Tree species growth prediction method, system and device based on multi-temporal point cloud and growth model, and medium

Abstract

The invention relates to the technical field of power line monitoring, and discloses a tree growth prediction method, system and device based on a multi-temporal point cloud and a growth model, and a medium, and the method comprises the steps: collecting the multi-temporal data of a target region, carrying out the vegetation region recognition and single tree segmentation of a vegetation point cloud, extracting the structural features of each tree, and carrying out the prediction of the growth of each tree; generating a time sequence characteristic data set; constructing a tree growth model, taking the time sequence characteristic data set as model input, introducing a point cloud structure parameter as a dynamic constraint item, and performing parameter optimization on the model to obtain a prediction curve of a multi-dimensional growth index of the tree; and based on the prediction curve of the multi-dimensional growth indexes, performing dynamic geometric reconstruction through original point cloud data, and generating an interactive visual three-dimensional model. According to the method, on the basis of multi-source heterogeneous point cloud data, the single-tree-level growth model can be constructed, the time sequence structure change trend of the single-tree-level growth model is analyzed, and the space error and the time stability of a prediction result are quantified.

Description

Technical Field

[0001] This invention relates to the field of power line monitoring technology, and in particular to a method, system, device and medium for predicting tree growth based on multi-temporal point clouds and growth models. Background Technology

[0002] With the continuous expansion of power transmission lines and the increasing complexity of the regions they cross, the dynamic growth and spatial expansion of vegetation within transmission corridors pose a persistent threat to the safe operation of these lines. The height of trees, the rate of canopy expansion, and changes in the spatial distance between trees and conductors are key parameters affecting the safe clearance of transmission lines, discharge risks, and the design of protective measures. Once tree growth exceeds safe distance limits, it can easily trigger safety accidents such as line discharge, tripping, grounding short circuits, and even large-scale power outages. Therefore, how to achieve continuous monitoring, dynamic prediction, and intelligent early warning of vegetation in transmission corridors has become an important technical direction in the field of power transmission operation and maintenance.

[0003] Traditional vegetation monitoring along power transmission corridors relies heavily on periodic manual inspections or single-phase aerial / satellite image recognition. Limited by manpower efficiency, observation accuracy, and temporal resolution, this approach struggles to continuously reflect the growth patterns and structural evolution of trees across different time scales. This method not only suffers from poor data timeliness and insufficient recognition accuracy but also fails to provide reliable data for growth trend prediction and future risk assessment, resulting in delayed risk prediction and limited early warning accuracy.

[0004] In recent years, with the rapid development of 3D laser scanning, UAV oblique photogrammetry, and multi-view stereo reconstruction technologies, acquiring high-precision, multi-temporal 3D point cloud data has become possible. Point cloud data possesses high spatial resolution and realistic geometric representation capabilities, accurately depicting the spatial structural features of trees, such as tree height, crown width, volume, branch and leaf density, and morphological distribution, providing data support for parametric modeling and structural change analysis of tree growth models. Simultaneously, by comparing point cloud data collected at different time points, the tree growth rate and spatial change trends can be quantified, achieving time-series-level dynamic monitoring. However, in practical applications, the fusion and analysis of multi-temporal point clouds still faces numerous technical bottlenecks: Point cloud data at different times are affected by the attitude of the acquisition platform, weather conditions, scanning angle and illumination changes, resulting in problems such as inconsistent coordinate reference, inconsistent point density, occlusion and noise interference, which leads to the accumulation of registration errors in time series data. The structural changes of the same tree at different times make it difficult for traditional matching algorithms based on geometric overlap to guarantee semantic consistency, affecting the stability of growth extraction; Current growth models are mostly based on statistical laws or empirical functions, which can only depict macroscopic growth trends and lack a deep integration mechanism with point cloud geometric features (such as canopy volume, branch and leaf density, and topological structure), making it difficult to achieve refined dynamic prediction and structural accuracy analysis of individual trees. Summary of the Invention

[0005] In view of the aforementioned existing problems, this invention is proposed. Therefore, this invention provides a tree species growth prediction method based on multi-temporal point clouds and growth models to solve the above problems.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a tree species growth prediction method based on multi-temporal point clouds and growth models, comprising: Collect multi-temporal data of the target area and perform data preprocessing; Based on the preprocessed multi-temporal data, vegetation area identification and single tree segmentation are performed on the vegetation point cloud, and the structural features of each tree are extracted to generate a temporal feature dataset. A tree growth model is constructed, using the aforementioned time-series feature dataset as the model input, and point cloud structure parameters are introduced as dynamic constraints. The model is then optimized to obtain the prediction curves of the multidimensional growth indicators of the trees. Based on the predicted curves of the multidimensional growth indicators, dynamic geometric reconstruction is performed using the original point cloud data to simulate the temporal growth of trees and generate an interactive, visualized 3D model.

[0007] As a preferred embodiment of the tree species growth prediction method based on multi-temporal point cloud and growth model described in this invention, the method includes: collecting multi-temporal data of the target area and performing data preprocessing, including: Acquire multi-temporal 3D spatial data of the target area at different time points, and perform standardized preprocessing on the raw point cloud data; The preprocessing includes noise removal, ground separation processing, unified acquisition coordinates, and point cloud sparsification and registration.

[0008] As a preferred embodiment of the tree species growth prediction method based on multi-temporal point clouds and growth models described in this invention, the method involves: identifying vegetation regions and segmenting individual trees within the vegetation point cloud, and extracting the structural features of each tree, including: Based on multi-temporal data of vegetation, vegetation areas are automatically identified and individual trees are separated. By combining the geometric features of the tree, feature extraction is performed at the individual tree level, key structural parameters of the tree are calculated, and a one-to-one correspondence between individual trees is established in multi-temporal point clouds. Similarity matching is performed on the key structural parameters to generate a continuous temporal feature dataset of individual trees. The beneficial effect of this preferred solution is that it enables the automatic segmentation and identification of individual trees in complex backgrounds, forming a tree growth dataset with time-series characteristics, and providing data support for dynamic prediction.

[0009] As a preferred embodiment of the tree species growth prediction method based on multi-temporal point clouds and growth models described in this invention, the construction of the tree growth model includes: Based on the variation patterns of multi-temporal characteristic data, a tree growth model integrating geometric structural parameters and physiological growth patterns is constructed. The model constructs the trend of tree morphology change over time through nonlinear functions, and uses the three-dimensional structural parameters of the tree as dynamic constraints to construct the relationship between tree structural growth and temporal evolution.

[0010] The advantage of this preferred scheme is that the model not only considers the physiological time sequence of tree growth, but also uses the three-dimensional structural features of trees as dynamic constraints, thereby achieving accurate simulation of the individual tree growth process.

[0011] As a preferred embodiment of the tree species growth prediction method based on multi-temporal point cloud and growth model described in this invention, the parameter optimization of the model includes: Based on the tree growth model, the least squares method is used to fit the initial parameters of the multi-temporal observation data, and a Bayesian optimization and confidence interval constraint mechanism is introduced to perform global optimization and adaptive update of the model parameters. By continuously iterating and correcting the model parameters, individualized growth predictions of trees of different species and under different habitat conditions are made. The model outputs prediction curves and confidence intervals of multidimensional growth indicators to assess the growth trend of trees in the future.

[0012] As a preferred embodiment of the tree species growth prediction method based on multi-temporal point cloud and growth model described in this invention, the generation of an interactive, visualized 3D model includes: Based on the multidimensional growth indicators predicted by the model, dynamic geometric reconstruction is performed by combining the original point cloud data. By mapping the temporal changes of tree structural parameter features to a spatial coordinate system, temporal three-dimensional reconstruction and dynamic visualization of individual trees are carried out, generating a continuous time series three-dimensional interactive model to show the structural changes and growth trends of trees at different growth stages.

[0013] As a preferred embodiment of the tree species growth prediction method based on multi-temporal point clouds and growth models described in this invention, the construction of the tree growth model further includes: The virtual point cloud predicted by the tree growth model is spatially compared with the measured point cloud in multiple time phases, and the accuracy of the model is comprehensively evaluated through multiple quantitative indicators. Based on the model evaluation results, a multi-factor error model is constructed, multi-dimensional influencing factors are introduced, and an adaptive weighting mechanism is used to correct the model error.

[0014] Secondly, the present invention provides a tree species growth prediction system based on multi-temporal point clouds and growth models, comprising: The acquisition and processing module is used to acquire multi-temporal data of the target area and perform data preprocessing; The identification and segmentation module is used to identify vegetation regions and segment individual trees in the vegetation point cloud based on the preprocessed multi-temporal data, extract the structural features of each tree, and generate a temporal feature dataset. The model building module is used to build a tree growth model. It takes the time-series feature dataset as the model input, introduces point cloud structure parameters as dynamic constraints, and optimizes the model parameters to obtain the prediction curve of the multidimensional growth index of the tree. The visualization module is used to perform dynamic geometric reconstruction based on the predicted curve of the multidimensional growth index using raw point cloud data, simulate the temporal growth of trees, and generate an interactive, visualized 3D model.

[0015] Thirdly, the present invention provides a computer device, comprising: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, they implement the steps of the tree species growth prediction method based on multi-temporal point cloud and growth model.

[0016] Fourthly, the present invention provides a computer-readable storage medium storing computer-executable instructions that, when executed by a processor, implement the steps of the tree species growth prediction method based on multi-temporal point cloud and growth model.

[0017] Compared with existing technologies, the beneficial effects of this invention are as follows: The tree species growth prediction and accuracy evaluation method based on multi-temporal point cloud and growth model proposed in this invention can construct a single-tree-level growth model on the basis of multi-source heterogeneous point cloud data, analyze its temporal structure change trend, and quantify the spatial error and temporal stability of the prediction results. This method can not only provide high-precision data support for vegetation safety monitoring, tree obstacle early warning and pruning decision-making in power transmission line corridors, but also can be extended to fields such as dynamic analysis of forestry resources, ecological environment monitoring, and vegetation succession research, and has significant engineering application value and scientific research significance. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a schematic diagram of the overall process of a tree species growth prediction method based on multi-temporal point cloud and growth model according to an embodiment of the present invention.

[0020] Figure 2 This is a schematic diagram of multi-temporal tree point cloud matching, which is a tree species growth prediction method based on multi-temporal point cloud and growth model according to an embodiment of the present invention.

[0021] Figure 3 This is a schematic diagram of the single-tree segmentation results and model construction of a tree species growth prediction method based on multi-temporal point cloud and growth model according to an embodiment of the present invention.

[0022] Figure 4 This is a schematic diagram of growth curve fitting based on a growth model, which is a tree species growth prediction method based on multi-temporal point cloud and growth model according to an embodiment of the present invention.

[0023] Figure 5 This is a schematic diagram illustrating the three-dimensional dynamic reconstruction and growth prediction visualization of trees based on a tree species growth prediction method using multi-temporal point clouds and growth models, as described in an embodiment of the present invention. Detailed Implementation

[0024] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0025] Example 1, referring to Figure 1 As an embodiment of the present invention, a tree species growth prediction method based on multi-temporal point clouds and growth models is provided, comprising: S101, Collect multi-temporal data of the target area and perform data preprocessing; S102, based on the preprocessed multi-temporal data, vegetation area identification and single tree segmentation are performed on the vegetation point cloud, the structural features of each tree are extracted, and a temporal feature dataset is generated. S103, construct a tree growth model, use the time series feature dataset as the model input, introduce point cloud structure parameters as dynamic constraints, optimize the model parameters, and obtain the prediction curve of the multidimensional growth index of the tree. S104, based on the prediction curve of multidimensional growth indicators, uses raw point cloud data to perform dynamic geometric reconstruction, simulates the temporal growth of trees, and generates an interactive, visualized 3D model.

[0026] In a preferred embodiment, multi-temporal data of the target area is acquired, and data preprocessing is performed, including: Acquire multi-temporal 3D spatial data of the target area at different time points, and perform standardized preprocessing on the raw point cloud data; Preprocessing includes noise removal, ground separation processing, acquisition coordinate unification, and point cloud sparsification and registration.

[0027] Specifically, UAVs equipped with lidar are used as the primary data acquisition method to acquire multi-temporal three-dimensional spatial data of the power transmission line corridor and surrounding vegetation areas. Repeated aerial surveys are conducted at different time periods under the same flight path and attitude parameters to achieve dynamic monitoring and temporal modeling of tree growth status. The UAV platform offers advantages such as maneuverability, short operation cycles, and high resolution, enabling rapid acquisition of vegetation point cloud data at seasonal or interannual scales. This provides a refined spatial foundation for subsequent analysis of tree growth changes. The UAV system possesses centimeter-level spatial resolution and multi-echo acquisition capabilities, simultaneously acquiring high-density three-dimensional point cloud information at the canopy, branch, and ground levels. During the acquisition process, high-precision GNSS and IMU systems are used to synchronously record attitude and position information, ensuring that the point cloud data has strict spatial reference consistency and temporal comparability.

[0028] After data acquisition is completed, the raw point cloud data undergoes standardized preprocessing, including: An outlier detection method based on statistical filtering and radius filtering algorithms is used to remove noise points, outliers and echo interference points, and abnormal points caused by laser echo error, illumination interference and wind speed changes, so as to ensure the overall quality and spatial continuity of the point cloud. The Progressive Morphological Filter (PMF) algorithm is used to extract ground points, enabling terrain modeling and non-ground point extraction, and extracting 3D point cloud data of pure vegetation areas. To address the differences in acquisition accuracy, point density, and attitude error among point clouds at different time phases, a coordinate unification method based on ground control points and attitude calculation is adopted to convert all time phase data to a unified coordinate reference and perform time-series calibration through timestamps. This achieves consistent management of point cloud data in both spatial and temporal dimensions, that is, unifying multi-time phase point clouds into the same coordinate system based on UAV GNSS / IMU data.

[0029] After coordinate unification and noise removal, the point cloud data is processed by voxel grid sparse processing to balance data volume and structural integrity, ensuring that point clouds of different time phases have the same point density and spatial distribution characteristics, avoiding deviations introduced by differences in point density in subsequent feature extraction and registration stages. To achieve high-precision alignment of multi-temporal point clouds, this invention adopts a registration method based on geometric features and iterative optimization. Initial attitude correction is performed by extracting stable feature points such as tree trunk bases or transmission tower bases to complete coarse registration. On this basis, an improved iterative nearest-point algorithm is used for fine registration. Global alignment with centimeter-level spatial accuracy is achieved by minimizing the Euclidean distance function of corresponding points. For areas with local occlusion or uneven point cloud density, a weighted ICP strategy is introduced, assigning higher weights to the canopy and trunk regions to enhance the overall fitting stability and tree structure continuity. That is, voxel grid downsampling is used to reduce the amount of computation, and high-precision temporal point cloud registration is performed by combining the ICP algorithm or feature matching algorithm to ensure spatial alignment and temporal continuity between multi-temporal point clouds.

[0030] It should be noted that this step utilizes a UAV equipped with a high-precision lidar system to perform 3D point cloud scanning along the power transmission channel at different time periods, acquiring high-density, multi-temporal spatial data. By performing noise filtering, ground point separation, sparsification, and coordinate unification processing on the point clouds of each temporal phase, and employing an improved ICP algorithm to achieve high-precision spatial registration, data acquired at different times can be seamlessly overlaid in a unified coordinate system. This step provides a highly consistent and accurate data foundation for subsequent time-series tree growth analysis and structural change modeling.

[0031] In a preferred embodiment, vegetation region identification and individual tree segmentation are performed on the vegetation point cloud, and the structural features of each tree are extracted, including: Based on multi-temporal data of vegetation, vegetation areas are automatically identified and individual trees are separated. By combining the geometric features of the tree, feature extraction is performed at the individual tree level, key structural parameters of the tree are calculated, and a one-to-one correspondence between individual trees is established in multi-temporal point clouds. Similarity matching is performed on the key structural parameters to generate a continuous temporal feature dataset of individual trees.

[0032] Specifically, the vegetation point cloud within the target area undergoes segmentation and feature extraction at the tree-level. Using algorithms based on Euclidean clustering, region growing, or deep learning semantic segmentation, the vegetation area is automatically identified and individual trees are separated, ensuring accurate extraction of individual trees even in complex environments. Subsequently, combining tree geometric features, the trunk axis, crown position, and canopy boundary are extracted, and key structural parameters are calculated, including tree height, crown width, canopy volume, trunk diameter distribution, canopy density, and spatial distribution characteristics of branches and leaves. To achieve time-series growth analysis, this embodiment establishes a one-to-one correspondence between individual trees in the multi-temporal point cloud. Matching is performed using the trunk base point, crown centroid, and geometric similarity to form a continuous temporal feature dataset. This process not only achieves a structured representation of individual trees but also provides highly consistent and reliable foundational data support for subsequent growth modeling and accuracy evaluation.

[0033] Key structural parameters include tree height, crown width, crown volume, and crown density. Tree height H is calculated using the height difference between the highest point in the non-ground point cloud and the ground point; crown width C is obtained based on the circumcircle or ellipse parameters of the horizontally projected point cloud; crown volume V is calculated using the Alpha Shape 3D envelope algorithm; trunk diameter DBH is obtained by fitting a circular model of the point cloud at breast height; and crown density D is calculated through statistical analysis of point density per unit volume. For multi-temporal point cloud data collected at different time points, the system automatically establishes correspondences between individual trees, constructs a time-series feature database, and enables continuous analysis and dynamic monitoring of tree growth.

[0034] In this embodiment, the gridded canopy modeling method is used to achieve automatic segmentation of individual trees and extraction of key structural parameters, as detailed below: (1) Construction of the canopy height model The preprocessed vegetation point cloud data, after ground point removal, is divided into gridded raster units, dividing the target area into grid cells of the same size. And extract the elevation value of the highest point within each grid. This generates a canopy height model. The calculation formula is as follows: in, This represents the highest elevation value of the point cloud within the corresponding grid. This represents the ground elevation value at the same location, obtained by fitting ground points.

[0035] The CHM model represents the height distribution on the tree canopy surface in the form of a two-dimensional matrix. It can intuitively reflect the shape and relative height difference of the tree canopy and is the core foundation for subsequent single-tree segmentation.

[0036] (2) Local extremum identification and single-tree segmentation Local extremum detection is performed on the generated CHM to identify the crown vertex (i.e., the local maxima) of each tree. Let the size of the local neighborhood window be... Then in the center of the window It is considered to be a single-tree crown when the following conditions are met, and is represented as: Based on the canopy distribution and elevation gradient, the watershed algorithm is used to segment the CHM into several non-overlapping canopy regions. , is represented as: in, , This represents the canopy spatial regions of different individual trees (a subset of the point cloud after region division), each region The canopy space of a single tree, and satisfying the condition that they do not overlap. This allows for individual tree segmentation. To avoid over-segmentation or under-segmentation caused by overlapping canopies, the segmentation parameters are dynamically adjusted based on the canopy height gradient and canopy width ratio to optimize the accuracy of individual tree boundaries.

[0037] (3) Extraction of tree structural parameters After segmentation, structural parameters are calculated for the point cloud of each tree region, mainly including tree height, crown width, crown volume, crown density, and spatial centroid, as detailed below: Compute tree height The difference in elevation between the highest point in the area containing the tree and the ground point is calculated and expressed as follows: in, This represents the elevation of the highest point within the point cloud region of the k-th tree, i.e., the Z-coordinate of the highest point of the tree crown. This represents the lowest point on the ground in the area where the k-th tree is located, serving as a reference point for calculating tree height.

[0038] Calculate crown width The minimum circumcircle diameter of the horizontal projection of the tree crown is extracted and expressed as: Calculate the canopy volume Using a three-dimensional voxelization method, the canopy region is discretized into voxel units with side length Δ, and the volume is calculated as follows: in, The number of non-empty voxels in the canopy is Δ, which is generally taken as 0.1 m.

[0039] Calculate the canopy density index The density of the tree canopy and foliage is represented as: in, This represents the number of point clouds in the canopy. The coordinates of the centroid of the tree are calculated by weighted averaging of the coordinates of all points, and are expressed as follows: It should be noted that in point cloud data processing, a hybrid segmentation algorithm combining region growing, Euclidean clustering, and deep learning is employed. This algorithm enables automatic segmentation and identification of individual trees in complex backgrounds, accurately extracting single trees even in power transmission corridor environments with dense understory vegetation and significant terrain undulations. For each tree, multidimensional geometric and structural parameters are further calculated, including tree height, crown width, crown volume, trunk diameter distribution, crown density, and volumetric biomass. Simultaneously, individual tree matching relationships are established at different time points, forming a tree growth dataset with time-series characteristics, providing data support for dynamic prediction.

[0040] In one alternative implementation, the structural features of each tree can be extracted by separating ground points through cloth simulation filtering or progressive morphological filtering, extracting vegetation point clouds based on color information or geometric features (normal, curvature), detecting tree crown vertices using a local maximum algorithm or an improved peak-valley detection algorithm, using the tree top as a seed point, performing region growth segmentation based on distance, normal, and elevation constraints, extracting horizontal slice contours, separating adhered tree crowns through ellipse fitting and overlap analysis, and then extracting structural features.

[0041] In another alternative implementation, the structural features of each tree can also be extracted using a deep learning network model, such as a hybrid CNN-graph neural network. After voxelization, 3D CNN is used to extract features, construct a point cloud structure, use graph convolution to aggregate contextual information, combine attention mechanisms to enhance important features, jointly optimize the loss function through multi-task learning, extract shared features of the backbone network, merge regions based on prediction results, and set constraints (minimum tree spacing, maximum tree height, etc.) to optimize the segmentation results.

[0042] In a preferred embodiment, constructing a tree growth model includes: Based on the variation patterns of multi-temporal characteristic data, a tree growth model integrating geometric structural parameters and physiological growth patterns is constructed. The model constructs the trend of tree morphology change over time through nonlinear functions, and uses the three-dimensional structural parameters of the tree as dynamic constraints to construct the relationship between tree structural growth and temporal evolution.

[0043] Specifically, a hybrid tree growth model integrating geometric structural features and physiological growth patterns is constructed to achieve individual-level dynamic growth prediction and adaptive parameter updates. This model takes the temporal changes of core geometric parameters such as tree height, crown width, crown volume, and trunk diameter as input, and combines time scales with physical growth laws to establish a coupled "morphological constraint-temporal evolution" growth model structure. The main body of the model employs nonlinear growth equations such as Logistic and Gompertz to reflect the overall trend of trees in the initial rapid growth, middle-term stable growth, and later slowing-down stages, which can be the trend of tree height or crown volume changes over time. Simultaneously, point cloud structural parameters are introduced as dynamic constraints, i.e., three-dimensional structural parameters such as tree height, crown width, and volume are used as dynamic constraints, enabling the model to reflect real spatial structural changes during the fitting process. This model drives the evolution of virtual trees in a three-dimensional scene over time through growth prediction, presenting visualized changes in tree height, crown width, and branch and leaf volume for dynamic visualization and three-dimensional simulation analysis.

[0044] The tree growth model is represented as follows: in, For time The height of the tree at any moment For the maximum potential height, , , These are the growth rate and structure correction coefficients. This is the canopy volume term.

[0045] In a preferred embodiment, optimizing the parameters of the model includes: Based on the tree growth model, the least squares method is used to fit the initial parameters of multi-temporal observation data, and Bayesian optimization and confidence interval constraint mechanism are introduced to perform global optimization and adaptive update of model parameters. By continuously iterating and correcting the model parameters, individualized growth predictions of trees of different species and under different habitat conditions are made. The model outputs prediction curves and confidence intervals of multidimensional growth indicators to assess the growth trend of trees in the future.

[0046] Specifically, the least squares method, Bayesian optimization, or genetic algorithms are used to globally fit the parameters, and the parameters are dynamically updated according to different tree species and growing environments to achieve predictive modeling of individual tree growth. To improve the model's prediction accuracy and generalization ability, the least squares method is used to fit the initial parameters to multi-temporal observation data. On this basis, Bayesian optimization and confidence interval constraint mechanisms are introduced to globally optimize and adaptively update the model parameters. By continuously iterating and correcting the model parameters, individualized growth predictions are made for different tree species and different habitat conditions. The model finally outputs prediction curves and confidence intervals for multi-dimensional growth indicators such as tree height, crown width, and volume, which can be used to quantitatively infer the growth trend of trees in the future.

[0047] The least squares method is used to fit the parameters K, a, b, α of the nonlinear model, and the optimal solution of the model is achieved by minimizing the sum of squared errors between the predicted and measured values. in, These are the measured characteristic values. The initial values ​​are the predicted values ​​of the model. To improve the generalization performance and robustness of the model, Bayesian optimization is introduced to search for and control the convergence of the initial parameter values. The hyperparameters of the model are dynamically adjusted through Gaussian process regression to avoid local optima and improve the fitting accuracy.

[0048] It should be noted that this embodiment combines traditional forestry growth functions and other nonlinear functions with point cloud geometric parameters to construct a coupled "morphological constraint-temporal evolution" model. The model not only considers the physiological time sequence of tree growth, but also uses the three-dimensional structural features of trees as dynamic constraints to achieve accurate simulation of the individual tree growth process. The model parameters are fitted and dynamically updated through least squares method, Bayesian optimization or genetic algorithm to achieve high-precision prediction of the future height, crown width and volume growth of trees.

[0049] In one alternative implementation, the tree growth model can also be based on machine learning to build a predictive model. By using multi-period point cloud, remote sensing and environmental data, machine learning algorithms (such as ensemble learning or deep learning) are used to mine the statistical relationship between growth patterns and features to accurately predict future growth. It has strong big data processing capabilities and high prediction efficiency.

[0050] In another alternative implementation, the tree growth model can also be constructed based on a process mechanism model. According to the physiological and ecological characteristics of trees, it simulates processes such as photosynthesis, respiration consumption, and carbon allocation to simulate growth, and describes morphological structure through rules, which can analyze the influence of environmental factors.

[0051] In a preferred embodiment, generating an interactive, visualized 3D model includes: Based on model-predicted multidimensional growth indicators, dynamic geometric reconstruction is performed using raw point cloud data. By mapping the temporal changes of tree structural parameters to a spatial coordinate system, temporal 3D reconstruction and dynamic visualization of individual trees are achieved, generating a continuous time-series 3D interactive model that displays the structural changes and growth trends of trees at different growth stages.

[0052] Specifically, based on the established tree growth model and multi-temporal point cloud data, the system uses model fitting parameters to drive the dynamic evolution of the tree's three-dimensional structure, predictively reconstructing the tree's morphology for future moments. By mapping the temporal changes of geometric features (such as tree height, crown width, and crown volume) to a spatial coordinate system, the system achieves temporal three-dimensional reconstruction and dynamic visualization of individual trees. It can generate continuous time-series three-dimensional animations or interactive models, intuitively presenting the structural changes and growth trends of trees at different growth stages.

[0053] After the growth model parameters are determined, the predicted structural parameter changes (such as tree height, crown width, and volume growth rate) are used to drive dynamic geometric reconstruction of the original point cloud data. Through point cloud interpolation and morphological constraint methods, virtual point cloud data for future timeframes is generated, enabling temporal simulation of tree growth. The predicted point cloud is then reconstructed into a grid to generate an interactive 3D model. Combined with a timeline playback function, this model visually displays the height increase, crown expansion, and volume changes during the tree's growth process. This embodiment can also overlay the predicted tree model with the point cloud of power transmission line conductors to calculate the changes in the spatial safety distance between the trees and conductors in real time, achieving dynamic risk prediction and 3D visualization early warning under growth trends.

[0054] Tree geometric features (including tree height H) extracted from multi-temporal point cloud data t Crown width C t Main direction vector Canopy volume V t (e.g., ...) serve as the structural driving parameters for the digital twin tree model. Each individual tree is generated in UE5 as a parametric tree template, with the template node structure including trunk nodes, branch nodes, and leaf cluster nodes. By matching the point cloud center coordinates (x, y, z) with the terrain elevation model (DTM), the tree model is accurately positioned and its attitude is calibrated in three-dimensional space.

[0055] Based on changes in the predicted parameters of the growth model, the system dynamically updates the growth status of the tree model in Unreal Engine. The growth model uses a nonlinear time evolution function to describe the growth patterns of tree height and crown width. Where Hmax and Cmax are the theoretical maximum values ​​of tree height and crown width, respectively, and kH and kC are growth rate coefficients. The growth inflection point is defined. The above parameters are updated recursively with a time step Δt to achieve continuous tree growth simulation along the time axis. In UE5, the morphological changes of trees are controlled by blueprint scripts to dynamically scale and deform the tree's skeletal structure. The growth ratio of each node (trunk, branches, leaves) is differentiated based on the point cloud feature calculation results. For example, the trunk elongation expands along the main axis according to the ratio H(t) / Ht0; the canopy radius is radially enlarged according to the ratio C(t) / Ct0; and the leaf density is adjusted based on changes in point cloud reflection intensity or canopy density. Simultaneously, the power transmission line point cloud data is imported into the same 3D scene, and the minimum distance between the tree canopy top and the power line is calculated using spatial index matching. , is represented as: in, This represents the i-th point in the tree point cloud set at time t. This represents the j-th point in the point cloud set of transmission lines. Let represent the set of three-dimensional point clouds of the entire tree predicted by the growth model at time t. This represents a 3D point cloud set of power transmission lines; and a dynamic, real-time, color-mapped visual warning is displayed in a virtual scene. When the tree growth prediction model indicates that the distance D(t) at a future time is less than the safety threshold D... safe At that time, the safety threshold D safe The values ​​are determined based on the line grade, tree species characteristics, and point cloud measurement errors. For 110kV transmission lines, 4.0–5.0 meters can be used. Risk areas are automatically marked, and early warning prompts are output. Furthermore, UE5's virtual geometry rendering and global illumination system enable reconstructed trees to have realistic physical lighting effects and high-precision shadow rendering, supporting immersive visualization playback of multi-temporal tree growth processes. Users can observe the growth evolution of trees from 2023 to 2025 through a time slider interaction, intuitively comparing changes in tree structure in different years.

[0056] It should be noted that this embodiment combines the fitted growth model parameters with the initial point cloud data to dynamically reconstruct the three-dimensional structure of trees at several future moments, generating a time-series virtual point cloud and mesh model. Through visualization, users can intuitively observe the growth and evolution process of trees in three-dimensional space, including dynamic trends such as changes in tree height, canopy expansion, and volume increment. It not only supports growth trend analysis but can also overlay a power transmission line spatial model to calculate the distance changes between trees and power lines in real time and predict potential tree obstacle risk areas in the future.

[0057] In a preferred embodiment, constructing a tree growth model further includes: The virtual point cloud generated by the tree growth model prediction is spatially compared with the measured point cloud in multiple time phases, and the accuracy of the model is comprehensively evaluated through multiple quantitative indicators. Based on the model evaluation results, a multi-factor error model is constructed, multi-dimensional influencing factors are introduced, and an adaptive weighting mechanism is used to correct the model error.

[0058] Specifically, by spatially comparing the virtual point cloud predicted by the growth model with multi-temporal measured point clouds, the model accuracy is comprehensively evaluated using quantitative indicators such as point cloud overlap, mean Euclidean distance, and Hausdorff distance. The reliability of the prediction results is verified from two dimensions: geometric consistency and spatial fit. To address the impact of different tree species characteristics, terrain undulations, and environmental factors on model accuracy, a multi-factor error analysis model is constructed to systematically diagnose model biases. Furthermore, adaptive weight adjustment and dynamic parameter correction strategies are employed to iteratively optimize the growth model, achieving high-precision fitting and robust prediction for various tree types in complex power transmission channel environments.

[0059] Among them, the mean Euclidean distance (MSE) reflects the overall point cloud error; the Hausdorff distance (HD) assesses the maximum local bias; and the point cloud overlap (OR) measures the spatial consistency between the predicted and measured results. Based on the evaluation, a multi-factor error model is constructed, incorporating influencing factors such as tree species type, terrain slope, environmental climate, and acquisition angle. An adaptive weighting mechanism is used for error correction, and combined with an iterative optimization strategy, the growth model parameters are automatically adjusted to achieve dynamic updates and self-learning optimization of the prediction model's accuracy.

[0060] It should be noted that, in order to ensure the reliability of the model prediction, a comprehensive accuracy evaluation system was established in this step. The predicted point cloud and the measured point cloud were spatially compared, and the error was quantitatively evaluated using indicators. At the same time, a multi-factor error correction model can be established based on multiple influencing factors such as different tree species, terrain slope, climate conditions and sampling accuracy. Through Bayesian optimization, adaptive weighted adjustment and feedback update mechanism, the model parameters are continuously optimized to improve the generalization ability and robustness of the algorithm.

[0061] This invention can deeply integrate high-precision UAV point clouds with tree growth models, breaking through the bottleneck of traditional static monitoring methods that cannot achieve time-series prediction. It realizes quantitative modeling, three-dimensional prediction and visualization of dynamic changes in vegetation in power transmission channels, and has extremely high engineering application value.

[0062] The above is an illustrative scheme of a tree species growth prediction method based on multi-temporal point clouds and growth models according to this embodiment. It should be noted that the technical solution of this tree species growth prediction system based on multi-temporal point clouds and growth models belongs to the same concept as the technical solution of the tree species growth prediction method based on multi-temporal point clouds and growth models described above. Details not described in detail in the technical solution of the tree species growth prediction system based on multi-temporal point clouds and growth models in this embodiment can be found in the description of the technical solution of the tree species growth prediction method based on multi-temporal point clouds and growth models described above.

[0063] Example 2: This example provides a tree species growth prediction system based on multi-temporal point clouds and growth models, including: The acquisition and processing module is used to acquire multi-temporal data of the target area and perform data preprocessing; The identification and segmentation module is used to identify vegetation regions and segment individual trees based on preprocessed multi-temporal data of vegetation point clouds, extract the structural features of each tree, and generate a temporal feature dataset. The model building module is used to build a tree growth model. It takes a time-series feature dataset as the model input, introduces point cloud structure parameters as dynamic constraints, and optimizes the model parameters to obtain the prediction curves of multidimensional tree growth indicators. The visualization module is used to predict curves based on multidimensional growth indicators. It performs dynamic geometric reconstruction using raw point cloud data to simulate the temporal growth of trees and generate interactive, visualized 3D models.

[0064] This embodiment also provides a computer device suitable for tree species growth prediction based on multi-temporal point clouds and growth models, including: The system includes a memory and a processor. The memory stores computer-executable instructions, and the processor executes these instructions to implement a tree species growth prediction method based on multi-temporal point clouds and growth models, as proposed in the above embodiments.

[0065] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements a tree species growth prediction method based on multi-temporal point cloud and growth model as proposed in the above embodiment.

[0066] The storage medium proposed in this embodiment and the tree species growth prediction method based on multi-temporal point cloud and growth model proposed in the above embodiment belong to the same inventive concept. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.

[0067] Based on the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part 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 a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0068] Example 3, referring to Tables 1-2 and Figures 2-5 As an embodiment of the present invention, a tree species growth prediction method based on multi-temporal point cloud and growth model is provided. In order to verify its beneficial effects, it is scientifically demonstrated through economic benefit calculation and simulation experiment.

[0069] This embodiment takes a 110kV Changhua power line in a certain province as the experimental object, and selects UAV lidar multi-temporal point cloud data from 2023 and 2025 for research. By registering, segmenting and extracting structural features of tree point cloud data in the same area, and combining tree growth models and dynamic reconstruction algorithms, the quantitative prediction and three-dimensional visualization of the growth process of individual trees are realized.

[0070] This method can dynamically analyze the temporal changes in tree height, canopy volume, and spatial morphology, and then calculate the minimum spatial distance from the treetop to the conductor, realizing a quantitative analysis framework based on "growth trend—safe distance—risk prediction". This not only provides precise support for vegetation risk early warning and inspection scheduling of transmission lines, but also provides scientific basis and technical assurance for power grid companies in refined management, ecological maintenance, and intelligent operation and maintenance decisions in transmission corridors.

[0071] The system operating environment in this embodiment mainly consists of two parts: a hardware environment and a software environment. It aims to support the high-performance computing needs for multi-temporal point cloud data processing, growth modeling, 3D dynamic reconstruction, and visualization analysis. The hardware environment includes a UAV platform equipped with a high-precision LiDAR, possessing centimeter-level ranging accuracy and multi-echo capability, capable of acquiring high-density 3D point clouds; a workstation / server configured with a high-performance CPU, GPU, large-capacity memory, and high-speed SSD storage for point cloud processing, model calculation, and visualization rendering; and auxiliary measurement equipment, including GNSS / IMU sensors, for precise UAV attitude and coordinate positioning. The software environment supports Windows 11 operating system to ensure the stability of large-scale point cloud processing; point cloud processing and modeling software PCL and CloudCompare are used for point cloud filtering, segmentation, registration, and 3D reconstruction; the point cloud processing framework R Studio is used for point cloud segmentation and tree feature recognition; and the visualization module supports 3D dynamic display, time-series playback, and spatial distance analysis between trees and guide lines.

[0072] To acquire high-precision 3D point cloud data within the corridor area, and to ensure spatial consistency and temporal comparability, the same UAV platform and LiDAR system were used for both time phases of data acquisition, following the same flight path planning and operational parameter settings. The UAV platform was a multi-rotor vertical takeoff and landing UAV equipped with a lightweight, high-precision LiDAR sensor, possessing centimeter-level ranging accuracy and multi-echo scanning capabilities, enabling efficient acquisition of surface and vegetation structure information in complex forest environments. During operations, a fixed altitude and constant speed were used, with the LiDAR scanning angle set within ±30°, and the point cloud density maintained at 200–300 points per square meter to ensure complete coverage of the tree canopy and guide wire structure. During flight operations, a GNSS / IMU integrated navigation system was used simultaneously to achieve high-precision calculation of the UAV's attitude and trajectory. Ground-based control points were used for exterior orientation element correction and accuracy verification to ensure consistency between the point cloud spatial coordinates and the geographic coordinate system. After data acquisition, trajectory optimization, point cloud synchronization, and waveform fusion processing were performed using LiDAR post-processing software to generate a unified format LAS point cloud file. Ultimately, two high-precision, multi-view, and fully covered 3D point cloud datasets of the power transmission corridor were obtained.

[0073] Furthermore, after completing the multi-temporal UAV lidar point cloud data acquisition, the raw point cloud data undergoes systematic preprocessing. The main objective of this step is to achieve noise suppression, ground point separation, temporal registration, and spatial standardization of the point cloud data, in order to construct a high-quality, multi-temporal comparable point cloud dataset. Ultimately, the registration accuracy of the point cloud is controlled within 5cm to ensure the temporal comparability of geometric features between individual trees.

[0074] Finally, all point cloud data were uniformly converted to the CGCS2000 coordinate system and subjected to region clipping and normalization. Using the centerline of the transmission line corridor as the reference axis, the corridor bandwidth was limited to generate a standardized multi-temporal point cloud dataset. The above preprocessing steps resulted in noise-reduced, accurately registered, and structurally clear point cloud data.

[0075] Furthermore, based on multi-temporal point cloud data acquired by UAV lidar, a gridded canopy modeling method is used to automatically segment individual trees and extract key structural parameters. The preprocessed vegetation point cloud data, after ground point removal, is divided into gridded raster units, dividing the target area into 0.5 m × 0.5 m grid cells to generate a canopy height model. Local extremum detection is performed on the generated CHM to identify the crown apex (i.e., local maxima) of each tree. After segmentation, structural parameters are calculated for the point cloud of each tree region, mainly including tree height, crown width, crown volume, crown density, and spatial centroid. The growth changes of trees in different years are analyzed, and the characteristics of individual trees at different time points are compared. A matching relationship is established through the spatial centroid distance and geometric feature differences, represented as: Where C represents the spatial centroid coordinates of a single tree point cloud, typically taken as the average position (x, y, z) of all points in the 3D point cloud; H represents the tree height, which is the difference between the maximum Z-value of a single tree in the point cloud and the ground height. , , For weighting coefficients, when At that time, it was determined to be the same tree. The matching threshold is typically set to 0.5–1.0 meters. It can be calibrated experimentally to balance matching accuracy and misjudgment rate. It effectively eliminates the influence of differences in point cloud density and flight attitude, enabling cross-temporal tracking of a single tree.

[0076] Finally, a multi-temporal structural parameter table for each tree is generated, providing accurate input for subsequent growth modeling, as shown in Table 1.

[0077] Table 1: Multi-temporal structural parameters of trees

[0078] Further, growth model construction and parameter fitting are performed. After extracting individual tree features from multiple time periods, this step constructs a growth model based on multi-time period feature parameters to achieve quantitative prediction of tree growth trends and dynamic structural analysis. Statistical optimization methods are then used to fit and update the model parameters, enabling dynamic prediction of key growth indicators such as tree height, crown width, and volume. First, using the time series feature dataset D={(t i H i Ci V i Based on )}, where t i H represents the observation time (in years). i C i V i Let Y(t) represent the tree height, crown width, and volume at time i, respectively. To characterize the growth pattern of trees over time, this embodiment uses the Logistic growth model as the basic function form. The spatial growth prediction model for a single tree point cloud, Y(t), is expressed as: Where Y(t) represents the growth of a tree at time t (such as tree height, crown width, or volume); K represents the maximum potential growth value (i.e., the theoretical upper limit of growth); a is the initial growth correlation coefficient, which determines the starting point of the growth curve; b is the growth rate parameter, which controls the growth speed and inflection point position. Based on multi-temporal point cloud data collected by UAV lidar, it quantifies the actual growth of a single tree in space, including changes in tree height, crown width, and branch and leaf density. This model is used to accurately calculate the spatial distance between the tree and the guide wire, and to conduct risk assessment and early warning.

[0079] To enhance the model's adaptability to different tree species and growing environments, point cloud geometric feature constraints are further introduced to construct a coupled "morphological constraint-temporal evolution" model, represented as: Where f(Gs,Ds) is a point cloud-based morphological feature function, including canopy density Gs and trunk diameter distribution Ds; α is a morphological constraint coefficient used to balance the growth trend term and structure correction term of the model.

[0080] The final model can output the predicted structural parameter sequence {H} for each tree at future time points. t+1 C t+1 V t+1 This provides precise input for subsequent 3D dynamic reconstruction and growth visualization.

[0081] Furthermore, three-dimensional dynamic reconstruction and prediction visualization are performed. After completing the parameter fitting and dynamic prediction of the tree growth model, this step realizes the three-dimensional dynamic reconstruction and visualization growth simulation of trees in the Unreal Engine 5 environment based on the extracted point cloud structural features and the output results of the growth model.

[0082] Furthermore, to verify the reliability and accuracy of the growth prediction model, this step involves quantitative analysis of the spatial differences between the measured point cloud and the model prediction results, and comparative evaluations are conducted between different tree species and different models.

[0083] The evaluation metric uses the point cloud Euclidean distance deviation rate to assess the model's prediction accuracy. The calculation formula is as follows: Wherein, Ntol represents the number of predicted points whose distance error between the predicted and measured points is less than the tolerance threshold, and Ntotal represents the total number of predicted points. The tolerance values ​​are set to 0.1m, 0.3m, 0.5m, 0.8m and 1.0m respectively to evaluate the applicability of the model under different accuracy requirements.

[0084] Based on this, error distribution analysis and Bayesian optimization algorithm are further used to adaptively correct the model parameters, and the optimization objective function is: in, For the model prediction points, For actual measurement points, This is the set of model parameters. Through multiple rounds of parameter iteration and error inversion, the average prediction accuracy of the model is improved by approximately 8%–12% under different terrains, tree species, and time scales.

[0085] The experiment took the 3-4 section of the 110kV long-line corridor as the research object, and selected typical tree species (fir and pine) for comparative analysis, including the model of this invention, the Gompertz growth model, and the Richards growth model. The results are shown in Table 2.

[0086] Table 2: Accuracy Statistics of Growth Models

[0087] The experimental results show that, under a relatively small tolerance of 0.1m, the prediction accuracy of each model is significantly affected by lidar ranging noise and point cloud resolution, with accuracy concentrated in the 20–30% range. When the tolerance is increased to 0.3m, the overall prediction accuracy of the models is significantly improved, with accuracy exceeding 55%, and the Gompertz model reaching the highest at 65.05%. Under a tolerance of 1.0m, the prediction accuracy of all three models is close to or exceeds 80%, indicating that the models can reflect the growth trend of trees at a macro scale relatively well.

[0088] A comprehensive comparison of the fitting performance of the three models reveals that the Gompertz model performs best for trees with low rates of change in the early stages of growth but tending towards stability later (such as pine trees); the Richards model demonstrates high flexibility and stability in the medium- to long-term prediction phase; while the Logistic model is suitable for overall trend assessment and rapid estimation scenarios. In subsequent systems, selecting the Gompertz model as the primary growth prediction model, combined with the Richards model for error correction and dynamic optimization, can effectively improve the overall prediction accuracy and robustness.

[0089] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A tree species growth prediction method based on multi-temporal point cloud and growth model, characterized in that, The method comprises the following steps: Collecting multi-temporal data of the target area and performing data preprocessing; Based on the preprocessed multi-temporal data, the vegetation point cloud is identified and the single tree is segmented, and the structural features of each tree are extracted to generate a time series feature dataset; Constructing a tree growth model, taking the time series feature dataset as the model input, introducing point cloud structure parameters as dynamic constraint terms, and optimizing the model parameters to obtain the prediction curve of the multi-dimensional growth index of the tree; Based on the prediction curve of the multi-dimensional growth index, dynamic geometric reconstruction is performed on the original point cloud data to simulate the time series growth of the tree and generate an interactive visual three-dimensional model.

2. The tree species growth prediction method based on multi-temporal point cloud and growth model of claim 1, wherein, Collecting multi-temporal data of the target area and performing data preprocessing, including: Obtaining multi-temporal three-dimensional spatial data of the target area at different time nodes, and performing standardization preprocessing on the original point cloud data; The preprocessing includes noise removal, ground separation processing, acquisition coordinate unification, and point cloud sparsification and registration.

3. The method of claim 1, wherein, The vegetation point cloud is identified and the single tree is segmented, and the structural features of each tree are extracted, including: Based on the multi-temporal data of the vegetation, the vegetation area is automatically identified and the single tree is separated; Combining the geometric features of the tree, the features of the single tree are extracted, the key structural parameters of the tree are calculated, and a one-to-one correspondence between the single tree and the multi-temporal point cloud is established. The similarity of the key structural parameters is matched to generate a continuous time series feature dataset of the single tree.

4. The method of claim 1, wherein, Constructing a tree growth model includes: Based on the change rule of multi-temporal feature data, a tree growth model is constructed which integrates geometric structure parameters and physiological growth rules. The model constructs the change trend of tree shape over time through a nonlinear function, and takes the three-dimensional structure parameters of the tree as dynamic constraint terms to construct the relationship between tree structure growth and time evolution.

5. The method of claim 4, wherein, Parameter optimization of the model includes: Based on the tree growth model, the least squares method is used to fit the initial parameters of the multi-temporal observation data, and the Bayesian optimization and confidence interval constraint mechanism are introduced to globally optimize and adaptively update the model parameters; Through continuous iteration and correction of the model parameters, the individual growth of trees under different tree species and different habitat conditions is predicted, and the model outputs the prediction curve of the multi-dimensional growth index and the confidence interval to evaluate the growth trend of the tree in the future period.

6. The method of claim 1, wherein, Generating an interactive visual three-dimensional model includes: Based on the multi-dimensional growth index predicted by the model, dynamic geometric reconstruction is performed on the original point cloud data, the time variation of the tree structure parameter features is mapped into the spatial coordinate system, and the time series three-dimensional reconstruction and dynamic visualization of the individual tree are performed to generate a continuous time series three-dimensional interactive model, showing the structural changes and growth trend of the tree at different growth stages.

7. The tree species growth prediction method based on multi-temporal point cloud and growth model as described in claim 4, characterized in that, Constructing a tree growth model also includes: Comparing the virtual point cloud generated by the tree growth model with the multi-temporal measured point cloud in space, and comprehensively evaluating the model accuracy through multiple quantitative indicators; Based on the model evaluation results, a multi-factor error model is constructed, multi-dimensional influence factors are introduced, and an adaptive weighting mechanism is used to correct the errors of the model.

8. A tree species growth prediction system based on multi-temporal point clouds and growth models, applying a tree species growth prediction method based on multi-temporal point clouds and growth models according to any one of claims 1 to 7, characterized in that, ​ The acquisition processing module is configured to acquire multi-temporal data of a target region and perform data preprocessing; The recognition and segmentation module is configured to perform vegetation region recognition and single tree segmentation on the vegetation point cloud based on the preprocessed multi-temporal data, extract structural features of each tree, and generate a time-series feature dataset; The model construction module is configured to construct a tree growth model, use the time-series feature dataset as model input, introduce point cloud structure parameters as dynamic constraint terms, and perform parameter optimization on the model to obtain a prediction curve of a multi-dimensional growth index of the tree; The visualization module is configured to perform dynamic geometric reconstruction on original point cloud data based on the prediction curve of the multi-dimensional growth index, simulate time-series growth of the tree, and generate an interactive visual three-dimensional model.

9. A computer device, comprising: Comprise: a memory and a processor; The memory is configured to store computer executable instructions, and the processor is configured to execute the computer executable instructions, and the computer executable instructions, when executed by the processor, implement the steps of the tree species growth prediction method based on multi-temporal point cloud and growth model according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The memory has stored computer executable instructions, and the computer executable instructions, when executed by the processor, implement the steps of the tree species growth prediction method based on multi-temporal point cloud and growth model according to any one of claims 1 to 7.

Images