Visual Modeling Of Single Wood Based On Multi-station Laser Point Cloud

Tree visualization modeling has important significance and broad prospects for tree growth simulation and forestry resource inventory.With the continuous upgrading of 3D laser scanning technology and the continuous innovation of visualization technology,the visual simulation of plants through point cloud information has gradually become a popular field of forestry visualization.Compared with traditional laser scanners,current laser scanning equipment has greatly improved scanning accuracy,scanning speed,and matching accuracy.On this basis,problems such as incomplete data and data loss caused by insufficient accuracy of scanning equipment in the past have been well resolved.The multi-station terrestrial laser scanner(TLS)method can collect more complete and high-precision tree point cloud data,thereby greatly improving the analysis ability of the internal structure of the tree.Therefore,this paper takes different types of single-wood point clouds obtained by multi-station laser scanning as the research object,analyzes the morphological structure and parameters of the tree point clouds,and builds a visual model to realize the virtual tree model display.The specific content is divided into the following parts:(1)Scheme planning and data preprocessing.It is mainly divided into four steps: data acquisition,registration,filtering and height normalization.Using the Trimble TX8 scanner and planning a multi-site collection plan,collecting different types of tree point cloud data in different seasons;using Trimble Realworks software to achieve the initial point cloud registration and secondary iterative correction;on the basis of acquiring the fusion point cloud,Using statistical filtering to achieve the initial denoising of the tree point cloud;on the basis of single tree segmentation,using the tree trunk centerline and the fitting ground plane intersection method,through separate analysis of the characteristics of the degree of inclination of each tree and the ground,achieving height normalization Correction.Through experimental analysis and selection of the optimal number of neighboring points and outlier threshold,the integrity of the tree morphology is ensured while removing interference points,which is very beneficial to the subsequent reconstruction process.(2)Extraction algorithm of single tree parameters.In this paper,algorithms such as local extremum,cubic B-spline curve fitting,and adaptive threshold voxel segmentation are used to extract single tree parameters.On the basis of height normalization,by finding the intersection of the tree trunk centerline and the fitted ground plane as the lowest point of tree height measurement,and then using the local extremum method to find the highest point of the trunk in the tree point cloud,according to the highest point and the lowest point The distance between the points calculates the tree height.In the process of extracting the diameter at breast height of the tree trunk,considering the irregular characteristics of the tree branches,the traditional least squares round fitting curve does not conform to the curve of the naturally growing tree diameter at breast height.Therefore,the B-spline curve fitting is introduced,and the curve fitting algorithm for tree diameter at breast height is innovatively constructed for the point cloud with large deviation in the circle fitting process,which improves the accuracy of tree diameter at breast height parameter extraction.In the process of extracting the height parameters of the tree branches,a 3cm*3cm*5cm voxel network is first constructed,and then an adaptive threshold segmentation method that is different from the traditional fixed threshold is used and multi-view analysis is added.The minimum value of the segmentation threshold is used as the final tree height parameter.Finally,the traditional taper model is used to calculate the tree taper equation.Experimental results show that,compared with the previous parameter extraction algorithms,the algorithm proposed in this paper reduces the extraction errors of DBH(Diameter at Breast Height),CBH(Crown Base Height)and tree height parameters respectively by 29.4%,15.6% and 31.7%,the corresponding RMSE is 0.532 cm,0.092 m and 0.327 m,respectively,and the method in this paper has high effectiveness and robustness for parameter extraction of different types of trees.(3)Skeleton extraction and visual modeling.In this paper,a support vector machine(SVM)algorithm based on neighborhood features is used to separate branches and leaves of a single tree point cloud.On this basis,skeleton extraction was realized and visual modeling was realized through generalized cylinder fitting.Firstly,the tree point cloud is pre-classified by KDtree,and then the neighborhood feature of the pre-classified point cloud is extracted as the feature factor for distinguishing branches and leaves in SVM,and the point cloud data after branch and leaf separation is obtained through SVM training.Then,based on the branch and leaf point cloud separation,the skeleton points are extracted,and the extracted skeleton points are connected and smoothly optimized.It uses cylindrical fitting centroid extraction based on least squares method,KDtree algorithm node combined with the shortest distance method of skeleton point connection,and spatial B-spline fitting algorithm-based skeleton connection curve optimization.Finally,on the basis of the obtained skeleton structure,the branch-stem relationship equation is introduced,and the single tree visualization model is reconstructed by the generalized cylindrical fitting method.Experimental results prove that when processing point cloud data of various tree types,the method in this paper can achieve better skeleton extraction and skeleton connection,with a segmentation accuracy of 84.2% and a connection accuracy of 97%.(4)Single wood visualization system construction and interface design.Based on the analysis of the basic functions and advantages of Matlab GUI and Open GL open graphics library,this paper uses Matlab and C++ as programming languages and Matlab GUI as the platform to combine the drawing functions and visualization enhancement processing in Open GL to achieve single-wood parameter extraction,Single wood visualization system and interface with structural analysis and 3D model reconstruction functions.