Tree wind load evaluation method based on point cloud model

Abstract

The present application relates to a kind of tree wind load evaluation methods based on point cloud model, by the image collected by unmanned aerial vehicle is carried out semantic segmentation, extract crown part image is carried out three-dimensional reconstruction, generate point cloud model, extract the point cloud data of crown and trunk, generate target tree three-dimensional point cloud model.The wind load influence area of crown is calculated by Alpha Shape algorithm.Combining wind speed, wind direction and tree geometric characteristics, the wind load distribution under different wind direction and wind speed conditions is simulated using ANSYS Fluent software, and the wind load result under wind load is calculated.Finally, based on the finite element analysis method, the stress analysis of tree structure is carried out combined with wind load data, and the fatigue life and damage risk of tree under the action of wind force are evaluated.The method can accurately evaluate the dynamic response of tree under the action of wind force, with the characteristics of high efficiency, accuracy and automation, widely applicable in the field of wind disaster warning, tree health monitoring and environmental protection.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of tree evaluation, and particularly relates to a tree wind load evaluation method based on a point cloud model. BACKGROUND

[0002] With climate change and the increase in extreme weather events, the threat of wind disasters to cities is gradually increasing. The behavior of trees under the action of wind, including bending, swaying and even breaking, has become an important content of research on wind disaster prevention. Especially in high wind weather, the wind resistance of trees is directly related to environmental safety, public facilities and the safety of life and property. Currently, tree wind load evaluation mainly relies on manual observation and traditional wind force models, but these methods usually have problems such as insufficient evaluation accuracy, poor real-time performance and limited scope of application.

[0003] Traditional tree wind load evaluation methods mainly use simple wind tunnel experiments or estimation methods based on past experience, which cannot accurately reflect the dynamic response of trees in complex environments. With the development of technology, image collection based on unmanned aerial vehicles and three-dimensional reconstruction technology provide new possibilities for tree wind load evaluation. However, in existing technologies, research on how to use images to generate accurate tree point cloud models and conduct automatic evaluation and structural analysis of wind load based on these models is still in its infancy, and comprehensive analysis of multi-directional wind load has not been effectively solved. SUMMARY

[0004] The purpose of the present application is to overcome the shortcomings of the prior art and provide a tree wind load evaluation method based on a point cloud model.

[0005] The purpose of the present application can be achieved by the following technical solutions:

[0006] The present application proposes a tree wind load evaluation method based on a point cloud model to overcome the shortcomings of the prior art. This method uses unmanned aerial vehicle image collection, semantic segmentation and point cloud reconstruction technology, combined with wind load calculation and tree structure stress analysis, to more accurately evaluate the wind resistance of trees under the action of wind, and to provide scientific basis for wind disaster warning and emergency response.

[0007] The present application provides a tree wind load evaluation method based on a point cloud model, comprising the following steps:

[0008] Use an unmanned aerial vehicle to take multiple-angle photos of the target tree to obtain images of the target tree;

[0009] Label the images using an image semantic segmentation algorithm to generate labels for each pixel in the image, including tree crown, tree trunk and background;

[0010] construct a preliminary three-dimensional point cloud model based on the labeled images, each point cloud data in the preliminary three-dimensional point cloud model comprising label information;

[0011] generate a target tree three-dimensional point cloud model according to the preliminary three-dimensional point cloud model and the label information thereof;

[0012] calculate a wind load influence area of a tree crown according to the target tree three-dimensional point cloud model;

[0013] simulate multi-directional wind loads based on the wind load influence area of the tree crown, a wind speed, and geometric features of the target tree, obtain a wind load simulation result, and the geometric features of the target tree are obtained according to the target tree three-dimensional point cloud model;

[0014] perform tree structure stress analysis in combination with the wind load simulation result, and predict a damage risk of the tree under the action of wind force.

[0015] Further, the labeling of the images by the image semantic segmentation algorithm generates a label for each pixel in the image, including the following steps:

[0016] perform a pretreatment operation on the collected images, including denoising and color correction;

[0017] input the pretreated images into a pre-trained convolutional neural network model for semantic segmentation, output a semantic segmentation image of each image, each pixel in the semantic segmentation image is assigned a label, and the convolutional neural network model comprises a MaskR-CNN model.

[0018] Further, the construction of the preliminary three-dimensional point cloud model based on the labeled images comprises the following steps:

[0019] perform multi-view three-dimensional reconstruction on the collected images using COLMAP three-dimensional reconstruction software to generate a point non-semantic three-dimensional point cloud model;

[0020] calculate three-dimensional coordinates of each pixel in the semantic segmentation image of each image in the non-semantic three-dimensional point cloud model, determine the point cloud data corresponding to the pixel according to the three-dimensional coordinates of each pixel, transfer the label of each pixel to the corresponding point cloud data, and generate a preliminary three-dimensional point cloud model.

[0021] Further, the multi-view three-dimensional reconstruction on the collected images using COLMAP three-dimensional reconstruction software to generate a point non-semantic three-dimensional point cloud model comprises the following steps:

[0022] extract feature points and their descriptors from each image using the SIFT algorithm, perform feature point matching between multiple images according to the feature points and their descriptors of each image using the nearest neighbor matching algorithm, and remove mis-matched feature points using the RANSAC algorithm;

[0023] Based on the matched feature points, a non-semantic three-dimensional point cloud model is constructed by an incremental reconstruction method using COLMAP.

[0024] Further, the target tree three-dimensional point cloud model is generated according to the preliminary three-dimensional point cloud model and the label information, and includes the following steps:

[0025] By the semantic segmentation map of each image and the geometric features of each point cloud data in the preliminary three-dimensional point cloud model, a geometric condition based on color, position and normal vector is used to remove irrelevant debris from the tree;

[0026] The point cloud data labeled as crown and trunk is extracted from the preliminary three-dimensional point cloud model;

[0027] The extracted point cloud data is simplified using a voxel grid algorithm, and the point cloud is divided into a grid in space to reduce the number of redundant points;

[0028] The SOR filtering algorithm is used to remove noise in the point cloud data to generate the target tree three-dimensional point cloud model.

[0029] Further, the wind load influence area of the crown is calculated according to the target tree three-dimensional point cloud model, and includes the following steps:

[0030] The point cloud data labeled as crown is extracted from the target tree three-dimensional point cloud model to generate a crown three-dimensional point cloud model;

[0031] The crown point cloud data in the crown three-dimensional point cloud model is projected along a predetermined direction to convert the crown three-dimensional point cloud model into two-dimensional plane data;

[0032] The three-dimensional coordinates in the crown three-dimensional point cloud model are converted into corresponding two-dimensional coordinates according to the predetermined direction to obtain a crown two-dimensional point set;

[0033] The Alpha Shape algorithm is used to calculate the boundary envelope of the crown two-dimensional point set to obtain the wind load influence area of the crown.

[0034] Further, the predetermined direction is dynamically adjusted according to the changes in wind speed and direction.

[0035] Further, the wind load influence area of the crown is calculated according to the following formula:

[0036]

[0037] Where A is the wind load influence area of the crown, (x i ,y i ) is the two-dimensional coordinate of the i-th point cloud data in the crown two-dimensional point set, and n is the total number of point clouds in the crown two-dimensional point set, (x n+1 ,yn+1 ) is considered as (x1, y1) to form a closed path.

[0038] Further, the wind load influence area of the crown, wind speed, and geometric characteristics of the target tree are combined to simulate multi-directional wind load and obtain wind load simulation results, including the following steps:

[0039] The wind resistance coefficients at the crown and trunk of the target tree are calculated according to the geometric characteristics of the target tree, the three-dimensional point cloud model of the crown is generated into a three-dimensional CAD model and imported into Fluent, the wind resistance coefficients at the crown and trunk are input into the ANSYS Fluent software, different wind speeds and wind pressure distributions generated in multiple directions are simulated using Fluent, the obtained flow field simulation results are analyzed to obtain the wind load simulation results, and the formula is:

[0040]

[0041] Where F is the wind load, C d is the wind resistance coefficient, A is the wind load influence area of the crown, p is the air density, and v is the wind speed.

[0042] Further, the wind load simulation results are combined to analyze the stress of the tree structure and predict the damage risk of the tree under the action of wind force, including the following steps:

[0043] Based on the three-dimensional point cloud model of the target tree, a finite element model of the tree is constructed, including the geometric shape, material properties, mechanical properties, and boundary conditions of the tree;

[0044] The wind load simulation results are input into the finite element model as external loads acting on each node or element of the finite element model, and the structural stress of the tree under the action of wind load is calculated, and the formula is:

[0045]

[0046] Where σ is the stress, F is the wind load, A cross is the cross-sectional area of the wind load action region;

[0047] According to the calculated structural stress, the fatigue strength and fatigue index of the target tree material are combined to predict the fatigue life of the target tree under the action of long-term wind load, and the formula is:

[0048]

[0049] Where N f is the fatigue life, Δσ is the stress amplitude, which is obtained according to the calculated structural stress, σ f is the fatigue strength of the target tree material, and β is the fatigue index of the target tree material.

[0050] Based on the predicted fatigue life of the target trees under long-term wind loads, the damage risk of the target trees under wind loads is assessed.

[0051] Compared with the prior art, the present invention has the following advantages:

[0052] (1) The present invention solves the shortcomings of the existing technology in wind load assessment and fatigue damage analysis of trees as a whole. Traditional methods often rely on static wind speed data and simplified mechanical models, which makes it difficult to accurately simulate the dynamic response of trees in complex wind environments and cannot comprehensively assess the stress state of trees and their potential risks. The present invention adopts advanced CFD simulation technology and tree geometric modeling methods to accurately simulate the dynamic response of trees under different wind speeds and wind directions, and evaluate the structural stability of trees in real time through stress analysis and fatigue calculation. Compared with the existing technology, the present invention can overcome the defects of traditional methods that ignore the dynamic response of trees and complex wind load effects, provide more accurate and comprehensive wind load assessment, and combine with urban emergency management systems, disaster prevention and mitigation systems, etc. This system can realize the prediction and analysis of tree damage in strong wind weather, provide a scientific basis for wind disaster warning and emergency response, and thus reduce the loss of public facilities and human life and property in wind disasters.

[0053] (2) This method uses drone image acquisition and image semantic segmentation algorithms to construct a high-precision three-dimensional point cloud model of the target tree. This method avoids the high cost and low efficiency of traditional manual measurement and can accurately extract tree structural information (such as crown and trunk), thereby providing more accurate geometric data for subsequent wind load analysis.

[0054] (3) This paper constructs point cloud models based on COLMAP 3D reconstruction software. Combined with image semantic segmentation results, this method extracts 3D tree geometry information through an automated algorithm, significantly improving the efficiency and accuracy of point cloud model construction. This method effectively removes background debris and noise, ensuring the accuracy and operability of the tree model.

[0055] (4) The present invention uses the Alpha Shape algorithm to calculate the wind load affected area of ​​the tree crown, which can dynamically adjust the wind load affected area of ​​the tree model under different wind direction conditions and accurately calculate the actual exposed area of ​​the tree crown, providing reliable data support for subsequent simulation and evaluation of wind loads.

[0056] (5) The application combines Fluent software to simulate multi-direction wind load, simulates multi-direction wind speed and wind pressure distribution through the input of three-dimensional CAD model of trees and wind resistance coefficient, and thus obtains accurate wind load simulation results. The technical means can comprehensively analyze wind load under different wind directions and wind speeds, and provides a more scientific basis for structural evaluation of trees under the action of wind force.

[0057] (6) The application combines finite element model to analyze the stress of tree structure, calculates the structural stress of trees under the action of wind load, and combines the fatigue strength and fatigue index of tree material to accurately predict the fatigue life of trees under the action of long-term wind force. The technology can provide a scientific basis for health evaluation and wind damage risk prediction of trees, and is helpful to realize early warning and management of wind damage risk of trees. BRIEF DESCRIPTION OF DRAWINGS

[0058] Figure 1 is the overall method flowchart of the application;

[0059] Figure 2 is the data acquisition and semantic segmentation flowchart of the application;

[0060] Figure 3 is the flowchart of point cloud model generation of the application;

[0061] Figure 4 is the flowchart of wind load simulation calculation of the application;

[0062] Figure 5 is the flowchart of trunk stress analysis and fracture prediction of the application. DETAILED DESCRIPTION

[0063] The technical solutions in the embodiments of the application will be clearly and completely described below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are part of the embodiments of the application, rather than all the embodiments of the application. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor should belong to the protection scope of the application.

[0064] Embodiment 1:

[0065] The embodiment provides a tree wind load evaluation method based on a point cloud model. Through technical means such as unmanned aerial vehicle image acquisition, image semantic segmentation, three-dimensional point cloud reconstruction, wind load simulation and structural stress analysis, the damage risk of trees under the action of wind force can be efficiently and accurately evaluated. As shown in Figure 1 , the following steps are included:

[0066] The target tree is photographed from multiple angles using an unmanned aerial vehicle to obtain the image of the target tree;

[0067] annotating the images using an image semantic segmentation algorithm to generate labels for each pixel in the images;

[0068] constructing a preliminary three-dimensional point cloud model based on the annotated images;

[0069] generating a target tree three-dimensional point cloud model according to the preliminary three-dimensional point cloud model and its label information;

[0070] calculating the wind load influence area of the tree crown according to the target tree three-dimensional point cloud model;

[0071] simulating multi-directional wind loads based on the wind load influence area of the tree crown, wind speed, and geometric characteristics of the target tree to obtain wind load simulation results;

[0072] conducting tree structure stress analysis in combination with the wind load simulation results to predict the damage risk of the tree under the action of wind force.

[0073] Specifically, the following contents are included:

[0074] The target tree is photographed from multiple angles by a drone to obtain image data of the tree. The images collected by the drone have high resolution and accuracy, and can comprehensively and meticulously obtain the shape information of the tree. The advantage of this step is that through the efficient shooting of the drone, a large amount of high-quality images can be obtained in a short time, avoiding the low efficiency and high cost of traditional manual measurement.

[0075] After obtaining the images, the images are annotated using an image semantic segmentation algorithm. Each pixel in the image is assigned label information, including tree crown, tree trunk, and background, etc. This process is carried out through a pre-trained convolutional neural network (such as MaskR-CNN), which can achieve accurate identification of different parts of the tree. Through this technology, the tree crown, trunk, and background can be effectively separated, and irrelevant area data can be removed, ensuring that the extraction of the tree part in the subsequent processing is more accurate. The advantage of this technical feature is that it efficiently processes images and achieves high-precision region annotation, reducing manual intervention and improving processing speed and accuracy.

[0076] Next, based on the annotated images, a preliminary three-dimensional point cloud model is constructed. COLMAP three-dimensional reconstruction software is used to perform multi-view three-dimensional reconstruction on the collected images to generate a point non-semantic three-dimensional point cloud model. This process extracts image feature points through the SIFT algorithm and uses the RANSAC algorithm to remove false matching points, finally realizing the fusion of multi-view point cloud data. Through this step, a relatively accurate three-dimensional point cloud model of the tree can be generated, providing basic data for subsequent wind load analysis. The advantage of this technical feature is that it utilizes the geometric relationship between images and automatically processes through algorithms, greatly improving the efficiency and accuracy of model construction, avoiding the tediousness and errors of manual modeling.

[0077] The parts information of the tree is extracted from the preliminary three-dimensional point cloud model to generate the target tree three-dimensional point cloud model. In this process, based on the color, position and geometric conditions of the normal vector, the sundries unrelated to the tree are removed, and the point cloud data labeled as the crown and the trunk are extracted. By using the voxel gridding algorithm to simplify the point cloud, the number of redundant points is further reduced, and the SOR filtering algorithm is used to remove the noise in the point cloud. The final generated target tree three-dimensional point cloud model accurately reflects the geometric shape of the tree, providing reliable data source for the calculation of the wind load influence area of the tree. The advantage of this technical feature is that by removing sundries, simplifying data and denoising, the accuracy and effectiveness of the point cloud data are ensured, and the errors that may occur in subsequent analysis are reduced.

[0078] Based on the target tree three-dimensional point cloud model, the wind load influence area of the crown is calculated. First, the point cloud data labeled as the crown is extracted from the target tree three-dimensional point cloud model to generate the crown three-dimensional point cloud model. Then, the crown three-dimensional point cloud model is projected along the preset direction to obtain the two-dimensional plane data of the crown. Next, the three-dimensional coordinates of the crown three-dimensional point cloud data are converted into two-dimensional coordinates to obtain the two-dimensional point set of the crown. The Alpha Shape algorithm is used to calculate the boundary envelope of the crown two-dimensional point set to obtain the wind load influence area of the crown. The specific calculation formula is:

[0079]

[0080] Wherein, A is the wind load influence area of the crown, (x i ,y i ) is the two-dimensional coordinates of the i-th point cloud data in the crown two-dimensional point set, n is the total number of point clouds in the crown two-dimensional point set, (x n+1 ,y n+1 ) is regarded as (x1, y1) to form a closed path. Through the calculation of two-dimensional projection and Alpha Shape algorithm, the effective wind load influence area of the crown can be accurately obtained, and then accurate basis is provided for wind load simulation.

[0081] In the wind load simulation stage, the wind resistance coefficient of the crown and the trunk of the target tree is calculated according to the geometric characteristics of the tree, and the crown three-dimensional point cloud model is generated into a three-dimensional CAD model and imported into ANSYS Fluent software. By simulating different wind speed and wind pressure distribution generated by multiple directions through Fluent, the wind load simulation result is finally obtained. The wind load calculation formula is:

[0082]

[0083] Wherein, F is the wind load, C dwhere C D is the wind drag coefficient, A is the wind load influence area of the tree crown, p is the air density, and v is the wind speed. By accurately simulating the wind load action on the tree under different wind speeds and directions, reliable data are provided for structural stress analysis. The use of Fluent to simulate wind field flow can consider multiple wind directions and different wind speeds, providing comprehensive and accurate results for wind load analysis of trees.

[0084] Combined with the wind load simulation results, the structural stress analysis of the tree is performed. Based on the three-dimensional point cloud model of the target tree, a finite element model of the tree is constructed, including the geometric shape, material properties, mechanical properties, and boundary conditions of the tree. The wind load simulation results are input into the finite element model as external loads acting on each node or element, and the structural stress of the tree under wind load is calculated. The structural stress calculation formula is:

[0085]

[0086] where s is the stress, F is the wind load, A cross is the cross-sectional area of the wind load action area; according to the calculated structural stress, combined with the fatigue strength and fatigue index of the tree material, the fatigue life of the tree under long-term wind load is predicted. The fatigue life calculation formula is:

[0087]

[0088] where N f is the fatigue life, and s is the stress amplitude, which is obtained according to the calculated structural stress, s f is the fatigue strength of the target tree material, and b is the fatigue index of the target tree material; through this step, the fatigue damage and damage risk of the tree under wind action can be accurately evaluated. The technical features have the advantages of providing scientific basis for the evaluation of the health status of the tree through the calculation of structural stress and fatigue life, and early warning of the damage risk caused by wind disasters.

[0089] In summary, the present embodiment systematically completes the damage risk assessment of trees under wind action through the technical means of unmanned aerial vehicle image acquisition, image semantic segmentation, three-dimensional point cloud reconstruction, wind load simulation, and finite element analysis. This method has the characteristics of high efficiency and precision, and can provide reliable technical support for wind disaster prevention and control, tree health monitoring, and other fields.

[0090] Example 2:

[0091] The parts not mentioned in this embodiment are the same as in Example 1.

[0092] The embodiment provides a tree wind load evaluation system based on a tree point cloud model, which comprises a data acquisition and semantic segmentation module, a point cloud model generation and processing module, a wind load simulation calculation module, a trunk stress analysis and fracture prediction module. Each module works cooperatively to provide a complete solution from data acquisition to final wind disaster prediction.

[0093] The data acquisition and semantic extraction module comprises a data acquisition submodule and an image semantic segmentation submodule, images are collected by a UAV and subjected to semantic segmentation by using a neural network, and a semantic segmentation image of each image is output, each pixel in the semantic segmentation image is assigned a label, and the label comprises a crown, a trunk and a background.

[0094] The data acquisition submodule uses a UAV to take multiple-angle photographs of a target tree by using a high-resolution camera, so that the image covers the entire tree area and is as high-quality as possible. The flight path of the UAV is accurately planned to ensure the overlap and coverage of the image, thereby facilitating subsequent three-dimensional reconstruction.

[0095] The image semantic segmentation submodule subjects the images taken by the UAV to semantic segmentation, and automatically identifies the crown area by using an open-source neural network model. This step can efficiently distinguish the crown from the trunk, environmental debris and other non-target areas, so as to avoid noise interference in the data. The model can process complex backgrounds and lighting conditions.

[0096] The point cloud model generation and processing module comprises a point cloud model generation submodule and a noise reduction processing submodule. The preliminary data obtained by the UAV image acquisition and semantic extraction module are further processed to generate a complete point cloud model.

[0097] The point cloud model generation submodule uses open-source three-dimensional reconstruction software such as COLMAP to perform multi-view three-dimensional reconstruction on the collected images to generate point cloud data. This process uses the SIFT algorithm to extract image feature points, combines the estimation of camera position and attitude, and performs incremental three-dimensional reconstruction by using methods such as nearest neighbor matching, relative pose constraint and five-point algorithm. The finally output point cloud data is a three-dimensional surface model of the tree, which contains all point cloud data of the crown, trunk and background environment.

[0098] The noise reduction processing submodule removes the debris irrelevant to the tree by using geometric conditions based on features such as color, position and normal vector, according to the semantic segmentation result and the geometric features of the point cloud data. The point cloud data is simplified by using a voxel gridding algorithm, the number of redundant points is reduced by dividing the point cloud into a grid in space, and the amount of calculation is reduced. In addition, the SOR filtering algorithm is used to remove noise in the point cloud data, so as to eliminate noise data caused by factors such as changes in light and weather, so as to obtain point cloud data of the crown area with uniform density and high accuracy.

[0099] The wind load simulation calculation module includes an influence area calculation submodule and a wind load simulation calculation submodule. Based on the point cloud model obtained through previous processing, the module calculates the effect of wind load on the tree in combination with wind speed, tree structure, and other parameters. The module calculates the wind load on the tree under different wind speeds and wind directions through physical modeling and computational fluid dynamics simulation, and evaluates the structural stability of the tree:

[0100] The influence area calculation submodule projects the crown point cloud data along a given direction. The projection process converts the three-dimensional point cloud into two-dimensional plane data, forming a point set. This point set represents the cross-sectional morphology of the crown in that direction. The projection direction is dynamically adjusted according to changes in wind speed and direction. In view of the irregular graph characteristics of the crown part, this module uses the Alpha Shape algorithm to calculate the envelope of the point set. By performing Alpha Shape processing on the crown point cloud data, the shape of the crown can be accurately captured, and its wind load influence area can be calculated.

[0101] The wind load simulation calculation submodule measures the complexity or roughness of the surface by analyzing the density of the point cloud, and calculates the wind resistance coefficient of the tree under different projection morphologies. For example, dense areas (such as leaves) have a higher wind resistance coefficient, while sparse areas (such as bare branches) have a lower wind resistance coefficient.

[0102] The extracted crown point cloud data is used to generate a three-dimensional CAD model and imported into Fluent. The function of the wind resistance coefficient is input into the ANSYS Fluent software, and Fluent is used to simulate different wind speeds and wind pressure distributions generated by multiple orientations. By analyzing the obtained flow field simulation results, the pressure distribution and wind load results of each region can be obtained, and the structural stress and fracture risk of each part of the tree can be further calculated.

[0103] Due to factors such as the shape and layout of the crown and leaves, the tree will create an obstacle to the airflow, changing the flow speed and pressure distribution. This local wind flow change will affect the calculation of wind load. This module uses Fluent to simulate the wind channel effect and provides visualization tools to display streamline, velocity vector diagram, and wind pressure distribution diagram, which can intuitively show the direction and speed of the wind flow, and thus can display the possible wind force intensification or weakening in local areas of the forest, helping to identify areas that may be severely affected.

[0104] The trunk stress analysis and fracture prediction module uses the wind pressure and other results obtained through the above CFD simulation, in combination with the ANSYS Mechanical structural analysis software, to realize deformation and displacement analysis and dynamic response analysis of the tree under wind load, and further judge the fracture risk of the tree when encountering strong winds.

[0105] The point cloud model of the tree and information such as material properties are input into the software, and the wind load data information obtained by CFD simulation is input to perform structural stress analysis.

[0106] Through fluid-structure coupling (FSI) analysis, the deformation and structural response of the tree under wind load are simulated. This method not only considers the static effect of wind load, but also can simulate the instantaneous response of the tree under the dynamic action of wind, such as vibration, sway, etc.

[0107] Fatigue fracture analysis: Trees will experience periodic wind load action in the wind, leading to fatigue damage. ANSYS Mechanical can perform fatigue analysis to evaluate the fatigue life of the tree under long-term wind load action and identify areas that may experience fatigue damage. In the weak parts of the tree, such as the connection between branches and the trunk, wind load is easy to cause the tree to break or be damaged. By simulating the fracture caused by wind load through fracture mechanics analysis, the risk of fracture of the tree when encountering strong wind can be judged.

[0108] Instability and collapse analysis: For taller trees, instability and collapse may occur when encountering strong winds. Through wind load simulation results, the risk of instability and collapse of the tree under strong wind can be evaluated.

[0109] Example 3:

[0110] The part not mentioned in this example is the same as in Example 1.

[0111] This example relates to a tree wind load evaluation system based on a tree point cloud model, mainly composed of the following parts:

[0112] The data acquisition and semantic analysis module includes a data acquisition sub-module and an image semantic segmentation sub-module.

[0113] The point cloud model generation and processing module includes a point cloud model generation sub-module and a noise reduction processing sub-module.

[0114] The wind load simulation calculation module includes an influence area calculation sub-module and a wind load simulation calculation sub-module.

[0115] The trunk stress analysis and fracture prediction module includes fluid-structure coupling and fatigue and instability analysis sub-modules.

[0116] The data acquisition and semantic segmentation process is shown in Figure 2 , including the following steps:

[0117] The UAV flies according to the predetermined trajectory and takes pictures. The flight trajectory should consider that the images in each direction have overlapping pixels, and should cover all angles of the tree in order to perform subsequent three-dimensional reconstruction.

[0118] The images extracted by the unmanned aerial vehicle are preprocessed, images that are unclear due to occlusion and light occlusion are removed, and the images are imported into a neural network;

[0119] The neural network performs semantic segmentation on the processed images, accurately distinguishes trees from backgrounds, extracts tree crown regions, and serves as a source image for subsequent point cloud model generation.

[0120] The process of generating a point cloud model is shown in Figure 3 , which includes the following steps:

[0121] The results of semantic segmentation are imported into three-dimensional reconstruction software, which converts images into three-dimensional point cloud data;

[0122] The point cloud data is simplified using voxel gridding and SOR filtering algorithms to remove redundancy and noise, ensuring the clarity and accuracy of the point cloud data;

[0123] The software uses the SIFT algorithm to extract key feature points and performs three-dimensional reconstruction based on the extracted feature points to generate a three-dimensional surface model of the tree. Geometric conditions based on color, position, and normal vector are used to remove irrelevant debris from the tree, ensuring that the point cloud model has high three-dimensional restoration and can accurately reflect the three-dimensional information of the tree for subsequent wind load calculation.

[0124] The process of wind load simulation calculation is shown in Figure 4 , which includes the following steps:

[0125] The tree crown point cloud data is projected along a given direction, and the data obtained by two-dimensional plane projection forms a point set representing the cross-sectional shape of the tree crown in that direction. The Alpha Shape algorithm is used to select an appropriate radius parameter to calculate the area of the irregular image, which is the area affected by wind load in that direction;

[0126] The complexity or roughness of the surface is measured by analyzing the density of the point cloud, and the wind resistance coefficient under different projection shapes of the tree is calculated for CFD simulation;

[0127] Three-dimensional point cloud data and wind resistance coefficient function information are imported into ANSYS Fluent for multi-directional wind speed and wind pressure distribution simulation. The software can simulate the equivalent effect of wind force and provide visualization tools to help identify areas susceptible to wind load. CFD simulation obtains pressure field and wind load distribution data, and can display the streamline and vortex structure formed by the fluid around the tree.

[0128] The process of trunk stress analysis and fracture prediction is shown in Figure 5 , which includes the following steps:

[0129] The wind pressure data obtained after simulating the wind load, together with the point cloud model and the material properties of the trees, are imported into the ANSYS Mechanical structural analysis software for analysis;

[0130] Through fluid-structure coupling (FSI) analysis, the deformation and structural changes of the trees under the wind load are simulated, and the displacement and vibration response of the trees under the effect of different wind forces can be analyzed;

[0131] Combined with the FSI analysis results, fatigue fracture risk analysis is performed on the parts of the trees that are prone to fatigue damage (such as the connection between the branches and the trunk), and the instability collapse risk of the higher trees under strong wind is evaluated. The risk assessment results can be reflected in the upper management system to further realize the risk warning function.

[0132] If the above functions are realized in the form of software function units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the present application or the parts that contribute to the prior art can be embodied in the form of a software product, which is stored in a storage medium and includes instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the methods described in various embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various program code storage media.

[0133] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present application, and these modifications or replacements should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method for tree wind load assessment based on point cloud model, characterized in that, The method comprises the following steps: Multi-angle shooting of target trees is performed using a UAV to obtain images of the target trees; An image semantic segmentation algorithm is used to label the images to generate labels for each pixel in the images, the labels including a crown, a trunk, and a background; A preliminary three-dimensional point cloud model is constructed based on the labeled images, each point cloud data in the preliminary three-dimensional point cloud model including label information; A target tree three-dimensional point cloud model is generated according to the preliminary three-dimensional point cloud model and the label information thereof; Wind load influence area calculation of the crown is performed according to the target tree three-dimensional point cloud model; Based on the wind load influence area of the crown, wind speed, and geometric characteristics of the target tree, multi-directional wind load is simulated to obtain wind load simulation results, the geometric characteristics of the target tree being obtained according to the target tree three-dimensional point cloud model; In combination with the wind load simulation results, tree structure stress analysis is performed to predict the damage risk of the tree under the action of wind force; The wind load influence area calculation of the crown according to the target tree three-dimensional point cloud model comprises the following steps: Point cloud data with a label of the crown is extracted from the target tree three-dimensional point cloud model to generate a crown three-dimensional point cloud model; The crown point cloud data in the crown three-dimensional point cloud model is projected along a preset direction to convert the crown three-dimensional point cloud model into two-dimensional plane data; Three-dimensional coordinates in the crown three-dimensional point cloud model are converted into corresponding two-dimensional coordinates according to the preset direction to obtain a crown two-dimensional point set; Alpha Shape algorithm is used to perform boundary envelope calculation on the crown two-dimensional point set to obtain the wind load influence area of the crown; The multi-directional wind load is simulated based on the wind load influence area of the crown, wind speed, and geometric characteristics of the target tree to obtain wind load simulation results, and the method comprises the following steps: Wind resistance coefficients at the crown and the trunk of the tree are calculated according to the geometric characteristics of the target tree, the crown three-dimensional point cloud model is converted into a three-dimensional CAD model and imported into Fluent, the wind resistance coefficients at the crown and the trunk are input into ANSYS Fluent software, Fluent is used to simulate different wind speed and wind pressure distribution generated in multiple directions, the obtained flow field simulation results are analyzed to obtain the wind load simulation results, and the formula is: wherein, is the wind load, is the wind drag coefficient, is the wind load influence area of the crown, is the air density, is the wind speed.

2. The tree wind load assessment method based on point cloud model according to claim 1, characterized in that, The image is labeled by using an image semantic segmentation algorithm to generate labels for each pixel in the image, and the method comprises the following steps: Preprocessing operations are performed on the collected images, including denoising and color correction; The preprocessed images are input into a pre-trained convolutional neural network model for semantic segmentation, and semantic segmentation images of the images are output, each pixel in the semantic segmentation images being assigned a label, and the convolutional neural network model comprises a Mask R-CNN model.

3. The tree wind load assessment method based on point cloud model according to claim 1 or 2, characterized in that, The preliminary three-dimensional point cloud model is constructed based on the labeled images, and the method comprises the following steps: Multi-view three-dimensional reconstruction is performed on the collected images using COLMAP three-dimensional reconstruction software to generate a non-semantic three-dimensional point cloud model; The three-dimensional coordinates of each pixel in the semantic segmentation images of each image in the non-semantic three-dimensional point cloud model are calculated, the three-dimensional coordinates of each pixel are used to determine the corresponding point cloud data, the labels of each pixel are transmitted to the corresponding point cloud data, and a preliminary three-dimensional point cloud model is generated.

4. The tree wind load assessment method based on point cloud model according to claim 3, characterized in that, The use of COLMAP three-dimensional reconstruction software for multi-view three-dimensional reconstruction of the collected images generates a non-semantic three-dimensional point cloud model, including the following steps: Using SIFT algorithm to extract feature points and their descriptors from each image, using nearest neighbor matching algorithm to match feature points between multiple images according to the feature points and their descriptors of each image, using RANSAC algorithm to remove mis-matched feature points; Based on the matched feature points, COLMAP is used to construct a non-semantic three-dimensional point cloud model by incremental reconstruction method.

5. The tree wind load assessment method based on point cloud model according to claim 1 or 2, characterized in that, The target tree three-dimensional point cloud model is generated according to the preliminary three-dimensional point cloud model and its label information, including the following steps: By the semantic segmentation map of each image and the geometric features of each point cloud data in the preliminary three-dimensional point cloud model, the geometric conditions based on color, position and normal vector are used to remove irrelevant debris from the tree; Extracting point cloud data labeled as tree crown and trunk from the preliminary three-dimensional point cloud model; Using voxel grid algorithm to simplify the extracted point cloud data, grid dividing the point cloud in space, reducing the number of redundant points; Using SOR filtering algorithm to remove noise in point cloud data, generating target tree three-dimensional point cloud model.

6. The tree wind load assessment method based on point cloud model according to claim 1, characterized in that, The preset direction is dynamically adjusted according to the changes of wind speed and direction.

7. The method of claim 1, wherein, The wind load influence area of the tree crown is calculated by the formula: wherein, is the wind load influence area of the tree crown, is the two-dimensional coordinate of the i-th point cloud data of the two-dimensional point set of the tree crown, is the two-dimensional coordinate of the i-th point cloud data of the two-dimensional point set of the tree crown, is the total number of point clouds of the two-dimensional point set of the tree crown, is considered as forms a closed path.

8. The tree wind load assessment method based on point cloud model according to claim 1, characterized in that, The wind load simulation results are combined to analyze the stress of the tree structure and predict the damage risk of the tree under the action of wind force, including the following steps: Based on the target tree three-dimensional point cloud model, a finite element model of the tree is constructed, which includes the geometric shape, material properties, mechanical properties and boundary conditions of the tree; The wind load simulation results are input into the finite element model as external loads acting on each node or element of the finite element model, and the structural stress of the tree under the action of wind load is calculated, the formula is: wherein, is the stress, is the wind load, is the cross-sectional area of the wind load action region; According to the calculated structural stress, combined with the fatigue strength and fatigue index of the target tree material, the fatigue life of the target tree under the action of long-term wind load is predicted, the formula is: wherein is the fatigue life, is the stress amplitude, obtained from the calculated structural stresses, is the fatigue strength of the target tree material, is the fatigue index of the target tree material; According to the predicted fatigue life of the target tree under the action of long-term wind load, the damage risk of the target tree under the action of wind load is evaluated.

Images