and height and calculate canopy volume based on empirical models for specific tree species. However, this method is technically crude, and its measurement accuracy is limited by human factors and tree canopy morphology, failing to meet the requirements for accuracy and efficiency in extracting forest parameters. Ground-based lidar measurement technology (LiDAR) uses a non-contact laser measurement method to quickly acquire high-precision, high-density three-dimensional spatial structure information of forest trees—point cloud data. It has the advantages of fast scanning speed and high degree of automation, opening up new technical means for extracting the crown volume of trees. Therefore, carrying out research on single-tree crown volume extraction based on laser point cloud data has important theoretical significance and engineering reference value.

[0003] Based on the different spatial scales of the smallest differential unit, laser point cloud single-tree crown extraction methods can be divided into three categories: ① Crown volume extraction method based on point cloud boundary detection. This method treats the tree crown as a continuous irregular geometric body with spatial boundaries. By detecting three-dimensional spatial boundary points and constructing an irregular spatial triangular network on the crown surface, the volume of the closed triangular network is directly calculated to obtain the crown volume. Typical examples include the three-dimensional convex hull extraction method and the spatial triangular network extraction method. This method identifies the boundary points of the crown point cloud outline and forms a closed spatial triangular network or convex hull set, which can express the spatial range of the crown relatively completely and can better take into account the differences in the spatial morphology distribution of crowns of different tree species. However, the detection of crown outline boundary point cloud and the reconstruction of spatial triangular network increase the complexity of the crown volume extraction process, which is not conducive to large-scale promotion and use. At the same time, factors such as the lack of point cloud and uneven density make the rationality of the crown boundary triangular network questionable; ② Crown volume extraction method based on point cloud layering. The canopy volume extraction method based on point cloud layering divides the laser point cloud into several layers along the canopy height, treating each layer as a spatial cylinder with an irregular base. The α-shape algorithm is used to detect the base boundary and calculate its area. The canopy volume is obtained by accumulating the volumes of each layer. This method reduces the difficulty of canopy volume extraction by equating the local canopy morphology to a regular frustum model; however, the selection of the point cloud layering scale reduces the automation level of canopy volume extraction. Some scholars have attempted to improve the adaptability and accuracy of laser point cloud canopy extraction to some extent by setting the optimal segmentation scale and improving the layered ground boundary point detection of the α-shape algorithm. ③ Canopy volume extraction method based on point cloud voxel segmentation. The canopy volume extraction method based on point cloud voxel segmentation mainly divides the canopy laser point cloud into several spatial voxel units of varying sizes based on the spatial morphology distribution of the canopy. The effectiveness of each voxel unit is detected by counting the number of points within it, and the canopy volume is indirectly extracted by integrating the effective voxels. This method, to some extent, compensates for the shortcomings of the aforementioned methods in effectively identifying canopy gaps. However, it also faces the dual challenges of voxel spatial scale affecting the efficiency and accuracy of canopy volume extraction. Furthermore, factors such as point cloud defects caused by occlusion and density also affect the universality of canopy volume extraction. In summary, existing methods for extracting single-tree canopy volume based on laser point cloud data suffer from low accuracy and limited universality.

[0004] Therefore, the present invention aims to use ground laser point cloud data of individual trees to project the crown points from the three-dimensional point cloud space to the three-dimensional spherical coordinate space, and use the spherical coordinate integration method to achieve high-precision extraction of the crown volume of a single tree. Summary of the Invention

[0005] This invention addresses the shortcomings of existing methods for extracting the volume of a single tree crown from laser point cloud data, such as low accuracy and limited universality. It provides a ground-based laser point cloud crown extraction method based on spherical coordinate integration. By projecting the crown points from the three-dimensional point cloud space to the three-dimensional spherical coordinate space, the efficiency of extracting the volume of a single tree crown from laser point cloud data is improved. Based on the analysis of the spatial distribution characteristics of the crown outline, the crown points are cut into several quadrangular pyramidal micro-elements, and integration is performed on these micro-elements, thereby improving the accuracy and stability of the extraction of the crown volume of a single tree from laser point cloud data.

[0006] This invention employs the following technical solution: a method for extracting the crown volume of a single tree from ground-based laser point clouds based on spherical coordinate integration, comprising:

[0007] Step (1): Obtain laser point cloud data of individual tree foundations;

[0008] Step (2) involves statistically analyzing the elevation information of the laser point cloud on the foundation of the individual tree and extracting the crown points of the individual tree using the optimal threshold elevation segmentation method to reduce misclassification and omission caused by unreasonable threshold settings.

[0009] Step (3) maps the tree crown points of the individual trees from the three-dimensional point cloud space to the three-dimensional spherical coordinate space, and determines the integral infinitesimal volume calculation model;

[0010] Step (4): In the three-dimensional spherical coordinate space, using the above-mentioned micro-element model as the integration function and the range of horizontal and vertical angles as the integration interval, the volume estimation result of a single tree is obtained by using the integration method.

[0011] Furthermore, step (1) acquires laser point cloud data of the foundation of individual trees. A multi-view station scanning method is used to obtain complete laser point cloud data of the individual tree crowns, compensating for the missing point cloud data of individual tree crowns caused by factors such as occlusion and sensor performance, thus improving the completeness of the point cloud data. The specific implementation steps are as follows:

[0012] (11) Three ground-based laser scanning stations are evenly arranged around a single tree, and a spherical target is fixedly placed at the line of sight of the three stations.

[0013] (12) Using a spherical target, the laser point cloud data of three different stations are registered to obtain complete laser point cloud data of a single tree.

[0014] Furthermore, in step (2), based on the spatial three-dimensional distribution of individual trees and the distribution characteristics that the elevation of the crown point is greater than that of the trunk point, threshold segmentation is used to separate the crown point and trunk point of individual trees; based on the distribution characteristics that the number of crown points is much greater than that of trunk points, an elevation-point count curve is drawn and the inflection point of the curve is selected as the optimal elevation threshold.

[0015] The specific implementation method is as follows:

[0016] 21) Divide the elevation values ​​of the laser point cloud of a single tree crown into intervals, count the number of laser point clouds in each interval, and draw an elevation-point count curve.

[0017] 22) Because individual tree structures are interconnected in space and the number of crown points is much greater than the number of trunk points, the elevation-point cloud distribution curve shows a trend of first increasing and then decreasing along the direction of increasing elevation value. At this point, the inflection point of the curve corresponds to the connection point between the crown and the trunk. Based on the elevation-point cloud curve, the inflection point of the curve and its corresponding elevation value are manually judged by visual inspection. The elevation value corresponding to the inflection point is the optimal segmentation threshold for extracting the crown point cloud.

[0018] 23) Points with elevation values ​​greater than or equal to the optimal threshold in the point cloud of a single tree are marked as crown points, and points with elevation values ​​less than the optimal threshold in the point cloud of a single tree are marked as trunk points, thereby achieving rapid and precise extraction of the crown of a single tree from the laser point cloud.

[0019] Furthermore, the specific implementation method of step (3) is as follows:

[0020] 31) In the three-dimensional point cloud space, the tree crown points are de-centrified to obtain the three-dimensional coordinates of the tree crown points relative to the centroid points;

[0021]

[0022] (x i ′,y i ′,z i Let P be any tree canopy point. i In the normalized coordinates of the centroid in the 3D point cloud space, (x i ,y i ,z i Let P be any tree canopy point. i In the Cartesian coordinates of the 3D point cloud space, i = 1, 2, ..., N, where N is the total number of canopy points.

[0023] 32) Select the centroid of the tree crown as the origin of the three-dimensional spherical coordinate system. Based on the spatial geometric transformation relationship, establish a projection model between the spatial coordinate system and the spherical coordinate system to map the tree crown point from the three-dimensional point cloud space to the three-dimensional spherical coordinate space.

[0024] Let the centroid of the tree crown be the origin O of the three-dimensional spherical coordinate system. Let the spatial Euclidean distance between the tree crown point P and the origin O be the radius R of the sphere. Let the angle between the projection of the line connecting the tree crown point P and the positive X-axis be the horizontal angle θ. Let clockwise be positive, then θ∈[0, 2π]. Let the angle between the projection of the line connecting the tree crown P and the positive Z-axis be the vertical angle. but

[0025] Based on spatial geometric relationships, the projection transformation model of a single tree crown point from three-dimensional point cloud space to three-dimensional spherical coordinate space is as follows:

[0026]

[0027] Let P be any tree canopy point i The coordinates in three-dimensional spherical coordinate space.

[0028] 33) In three-dimensional spherical coordinate space, the tree crown point is cut into several quadrangular pyramidal micro-elements along the horizontal and vertical angles. Search all points in each quadrangular pyramid and take the maximum value of the spherical coordinate radius of the point in the micro-element as the height of the quadrangular pyramidal micro-element. Calculate the volume of each micro-element according to the spatial geometric volume calculation model of the quadrangular pyramid.

[0029] The quadrangular pyramidal microelement contains a set of canopy points. And jλ1≤θ i <(j+1)λ1, Where j is the canopy point P i The corresponding number of horizontal segments, where l is the canopy point P. i The corresponding vertical division numbers are j = 1.2...m, l = 1.2...n. Let λ1 and λ2 be the maximum radius of the point sphere within the pyramidal micro-element, respectively, and let m and n be the total number of horizontal and vertical segmentations of the canopy points, respectively. Based on the spatial rule-based method for calculating the volume of a pyramidal micro-element, the volume dV calculation model is as follows:

[0030]

[0031] Furthermore, the specific implementation method for obtaining the volume estimate of a single tree using the integration method in step (4) is as follows:

[0032] 41) The volumes of each micro-element are summed, and the resulting total is the volume extraction result of a single tree.

[0033]

[0034] Beneficial effects:

[0035] This invention provides a complete method for extracting the canopy volume of a single tree from ground-based laser point clouds based on spherical coordinate integration. The method includes a point cloud data acquisition scheme, canopy point extraction with an optimal threshold, and canopy volume calculation using spherical coordinate integration. By adopting a visual optimal threshold selection principle, the operability and reliability of canopy point extraction are improved. By projecting canopy points from the three-dimensional point cloud space to the three-dimensional spherical coordinate space, a square pyramidal micro-element model can accurately describe the three-dimensional spatial morphology within a local area of ​​the canopy, improving the efficiency and accuracy of laser point cloud canopy volume extraction. Furthermore, this method has good universal applicability to different tree species, point cloud densities, and other background environments. Attached Figure Description

[0036] Appendix Figure 1 A flowchart for extracting the crown volume of a single tree from a ground-based laser point cloud based on spherical coordinate integration, provided in an embodiment of the present invention;

[0037] Appendix Figure 2 This is a schematic diagram of laser point cloud data acquisition for a single-wood foundation provided in an embodiment of the present invention;

[0038] Appendix Figure 3 A single-tree laser point cloud segmentation effect diagram provided for the visualization of optimal threshold selection in an embodiment of the present invention;

[0039] Appendix Figure 4 This is a schematic diagram illustrating the relationship between a three-dimensional point cloud and a spherical coordinate system provided in an embodiment of the present invention.

[0040] Appendix Figure 5 A schematic diagram of an integral infinitesimal element provided in an embodiment of the present invention;

[0041] Appendix Figure 6 The following are laser point cloud data of individual trees provided in this embodiment of the invention: (a) Osmanthus laser point cloud; (b) Cedar laser point cloud; (c) Camphor laser point cloud; (d) Celtis sinensis laser point cloud; (e) Ginkgo laser point cloud; (f) Cherry blossom laser point cloud;

[0042] Appendix Figure 7 The absolute error distribution diagram of the single tree crown extraction results provided in the embodiments of the present invention;

[0043] Appendix Figure 8 The relative error distribution diagram of the single tree crown extraction results provided in the embodiments of the present invention. Detailed Implementation

[0044] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and experimental embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0045] In its natural state, the crown of a single tree extends into space from the trunk. In point cloud space, the outer contour of the crown appears as an irregular spatial volume with closed boundaries, but the local area can be regarded as a regular spherical surface. If the crown laser point cloud is converted from a three-dimensional rectangular coordinate space to a three-dimensional spherical coordinate space, and the crown layer is cut into several closely connected conical micro-elements, the crown volume can be obtained by the volume integral of the conical micro-elements. First, the elevation distribution of the single tree laser point cloud is statistically analyzed and an elevation-point count curve is plotted. The crown laser point cloud is extracted using a visualization optimal threshold elevation segmentation method. Then, a three-dimensional spherical coordinate system is defined, and the single tree crown laser point cloud is projected onto the three-dimensional spherical coordinate space. Finally, the single tree crown point cloud is cut into several conical micro-elements along horizontal and vertical angles, and the volume integral of the micro-elements determines the crown volume of the individual tree. The micro-element cutting process can better express the spatial distribution characteristics of the local area of ​​the tree crown outline. The micro-element volume integration process can better solve the irregular description of the overall area of ​​the tree crown outline. Compared with the existing laser point cloud single tree volume extraction method, it has the advantages of high single tree volume accuracy and good universality. It has important theoretical significance and engineering application reference value for research such as three-dimensional morphology simulation of trees, forest resource survey, and aboveground biomass estimation.

[0046] As attached Figure 1 As shown, in one embodiment, a method for extracting the crown volume of a single tree from ground-based laser point clouds based on spherical coordinate integration is proposed, specifically including the following steps:

[0047] Step (1): Obtain laser point cloud data of individual tree foundations;

[0048] Step (2): Perform elevation information statistics on the laser point cloud of the individual tree foundation, and extract the tree crown points of the individual tree by visual optimal threshold elevation segmentation;

[0049] Step (3) maps the tree crown points of the individual trees from the three-dimensional point cloud space to the three-dimensional spherical coordinate space, and determines the integral infinitesimal volume calculation model;

[0050] Step (4): In the three-dimensional spherical coordinate space, the integral method is used to obtain the estimated result of the crown volume of a single tree by taking the micro-element model as the integral function and the range of horizontal and vertical angles as the integration interval.

[0051] Step (1) Acquire laser point cloud data of the foundation of a single tree. In this embodiment of the invention, the ground laser scanning system is affected by factors such as ground cover obstruction, resulting in missing point cloud data for single tree canopies. To improve the integrity of the laser point cloud, a multi-station uniform deployment scheme is adopted for acquiring laser point cloud data of the foundation of a single tree: three scanning stations are evenly deployed around the single tree as the center, and spherical targets are placed at the line of sight of each station. The spherical targets are used for registration operations of the point cloud at each station. The multi-station uniform deployment scheme is attached. Figure 2 As shown.

[0052] Step (2) involves extracting canopy points using an elevation segmentation method with a visually optimal threshold selection. Individual trees have a spatial three-dimensional structure; the elevation of the canopy points is greater than the elevation of the trunk points, and the number of trunk points is much greater than the number of trunk points. The point cloud of a single tree can be segmented into canopy points and trunk points using the elevation values ​​of the connection points between the canopy and trunk. The optimal threshold is determined using a visually optimal threshold selection principle. First, the elevation values ​​of the laser point cloud of a single tree are statistically analyzed in intervals of 0.1m increments, and an elevation-point count curve is plotted. The inflection point of the elevation-point count curve is visually identified, and the elevation value corresponding to the inflection point is set as the optimal threshold, as shown in the attached figure. Figure 3 As shown. The optimal threshold H is... threshold =2.49m. Based on this threshold, the individual data point cloud is divided into canopy points and trunk points, as shown in formulas (1-2).

[0053] The principle for determining the canopy point is: p i ={(x i y i , z i )|z i ≥H threshold} (1) The principle for determining the trunk point is: p i ={(x i y i , z i )|z i <H threshold} (2)

[0054] Step (3): Project the tree canopy points from the 3D point cloud space to the 3D spherical coordinate space to determine the integral infinitesimal volume calculation model. Definition of the 3D spherical coordinate system: The centroid of the tree canopy points is the origin O of the 3D spherical coordinate system, the spatial Euclidean distance between the tree canopy point P and the origin is the radius R of the sphere, the angle between the projection of the line connecting the tree canopy points P and the positive X-axis on the horizontal plane is the horizontal angle θ, and clockwise is defined as positive, then θ∈[0, 2π], the angle between the projection of the line connecting the tree canopy points P and the positive Z-axis on the vertical plane is the vertical angle. but The positional relationship between the 3D point cloud spatial rectangular coordinate system and the 3D spatial spherical coordinate system is shown in the appendix. Figure 4 As shown. According to the spatial geometric transformation relationship, any tree crown point P i The three-dimensional point cloud rectangular coordinates (x i ,y i ,z i Projected to three-dimensional spherical coordinates The specific steps for determining the integral infinitesimal element model are as follows:

[0055] 31) Calculate the coordinates of the centroid of the tree crown. The purpose of calculating the coordinates of the centroid of the tree crown is to determine the origin of the three-dimensional spherical coordinate system of the tree crown. The coordinates of the centroid of the tree crown can be calculated from the average value of the three-dimensional spatial coordinates of the tree crown points, as shown in formula (3), where (x0, y0, z0) are the coordinates of the centroid of the tree crown, and N is the number of tree crown points.

[0056]

[0057] 32) Centroidalization of tree crown points. The purpose of centroidalization of tree crown points is to convert the three-dimensional spatial coordinates of tree crown points into coordinate values ​​relative to the centroid. The process of centroidalization of tree crown points is shown in formula (4).

[0058] (x i ′,y i ′,z i ′)=(x0-x i y0-y i , z0-z i (4)

[0059] 33) Projection transformation relationship of tree canopy points. According to the spatial geometric transformation relationship, the projection transformation model of tree canopy points from three-dimensional point cloud space to three-dimensional spherical coordinate space is shown in formula (5).

[0060]

[0061] 34) Tree Canopy Micro-element Model. In three-dimensional spherical coordinate space, the tree canopy points are divided into several square pyramidal micro-elements along the horizontal and vertical directions. Each square pyramidal micro-element contains a set of tree canopy points. And jλ1≤θ i <(j+1)λ1, Where j is the canopy point P i The corresponding number of horizontal segments, where l is the canopy point P. i The corresponding vertical division numbers are j = 1.2...m, l = 1.2...n. Let λ1 and λ2 be the maximum radius of the point sphere coordinates within the quadrangular pyramidal infinitesimal element, respectively, and let m and n be the total number of horizontal and vertical segmentations of the tree crown points, respectively. (See attached diagram.) Figure 5 As shown. According to the spatial rule method for calculating the volume of a square pyramid, the calculation model for the volume dV of a square pyramidal infinitesimal element can be expressed by formula (6):

[0062]

[0063] Step 4: Determine the crown volume of a single tree using the spherical coordinate integration method. This is because the number of crown points and the radius R within different square pyramidal infinitesimal elements vary. jlSince the tree canopy volume is a non-continuous variable, the formula for calculating the volume of the tree canopy is used, as shown in formula (7).

[0064]

[0065] Angular resolutions λ1 and λ2 are input parameters for extracting tree canopy volume from laser point clouds. Choosing values ​​that are too large results in coarse extraction of the canopy volume, while choosing values ​​that are too small results in low efficiency for extracting the volume of individual trees. Considering both accuracy and efficiency, the angular resolutions for the horizontal and vertical angles are set to be equal and 3 to 5 times the average spacing of the point cloud.

[0066] To verify the effectiveness and feasibility of the ground-based laser point cloud single-tree crown volume extraction method based on spherical coordinate integration provided by this invention, experimental examples were used for verification.

[0067] 1. Experimental Data. The experimental subjects were individual trees of six different species (Osmanthus, Cedrus deodara, Camphor, Celtis sinensis, Ginkgo, and Cherry Blossom) with varying crown morphologies, located at the Chenggong Campus of Yunnan Normal University. The laser point cloud acquisition equipment was a Leica P40 ground-based lidar system. Key parameters were: ranging accuracy 1.2mm + 10ppm; angular accuracy 8″; point accuracy 3mm@50m; scanning rate 1 million points / second; horizontal and vertical field of view 360° and 270°, respectively. A multi-station scanning scheme was used for data acquisition. Three scanning stations were evenly distributed around the research object, and a planar target was placed in the common visible area of ​​the three stations. The final experimental data consisted of single-tree laser point clouds obtained using the Leica commercial software Cyclic One, which performed multi-station point cloud data registration, denoising, and cropping. The maximum registration error between adjacent stations was 0.09m. The experimental laser point cloud data is shown below. Figure 6 As shown in (a~f).

[0068] The six selected sets of ground-based laser point cloud experimental data have different average laser point cloud distances, point cloud numbers, and individual tree heights, representing five different canopy spatial morphologies: spatial sphere, spatial triangular pyramid, spatial ellipsoid, spatial cylinder, and spatial inverted triangular pyramid. These data are universally applicable in urban green space resource surveys and carbon storage engineering studies, and can meet the experimental data requirements for the feasibility and reliability analysis of the spherical coordinate integration method for extracting the canopy volume of individual trees in this paper. The basic information statistics of the laser point cloud experimental data are shown in Table 1.

[0069] Table 1. Statistical Table of Basic Information of Experimental Data

[0070]

[0071] 2. Results and Analysis. Since the outline of a single tree crown is an irregular spatial volume, its overall spatial morphology cannot be precisely described by a mathematical model, thus the theoretical value of the crown volume cannot be directly determined. Wei Xuehua et al. (A method for calculating crown volume based on three-dimensional laser scanning point clouds [J]. Transactions of the Chinese Society for Agricultural Machinery, 2013, 44(07):235-240) proposed a method for extracting crown volume from laser point clouds based on voxel segmentation. They analyzed the effectiveness and reliability of the algorithm using multiple sets of experimental data and discussed the influence of voxel size selection on the accuracy and efficiency of crown volume extraction. The results showed that smaller voxels result in higher crown extraction accuracy and have become a standard reference for evaluating the performance of crown volume extraction from laser point cloud data. This paper uses the voxel segmentation method for extracting crown volume from laser point cloud data as described in this paper. When the voxel size is 0.05m, the estimated result is used as a theoretical reference for the crown volume. The absolute error and relative error are selected as the evaluation indexes for the accuracy of canopy volume extraction, as shown in formula (8). In the formula, represents the canopy volume extracted by laser point cloud, and represents the theoretical reference value of canopy volume. The absolute error describes the magnitude of the deviation between the canopy extraction result and the theoretical value, and the relative error % describes the proportion of the deviation between the canopy extraction result and the theoretical value.

[0072]

[0073] The canopy volume extraction results of point cloud data using the spherical coordinate integration method of this invention are compared with those of Method 1 (Liu Fang, Feng Zhongke, Yang Liyan, et al. Research on canopy volume estimation based on three-dimensional laser point cloud data [J]. Transactions of the Chinese Society for Agricultural Machinery, 2016, 47(03):328-334.) and Method 2 (Lin Song, Tian Linya, Bi Jixin, et al. Accurate calculation of single tree canopy volume from three-dimensional laser scanning data [J]. Science of Surveying and Mapping, 2020, 45(08):115-122). The two existing methods represent canopy volume extraction methods based on point cloud boundary detection and point cloud layering, respectively. Table 2 shows the extracted values ​​and accuracy statistics of laser point cloud canopy volume based on different methods. Comparing different laser point cloud experimental data, when the canopy spatial morphology is spherical or triangular pyramidal with a relatively regular outline, such as in the experimental data of osmanthus, cedar, and camphor trees, different methods for canopy volume extraction all showed good results. The accuracy index of canopy volume extraction in the experimental data remained within a certain range, and the maximum absolute error of canopy volume extraction was 1.02m. 3 The maximum relative error was 7.32%, as shown in bold in Table 2. When the canopy outline was irregular, such as in the experimental data of *Celtis yunnanensis* and *Ginkgo biloba*, the accuracy of canopy volume extraction fluctuated significantly. This was most evident in the extraction results of laser point cloud data from *Celtis yunnanensis*, with a maximum absolute error of 3.46 m. 3 Compared with different laser point cloud canopy volume extraction methods, the maximum absolute error of canopy extraction based on the method presented in this paper is 2.33m. 3The maximum relative error is 3.40%, which shows higher extraction accuracy compared to existing methods, and maintains good applicability to different canopy morphologies and tree species. The reason for the discrepancy is that the point cloud boundary detection method, by detecting the canopy boundary points and generating a 3D convex hull, effectively takes into account the irregularity of the canopy's outer contour. However, when the canopy shape has large "spurs," the unreasonable convex hull causes the canopy volume extraction result to be distorted. The canopy volume extraction of point cloud layers cuts the canopy into several faults and treats the faults as regular frustums. Its canopy volume calculation only considers the shape of the fault sampling surface and does not take into account the canopy spatial characteristics between the two sampling surfaces. This results in the volume of elliptical canopies being too small and the volume of irregularly shaped canopies being too large. The present invention projects discrete 3D laser point clouds onto 3D spherical coordinate space and cuts the canopy layer into several conical micro-elements, which can more accurately describe the differences in the canopy edge contour. By integrating the volume of the micro-elements in spherical coordinate space, it better takes into account factors such as the internal missing parts of the point cloud and the differences in canopy shape, showing higher extraction accuracy and universality.

[0074] Table 2. Laser point cloud canopy volume extraction and accuracy analysis

[0075]

[0076] Appendix Figure 7-8 The distribution curves of accuracy evaluation indices for canopy volume extraction results based on experimental data from the present invention and existing methods are shown. The diamond-shaped curves represent the accuracy evaluation curves for canopy extraction from point cloud boundary detection experimental data, the square-shaped curves represent the accuracy evaluation curves for canopy volume extraction from point cloud layering experimental data, and the triangular-shaped curves represent the accuracy evaluation curves for canopy extraction from experimental data using the present invention. For different canopy laser point cloud experimental data, the absolute and relative error distribution curves of the canopy extraction method of the present invention show relatively gentle changes, and the accuracy indices are distributed within a certain range. Analysis indicates that the laser point cloud single-tree canopy extraction method based on the present invention has the advantage of high stability.

[0077] 3. Conclusion. Canopy volume is a key indicator parameter for forest surveys, forestry resource monitoring, and management. Addressing the shortcomings of laser point cloud canopy volume extraction methods, such as limited universality, this application proposes a ground-based laser point cloud canopy volume extraction method based on spherical coordinate integration, after analyzing the structure of individual trees. Experiments were conducted using multiple sets of ground-based laser point cloud data, involving factors such as point cloud density, canopy morphology, and tree species. Comparison with two existing laser point cloud canopy volume extraction methods—point cloud boundary detection and point cloud layering—shows that the proposed method projects the laser point cloud into a three-dimensional spherical coordinate space, changing the traditional canopy volume extraction approach based on disordered point cloud layering and boundary detection. By treating the local area of ​​the canopy outline as a spherical model, a spherical coordinate integration strategy for canopy volume calculation based on conical micro-elements is designed, which better considers factors such as missing points within the point cloud, irregular canopy outline morphology, and differences in tree growth. Compared with other existing methods, this method has better accuracy and stability in canopy extraction, with a maximum absolute error of 2.33 m for individual tree canopy volume extraction. 3 The maximum relative error was 3.40%.

[0078] The specific embodiments described herein are merely illustrative of the feasibility of the present invention. Those skilled in the art to which this invention pertains may make various modifications or additions to the described specific embodiments or use similar methods to replace them, but all equivalent implementations or changes that do not depart from the scope of the present invention should be included within the scope of the present invention.

Claims

1. A method for extracting single-tree crown volume based on spherical coordinate integration of ground-based laser point cloud, characterized in that, The method comprises the following steps: Step (1), acquiring monomer tree ground laser point cloud data; Step (2), performing elevation information statistics on the monomer tree ground laser point cloud, and extracting monomer tree crown points by using visual optimal threshold elevation segmentation; Step (3), mapping the monomer tree crown points from a three-dimensional point cloud space to a three-dimensional spherical coordinate space, and determining an integral microelement body calculation model; Step (4), in the three-dimensional spherical coordinate space, taking the integral microelement body calculation model as an integral function, and taking a horizontal angle and a vertical angle range as an integral interval, and obtaining monomer tree crown volume estimation results by using an integral method; Step (3) specifically comprises: A three-dimensional space spherical coordinate system is established: taking the crown gravity center point as the three-dimensional space spherical coordinate system origin O, the space Euclidean distance of the crown point P and the origin O is the spherical radius R, the projection of the OP connecting line on the horizontal plane and the X axis positive direction angle is the horizontal angle θ, and the projection of the OP connecting line on the vertical plane and the Z axis positive direction angle is the vertical angle then Point cloud projection transformation: according to the spatial geometric relationship, the projection transformation model of the monomer tree crown points from the three-dimensional point cloud space to the three-dimensional spherical coordinate space is as follows: where (x i ′,y i ′,z i ′) are arbitrary crown points P i In the three-dimensional point cloud space, the center of gravity normalization coordinate value is where (x i In the three-dimensional spherical coordinate space, the coordinate value is i = 1, 2,..., N, N is the total number of crown points; The crown micro-element model: in the three-dimensional spherical coordinate space, the crown points are cut into several four-prism micro-elements along the horizontal direction and the vertical direction, and the four-prism micro-element contains the crown point set and jλ1≤θ i <(j+1)λ1, wherein j is the crown point P i corresponding to the horizontal segmentation number, l is the crown point P i corresponding to the vertical segmentation number, j=1,2,...,m, l=1,2,...,n, is the maximum value of the point spherical coordinate radius in the four-prism micro-element, λ1 and λ2 are the horizontal angle resolution and the vertical angle resolution respectively, and m and n are the total number of the horizontal segmentation and the vertical segmentation of the crown point respectively, according to the space rule four-prism volume calculation method, the four-prism micro-element volume dV calculation model is:

2. The ground-based laser point cloud single-tree crown volume extraction method based on spherical coordinate integration according to claim 1, characterized in that, The step (1) specifically comprises: A monomer tree is taken as a center, and three ground laser scanning stations are uniformly arranged around the monomer tree; and a spherical target is installed at a line-of-sight position of the three ground laser scanning stations. The laser point cloud data of the three different ground laser scanning stations are registered by using the spherical target, and complete monomer tree laser point cloud data are acquired.

3. The ground-based laser point cloud single-tree crown volume extraction method based on spherical coordinate integration according to claim 1, characterized in that, The step (2) specifically comprises: According to the elevation value distribution of the monomer tree laser point cloud, an elevation-point number statistical curve is drawn, and an elevation value corresponding to an inflection point of the statistical curve is set as an optimal threshold value; The optimal threshold value is used for segmenting the point cloud, so that the monomer tree crown extraction is realized; wherein the points of the monomer tree point cloud with an elevation value greater than or equal to the optimal threshold value are segmented as crown points, and the points of the monomer tree point cloud with an elevation value less than the optimal threshold value are segmented as stem points.

4. The ground-based laser point cloud tree crown volume extraction method based on spherical coordinate integration according to claim 1, characterized in that, The crown point P i The gravity center normalization coordinate value in the three-dimensional point cloud space is: (x i ,y i ,z i ) are the Cartesian coordinates of the crown point P i in the three-dimensional point cloud space.

5. The ground-based laser point cloud tree crown volume extraction method based on spherical coordinate integration according to claim 1, characterized in that, The step (4) specifically comprises: According to the crown microelement radius R jl As a non-continuous variable, the crown volume V is calculated using a sum of discrete variables:

6. The ground-based laser point cloud tree crown volume extraction method based on spherical coordinate integration according to claim 5, characterized in that, The horizontal angle resolution λ1 and the vertical angle resolution λ2 are equal in value, and are 3-5 times of the average spacing of the point cloud.

Images