Inversion Of Forest Characteristics Using UAV LiDAR Single Tree Point Cloud

Forests play a key role in terrestrial ecosystems,and the variables extracted from single trees can be used in various fields and applications for evaluating forest production and assessing forest ecosystem services.Unmanned aerial vehicles using light detection and ranging(UAV LiDAR)with high spatial resolution have shown great potential in forest applications because it can penetrate the forest canopy,map the three-dimensional structure of the forest,and map the terrain information under the forest.Individual tree segmentation using LiDAR data has become one of the hotspots of forestry research.In this study,we proposed two methods of individual tree segmentation directly based on point cloud,and evaluated the proposed method using datasets acquired by a Velodyne 16 E LiDAR system mounted on a multi-rotor UAV.(1)Mean shift-NCut individual tree segmentation: For the normalized non-ground point clouds,the mean shift algorithm was used for rough segmentation,and the iterative NCut algorithm was used for fine segmentation.The results showed that the proposed method achieves DET,OM and COM of 0.90,0.10,and 0.11,respectively,in delineating single trees.(2)Self-adaptive mean shift tree segmentation: The optimal kernel bandwidth was determined according to the corresponding crown diameter,then the single tree point cloud was segmented by mean shift.The method was iteratively implemented within a given area until all trees are segmented.The results showed that the proposed method achieves DET,OM and COM of 0.87,0.13,and 0.09,respectively,in delineating single trees.(3)In this paper,the two methods are compared with the watershed segmentation and the fixed bandwidth mean shift.Comparative analysis demonstrated that our methods both provided a promising solution to reliable and robust segmentation of single trees from UAV LiDAR data with high point cloud density.Mean shift – NCut shows the highest segmentaion accuracy but takes the longest time.The segmentaion accuracy of self-adaptive mean shift is slightly lower than mean shift-NCut but greatly improves computational efficiency.(4)Based on the individual tree point cloud,we extracted the height quantile,density quantile and canopy variables and estabilshed the prediction model of forest characteristic parameters based on stepped-regression model.The research shows that the inversion of forest characteristic parameters based on UAV LiDAR single wood point cloud has shown great potential...