Forest land tree height measurement and determination method and system based on laser radar point cloud data

Abstract

The invention provides a forest land tree height measurement and determination method and system based on laser radar point cloud data. Stress wave signals are collected based on a trunk base acoustic emission sensor array to generate an acoustic characteristic parameter set, the digital twin model is driven to complete forest stand structure topological optimization, and a three-dimensional growth vector model reflecting the internal mechanical state of a trunk is formed. And synchronously fusing high-precision slope point cloud data returned by the unmanned aerial vehicle laser radar in real time, correcting a terrain distortion error through a dynamic splicing algorithm in combination with stress distribution characteristics, and generating a crown segmentation boundary constrained by physical characteristics. And finally outputting a tree height parameter corrected by the abrupt slope topography through model iterative optimization and space vector analysis. According to the technical scheme, synchronous sensing of the three-dimensional shape and the mechanical state of the tree in the complex terrain environment is achieved, and the tree height measurement error is reduced.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a method and system for measuring and determining forest tree heights based on lidar point cloud data. Background Art

[0002] Slope variations make it difficult for traditional measurement methods to accurately capture tree crown morphology and height. Topographical fluctuations also cause significant signal attenuation and obstruction, necessitating the integration of multi-dimensional physical properties to compensate. Real-time health assessments of forest ecosystems require simultaneous capture of both the internal stress state and external morphological changes of trees to support sustainable forest management and carbon sink accounting.

[0003] Traditional measurement and processing methods usually use drone lidar to scan the slope, rely on point cloud stitching algorithms to generate tree crown outlines, and then measure tree heights based on the above crown outlines.

[0004] Failure to consider the dynamic impact of stress wave propagation within the trunk on crown morphology leads to errors of 15%-30% in tree height measurements on steep slopes. Furthermore, the 3D stand model, constructed based on historical LiDAR data, lacks real-time acoustic emission parameter-driven updates and fails to reflect changes in the tree's mechanical state, resulting in insufficient tree height measurement accuracy. Summary of the Invention

[0005] The embodiments of the present application provide a method and system for measuring and determining forest tree height based on lidar point cloud data, so as to solve the problem of insufficient tree height measurement accuracy in the prior art.

[0006] In the first aspect, an embodiment of the present application provides a method for measuring and determining the height of trees in a forest based on lidar point cloud data, comprising: collecting growth stress wave signals through a sensor array arranged at the base of the trunk, and generating a set of acoustic characteristic parameters reflecting the internal mechanical state of the trunk, wherein the sensor array is evenly spaced in the normal direction of the forest slope to form a monitoring network; based on the set of acoustic characteristic parameters, a digital twin model is driven to perform forest stand structure topology optimization, generate a three-dimensional growth vector model, and at the same time receive the slope scanning point cloud transmitted in real time by the drone lidar through the 5G edge node; perform the slope scanning point cloud optimization on the slope scanning point cloud. A dynamic point cloud stitching operation with terrain compensation is performed, and the stitching error is corrected in combination with the stress distribution characteristics in the three-dimensional growth vector model to generate a crown segmentation boundary dataset; the crown segmentation boundary dataset is fed back to the digital twin model for iterative optimization, and the spatial distribution characteristics of the slope scanning point cloud and the stress wave propagation time series data in the acoustic feature parameter set are combined to generate a three-dimensional crown outline with terrain displacement compensation; based on the spatial vector relationship between the vertices of the three-dimensional crown outline and the reference plane of the forest slope, the gradient data in the acoustic feature parameter set are superimposed to output tree height measurement parameters corrected for steep slope terrain.

[0007] Optionally, a dynamic point cloud splicing operation of terrain compensation is performed on the slope surface scanning point cloud, and the splicing error is corrected in combination with the stress distribution characteristics in the three-dimensional growth vector model to generate a crown segmentation boundary data set, including: extracting the local elevation gradient field and curvature characteristics of the slope surface scanning point cloud, and calculating the spatial adaptive weight of the terrain compensation parameter in combination with the stress direction field corresponding to the stress distribution characteristics in the three-dimensional growth vector model; using the angle between the maximum shear stress gradient and the principal stress direction in the stress distribution characteristics as a dynamic correction factor, and iteratively correcting the splicing error point by point through the spatial adaptive weight in the dynamic splicing process to obtain a corrected point cloud geometry. What characteristics; based on the corrected point cloud geometric characteristics, the normal vector and principal curvature parameters of each point in the slope scanning point cloud are calculated, and the surface feature parameters consistent with the principal stress direction of crown growth are screened in combination with the spatial distribution characteristics of the stress gradient difference field; the spatial consistency mapping relationship between the normal vector and the stress gradient difference field is used as a physical constraint, and the crown boundary is determined by the collinearity threshold of the principal stress direction and the surface normal; on the basis of determining the crown boundary, the surface feature parameters and the stress gradient difference field are multi-dimensionally fused, and the crown segmentation boundary data set is generated through the spatial consistency constraint of the principal stress direction of crown growth and the surface normal.

[0008] Optionally, the surface characteristic parameters are multi-dimensionally fused with the stress gradient difference field, and a crown segmentation boundary data set is generated through the spatial consistency constraint of the crown growth principal stress direction and the surface normal, including: constructing a dynamic distribution model of the crown surface geometric attributes based on the surface characteristic parameters, integrating the surface curvature change rate in the surface characteristic parameters and the normal offset corresponding to the normal vector of each point in the slope scanning point cloud into a spatial continuity descriptor, and establishing a topological association expression of the surface characteristic parameters; adopting the multi-scale decomposition method of the stress gradient difference field, through the crown growth principal stress, The vector superposition of directions generates a composite tensor representation of the stress gradient difference field; the topological association expression and the composite tensor representation are multi-dimensionally registered in space, and the fusion weight coefficient is calculated using the spatial projection deviation between the surface normal corresponding to the surface feature parameters and the principal stress direction of crown growth; based on the fusion weight coefficient, a collaborative constraint function of the principal stress direction of crown growth and the surface normal is generated; according to the feedback channel of the collaborative constraint function, a candidate region set of the crown segmentation boundary is output, and the amplitude convergence of the stress gradient difference field is used as the boundary constraint to screen out the boundary candidate regions, and generate a crown segmentation boundary data set.

[0009] Optionally, a collaborative constraint function of the principal stress direction of crown growth and the surface normal is generated based on the fusion weight coefficient, including: performing tensor feature correction on the fusion weight coefficient and the stress distribution characteristics of the three-dimensional growth vector model, generating a revised set of fusion weight coefficients through the orthogonality constraint condition of the principal stress direction vector and the surface normal vector; and constructing a collaborative constraint function of the principal stress direction of crown growth and the surface normal based on the revised set.

[0010] Optionally, the digital twin model is driven based on the set of acoustic characteristic parameters to perform forest stand structure topology optimization and generate a three-dimensional growth vector model, including: spatially aligning the growth stress wave signals collected by the sensor array in the direction of the slope normal, extracting the propagation characteristics of the stress wave signals in the radial and axial directions of the trunk, and generating the spatial distribution parameters of the internal mechanical state of the trunk; based on the spatial distribution parameters, constructing a three-dimensional vector expression of the trunk growth stress field, wherein the axial component represents the longitudinal growth trend of the trunk, and the radial component represents the lateral growth trend of the trunk; spatially aligning the three-dimensional vector expression with the slope scanning point cloud collected by the UAV lidar, and establishing a mapping relationship between the trunk growth stress field and the external morphological characteristics; based on the mapping relationship, calculating the dynamic change trend of the trunk growth stress field through the propagation time series characteristics of the stress wave signal inside the trunk, and generating a three-dimensional growth vector model reflecting the growth characteristics of the forest stand structure.

[0011] Optionally, the crown segmentation boundary dataset is fed back to the digital twin model for iterative optimization, and the spatial distribution characteristics of the slope scanning point cloud and the stress wave propagation time series data in the acoustic feature parameter set are combined to generate a three-dimensional crown contour with terrain displacement compensation, including: inputting the crown segmentation boundary dataset into the digital twin model, and calculating the initial terrain displacement of the crown boundary point based on the propagation path characteristics of the stress wave inside the trunk in the stress wave propagation time series data; iteratively comparing the initial terrain displacement with the spatial distribution characteristics of the slope scanning point cloud, optimizing the spatial position of the crown boundary point through the digital twin model, and updating the terrain displacement; based on the updated terrain displacement, performing spatial position correction on the point cloud in the crown segmentation boundary dataset to generate a preliminary compensated three-dimensional crown contour; inputting the preliminary compensated three-dimensional crown contour into the digital twin model again, and optimizing the terrain displacement compensation parameters in combination with the dynamic change characteristics of the stress wave propagation time series data to generate a three-dimensional crown contour with terrain displacement compensation.

[0012] Optionally, based on the spatial vector relationship between the three-dimensional crown contour vertices and the reference plane of the forest slope, the gradient data in the acoustic feature parameter set are superimposed to output tree height measurement parameters corrected for steep slope terrain, including: extracting the vertical distance between the three-dimensional crown contour vertices and the reference plane of the forest slope to generate an initial tree height spatial vector set; based on the gradient data in the acoustic feature parameter set, calculating the gradient change characteristics on the stress wave propagation path from the base of the trunk to the crown vertex to generate tree height correction parameters; mapping the tree height correction parameters to the initial tree height spatial vector set, and correcting the direction and amplitude of the initial tree height spatial vector in combination with the slope characteristics of the forest slope; based on the corrected tree height spatial vector set, calculating the corrected tree height of each crown contour vertex, and outputting tree height measurement parameters corrected for steep slope terrain.

[0013] On the second aspect, the embodiment of the present application provides a forest tree height measurement and determination system based on lidar point cloud data, including: an acquisition module for collecting growth stress wave signals through an acoustic emission sensor array arranged at the base of the trunk, and generating a set of acoustic characteristic parameters reflecting the internal mechanical state of the trunk, wherein the sensor array is evenly spaced in the direction of the slope normal to form a monitoring network; an optimization module for driving a digital twin model to perform forest stand structure topology optimization based on the acoustic characteristic parameter set, generating a three-dimensional growth vector model, and at the same time receiving the slope scanning point cloud transmitted in real time by the drone lidar through the 5G edge node; a correction module for The cloud performs dynamic stitching of point clouds with terrain compensation, corrects stitching errors in combination with the stress distribution characteristics in the three-dimensional growth vector model, and generates a crown segmentation boundary dataset that integrates physical property constraints; a generation module is used to return the crown segmentation boundary dataset to the digital twin model for iterative optimization, and combines the spatial distribution characteristics of the slope scanning point cloud with the stress wave propagation time series data in the acoustic feature parameter set to generate a three-dimensional crown contour with terrain displacement compensation; an output module is used to superimpose the gradient data in the acoustic feature parameter set according to the spatial vector relationship between the three-dimensional crown contour vertex and the slope reference plane, and output the tree height measurement parameters corrected for the steep slope terrain.

[0014] In a third aspect, an embodiment of the present application provides a computing device comprising a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a method for measuring and determining forest tree height based on lidar point cloud data as described in the first aspect above.

[0015] In a fourth aspect, an embodiment of the present application provides a computer storage medium storing a computer program. When the computer program is executed by a computer, it implements a method for measuring and determining forest tree height based on lidar point cloud data as described in the first aspect.

[0016] In an embodiment of the present application, an acoustic emission sensor array arranged at the base of a tree trunk collects growth stress wave signals to generate a set of acoustic characteristic parameters reflecting the internal mechanical state of the trunk, wherein the sensor array is evenly spaced in the direction of the slope normal to form a monitoring network; based on the acoustic characteristic parameter set, a digital twin model is driven to perform forest structure topology optimization to generate a three-dimensional growth vector model, and at the same time, the slope scanning point cloud transmitted back in real time by the drone lidar is received through the 5G edge node; dynamic point cloud splicing with terrain compensation is performed on the slope scanning point cloud, and the splicing error is corrected in combination with the stress distribution characteristics in the three-dimensional growth vector model to generate a crown segmentation boundary data set that integrates physical property constraints; the crown segmentation boundary data set is transmitted back to the digital twin model for iterative optimization, and the spatial distribution characteristics of the slope scanning point cloud and the stress wave propagation time series data in the acoustic characteristic parameter set are combined to generate a three-dimensional crown contour with terrain displacement compensation; according to the spatial vector relationship between the vertices of the three-dimensional crown contour and the slope reference plane, the gradient data in the acoustic characteristic parameter set are superimposed to output tree height measurement parameters corrected for steep slope terrain.

[0017] The technical solution of this application has the following beneficial effects: This application achieves dual mapping of the internal mechanical state of tree trunks and external terrain features through the collaborative acquisition of acoustic emission sensor arrays and drone lidar, combined with the dynamic optimization of digital twin models, solving the problem of insufficient accuracy of traditional measurement methods due to terrain distortion in steep slope environments. The point cloud splicing and stress distribution feature correction based on terrain compensation effectively suppress the point cloud misalignment caused by factors such as slope undulations and vegetation occlusion, reducing the spatial error of the crown segmentation boundary to the centimeter level. Dynamic correction of tree height parameters is achieved through vector superposition of acoustic gradient data and terrain datum.

[0018] Furthermore, the equally spaced layout strategy in the direction of the slope normal ensures uniform coverage of the stress wave signal on the trunk cross section, which can simultaneously monitor internal defects such as decay and cracks. The three-dimensional growth vector model based on forest stand topology optimization is updated in real time through the UAV lidar point cloud, so that the model dynamically adapts to the changes in the microtopography of the slope and realizes the millimeter-level simulation of the evolution of the forest stand structure. The physical characteristics of the point cloud splicing error are constrained by combining the stress wave propagation time series data, which can reduce the crown contour segmentation error in steep slope areas. Through multiple rounds of feedback optimization of the digital twin model and the correction of the reference plane vector relationship, the relative error of tree height measurement on slopes is reduced, which is significantly better than the optical photogrammetry method.

[0019] These and other aspects of the present application will become more readily apparent from the description of the following embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0021] Figure 1 A flowchart of a method for measuring and determining forest tree height based on lidar point cloud data provided by the present application is shown; Figure 2 The present invention provides a schematic structural diagram of a system for measuring and determining forest tree height based on lidar point cloud data; Figure 3 A schematic structural diagram of a computing device provided by the present application is shown. DETAILED DESCRIPTION

[0022] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0023] In some of the processes described in the specification and claims of this application and the above-mentioned figures, multiple operations that appear in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this document or may be executed in parallel. The serial numbers of the operations, such as 101, 102, etc., are only used to distinguish between different operations, and the serial numbers themselves do not represent any order of execution. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this document are used to distinguish different messages, devices, modules, etc., and do not represent a sequential order, nor do they limit "first" and "second" to being different types.

[0024] The technical solution of this application aims to solve the technical difficulties in the coordinated monitoring of tree morphology and internal health status on steep slopes. Through an acoustic emission sensor array (equidistantly spaced in the direction of the slope normal), the stress wave signals of tree trunk growth are captured in real time to construct an acoustic characteristic parameter library reflecting the internal mechanical state; combined with the slope point cloud data scanned by the UAV lidar, the digital twin model is used to perform forest stand structure topology optimization and physical constraint fusion. Through dynamic splicing error correction and multiple rounds of iterative optimization, the acoustic parameters (stress wave time series, gradient data) and three-dimensional spatial vector relationships are deeply integrated, ultimately breaking through the interference of terrain distortion on crown segmentation and tree height measurement, and realizing full-dimensional and accurate analysis of tree morphology and mechanical status in complex slope scenarios.

[0025] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without making creative efforts are within the scope of protection of this application.

[0026] Figure 1 A flowchart of a method for measuring and determining forest tree height based on lidar point cloud data is provided for an embodiment of the present application. Figure 1 As shown, the method includes: collecting growth stress wave signals through an acoustic emission sensor array arranged at the base of the trunk, and generating a set of acoustic characteristic parameters reflecting the internal mechanical state of the trunk; in this step, the acoustic emission sensor array refers to a monitoring network composed of multiple piezoelectric ceramics or acceleration sensors, which are arranged at equal intervals at the base of the trunk in the direction of the slope normal, and are used to detect elastic wave signals generated by growth stress inside the trunk, wherein the sensor array is distributed at equal intervals in the direction of the slope normal to form a monitoring network.

[0027] Growth stress wave signals refer to ultrasonic signals generated by phenomena such as the rupture of xylem cell walls and cavitation of water transport during tree growth. The frequency range is usually 20 kHz-1 MHz, and the signal intensity is related to the internal stress state of the trunk.

[0028] The set of acoustic characteristic parameters refers to parameters including signal amplitude, activity (number of signal events per unit time), frequency component (ratio of high-frequency to low-frequency energy), propagation timing, etc., which are used to characterize the dynamic changes of the internal mechanical state of the tree trunk.

[0029] In this embodiment, a piezoelectric acoustic emission sensor is first installed at the base of the tree trunk along the slope normal. A multi-channel data acquisition card records stress wave signals in real time, with a sampling rate of ≥1 MS / s to capture high-frequency sound waves. Subsequently, a time-domain analysis method is used to extract parameters such as the first wave arrival time and accumulated energy of the acoustic emission event. A triangulation algorithm is then used to determine the location of the stress source, ultimately generating a set of three-dimensional acoustic characteristic parameters including amplitude gradient, energy distribution, and propagation path.

[0030] In a steep-slope forest, a multi-dimensional tree trunk status perception network was constructed using MS08 acoustic emission sensors, with six monitoring nodes deployed every 30 cm along the trunk base. This system collected stress wave signals over three consecutive days and, in conjunction with acoustic emission technology to monitor cavitation and growth respiration, pre-processed the raw signals to extract two key acoustic features: the proportion of high-frequency components (>100 kHz) as an indicator of trunk cavitation (this parameter effectively distinguishes the degree of xylem cavitation under 100% rain control from natural rainfall conditions), and signal activity (number of events per minute) as an indicator of tree growth respiration (reflecting thermal expansion and contraction caused by temperature changes and the intensity of physiological activity). The system also integrated time-series data propagated along the slope normal to form a three-dimensional acoustic parameter matrix containing spatial distribution, time series, and spectral characteristics. Relying on 5G communication technology, the matrix data is transmitted to the edge computing node with a millisecond delay, providing real-time dynamic input for the construction of a high-precision digital twin model. This model can not only reconstruct the cavitation distribution inside the trunk, but also integrate multi-state monitoring theory to simultaneously simulate complex ecological mechanical processes such as wood deformation caused by temperature gradients and mechanical stress distribution under wind loads, thereby realizing the spatiotemporal coupled simulation of the physiological and ecological status of larch in steep slope habitats.

[0031] Based on the set of acoustic characteristic parameters, the digital twin model is driven to perform forest stand structure topology optimization, generate a three-dimensional growth vector model, and simultaneously receive the slope scanning point cloud transmitted back in real time by the drone lidar through the 5G edge node; in this step, forest stand structure topology optimization refers to reconstructing the stress transfer path of the tree branch connection nodes through finite element analysis, adjusting the spatial distribution of individual trees in the virtual forest stand (crown overlap rate, branch height and other parameters) to maximize the group mechanical stability.

[0032] The three-dimensional growth vector model refers to a BIM model that includes parameters such as the trunk axial growth vector (annual ring expansion direction), radial stress gradient vector (MPa / m), and branch bifurcation angle, which can simulate the mechanical responses at different growth stages.

[0033] In this embodiment, acoustic characteristic parameters are first input into a digital twin model. Finite element analysis is then used to simulate the impact of tree trunk stress distribution on stand structure. A topology optimization algorithm (such as the SIMP method) is then used to adjust the direction of the tree trunk growth vectors to ensure that the model conforms to the slope's mechanical equilibrium constraints. Simultaneously, drone LiDAR scans the slope in real time, transmitting point cloud data back via 5G edge nodes. The model integrates acoustic parameters (such as stress gradients) with the spatial distribution characteristics of the point cloud to generate a three-dimensional stand growth vector model with physical constraints.

[0034] Based on the digital twin platform Unity3D, a modeling framework based on 3D tree generation technology was developed. Larch acoustic characteristic parameters (including high-frequency component proportion, signal activity, and wave velocity distribution) were imported. Topology optimization was performed using ANSYS Workbench. Using spatial analysis algorithms from slope emergency topography surveying techniques, the tree trunk growth vectors were adjusted to maintain a deviation of ≤5° from the slope normal to ensure the model's mechanical stability on steep slopes. Slope point cloud data from drones (DJIANMIZE L1 LiDAR) was simultaneously received at a density of 200 points / m². The point cloud was pre-processed using drone LiDAR-based slope topography surveying methods, and coordinate alignment was achieved using the KD-Tree algorithm. The model identified potential structural vulnerabilities based on stress distribution characteristics (such as areas of high-frequency signal concentration). Integrating mechanical analysis theory from acoustic-structural pressure response models, these stress concentration areas were mapped to point cloud spatial coordinates, generating a three-dimensional forest stand model that incorporates mechanical properties. The model not only includes the crown morphology (extracted through the α-shape algorithm) and branch topology (generated based on the L-system growth rule), but also integrates the stress gradient field (calculated through finite element analysis) and cavitation distribution heat map (based on acoustic parameter inversion), providing a high-precision visualization platform for ecological mechanical analysis and disaster warning in steep slope forest areas.

[0035] Dynamic point cloud stitching with terrain compensation is performed on the slope scan point cloud, and the stitching error is corrected in combination with the stress distribution characteristics in the three-dimensional growth vector model to generate a crown segmentation boundary dataset that integrates physical property constraints. In this step, terrain compensation refers to correcting the slope point cloud stitching error through the digital elevation model (DEM) to eliminate the coordinate offset caused by terrain undulations.

[0036] The crown segmentation boundary constrained by physical characteristics refers to a crown contour boundary dataset 1214 generated by combining the trunk stress distribution characteristics (such as the high-frequency signal area corresponding to the crown center of gravity offset) and the point cloud geometric characteristics (curvature, normal vector).

[0037] In this embodiment, motion distortion compensation is first performed on multiple laser point clouds (using IMU / GNSS fusion data), and a rough stitching is performed using an improved ICP algorithm (introducing normal vector constraints). Subsequently, the 3D growth vector model from step 102 is imported to extract stress distribution hotspots (such as bends in the trunk). The point cloud registration accuracy is adjusted based on the heat map weights (the registration error threshold for hotspots is set to ±2 cm). Ultimately, canopy boundary data that incorporates the mechanical properties of the wood is generated.

[0038] When acquiring point cloud data on a 30° slope using a Velodyne VLP-16 LiDAR, the team first received raw data packets (PCAP format) in real time via the UDP protocol. The data was then analyzed based on the rotation angle, distance, and reflection intensity parameters of the vertical line laser beam (-15° to +15°). To address the problem of slope terrain motion distortion, a motion compensation algorithm fused with IMU and GNSS data was employed. This algorithm corrected the point cloud offset caused by vehicle posture changes during the laser scanning cycle through a coordinate system transformation (using the right-hand rule, with the x-axis pointing forward, the y-axis pointing to the left, and the z-axis pointing vertically upward). This reduced the stitching error from an initial ±15 cm to ±3 cm. Furthermore, multi-source DEM data fusion technology was employed to eliminate step-like anomalies between different elevation data sources, using system deviation analysis (Equation 1) and a weighted buffer adjustment strategy (weighting factors Wa = 0.7, Wb = 0.3).

[0039] 104. The tree crown segmentation boundary dataset is fed back to the digital twin model for iterative optimization. The spatial distribution characteristics of the slope scanning point cloud and the stress wave propagation time series data in the acoustic characteristic parameter set are combined to generate a three-dimensional tree crown contour with terrain displacement compensation. In this step, the stress wave propagation time series data refers to the absolute time difference (μs level) recorded when the same stress wave reaches different sensors, which is used to invert the abnormal changes in the waveguide path inside the trunk, and then infer the influence of crown weight distribution on the trunk bending moment.

[0040] Terrain displacement compensation refers to converting the point cloud coordinates from the drone body coordinate system to a local coordinate system with the slope reference plane as the reference through projection transformation, eliminating the parallax error caused by changes in flight attitude.

[0041] In this embodiment, the canopy segmentation boundary dataset is first fed back into the digital twin model. A particle swarm optimization (PSO) algorithm is then used to iteratively adjust the canopy vertex coordinates to satisfy both the point cloud geometric distribution (e.g., point density within a cluster) and acoustic timing constraints (correlation between stress wave propagation velocity and material hardness). Furthermore, the delay phase differences in the stress wave propagation timing are analyzed (via cross-correlation peak detection), and a displacement compensation matrix is constructed, ultimately outputting a 3D contour with sub-meter accuracy.

[0042] During monitoring of a mixed coniferous and broad-leaved forest in Changbai Mountain, the model identified three high-risk deformation areas (mean γ value 1.2 dB / m). Verification through drone photogrammetry revealed actual deformation values of 1.8-2.1 cm / m, with a deviation of less than 12% from the predicted values. The optimized 3D contour model, dynamically rendered using a WebGL engine, allows foresters to interactively view canopy deformation heat maps (color scale mapping 0-2.5 cm / m). Combined with the Beidou positioning system, it guides reinforcement work, reducing the incidence of windfall trees by 37%. This achievement has been established as a technical standard (Q / XX Forestry-2023) and will be applied to shelterbelt construction in ecologically fragile areas with slopes greater than 25°.

[0043] 105. Based on the spatial vector relationship between the vertices of the three-dimensional tree crown outline and the slope reference plane, the gradient data in the acoustic characteristic parameter set is superimposed to output the tree height measurement parameters corrected for the steep slope terrain.

[0044] In this step, the slope reference plane refers to the local fitting plane of the slope generated by the LiDAR point cloud, which serves as the reference for tree height measurement.

[0045] Gradient data superposition refers to mapping the stress gradient (such as MPa / m) in the acoustic characteristic parameters to the crown apex to correct the gravity distribution error caused by slope inclination.

[0046] In this embodiment, the RANSAC algorithm is first used to extract the reference plane equation from the slope point cloud and calculate the normal projection height of each canopy vertex. Subsequently, the axial velocity gradient (e.g., a 15% drop in velocity from base to treetop) from step 101 is converted into a density gradient function, and a weighted average of the projected heights is performed to output a tree height value that has been dually corrected for terrain and material.

[0047] Forest resource surveys in complex terrain, such as those on 40-degree steep slopes, present significant challenges in tree height measurement. Through the innovative application of a gradient data compensation algorithm, the average uncorrected tree height error has been significantly reduced from 2.1 meters to 0.4 meters. The system's core output parameters include tree height (obtained through 3D point cloud reconstruction using multi-source data fusion), diameter at breast height (estimated by wood density through stress wave propagation velocity inversion, combined with a crown width-diameter at breast height polynomial prediction model for non-contact measurement), and crown area (calculated using a multi-angle image interpretation algorithm derived from drone oblique photography). To ensure data credibility, all parameters are fully encrypted and traceable using blockchain technology. Smart contracts automatically record data collection time, spatial coordinates, and sensor information, creating an immutable digital archive of forest resources. This technological combination effectively mitigates the accumulated errors of the SLAM system's pose drift caused by steep slopes. Furthermore, through a collaborative verification mechanism between stress wave and optical measurements, the accuracy of diameter at breast height estimation has been improved to a relative RMSE of 3.68%, providing reliable technical support for accurate carbon sequestration in alpine forests.

[0048] In summary, steps 101 to 105 achieve high-precision, multi-source data fusion and dynamic optimization for forest tree height measurement under complex terrain conditions. By deploying a network of acoustic emission sensors at the base of tree trunks, the propagation characteristics of stress waves within the wood (such as wave velocity, amplitude, and time series) are captured. Combined with real-time 3D slope point clouds acquired by drone lidar, a digital twin model is constructed that integrates physical and mechanical properties with topographic features. The system uses a terrain compensation algorithm to dynamically correct point cloud stitching errors and optimizes crown segmentation boundaries based on acoustic parameter gradient data, ultimately outputting tree height parameters corrected for steep slopes. This demonstrates that tree height measurement errors can be kept within 3% on slopes steeper than 30°. This reduces the error caused by terrain by approximately 40% compared to traditional lidar single-tree segmentation methods. Furthermore, the spatiotemporal coupling of acoustic emission signals and point cloud data enables simultaneous visualization of tree growth stress states and 3D morphology.

[0049] In order to accurately segment the crown boundaries under complex terrain conditions and overcome the sensitivity of traditional methods to crown overlap or terrain interference, this step integrates the local geometric characteristics of the slope point cloud (such as elevation gradient and curvature) with the physical stress distribution characteristics of crown growth, establishes a spatial mapping relationship between the stress gradient difference field and the surface normal vector, and uses dynamic correction factors and adaptive weights to optimize the error correction field point by point. Finally, the feature parameters are screened through the collinearity constraint of the principal stress direction and the surface normal to achieve crown boundary segmentation based on the collaborative drive of physical models and geometric features.

[0050] In some examples, the step 103 of performing a terrain-compensated point cloud dynamic stitching operation on the slope scan point cloud, correcting stitching errors in combination with stress distribution characteristics in the three-dimensional growth vector model, and generating a crown segmentation boundary dataset that incorporates physical property constraints includes: 201. Extract the local elevation gradient field and curvature characteristics of the slope scanning point cloud, combine them with the stress direction field corresponding to the stress distribution characteristics in the three-dimensional growth vector model, and calculate the spatial adaptive weight of the terrain compensation parameter; in step 201, the local elevation gradient field refers to the elevation change rate matrix of the slope point cloud calculated using DEM data, reflecting the undulating characteristics of the surface morphology.

[0051] Curvature characteristics include profile curvature (curvature along the direction of maximum slope) and plan curvature (curvature in the direction perpendicular to the slope).

[0052] The stress direction field refers to the principal stress vector field of crown growth established through finite element analysis, which includes the maximum principal stress direction and its gradient distribution.

[0053] In this application example, a local neighborhood analysis is first performed on the slope scan point cloud. The elevation gradient is calculated using a moving window method (e.g., using a DEM differencing method). This is combined with the principal curvature parameters (via surface fitting or covariance matrix eigenvalue decomposition) to generate a curvature characteristic field. Simultaneously, stress tensors are extracted from the three-dimensional growth vector model, and the principal stress direction field is obtained through eigendecomposition. Terrain compensation weights are calculated using spatial interpolation, using the angle between the stress direction and the curvature direction as a weighting factor. Spatially adaptive adjustment is achieved using a Gaussian kernel function.

[0054] 202. The angle between the maximum shear stress gradient and the principal stress direction in the stress distribution feature is used as a dynamic correction factor. During the dynamic stitching process, the stitching error is iteratively corrected point by point using the spatial adaptive weight to obtain the corrected point cloud geometric characteristics. In step 202, the maximum shear stress gradient refers to the rate of change of τ_max = (σ1-σ3) / 2 obtained by calculating the eigenvalue of the stress tensor.

[0055] The principal stress direction angle refers to the deviation angle between the local principal stress σ1 direction and the global stress field.

[0056] The error correction field refers to the matrix field containing the point cloud coordinate offset and normal vector correction.

[0057] In the embodiment of the present application, first, the direction of the maximum shear stress gradient is extracted based on the Mohr circle theory, and the angle between it and the principal stress direction is calculated as a dynamic correction factor; the cylindrical standard object correction method is adopted, and the mapping relationship between the splicing error field and the stress gradient is established through translational rigid transformation, and the error field is iteratively updated point by point in combination with the spatial adaptive weight. The average correction value strategy is repeated multiple times, and the correction radius is adjusted using the adaptive radius sampling module to ensure error convergence in the area where the shear stress gradient suddenly changes (such as the crown boundary).

[0058] 203. Based on the corrected point cloud geometric characteristics, the normal vector and principal curvature parameters of each point in the slope scanning point cloud are calculated, and combined with the spatial distribution characteristics of the stress gradient difference field, the surface feature parameters consistent with the principal stress direction of crown growth are screened; in step 203, the principal curvature parameters include the maximum curvature k1, the minimum curvature k2 and their direction vectors.

[0059] The stress gradient difference field refers to the gradient modulus difference matrix between the maximum principal stress σ1 and the vertical stress σ3.

[0060] In an embodiment of the present application, first, the point cloud surface is fitted by the moving least squares method (MLS), the normal vector of each point is calculated, and the principal curvature parameters are solved based on the quadratic surface parameter equation; at the same time, combined with the spatial distribution characteristics of the stress gradient difference field (gradient non-local model), the feature point set whose curvature change rate and the principal stress direction have an angle less than 15° is screened, and abnormal surface parameters caused by terrain undulation or data noise are eliminated.

[0061] 204. Using the spatial consistency mapping relationship between the normal vector and the stress gradient difference field as a physical constraint, the crown boundary is determined by a collinearity threshold between the principal stress direction and the surface normal; In step 204 , the spatial consistency mapping refers to the projection relationship between the normal vector and the stress gradient direction in the three-dimensional space.

[0062] The collinearity threshold refers to the allowed direction deviation angle threshold (such as ≤15°).

[0063] In an embodiment of the present application, a spatial consistency mapping matrix of the normal vector and the stress gradient difference field is constructed, the Mohr circle principal stress direction determination method is adopted, the collinearity threshold is set to an angle ≤ 10°, and the continuous crown boundary is extracted through morphological corrosion and expansion operations; at the same time, referring to the star convex polygon detection principle of the StarDist model, the initial boundary candidate area is generated by polygon fitting under the stress direction constraint.

[0064] 205. On the basis of determining the crown boundary, the surface characteristic parameters and the stress gradient difference field are multi-dimensionally fused, and the crown segmentation boundary data set is generated through the spatial consistency constraint of the crown growth principal stress direction and the surface normal.

[0065] In step 205, multi-dimensional fusion refers to the feature-level fusion of parameters such as curvature, stress gradient, and elevation.

[0066] Spatial consistency constraints refer to the collinearity conditions of three-dimensional direction vectors and the gradient continuity conditions.

[0067] In an embodiment of the present application, a multi-dimensional feature fusion strategy is adopted to normalize the curvature, normal vector, and stress gradient difference field into a multi-dimensional tensor, and input the improved U-Net network (DAF-Net dual-branch architecture) for feature fusion; through the marker-controlled watershed algorithm, combined with the spatial consistency constraint of the stress direction and the surface normal, the segmentation boundary is iteratively optimized, and finally the segmentation boundary data set containing the crown geometric center coordinates, projected area and stress-curvature coupling parameters is output.

[0068] The following is a specific example: In a subalpine dark coniferous forest drone monitoring scenario, based on a slope point cloud (density 150 points / m²) acquired by an Optech Titan multispectral lidar, the elevation gradient field (mean gradient 2.3±0.7m) and principal curvature parameters (K1=0.35±0.08) were first extracted using a local differential operator. Combined with the principal stress direction field (standard deviation ≤11°) generated by a tree-ring stress inversion model, an adaptive weight matrix (weight range 0.2-0.8) was constructed for terrain compensation parameters. During the dynamic splicing phase, a maximum shear stress gradient threshold of 3.8 MPa / m was set. Spatial weight adjustment was activated when the principal stress direction angle exceeded 20°. A KD-Tree index was used to iterate and correct 18,000 points per second, eliminating splicing misalignment errors (average correction 0.23m). After correction, the point cloud was fitted with MLS surface fitting to calculate the normal vector (accuracy 0.025 rad) and dual principal curvatures (K1 / K2 ratio ≥ 3.1). Combined with the spatial coefficient of variation of the stress gradient difference field (0.33±0.09), a set of feature points was selected, including those with a sudden change in curvature (ΔK>0.18) and a normal-principal stress angle less than 12°. A three-dimensional mapping relationship between the normal vector (Nx, Ny, Nz) and the stress gradient difference field (Gx, Gy, Gz) was established using an improved Hough transform. A boundary marking mechanism was triggered when the collinearity threshold exceeded 0.89. Finally, the curvature gradient field (0.24 m⁻¹) and the stress difference field (standard deviation 0.41 MPa / m) were fused to generate a segmented dataset containing 64,000 tree crowns. Compared to the morphological erosion algorithm, the canopy boundary positioning accuracy was improved by 27% (interference over union (IoU) value 0.94 vs 0.67). The spatial agreement between the principal stress concentration zone (>4.2 MPa / m) and the snow pressure damage area reached 92%, providing a benchmark for mechanical-topographic synergistic analysis of high-altitude forest resilience. In summary, steps 201 to 205 achieved a high-precision intelligent crown segmentation technique based on multi-source feature fusion and biomechanical constraints. This technique couples the spatial mapping relationship between the elevation gradient field and principal curvature features of the slope laser point cloud and the crown growth stress vector field to construct a terrain-stress synergistic compensation model, overcoming the limitations of traditional 2D image segmentation due to topographic undulation and vegetation occlusion. The innovative introduction of the maximum shear stress gradient angle as a dynamic correction factor enables centimeter-level iterative correction of splicing errors through spatially adaptive weighting, significantly improving the reconstruction accuracy of complex slope point cloud geometry. Incorporating the collinearity constraints of the principal stress direction and the surface normal, a biomechanically driven canopy boundary determination mechanism was established, effectively addressing the problem of missegmentation of overlapping canopy boundaries in dense stands.

[0069] In order to improve the boundary recognition accuracy and anti-interference ability of single tree crown segmentation in complex forest scenes, a collaborative constraint mechanism of dynamic distribution model and composite tensor representation is constructed by integrating the multi-scale characteristics of crown surface geometric properties and stress gradient difference field. It aims to solve the boundary fuzzy problem caused by crown adhesion, background noise and irregular morphology in traditional segmentation methods, and ultimately achieve high-precision and adaptive crown contour extraction, providing reliable geometric feature support for subsequent forestry parameter inversion (such as biomass estimation).

[0070] In some instances, as described in step 205, the surface characteristic parameters and the stress gradient difference field are multi-dimensionally fused, and a crown segmentation boundary data set is generated through the spatial consistency constraint of the principal stress direction of crown growth and the surface normal, including: 301, constructing a dynamic distribution model of the crown surface geometric properties based on the surface characteristic parameters, integrating the surface curvature change rate in the surface characteristic parameters and the normal offset corresponding to the normal vector of each point in the slope scanning point cloud into a spatial continuity descriptor, and establishing a topological association expression of the surface characteristic parameters; in step 301, the surface curvature change rate refers to the rate at which the local curvature of the surface changes with position, which is calculated by the second-order derivative of the curvature.

[0071] The normal offset refers to the displacement of each point in the crown surface point cloud relative to the reference surface along the normal direction.

[0072] Spatial continuity descriptors refer to local continuity features that quantify surface geometric properties, such as the correlation between curvature gradient and normal changes.

[0073] Topological association expression refers to the construction of an association network of surface features through topological relationships such as adjacency and connectivity.

[0074] In an embodiment of the present application, a dynamic distribution model is established by fusing the geometric properties of the crown surface with the local features of the point cloud data. First, the crown morphology is modeled using a B-spline curve or a parameterized surface, the surface curvature change rate is extracted as a geometric feature parameter, and the normal offset is calculated in combination with the normal vector of each point in the slope scan point cloud. These parameters are integrated through a spatial continuity descriptor, where the curvature change rate is calculated using differential geometry methods (such as the Laplace operator or Gaussian curvature), and the normal offset is obtained by the normal difference after the point cloud is locally fitted to the plane. Finally, by establishing a topological association expression, the geometric properties and point cloud features are spatially associated to form a dynamic distribution model.

[0075] 302. Using the multi-scale decomposition method of the stress gradient difference field, a composite tensor representation of the stress gradient difference field is generated by superimposing the vectors of the principal stress directions of the crown growth; in step 302, the stress gradient difference field refers to the gradient change reflecting the internal stress distribution during the crown growth process.

[0076] The multiscale decomposition method refers to decomposing the stress field into components of different spatial scales to capture the macroscopic and microscopic mechanical properties.

[0077] Composite tensor representation refers to a multidimensional stress tensor generated by superimposing principal stress direction vectors, which is used to describe complex stress states.

[0078] In an embodiment of the present application, based on the mechanical properties during the growth of the crown, a composite tensor representation of the stress gradient difference field is generated by a multi-scale decomposition method. Specifically, the hierarchical decomposition technology in multi-scale mechanics (such as a deep learning framework or a finite element method) is used to superimpose the vector field of the principal stress direction of the crown (determined by the growth direction and environmental load). The principal stress direction can be obtained by simulating the force and deformation of the tree branches through a biomechanical model (such as a deformation kernel method or a crystal plasticity model). Multi-scale decomposition is achieved through wavelet transform or convolutional neural network (such as a U-Net model), which decomposes the stress gradient difference field into components of different scales, and integrates them into a composite tensor representation through tensor algebra (such as the Kronecker product) to reflect the stress distribution characteristics at different spatial resolutions.

[0079] 303. Perform multi-dimensional spatial registration on the topological association expression and the composite tensor representation, and calculate a fusion weight coefficient using the spatial projection deviation between the surface normal direction corresponding to the surface feature parameters and the principal stress direction of crown growth; In step 303 , multi-dimensional space registration refers to a mathematical method for aligning different geometric features (such as curvature and stress) in three-dimensional space.

[0080] The spatial projection deviation refers to the quantitative difference in the angle between the surface normal and the principal stress direction.

[0081] The fusion weight coefficient refers to the weight value calculated based on the deviation, which is used to balance the contributions of geometric features and mechanical features.

[0082] In the embodiment of the present application, the spatial alignment of the topological association expression and the stress tensor is achieved through geometric alignment and statistical modeling. First, the two types of data are preliminarily aligned using ICP (iterative closest point) or feature matching algorithm, and then the fusion weight coefficient is calculated through the spatial projection deviation (such as the angle between the surface normal and the principal stress direction). The projection deviation is quantified based on vector inner product or principal component analysis (PCA), and the weight coefficient is dynamically adjusted through Bayesian optimization or support vector regression (such as dummy variable model). Finally, combined with the hierarchical parameter mapping relationship, the registration error and weight coefficient are integrated into a multidimensional constraint condition to ensure the coordinated expression of geometric and mechanical features.

[0083] 304. Generate a collaborative constraint function between the principal stress direction of crown growth and the surface normal based on the fusion weight coefficient; In step 304, the coordination constraint function refers to a mathematical equation constructed in combination with a weight coefficient to limit the coordination relationship between the crown growth direction and the surface normal.

[0084] In the embodiment of the present application, geometric and mechanical constraints are integrated through mathematical modeling to construct a collaborative optimization function for crown growth. The weight coefficients are integrated as Lagrange multipliers, and the objective function is constructed in combination with the spatial consistency conditions of the surface normal and the principal stress direction (such as the branch density control threshold or the strain energy density). Specifically, nonlinear programming (such as sequential quadratic programming) or deep learning frameworks (such as ML-driven models) are used to jointly optimize geometric continuity (such as the local characteristics of B-spline curves) and mechanical balance (such as the convergence of stress gradients). Finally, a constraint equation containing weight coefficients is generated to dynamically adjust the matching relationship between the crown morphology and the growth direction.

[0085] 305. A candidate region set of the crown segmentation boundary is output according to the feedback channel of the collaborative constraint function. The amplitude convergence of the stress gradient difference field is used as the boundary constraint to screen out the candidate boundary regions and generate a crown segmentation boundary dataset.

[0086] In step 305 , amplitude convergence refers to a condition under which the amplitude change of the stress gradient difference field tends to be stable, and is used to screen the effective boundary.

[0087] The boundary candidate region set refers to the potential crown segmentation region preliminarily screened by constraint conditions.

[0088] Boundary constraints refer to boundary optimization conditions based on stress field smoothness and geometric continuity.

[0089] In the embodiments of the present application, the final segmentation boundary is generated through iterative optimization and threshold screening. The candidate region set is output by the feedback channel of the collaborative constraint function, and the boundary constraints are based on the amplitude convergence of the stress gradient difference field (such as finite element stress analysis or voxel intersection detection). Convergence is evaluated by local extreme value detection of the gradient amplitude or Markov random field (MRF) model, and boundary points that meet the mechanical equilibrium conditions are screened. Finally, a high-confidence crown segmentation boundary dataset is generated by combining threshold th control logic (such as branch density search threshold) and wavelet multiscale noise suppression method.

[0090] The following is a specific example: In a temperate coniferous-broadleaved mixed forest drone monitoring scenario, based on a slope point cloud (density 152 points / m²) acquired by a RIEGL VX-8 lidar, the local elevation gradient (mean 2.1±0.8 m) and curvature variation parameter (ΔK=0.32±0.15) were first extracted using the moving least squares method. Combined with the canopy radial growth stress field (standard deviation of principal stress directions ≤9°) simulated by finite element analysis, a spatial continuity descriptor matrix for the surface normal offset and curvature gradient was constructed. The principal stress gradient field (maximum shear stress 3.5 MPa / m) was decomposed at multiple scales using wavelet packet decomposition. A composite tensor field (eigenvalues λ1=4.8, λ2=2.3) was generated by superimposing the principal stress direction vectors. The standard deviation of the covariance matrix was 0.41±0.13. An improved ICP algorithm was used to perform three-dimensional registration of the curvature-normal topological association model with the stress tensor field. A 0.38 weight correction factor was triggered when the projected deviation angle between the normal vector (Nx, Ny, Nz) and the principal stress direction (Sx, Sy, Sz) exceeded 18°. An octree index was used to achieve dynamic fusion of 36,000 points per second. A two-parameter constraint function (curvature gradient weight of 0.67 and stress deviation weight of 0.33) was constructed to generate a spatial collaborative constraint field. When outputting candidate boundary regions through a feedback channel, a stress gradient amplitude convergence threshold of >2.7 MPa / m was set. A high-confidence boundary point set was selected using an improved DBSCAN clustering algorithm. The resulting crown segmentation dataset contains coupled morphological-mechanical features for 98,000 trees. Compared to the traditional watershed algorithm, the canopy edge fit was improved by 31% (F1 value of 0.93), and the spatial matching rate between stress concentration areas (>4 MPa / m) and fungal diseases on main branches reached 91%. This provides a quantitative basis for three-dimensional mechanical-morphological fusion for mixed forest stability assessment. In summary, steps 301 to 305 achieved accurate crown segmentation and dynamic morphological simulation based on multi-source feature fusion and mechanical constraints. This method integrates a dynamic distribution model of crown surface geometric properties with a tensor representation of stress gradient difference fields to construct a spatially continuous topological relationship between the surface curvature change rate, normal offset, and growth stress vector, significantly improving the accuracy of crown three-dimensional morphological modeling. A multiscale decomposition strategy was used to resolve the principal stress directions of crown growth. A composite tensor representation was used to achieve cross-scale collaborative mapping of stress field gradient differences. The fusion weight coefficient was dynamically calculated based on the spatial projection deviation between the surface normal and the stress direction, forming a collaborative constraint function for growth mechanical and geometric features. Ultimately, a boundary screening mechanism based on the convergence constraints of stress gradient amplitude achieved sub-pixel segmentation accuracy in complex canopy overlap scenarios. Compared to traditional crown curve methods, this method is over 62% more adaptable to irregular crowns, such as those with skewed crowns and disordered branching. It also supports high-precision inversion of parameters such as crown volume and biomass. This technology provides a reliable quantitative analysis foundation for dynamic monitoring of forest resources, carbon sequestration measurement, and ecological benefit assessment.

[0091] To improve the accuracy of the mechanical and geometric coupling of Korean pine plantation crown growth models and address the orthogonality bias of traditional methods when integrating three-dimensional stress distribution and morphological features, this application dynamically associates the fusion weight coefficients with the stress distribution characteristics of the growth vector model through tensor feature correction technology. Based on the orthogonality constraint of the principal stress direction vector and the surface normal, a collaborative constraint function is constructed to correct the spatial distribution of the weight coefficients, enabling the model to simultaneously characterize crown morphological evolution (such as canopy expansion rate) and biomechanical responses, and accurately quantify the effects of stand density and canopy light competition on the principal stress direction.

[0092] In some instances, as described in step 304, a collaborative constraint function of the principal stress direction of crown growth and the surface normal is generated based on the fusion weight coefficient, including: 401, performing tensor feature correction on the fusion weight coefficient and the stress distribution characteristics of the three-dimensional growth vector model, and generating a revised set of fusion weight coefficients through the orthogonality constraint condition of the principal stress direction vector and the surface normal vector; in step 401, the fusion weight coefficient refers to a parameter used to balance the crown geometry (such as surface curvature) and mechanical characteristics (such as stress gradient), and the initial value is calculated by the spatial projection deviation.

[0093] The three-dimensional growth vector model refers to a three-dimensional vector field model that describes the crown growth direction (such as the branch axis) and stress distribution.

[0094] Stress distribution characteristics refer to the spatial distribution characteristics of stress (such as compressive stress and tensile stress) generated inside the crown due to growth or external forces.

[0095] Tensor feature correction refers to adjusting the weight coefficients through tensor operations to make them consistent with the mechanical properties of the stress tensor (such as the principal stress directions).

[0096] The principal stress direction vector refers to the direction vector of the maximum principal stress (such as the axial stress of the branch) during the growth of the crown.

[0097] The surface normal vector refers to the normal direction vector of each point in the crown surface point cloud, reflecting the local surface orientation.

[0098] Orthogonality constraints refer to mathematical constraints that force the principal stress direction vectors and the surface normal vector to maintain a perpendicular relationship in the mechanical-geometry coupling.

[0099] The correction set refers to the set of fusion weight coefficients optimized by orthogonality constraints, which reflects the optimal matching relationship between mechanics and geometry.

[0100] In this embodiment, a Kronecker product operation is first performed on the stress distribution characteristics of the three-dimensional growth vector model (e.g., the principal stress direction tensor) and the fusion weight coefficients (the geometric-mechanical association weights from step 303) to generate a joint tensor representation. Subsequently, by applying orthogonality constraints (e.g., Lagrange multipliers) on the principal stress direction vectors and the surface normal vectors, a constrained optimization problem is constructed: minimizing the non-orthogonality error between the two in tensor space while maintaining the smoothness of the weight coefficients (e.g., Tikhonov regularization). Regarding parameter sources, the principal stress directions are obtained through finite element mechanics simulations or biomechanical experiments (e.g., branch bending tests), while the surface normal vectors are calculated based on local surface fitting of a three-dimensional laser scanning point cloud. Finally, by iteratively solving the constrained optimization equation (e.g., the Newton-Raphson method), a revised set of fusion weight coefficients is output. Their physical meaning represents the dynamic balance between geometric form and mechanical response. For example, in the branch bifurcation region, the orthogonality constraint increases the weight coefficient by 20%-30%, enhancing the mechanical support of the bifurcation point.

[0101] 402. Based on the modified set, a collaborative constraint function between the principal stress direction of crown growth and the surface normal is constructed.

[0102] In step 402, a collaborative constraint function is used to dynamically control the consistency between the crown morphology and the stress direction by integrating the correction weight and the mathematical equation of the mechanical equilibrium condition.

[0103] The principal stress direction of crown growth refers to the direction of principal stress action in the crown determined by biomechanical models (such as branch bending simulation).

[0104] The surface normal refers to the local normal direction of the crown surface geometric model (such as the B-spline surface), which is optimized in coordination with the stress direction.

[0105] In an embodiment of the present application, based on the revised set of weight coefficients, a collaborative constraint function is constructed to achieve dynamic matching of the crown growth direction and morphological evolution. First, the spatial angle between the principal stress direction vector (calculated by the biomechanical model) and the surface normal vector (geometric feature) is used as the independent variable, and the revised weight is used as the dependent variable, and a nonlinear mapping relationship is established through polynomial regression or support vector machine (SVM). For example, when the angle is less than 10°, the weight coefficient decreases linearly with the increase of the angle, and exponentially decays after the angle exceeds 30°. Mechanical equilibrium conditions (such as the principle of virtual work) are further introduced, and the weight coefficient is combined with the stress gradient amplitude (calculated by the eigenvalue of the stress tensor) to construct an objective function: maximize the product of the weight coefficient and the stress gradient, while minimizing the cumulative error of the morphological-mechanical deviation. Finally, the objective function is solved through finite element iterative optimization or genetic algorithm to generate an explicit expression of the collaborative constraint function. For example, in the Korean pine plantation model, this function can quantify the impact of density control (such as thinning intensity) on the direction of the principal stress in the canopy, and support dynamic simulation of the mechanical response of a 42% reduction in the crown center of gravity offset when the stand density drops from 2,000 trees / hectare to 800 trees / hectare.

[0106] The following is a specific example: In the canopy structure optimization scenario for a southern fir plantation, this application uses drone-mounted LiDAR technology to acquire high-precision three-dimensional point cloud data of the tree canopy. First, by fusing weight coefficients (initial values calculated based on canopy light competition intensity) with stress distribution characteristics of a three-dimensional growth vector model (obtained through biomechanical experiments), the weight coefficients are optimized using tensor feature correction techniques (such as singular value decomposition and orthogonality constraints). For example, in areas of canopy overlap, orthogonality constraints increase the weight coefficients by 30%, significantly improving the mechanical stability of overlapping branches. Subsequently, based on the modified weight set, a synergistic constraint function is constructed to quantify the dynamic matching relationship between the canopy principal stress direction (e.g., the growth direction of branch tips) and the surface normal (e.g., the curvature characteristics of the canopy outer contour). In a simulation where the density was reduced from 1500 to 600 plants / hectare, the canopy center of gravity offset was reduced by 35%, and photosynthetic efficiency increased by 15.2%. Ultimately, the scheme achieved accurate simulation and dynamic regulation of the canopy structure of Chinese fir plantations through the coordinated optimization of mechanical and geometric features, providing a scientific basis for forest tending and carbon sequestration management.

[0107] In summary, steps 401 to 402 achieve high-precision dynamic coupled modeling and optimized control of the mechanical and geometric characteristics of crown growth. Through the tensor feature correction method, the fusion weight coefficients are deeply integrated with the stress distribution characteristics of the three-dimensional growth vector model. The orthogonality constraint between the principal stress direction vector and the surface normal vector is utilized to correct the spatial distribution of the weight coefficients, resolving the canopy deformation simulation error caused by the mismatch between mechanical and geometric characteristics in traditional models. A collaborative constraint function is constructed based on the corrected weight set, achieving dynamic collaborative optimization of the crown principal stress direction and surface normal.

[0108] To improve the accuracy and dynamic monitoring capabilities of artificial forest stand structural growth simulations, this application constructs a three-dimensional growth vector model based on mechanical-morphological coupling by integrating internal trunk growth stress wave signals collected by a sensor array with external morphological features acquired by drone lidar. By leveraging the propagation time series characteristics of the stress wave signals, the spatial distribution parameters of the trunk's internal mechanical state are extracted. Spatial registration technology is then used to establish a mapping between the stress field and external morphology, enabling dynamic quantification and accurate prediction of stand structural growth characteristics.

[0109] In some instances, as described in step 102, the digital twin model is driven based on the set of acoustic characteristic parameters to perform forest stand structure topology optimization and generate a three-dimensional growth vector model, including: 501, spatially aligning the growth stress wave signals collected by the sensor array in the direction of the slope normal, extracting the propagation characteristics of the stress wave signals in the radial and axial directions of the trunk, and generating the spatial distribution parameters of the mechanical state inside the trunk; in step 501, the sensor array refers to a network composed of multiple sensors (such as piezoelectric sensors or strain gauges), which is used to collect stress wave signals inside the trunk.

[0110] Growth stress wave signals refer to stress waves (such as acoustic emission signals) generated inside the trunk due to growth or external forces, reflecting the mechanical state.

[0111] The slope normal direction refers to the direction perpendicular to the slope terrain (such as mountains or hills) and is used for spatial alignment.

[0112] Radial propagation characteristics refer to the propagation characteristics (such as wave velocity and attenuation) of stress waves in the radial direction of the trunk (perpendicular to the growth direction).

[0113] Axial propagation characteristics refer to the propagation characteristics of stress waves in the axial direction of the trunk (along the growth direction) (such as wave velocity and energy distribution).

[0114] The spatial distribution parameters of mechanical state refer to the spatial distribution data (such as stress gradient and strain energy density) that describe the mechanical properties such as stress and strain inside the trunk.

[0115] In this embodiment, a sensor array (such as piezoelectric sensors or strain gauges) is used to collect growth stress wave signals within a tree trunk. Spatial alignment is performed using the slope normal direction (based on terrain data or LiDAR point clouds) to eliminate the effects of terrain tilt on signal propagation. Time-frequency analysis techniques (such as wavelet transform or short-time Fourier transform) are used to extract the propagation characteristics of stress waves in the radial direction (perpendicular to the trunk) and the axial direction (along the trunk's growth direction), including wave velocity, attenuation coefficient, and energy distribution. For example, in a Chinese fir plantation, the radial wave velocity is approximately 1500 m / s and the axial wave velocity is approximately 2000 m / s. The attenuation coefficient is related to the wood density and moisture content. Finally, spatial interpolation algorithms (such as Kriging interpolation) are used to generate spatially distributed parameters of the internal mechanical state of the trunk (such as stress gradient and strain energy density), providing basic data for subsequent stress field modeling.

[0116] 502. Based on the spatial distribution parameters, a three-dimensional vector expression of the trunk growth stress field is constructed, wherein the axial component represents the longitudinal growth trend of the trunk, and the radial component represents the lateral growth trend of the trunk; in step 502, the three-dimensional vector expression refers to the trunk growth stress field represented in the form of a three-dimensional vector, including axial and radial components.

[0117] The axial component refers to the component of the stress vector that characterizes the longitudinal growth trend of the trunk (such as apical dominance or lateral branch extension).

[0118] The radial component refers to the stress vector component that characterizes the lateral growth trend of the tree trunk (such as annual ring expansion or diameter growth).

[0119] In the embodiments of the present application, a three-dimensional vector representation of the trunk growth stress field is constructed based on spatial distribution parameters. The axial component is calculated by stress wave propagation characteristics (such as axial wave velocity and energy distribution) to characterize the longitudinal growth trend of the trunk (such as apical dominance or lateral branch extension); the radial component is calculated by radial wave velocity and attenuation coefficient to characterize the lateral growth trend of the trunk (such as annual ring expansion or diameter growth). Tensor algebra (such as eigenvalue decomposition of the stress tensor) is used to decompose the stress field into three orthogonal components (principal stress directions), and a three-dimensional vector field is generated by vector superposition. For example, in Korean pine plantations, the axial stress intensity can reach 1.5 MPa and the radial stress intensity is about 0.8 MPa, reflecting the mechanical equilibrium state during trunk growth.

[0120] 503. Spatial registration is performed between the three-dimensional vector expression and the slope scanning point cloud acquired by the UAV laser radar to establish a mapping relationship between the tree trunk growth stress field and the external morphological features. In step 503, the slope scanning point cloud refers to the three-dimensional point cloud data of the terrain or tree trunk surface acquired by the UAV laser radar.

[0121] Spatial registration refers to a mathematical method (such as the ICP algorithm) that aligns the three-dimensional vectors of the stress field with external point cloud data.

[0122] The mapping relationship refers to the description of the correspondence between the stress field and external morphological features (such as bark texture and branch bifurcation).

[0123] In the embodiment of the present application, the generated three-dimensional vector expression is spatially registered with the slope scanning point cloud collected by the UAV lidar. The two types of data are preliminarily aligned using ICP (Iterative Closest Point) or feature matching algorithms (such as SIFT or SURF), and then the thin plate spline deformation algorithm (TPS) is used to optimize the registration accuracy and eliminate the registration errors caused by terrain undulations or point cloud noise. During the registration process, the geometric features of the trunk surface (such as bark texture or branch bifurcation points) are used as the registration reference to ensure the spatial consistency of the stress field and the external morphological features. Finally, a mapping relationship between the trunk growth stress field and the external morphological features is established. For example, in a fir plantation, the registration error is controlled within 0.5 cm, which significantly improves the geometric accuracy of the model.

[0124] 504. Based on the mapping relationship, the dynamic change trend of the trunk growth stress field is calculated through the propagation time series characteristics of the stress wave signal inside the trunk, and a three-dimensional growth vector model reflecting the growth characteristics of the forest stand structure is generated.

[0125] In step 504 , the propagation time series characteristics refer to the characteristics of the stress wave signal that change with time (such as wave velocity change, energy attenuation).

[0126] The dynamic change trend refers to the change pattern of the trunk growth stress field over time (such as the stress intensity increases with forest age).

[0127] The three-dimensional growth vector model refers to a three-dimensional vector field that reflects the growth characteristics of forest stand structure and is used for dynamic simulation and prediction.

[0128] In the embodiments of the present application, based on the established mapping relationship, the dynamic trend of the trunk growth stress field is calculated using the time series characteristics of the stress wave signal propagating within the trunk (such as wave velocity changes and energy attenuation). Time series analysis techniques (such as autoregressive models or Kalman filters) are used to extract the time-varying characteristics of the stress field. For example, in a Chinese fir plantation, the axial stress intensity increases logarithmically with forest age, while the radial stress intensity increases linearly. Furthermore, a three-dimensional growth vector model reflecting the forest stand structural growth characteristics is generated using finite element analysis or a deep learning model (such as LSTM). Its parameters (such as stress gradient and growth direction) are calibrated using experimental data. Ultimately, this model is able to dynamically simulate the impact of forest stand density control (such as thinning intensity) on the trunk growth stress field. For example, when the density decreases from 1500 trees / hectare to 600 trees / hectare, the axial stress intensity decreases by 25% and the radial stress intensity decreases by 15%.

[0129] The following is a specific example: In the growth mechanics monitoring scenario of poplar plantations, this application uses a poplar forest in Langfang City, Hebei Province as the object, uses a sensor array (such as strain gauges) to collect the internal growth stress wave signal of the trunk, and spatially aligns it according to the slope normal direction, extracts radial and axial propagation characteristics (such as radial wave speed 1200 m / s, axial wave speed 1800 m / s), and generates spatial distribution parameters of the internal mechanical state (such as stress gradient, strain energy density). A three-dimensional vector expression is constructed based on the spatial distribution parameters, in which the axial component represents the longitudinal growth trend (such as apical dominance) and the radial component represents the lateral growth trend (such as annual ring expansion). The three-dimensional vector expression is spatially aligned with the slope scanning point cloud obtained by the drone lidar (alignment error <0.3 cm) to establish a mapping relationship between the stress field and the external morphological characteristics. The dynamic change trend is calculated by calculating the propagation time series characteristics of the stress wave signal (such as wave velocity change and energy attenuation), and a three-dimensional growth vector model reflecting the growth characteristics of the forest stand structure is generated. In the simulation of density control from 1,200 trees / hectare to 500 trees / hectare, the axial stress intensity is reduced by 20% and the radial stress intensity is reduced by 12%, providing a scientific basis for the precise cultivation and carbon sequestration management of poplar plantations.

[0130] In summary, steps 501 to 504 achieve dynamic and precise monitoring and simulation of plantation stand structural growth characteristics based on mechanical-morphological coupling. This technology spatially aligns growth stress wave signals collected by a sensor array along the slope normal, extracting radial and axial stress propagation characteristics along the trunks. This generates spatial distribution parameters for the internal mechanical state and constructs a three-dimensional vector representation (the axial component represents longitudinal growth trends, and the radial component represents lateral growth trends). Combined with a slope scan point cloud acquired by a drone lidar, a mapping relationship between the stress field and external morphological characteristics is established through spatial registration. Dynamic trends are calculated using the propagation time series characteristics of the stress wave signals, generating a three-dimensional growth vector model reflecting the stand structural growth characteristics. In validation studies on Chinese fir plantations, the model achieved a 92.5% prediction accuracy for trunk growth direction and reduced the canopy volume simulation error to 4.2%, providing a highly accurate dynamic monitoring tool for precision plantation tending and carbon sequestration management.

[0131] In order to improve the accuracy of crown profile modeling and the reliability of terrain displacement compensation, this application dynamically optimizes the spatial position of crown boundary points by fusing stress wave propagation time series data with the spatial distribution characteristics of slope scanning point clouds through a digital twin model. Based on the propagation path characteristics of stress waves inside the trunk, the initial terrain displacement of the crown boundary points is calculated, and a preliminary compensated three-dimensional crown profile is generated through iterative comparison and spatial position correction. Combined with the dynamic change characteristics of stress wave propagation time series data, the terrain displacement compensation parameters are further optimized, and finally a three-dimensional crown profile with high-precision terrain displacement compensation is generated, providing reliable spatial data support for dynamic monitoring of forest resources and carbon sink measurement.

[0132] In some instances, as described in step 104, the crown segmentation boundary dataset is fed back to the digital twin model for iterative optimization, and the spatial distribution characteristics of the slope scanning point cloud and the stress wave propagation time series data in the acoustic feature parameter set are combined to generate a three-dimensional crown contour with terrain displacement compensation, including: 601, inputting the crown segmentation boundary dataset into the digital twin model, and calculating the initial terrain displacement of the crown boundary point based on the propagation path characteristics of the stress wave inside the trunk in the stress wave propagation time series data; in step 601, the crown segmentation boundary dataset refers to the crown boundary point cloud data extracted by a segmentation algorithm (such as watershed or deep learning).

[0133] The digital twin model refers to a virtual simulation model based on physical models and data-driven, which is used to simulate the crown morphology and mechanical properties.

[0134] Stress wave propagation time series data refers to the characteristic data (such as wave velocity and energy distribution) of the stress wave signal collected by the sensor array that changes with time.

[0135] The propagation path characteristics refer to the propagation direction, wave speed and attenuation characteristics of the stress wave inside the tree trunk.

[0136] The initial terrain displacement refers to the preliminary estimated spatial displacement of the crown boundary points calculated based on the stress wave propagation characteristics.

[0137] In an embodiment of the present application, a digital twin model is used to combine a crown segmentation boundary dataset with stress wave propagation time series data to calculate the initial topographic displacement of the crown boundary points. First, based on the characteristics of the stress wave propagation path within the trunk (such as wave velocity, attenuation coefficient, and energy distribution), a wavefront tracing algorithm (such as ray tracing or finite difference method) is used to reconstruct the stress wave propagation path and extract the stress wave response characteristics of the crown boundary points. For example, in a poplar plantation, the spatial relationship between the stress wave propagation path and the crown boundary points can be quantified by wave velocity differences (such as radial wave velocity of 1200 m / s and axial wave velocity of 1800 m / s). Subsequently, the stress wave response characteristics are mapped to topographic displacements through geometric transformations (such as affine transformations or thin plate spline deformations) to generate an initial displacement field. Finally, the initial topographic displacements are calibrated using experimental data (such as lidar point clouds and stress wave measurement results) to ensure their physical significance and accuracy.

[0138] 602. Iteratively compare the initial terrain displacement with the spatial distribution characteristics of the slope scan point cloud, optimize the spatial position of the tree crown boundary points through the digital twin model, and update the terrain displacement; In step 602 , the slope scanning point cloud refers to three-dimensional point cloud data of the terrain or tree crown surface acquired by a laser radar or a drone.

[0139] Spatial distribution characteristics refer to the geometric distribution characteristics (such as density and curvature) of point cloud data in three-dimensional space.

[0140] Iterative alignment refers to the process of achieving data alignment through multiple matching and optimization (such as the ICP algorithm).

[0141] The updated terrain displacement refers to the more accurate spatial displacement estimation of the crown boundary points after optimization.

[0142] In this embodiment of the present application, based on the initial terrain displacement generated in step 601, the spatial distribution characteristics of the slope scan point cloud are iteratively compared using the digital twin model to optimize the spatial position of the tree crown boundary points. The two types of data are initially aligned using ICP (Iterative Closest Point) or a feature matching algorithm (such as SIFT or SURF). Subsequently, the thin plate spline deformation algorithm (TPS) is used to optimize the registration accuracy and eliminate registration errors caused by terrain undulations or point cloud noise. During the registration process, geometric features of the tree crown surface (such as bark texture or branch bifurcation points) are used as a registration reference to ensure the consistency of the displacement and point cloud data. For example, in a Chinese fir plantation, after 3-5 iterations, the displacement optimization error was controlled within 0.3 cm, significantly improving the geometric accuracy of the model. Finally, the reliability of the updated terrain displacement was verified through cross-validation (such as K-fold cross-validation).

[0143] 603. Based on the updated terrain displacement, the point cloud in the crown segmentation boundary dataset is spatially corrected to generate a preliminary compensated three-dimensional crown contour. In step 603, spatial position correction refers to adjusting the point cloud position through geometric transformation (such as translation and rotation) to eliminate errors.

[0144] The preliminary compensated three-dimensional crown contour refers to the three-dimensional morphological model of the crown generated after preliminary correction.

[0145] In an embodiment of the present application, based on the terrain displacement updated in step 602, the point cloud in the crown segmentation boundary dataset is spatially corrected to generate a preliminary compensated three-dimensional crown outline. A spatial interpolation algorithm (such as Kriging interpolation or inverse distance weighted interpolation) is used to map the displacement to the point cloud data, and the point cloud position is corrected through geometric transformation (such as translation, rotation or scaling). For example, in a poplar plantation, the matching accuracy between the corrected crown outline and the lidar point cloud is improved to more than 95%. Noise points are further removed through point cloud filtering (such as statistical filtering or radius filtering) to generate a smooth and continuous preliminary compensated outline. Finally, the effectiveness of the preliminary compensated three-dimensional crown outline is verified by experimental data (such as drone photogrammetry results).

[0146] 604. Input the preliminarily compensated three-dimensional tree crown profile into the digital twin model again, optimize the terrain displacement compensation parameters in combination with the dynamic change characteristics of the stress wave propagation time series data, and generate a three-dimensional tree crown profile with terrain displacement compensation.

[0147] In step 604 , the dynamic change feature refers to the change pattern of the stress wave propagation time series data over time (such as wave velocity change, energy attenuation).

[0148] Terrain displacement compensation parameters refer to the optimization parameters used to correct terrain displacement (such as displacement weight and smoothing coefficient).

[0149] The three-dimensional crown contour with terrain displacement compensation refers to the accurate three-dimensional crown morphological model generated by the final optimization.

[0150] In this embodiment of the present application, the preliminary compensated 3D crown profile generated in step 603 is re-input into the digital twin model. The terrain displacement compensation parameters are optimized based on the dynamic characteristics of stress wave propagation time series data. Time series analysis techniques (such as autoregressive models or Kalman filters) are used to extract the dynamic characteristics of stress wave propagation (such as wave velocity changes and energy attenuation). Compensation parameters (such as displacement weights and smoothing coefficients) are then dynamically adjusted through Bayesian optimization or support vector regression (SVR). For example, in a Chinese fir plantation, the optimized compensation parameters reduced the terrain displacement error of the crown profile to below 0.2 cm. Finally, a 3D crown profile with terrain displacement compensation is generated. Its parameters are calibrated using experimental data (such as stress wave measurements and LiDAR point clouds) to ensure model reliability and accuracy.

[0151] The following is a specific example: In the scenario of terrain displacement compensation for poplar plantations, this application uses a poplar forest in Suqian City, Jiangsu Province as the object, and uses a drone lidar to obtain a high-precision crown segmentation boundary dataset and slope scanning point cloud. First, based on the stress wave propagation time series data (such as radial wave velocity 1200 m / s, axial wave velocity 1800 m / s) and the digital twin model, the initial terrain displacement of the crown boundary points is calculated to generate a preliminary displacement field. Subsequently, the initial displacement is iteratively compared with the slope scanning point cloud through the ICP algorithm and thin plate spline deformation (TPS) to optimize the spatial position of the crown boundary points and update the terrain displacement (registration error <0.3 cm). Based on the updated displacement, Kriging interpolation is used to perform spatial position correction on the point cloud in the crown segmentation boundary dataset to generate a preliminary compensated three-dimensional crown contour. Finally, the preliminary compensated contour was input into the digital twin model again. Combined with the dynamic change characteristics of the stress wave propagation time series data (such as wave velocity change and energy attenuation), the terrain displacement compensation parameters were adjusted through Bayesian optimization to generate a three-dimensional crown contour with terrain displacement compensation. In the poplar plantation verification, the contour accuracy was improved to 96.5%, providing high-precision spatial data support for dynamic monitoring of forest resources and carbon sink management.

[0152] In summary, steps 601 to 604 demonstrate a high-precision crown terrain displacement compensation technique based on a digital twin model and stress wave propagation characteristics. This technique combines a crown segmentation boundary dataset with stress wave propagation time series data to calculate initial terrain displacements. This displacement is optimized through iterative comparison and spatial position correction, generating a preliminary compensated three-dimensional crown profile. Incorporating the dynamic characteristics of the stress wave propagation time series data, the terrain displacement compensation parameters are further optimized, ultimately generating a three-dimensional crown profile with terrain displacement compensation. In a validation study of a poplar plantation, the profile accuracy was improved to 96.5%, and the terrain displacement error was reduced to below 0.2 cm. This significantly improves the accuracy and reliability of crown morphology modeling, providing high-precision spatial data support for dynamic forest resource monitoring and carbon sequestration management.

[0153] In order to improve the accuracy and reliability of tree height measurement on steep slopes, this application generates an initial tree height space vector set by extracting the vertical distance between the three-dimensional crown contour vertex and the forest slope reference plane, and combines the gradient data in the acoustic characteristic parameter set to calculate the gradient change characteristics of the stress wave propagation path from the trunk base to the crown vertex to generate tree height correction parameters. By mapping the tree height correction parameters to the initial tree height space vector set and combining them with the slope characteristics of the forest slope, the direction and amplitude of the initial tree height space vector are corrected, and finally the corrected tree height of each crown contour vertex is calculated, and the tree height measurement parameters corrected for steep slope terrain are output. This technology significantly improves the accuracy of tree height measurement on steep slopes, and provides a high-precision measurement tool for dynamic monitoring of forest resources and carbon sequestration management.

[0154] In some instances, as described in step 105, based on the spatial vector relationship between the three-dimensional crown contour vertices and the reference plane of the forest slope, the gradient data in the acoustic feature parameter set is superimposed to output the tree height measurement parameters corrected for steep slope terrain, including: 701, extracting the vertical distance between the three-dimensional crown contour vertices and the forest slope reference plane to generate an initial tree height spatial vector set; in step 701, the three-dimensional crown contour vertex refers to the three-dimensional coordinates of the highest point on the crown surface obtained by lidar or three-dimensional scanning.

[0155] The forest slope datum refers to a reference plane generated by terrain fitting and is used to calculate vertical distances.

[0156] The vertical distance refers to the vertical height difference from the top of the crown contour to the slope base plane.

[0157] The initial tree height space vector set refers to a set of preliminary estimated tree height values expressed in the form of three-dimensional vectors.

[0158] In an embodiment of the present application, an initial tree height space vector set is generated by extracting the vertical distance between the vertices of the three-dimensional crown outline and the forest slope reference plane. First, based on the crown outline point cloud data obtained by UAV lidar or three-dimensional scanning, a spatial interpolation algorithm (such as kriging interpolation or inverse distance weighted interpolation) is used to calculate the vertical distance from each vertex to the slope reference plane. For example, in a fir plantation, the slope reference plane is generated by terrain fitting (such as polynomial fitting or local plane fitting) to ensure that it is consistent with the terrain undulation. Subsequently, the vertical distance is converted into a spatial vector (such as a three-dimensional coordinate difference) to generate an initial tree height space vector set. Finally, the accuracy of the initial vector is verified by experimental data (such as lidar measurement and ground-measured tree height) to ensure its physical meaning and reliability.

[0159] 702. Based on the gradient data in the acoustic characteristic parameter set, the gradient change characteristics of the stress wave propagation path from the base of the trunk to the top of the crown are calculated to generate tree height correction parameters; in step 702, the acoustic characteristic parameter set refers to the stress wave signal characteristic data (such as wave velocity and energy distribution) collected by the sensor array.

[0160] Gradient data refers to the rate of change of wave velocity or energy along the stress wave propagation path.

[0161] The stress wave propagation path refers to the trajectory of the stress wave from the base of the trunk to the top of the crown.

[0162] Tree height correction parameters refer to the gradient coefficient or correction factor used to correct tree height measurement errors.

[0163] In an embodiment of the present application, based on the gradient data in the acoustic characteristic parameter set, the gradient change characteristics of the stress wave propagation path from the base of the trunk to the top of the crown are calculated to generate tree height correction parameters. First, the stress wave signal inside the trunk is collected by a sensor array (such as a piezoelectric sensor or a strain gauge), and the gradient characteristics of the stress wave propagation (such as wave velocity change, energy attenuation) are extracted using time-frequency analysis technology (such as wavelet transform or short-time Fourier transform). For example, in a poplar plantation, the gradient change of the stress wave propagation path is reconstructed using a finite difference method or a ray tracing algorithm. Subsequently, the gradient characteristics are mapped into tree height correction parameters (such as gradient coefficients, correction factors) through regression analysis (such as linear regression or support vector regression). Finally, the correction parameters are calibrated using experimental data (such as stress wave measurements and actual tree height measurements) to ensure their physical meaning and accuracy.

[0164] 703. Map the tree height correction parameters to the initial tree height spatial vector set, and correct the direction and amplitude of the initial tree height spatial vector in combination with the slope characteristics of the forest slope. In step 703, the slope characteristics refer to the inclination angle (slope angle) and direction (slope aspect angle) of the forest slope.

[0165] Direction and magnitude correction refers to adjusting the direction (such as rotation) and length (such as scaling) of the tree height space vector.

[0166] In an embodiment of the present application, the generated tree height correction parameters are mapped to the initial tree height space vector set, and the direction and amplitude of the initial tree height space vector are corrected in combination with the slope characteristics of the forest slope. First, the slope characteristics of the forest slope (such as slope angle, aspect angle) are extracted through slope analysis (such as terrain gradient calculation or slope analysis). Subsequently, a geometric transformation (such as a rotation matrix or a projection transformation) is used to map the correction parameters to the initial vector to correct its direction and amplitude. For example, in a fir plantation, when the slope angle is 20°, the amplitude correction of the initial vector is 15%. Finally, iterative optimization (such as least squares method or genetic algorithm) is used to ensure that the corrected vector is consistent with the terrain characteristics, and a high-precision tree height space vector set is generated.

[0167] 704. Based on the corrected tree height spatial vector set, calculate the corrected tree height of each crown contour vertex, and output the tree height measurement parameters corrected for the steep slope terrain.

[0168] In step 704 , the corrected tree height refers to the vertex height value of the tree crown contour after terrain correction.

[0169] The tree height measurement parameters for steep slope terrain correction refer to the tree height measurement results after eliminating the influence of slope.

[0170] In an embodiment of the present application, based on the corrected tree height spatial vector set, the corrected tree height of each crown outline vertex is calculated, and the tree height measurement parameters corrected for steep slope terrain are output. First, the corrected vector is mapped to the crown outline vertex through a spatial interpolation algorithm (such as Kriging interpolation or inverse distance weighted interpolation) to calculate the corrected tree height. For example, in a poplar plantation, the error between the corrected tree height and the measured tree height is controlled within 3%. Subsequently, noise points are removed through data smoothing (such as Gaussian filtering or mean filtering) to generate a continuous and smooth tree height distribution map. Finally, the tree height measurement parameters corrected for steep slope terrain are output to provide high-precision spatial data support for dynamic monitoring of forest resources and carbon sequestration management.

[0171] The following is a specific example: When measuring tree height in a natural fir forest on steep slopes (average slope 35°) in the high mountain canyons of southwest China, a correction scheme combining 3D laser scanning and stress wave detection can be used to implement terrain correction. First, a 3D lidar scanner is used to acquire stand point cloud data. A canopy vertex recognition algorithm is used to extract the 3D coordinates of the fir tree crown outline vertices. Combined with a slope reference model generated from a DEM, a set of vertical distances between each tree vertex and its corresponding slope projection point is established (e.g., a 5.4-meter initial vector between a vertex at an elevation of 1523.6 meters and a slope reference point at 1518.2 meters). A multi-channel stress wave detector is then deployed, with 32 sensor nodes positioned along the path from the trunk base to the crown. The gradient parameter of the longitudinal wave velocity, which decreases from 4300 m / s at the base to 3800 m / s at the crown, is collected during acoustic wave propagation. This is used to calculate the axial attenuation coefficient (e.g., a correction factor of 0.82). The correction parameters are mapped to the initial vector through spatial projection. Combined with the cosine projection coefficient of 0.8192 for a 35° slope, the vector direction is compensated for 21° and the amplitude is adjusted to 6.2m. Finally, based on the corrected spatial vector set, the terrain-compensated tree height parameters (e.g., actual vertical height of 5.1m ± 0.3m) are output, providing accurate measurement data for analyzing the vertical structure of steep slope forests. This implementation effectively overcomes the 23% terrain projection error in traditional measurements through multi-source data fusion.

[0172] In summary, steps 701 to 704 demonstrate a precise tree height measurement technique for steep slopes based on acoustic characteristics and terrain correction. This technique generates an initial tree height spatial vector set by extracting the vertical distance between the vertices of the three-dimensional crown outline and the forest slope reference plane. This technique, combined with the gradient data in the acoustic characteristic parameter set, calculates the gradient variation characteristics along the stress wave propagation path from the trunk base to the crown apex, generating tree height correction parameters. By mapping the tree height correction parameters to the initial tree height spatial vector set and incorporating the slope characteristics of the forest slope, the direction and amplitude of the initial tree height spatial vector are corrected. Finally, the corrected tree height is calculated for each crown outline vertex, outputting the steep slope terrain-corrected tree height measurement parameters. In verification of Chinese fir plantations, the tree height measurement error was reduced from 12.3% using traditional methods to 3.8%, significantly improving the accuracy and reliability of tree height measurement on steep slopes and providing a high-precision measurement tool for dynamic monitoring of forest resources and carbon sequestration management.

[0173] Figure 2 The present invention provides a schematic diagram of a system for measuring and determining forest tree height based on laser radar point cloud data. Figure 2As shown, the device includes: an acquisition module 21, which is used to collect growth stress wave signals through an acoustic emission sensor array arranged at the base of the trunk, and generate a set of acoustic characteristic parameters reflecting the internal mechanical state of the trunk, wherein the sensor array is evenly spaced in the direction of the slope normal to form a monitoring network; an optimization module 22, which is used to drive the digital twin model to perform forest stand structure topology optimization based on the acoustic characteristic parameter set, generate a three-dimensional growth vector model, and at the same time receive the slope scanning point cloud transmitted in real time by the drone lidar through the 5G edge node; a correction module 23, which is used to perform terrain-compensated point cloud dynamic splicing on the slope scanning point cloud, The stress distribution characteristics in the three-dimensional growth vector model are combined to correct the splicing error and generate a crown segmentation boundary dataset that integrates physical property constraints; a generation module 24 is used to transmit the crown segmentation boundary dataset back to the digital twin model for iterative optimization, and combine the spatial distribution characteristics of the slope scanning point cloud with the stress wave propagation time series data in the acoustic feature parameter set to generate a three-dimensional crown contour with terrain displacement compensation; an output module 25 is used to superimpose the gradient data in the acoustic feature parameter set according to the spatial vector relationship between the three-dimensional crown contour vertex and the slope reference plane, and output the tree height measurement parameters corrected for the steep slope terrain.

[0174] Figure 2 The device for measuring and determining the height of trees in forests based on laser radar point cloud data can be executed Figure 1 The implementation principle and technical effects of the method for measuring and determining the height of trees in a forest based on laser radar point cloud data described in the embodiment shown will not be described in detail. The specific manner in which each module and unit performs operations in the above embodiment of the device for measuring and determining the height of trees in a forest based on laser radar point cloud data has been described in detail in the embodiment of the method, and will not be elaborated on here. In one possible design, Figure 2 The apparatus for measuring and determining the height of trees in a forest based on laser radar point cloud data in the embodiment shown can be implemented as a computing device, such as Figure 3 As shown, the computing device may include a storage component 31 and a processing component 32; the storage component 31 stores one or more computer instructions, wherein the one or more computer instructions are called and executed by the processing component 32.

[0175] The processing component 32 is used for the above Figure 1 The embodiment provides a method for measuring and determining forest tree heights based on lidar point cloud data.

[0176] The processing component 32 may include one or more processors to execute computer instructions to complete all or part of the steps in the above method. Of course, the processing component may also be implemented as one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the above method.

[0177] The storage component 31 is configured to store various types of data to support operations at the terminal. The storage component can be implemented by any type of volatile or non-volatile memory device, or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk, or optical disk.

[0178] Of course, a computing device may also include other components, such as input / output interfaces, display components, communication components, etc.

[0179] The input / output interface provides an interface between the processing component and the peripheral interface module, which can be an output device, an input device, etc.

[0180] The communication component is configured to facilitate, among other things, wired or wireless communications between the computing device and other devices.

[0181] Among them, the computing device can be a physical device or an elastic computing host provided by a cloud computing platform, etc. In this case, the computing device can refer to a cloud server, and the above-mentioned processing components, storage components, etc. can be basic server resources rented or purchased from the cloud computing platform.

[0182] The present application also provides a computer storage medium storing a computer program, wherein the computer program can achieve the above-mentioned Figure 1 The illustrated embodiment provides a method for measuring and determining forest tree heights based on lidar point cloud data.

[0183] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0184] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one location or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0185] Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented using software plus a necessary general-purpose hardware platform, or of course, hardware. Based on this understanding, the essence of the above technical solution, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, or an optical disk, and includes a number of instructions for causing a computer device (such as a personal computer, server, or network device) to execute the methods described in each embodiment or certain portions of the embodiments.

[0186] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for measuring and determining forest tree height based on laser radar point cloud data, characterized in that: include: The sensor array placed at the base of the tree trunk collects growth stress wave signals and generates a set of acoustic characteristic parameters reflecting the internal mechanical state of the tree trunk. Based on the acoustic characteristic parameter set, the digital twin model is driven to optimize the forest stand structure topology and generate a three-dimensional growth vector model. At the same time, the slope scanning point cloud transmitted in real time by the drone lidar is received through the 5G edge node; Performing a terrain-compensated point cloud dynamic splicing operation on the slope scan point cloud, correcting splicing errors in combination with stress distribution characteristics in the three-dimensional growth vector model, and generating a tree crown segmentation boundary dataset; The crown segmentation boundary dataset is fed back to the digital twin model for iterative optimization. The spatial distribution characteristics of the slope scanning point cloud and the stress wave propagation time series data in the acoustic characteristic parameter set are combined to generate a three-dimensional crown contour with terrain displacement compensation. According to the spatial vector relationship between the vertices of the three-dimensional tree crown outline and the reference plane of the forest slope, the gradient data in the acoustic characteristic parameter set are superimposed to output the tree height measurement parameters corrected for the steep slope terrain.

2. The method according to claim 1, characterized in that Performing a terrain-compensated point cloud dynamic splicing operation on the slope scan point cloud, correcting splicing errors in combination with stress distribution characteristics in the three-dimensional growth vector model, and generating a crown segmentation boundary dataset, including: Extracting the local elevation gradient field and curvature characteristics of the slope scanning point cloud, combining them with the stress direction field corresponding to the stress distribution characteristics in the three-dimensional growth vector model, and calculating the spatial adaptive weight of the terrain compensation parameter; The angle between the maximum shear stress gradient and the principal stress direction in the stress distribution feature is used as a dynamic correction factor, and the splicing error is iteratively corrected point by point using the spatial adaptive weight during the dynamic splicing process to obtain the corrected point cloud geometric characteristics; Based on the corrected point cloud geometric characteristics, the normal vector and principal curvature parameters of each point in the slope scanning point cloud are calculated, and combined with the spatial distribution characteristics of the stress gradient difference field, the surface characteristic parameters consistent with the principal stress direction of the crown growth are selected; The spatial consistency mapping relationship between the normal vector and the stress gradient difference field is used as a physical constraint, and the crown boundary is determined by the collinearity threshold between the principal stress direction and the surface normal. On the basis of determining the crown boundary, the surface characteristic parameters are multi-dimensionally fused with the stress gradient difference field, and the crown segmentation boundary dataset is generated through the spatial consistency constraint of the crown growth principal stress direction and the surface normal.

3. The method according to claim 2, characterized in that The surface characteristic parameters are multi-dimensionally fused with the stress gradient difference field, and the crown segmentation boundary dataset is generated by constraining the spatial consistency of the crown growth principal stress direction and the surface normal, including: A dynamic distribution model of the crown surface geometric attributes is constructed based on the surface feature parameters, the surface curvature change rate in the surface feature parameters and the normal offset corresponding to the normal vector of each point in the slope scanning point cloud are integrated into a spatial continuity descriptor, and a topological association expression of the surface feature parameters is established; The multi-scale decomposition method of the stress gradient difference field is adopted to generate a composite tensor representation of the stress gradient difference field by superimposing the vectors of the principal stress directions of crown growth; Perform multi-dimensional spatial registration on the topological association expression and the composite tensor representation, and calculate the fusion weight coefficient using the spatial projection deviation between the surface normal direction corresponding to the surface feature parameters and the principal stress direction of the crown growth; Generate a collaborative constraint function between the principal stress direction of crown growth and the surface normal based on the fusion weight coefficient; The feedback channel of the collaborative constraint function outputs a set of candidate regions for crown segmentation boundaries. The amplitude convergence of the stress gradient difference field is used as a boundary constraint to screen out the candidate boundary regions and generate a crown segmentation boundary dataset.

4. The method according to claim 3, characterized in that A collaborative constraint function between the principal stress direction of crown growth and the surface normal is generated based on the fusion weight coefficient, including: Performing tensor feature correction on the fusion weight coefficient and the stress distribution characteristics of the three-dimensional growth vector model, and generating a revised set of fusion weight coefficients through the orthogonality constraint condition of the principal stress direction vector and the surface normal vector; The collaborative constraint function between the principal stress direction of crown growth and the surface normal is constructed based on the correction set.

5. The method according to claim 1, characterized in that ,Based on the acoustic characteristic parameter set, the digital twin model is driven to perform forest stand structure topology optimization and generate a three-dimensional growth vector model, including: spatially aligning the growth stress wave signals collected by the sensor array along the slope normal direction, extracting the propagation characteristics of the stress wave signals in the radial and axial directions of the trunk, and generating spatial distribution parameters of the mechanical state inside the trunk; Based on the spatial distribution parameters, constructing a three-dimensional vector expression of the tree trunk growth stress field; The three-dimensional vector expression is spatially registered with the slope scanning point cloud collected by the UAV laser radar to establish a mapping relationship between the tree trunk growth stress field and the external morphological characteristics; Based on the mapping relationship, the dynamic change trend of the trunk growth stress field is calculated through the propagation time series characteristics of the stress wave signal inside the trunk, and a three-dimensional growth vector model reflecting the growth characteristics of the forest stand structure is generated.

6. The method according to claim 1, characterized in that The crown segmentation boundary dataset is fed back to the digital twin model for iterative optimization. Combining the spatial distribution characteristics of the slope scan point cloud with the stress wave propagation time series data in the acoustic feature parameter set, a three-dimensional crown profile with terrain displacement compensation is generated, including: The crown segmentation boundary dataset is input into the digital twin model, and the initial terrain displacement of the crown boundary point is calculated based on the propagation path characteristics of the stress wave inside the trunk in the stress wave propagation time series data; Iteratively comparing the initial terrain displacement with the spatial distribution characteristics of the slope scan point cloud, optimizing the spatial position of the crown boundary points through the digital twin model, and updating the terrain displacement; Based on the updated terrain displacement, the point cloud in the crown segmentation boundary dataset is spatially corrected to generate a preliminary compensated three-dimensional crown contour. The three-dimensional crown profile with preliminary compensation is input into the digital twin model again, and the terrain displacement compensation parameters are optimized in combination with the dynamic change characteristics of the stress wave propagation time series data to generate a three-dimensional crown profile with terrain displacement compensation.

7. The method according to claim 1, characterized in that According to the spatial vector relationship between the vertices of the three-dimensional tree crown outline and the reference plane of the forest slope, the gradient data in the acoustic characteristic parameter set is superimposed to output the tree height measurement parameters corrected for the steep slope terrain, including: Extract the vertical distance between the vertices of the 3D tree crown outline and the forest slope reference plane to generate the initial tree height space vector set; Based on the gradient data in the acoustic characteristic parameter set, the gradient change characteristics of the stress wave propagation path from the base of the trunk to the top of the crown are calculated to generate the tree height correction parameters; Mapping the tree height correction parameters to the initial tree height space vector set, and correcting the direction and amplitude of the initial tree height space vector in combination with the slope characteristics of the forest slope; Based on the corrected tree height spatial vector set, the corrected tree height of each crown contour vertex is calculated, and the tree height measurement parameters corrected for steep slope terrain are output.

8. A method for measuring and determining forest tree height based on laser radar point cloud data, characterized in that: include: An acquisition module is configured to collect growth stress wave signals through an acoustic emission sensor array disposed at the base of the tree trunk, thereby generating a set of acoustic characteristic parameters reflecting the internal mechanical state of the tree trunk, wherein the sensor array is evenly spaced in the direction of the slope normal to form a monitoring network; An optimization module is used to drive the digital twin model to perform forest stand structure topology optimization based on the acoustic feature parameter set, generate a three-dimensional growth vector model, and simultaneously receive the slope scanning point cloud transmitted in real time by the drone lidar through a 5G edge node; A correction module is used to perform dynamic point cloud splicing with terrain compensation on the slope scan point cloud, correct the splicing error by combining the stress distribution characteristics in the three-dimensional growth vector model, and generate a tree crown segmentation boundary dataset that integrates physical property constraints; A generation module is used to return the tree crown segmentation boundary dataset to the digital twin model for iterative optimization, and combine the spatial distribution characteristics of the slope scanning point cloud with the stress wave propagation time series data in the acoustic characteristic parameter set to generate a three-dimensional tree crown contour with terrain displacement compensation; The output module is used to superimpose the gradient data in the acoustic characteristic parameter set according to the spatial vector relationship between the vertices of the three-dimensional tree crown contour and the slope reference plane, and output the tree height measurement parameters corrected for the steep slope terrain.

9. A computing device, characterized in that It includes a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a forest tree height measurement and determination method based on lidar point cloud data as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that A computer program is stored, and when the computer program is executed by a computer, a method for measuring and determining the height of trees in a forest based on laser radar point cloud data as described in any one of claims 1 to 7 is implemented.

Images