Methods On Broadleaf Forests Individual Tree Segmentation And Leaf-Wood Separation Using Terrestrial LiDAR Data

Forests play a crucial role in the global carbon,water,energy cycle,biodiversity,and climate change.Forest structural and functional parameters,such as tree height,diameter at breast height(DBH),crown width,aboveground biomass(AGB),and leaf area index(LAI),characterizing the spatial strcture and biophysical status of forests,and are of great significance for forest resource management and vegetation ecology research.Through high-density and high-precision point cloud data sampling,terrestrial Li DAR can record the details of branches and leaves for every single tree,and has become an important tool for estimating forest structural and functional parameters.Individual tree segmentation and leaf-wood separation are key prerequisites for accurately estimating various forest parameters using terrestrial Li DAR.In response to the weak crowns delineation ability of current individual tree segmentation methods for dense broadleaf forests and the poor robustness of current leaf-wood separation methods to tree species and size,this thesis proposed a individual tree segmentation method for broadleaf forests that utilizes branch information to guide crown segmentation,and a leaf-wood separation method for individual trees that fully utilizes the shortest path information.The main research contents and results of this thesis are as follows:(1)Proposed a region grow-based individual tree segmentation method(GrowSeg):According to the truth that woods form the basic framework of a tree with leaves attached to them,GrowSeg segmented the crowns of dense broadleaf forest accurately through four major steps: trunk detection,undershoot trunk extraction,branch extraction,and crown segmentation.Two plots(Magnolia plot and Phoebe plot)were used for method testing,which involved different tree species composition,understory vegetation density,terrain conditions,and point cloud quality.The mean Intersection over Union(m Io U)was greater than 0.8,and the relative root mean square error(r RMSE)of tree height and crown width estimation were less than 5% and 12%,respectively.GrowSeg outperformed the classical method CSP in both trunk detection and crown segmentation,especially in handling situations where large and small tree crowns are intersected or multiple tree crowns are tightly enclosed.(2)Developed a graph-based leaf-wood separation method(GBS): As the shortest path can effectively reflect the structures for trees of different species and sizes,by fully utilizing the shortest path-based features,GBS is compact and robust.Ten types of tree data—covering tropical,temperate,and boreal species—with heights ranging from 5.4 to43.7 m,were used to test the method performance.The mean accuracy and kappa coefficient at the point level were 94% and 0.78,respectively,and our method outperformed two other state-of-the-art methods(Le Wo S and RF model).Through further analysis and testing,the GBS method exhibited a strong ability for detecting small and leaf-surrounded branches,and was also sufficiently robust in terms of data subsampling.This method further demonstrated the potential of the shortest path-based features in leafwood separation,and future work should focus on its utility in forest plots.