Reconstruction Of Individual Trees By Using Terrestrial Laser Scanning Point Clouds

3D geometric tree models are of great interest to many applications,such as digital city and digital forestry,among others.It can enhance the reality and immersion of the digital scenes.However,reconstructing 3D geometric models of individual trees with high precision and fidelity is one of the difficult problems in the digitization process of forestry.Terrestrial laser scanning technology has been widely used in many fields because of its advantages of short data acquisition period,high accuracy,large amout of points and high-efficiency.Applying it to reconstructing tree models can provide important technical support for extracting structure parameters and protecting some ancient trees.For the serious incompleteness and dramatically varying point density of the terrestrial laser scanning point clouds,a discriminative and robust method need to address the problems of uneven point density and point missing and is used to reconstructe 3D tree models.In this paper,the particularity of the terrestrial laser scanning point clouds and data acquisition are considered to reconstructe individual tree models.The main research content and contributions are as follows:（1）Aiming to address the problem of points missing,a missing point clouds restoration method based on local point cloud features is proposed,which can successfully extract the skeleton points and efficiently repair missing poins.Specific,the initial skeleton points of missing single tree point clouds were extracted by L₁-Median algorithm firstly,and then the dominant direction and local weight density of the point cloud were calculated based on the skeleton points.Besides,the contraction force and gravitational force were calculated to guide the relevant neighboring points moving freely toward the missing region.When the force balance between gravitational force and contraction force is reached,the motion and iterative repairing process is terminated,and the repairing optimization of the missing branch point cloud is completed.（2）In order to address the problem of reconstructing individual trees,a skeleton extraction algorithm coupled with L₁-Median and minimum spanning tree（MST）is proposed,which can successfully abstract tree skeleton structure and realize tree modeling.Specific,the skeleton points of single tree were extracted by MST,and then the L₁-Median branch points were used to correct the skeleton extension.This method not only effectively maintains the local characteristics of canopy-scale twigs,but also maintains the correct extension of trunk skeleton topology.Aiming to ensure the lightweight and natural extension of the skeleton structure,some measuers such as skeleton points weight adjustment,skeleton simplification and skeleton smoothing were refined.Finally,the algorithm combined by the local cylinder fitting and allometric growth theory was proposed to calculate the radius of skeleton points,and the tree models were renconstructed.The results show that the algorithm proposed in this paper can construct 3D models of single trees with different tree species and structures and retain more detailed information of twigs.（3）Taking individual trees of 7 different spscies as examples,Qualitative and quantitative analysis of the methods we proposed was carried out.The results show that the repairing algorithm proposed in this paper can effectively repair the missing branch points.Besides,with the increase of the number of iterations,the effect of the repair point cloud is gradually better.The three-dimensional model constructed by the repaired point clouds has a high degree of agreement with the model under the complete point cloud,and the repairing algorithm is not sensitive to the noise and density.In addition,the reconstruction method can successfully reconstruct a high precision and high fidelity tree model.Compared with measured DBH,the accuracy of DBH obtained by the model can reach 3cm.Moreover,the average distance between the point cloud and the model surface can reach 3cm,which proved that the point cloud has a high degree of agreement with the model.In terms of algorithm performance,the reconstruction algorithm proposed in this paper is robust to parameter controling and density,and it can abstract more detail information of the twigs comparing to other methods.