Development And Application Of Tree Quantitative Structure Models Based On LiDAR Point Clouds

The reconstruction of tree structure not only enhances the sense of reality of 3D scene,but also provides an effective method for scientific management of forest vegetation.Traditional tree modeling methods lack detailed information to describe tree branch geometry.Light Detection and Ranging(Li DAR)point clouds contain spatial information such as the relative positions of points,local topologies,and global geometry,providing detailed information to reconstruct the precise tree structure.At present,the research of tree modeling lacks attention to the fine degree and geometric precision of reconstructed tree structure.The algorithm is poor in generality,easy to produce a large number of deformed structures,and unable to accurately extract tree factors.Quantitative Structure Model(QSM)is a fine modeling method of individual tree structure based on the Terrestrial Laser Scanner(TLS)point clouds.It is also a geometric model of individual tree that describes complete woody parts of trees on the ground in a hierarchical order.The research of QSM has important theoretical and application value.This paper will provide a new,Accurate and Detailed QSM-Ad QSM.It is very suitable for reconstruction of TLS point clouds,and is also applicable to UAV-based Laser Scanner(UAVLS)point clouds and Mobile Laser Scanner(MLS)point clouds.The main research contents and research results are as follows:(1)A general method for automatic reconstruction of fine geometric structure of individual tree branch was proposed.To accurately fit the geometric structure and topological relations of tree branches from different types(TLS,ALS,MLS)of point clouds,this paper based on the graph theory method,combined with the advantages of prior hypothesis modeling and lightweight expression to reconstruct the geometric structure and topology of individual tree branches,and proposed a comprehensive and high precision method for the reconstruction of fine geometric structure of individual tree.?1 Firstly,the input point clouds was triangulated based on 3D Delaunay,and the shortest path of the directed weighted graph was calculated based on Dijkstra algorithm.Then,the minimum spanning tree(MST)algorithm framework was used to extract the initial skeleton of tree,and the merging algorithm of redundant vertices and edges was designed to simplify and optimize the initial skeleton.Finally,based on the tree skeleton,the cylinder fitting and optimization of individual tree structure were.?2 Multiple tree point clouds of different sizes and shapes in different regions(tropical,subtropical,and temperate)were collected based on TLS,MLS,and UAV-ALS.The accuracy evaluation method of single tree geometric structure reconstruction was studied,and the accuracy of Ad QSM reconstruction was verified.The Average distance(AD)and Standard Deviation(SD)from point clouds to the surface of geometric structure were less than 3cm and 8cm,respectively.?3 Compared with the mainstream method,the multi-dimensional Ad QSM reconstruction of individual tree geometry structure was studied.The level of Ad QSM reconstruction branches can be increased by1?3 levels.The modeling efficiency of Ad QSM was 2.5?3 times that of Tree QSM and 3?8.3 times that of Simpletree.(2)A quantitative tree factor model based on the fine geometric structure of single tree branch was constructedThe conventional tree modeling method is not accurate enough to reconstruct the branch structure,which makes it impossible to quantitatively extract the tree factors.In this paper,based on the reconstruction of the fine geometric structure of single tree,combined with the theory of tree measurement,an automatic measurement model of tree factors was constructed.The biggest advantage of Ad QSM is that it can automatically extract tree factors such as the volume of branches and trunks,which are difficult to be measured directly by traditional tree measurement methods.?1 A series of quantitative calculation methods for tree factors were designed on the basis of the fine geometric structure reconstruction of single tree.In addition to automatically extracting basic tree measurement factors such as DBH and tree height,it can also automatically extract total tree volume,branch volume(length,number,grade and angle),trunk volume(length and diameter),etc.?2 Based on the TLS point cloud,71 large tropical trees from different regions were reconstructed,and the measured data after felling of the trees were used as reference values to verify the accuracy of the total tree volume,branch volume,DBH and tree height extracted by Ad QSM.Comparing the DBH and tree height extracted by Ad QSM with the reference values,the linear fitting R~2were 0.98 and 0.93,and the r RMSE were 16.86%and 10.35%,respectively.The total tree volume and branch volume extracted by Ad QSM were compared with the reference values.The linear fitting R~2 were0.98 and 0.96,and the r RMSE were 24.63%and 38.32%.A method to estimate the aboveground biomass(AGB)of individual trees by combining the volume of trees extracted from Ad QSM with the basic density of wood was studied.The AGB reference value obtained from tree drying and weighing ranged from0.02629Mg to 43.94535Mg,the consistent correlation coefficient(CCC)between the estimated AGB value and the reference value was 0.98,and the r RMSE was 26.26%.?3 Compared with the mainstream methods in multi factor and multi dimension,the r RMSE of tree height,whole tree volume,branch volume and AGB extracted by Ad QSM is 1%?4%higher than the mainstream methods.(3)Researched a new method based on Ad QSM in forestry1 A new method for establishing biomass equation without felling trees was proposed.The traditional method of establishing the biomass equation requires cutting down trees,which is time-consuming,laborious and destructive.When it is impossible to cut down trees due to policy or economic cost,we can use Ad QSM combined with TLS point clouds to develop biomass equation.This method has the advantage of non-destructiveness without cutting down trees,and saves the cost of manpower and material resources.The biomass equations were established for 15 different species of trees.The results showed that there was little difference between the non-destructive TLS-Ad QSM method and the traditional destructive method.The slope was 1.12 and 1.05,intercept was-4.83 and-4.65,R~2 was 0.95and 0.98.2 This paper realizes the leap of QSM research from single-tree level modeling to plot level modeling.This paper proposes the concept of"Plot Quantitative Structure Model"for the first time,which is a quantitative structure model realized by combining single tree segmentation under the point clouds of sample plot level.A large scale tree modeling method based on Ad QSM was studied,and the real three-dimensional geometric structure of a circular sample plot with a radius of 20m was reconstructed after the realization of single tree segmentation.The volume of 181 trees was extracted automatically and the AGB of the whole plot was estimated to be 15889.3kg combined with the wood density.Using the sample site survey data as the reference value,AGB was overestimated by 17.12%.At the 95%confidence interval level,the CCC between reference value and estimated value based on"Plot Quantitative Structure Model"was 0.87,and the LLCI and UICI were 0.83 and 0.90,respectively.3 A new low-cost modeling method based on Ad QSM combined with terrestrial close-range photogrammetry point clouds was proposed.Based on forest terrestrial photogrammetry technology,the photogrammetric point clouds of a circular plot with a radius of 30m was collected.Ad QSM reconstructed the tree structure model of 44 dry willows and extracted forest parameters such as trunk volume,DBH and tree height.Taking the plot survey data as the reference value,the DBH extracted by Ad QSM was compared with the reference value,and the linear fitting R~2 and CV(RMSE)were 0.94 and 6.60%,respectively.Comparing the tree height extracted by Ad QSM with the reference value,the linear fitting R~2 and CV(RMSE)were 0.95 and 12.34%.Comparing the trunk volume extracted by Ad QSM with the reference value,the linear fitting R~2 was 0.74,the RMSE was 0.12369m~3,and the CV(RMSE)was 32.78%.Relying on the relatively low cost of obtaining terrestrial photogrammetric point clouds,it not only expands the scope of application of Ad QSM,but also provides a low-cost tree structure modeling technical solution for forest surveys.In summary,this paper provides a new model of tree structure modeling(Ad QSM)and new methods based on Li DAR point clouds.