Research Of Forest Parameter Estimation Based On UAV And TLS LiDAR Point Cloud Data

Lidar technology,as an emerging active remote sensing technology,can obtain forest structure parameters and understory topography information by transmitting laser pulses and receiving return information.At present,LiDAR data from different platforms have been applied to the inversion studies of forest parameter information such as tree height,diameter at breast height,crown width,wood volume,biomass,and stocking volume.However,the different working modes of UAV LiDAR and TLS LiDAR lead to the lack of intra-forest information in UAV point clouds and the lack of forest canopy information in TLS point clouds.It is difficult for a single platform LIDAR to obtain a complete 3D point cloud of forest trees,and the fusion of the two is beneficial to eliminate the blind areas of their respective scans and estimate more accurate information of forest parameters.Based on this,this paper proposes a point cloud alignment method based on the relationship between ground features and tree positions for fusing near-ground multi-source LiDAR point clouds.The main research of this paper is as follows:(1)An unmarked automated alignment method for UAV and TLS LiDAR point clouds is proposed based on the relationship between ground features and tree locations in forest areas.Firstly,the ground point clouds are extracted from the UAV point cloud and the TLS point cloud respectively by using the improved progressive encrypted triangular mesh filtering algorithm,and the initial alignment parameters are obtained by using the random sampling consistency algorithm based on the similar fast point feature histogram features of both ground point clouds to complete the initial alignment.Then the tree position points are extracted from the initial alignment UAV point cloud and TLS point cloud as the alignment primitive to build Delaunay triangle network,and find the triangle pairs with the same name based on the principle of angular similarity of triangles,and finally use the singular value decomposition method to get the rotation translation parameters to complete the fine alignment.The experimental results showed that the average horizontal offset distance of trees in the Quercus mongolica plot was 0.173 m.The average horizontal offset distance of trees in Pinus sylvestris plot was 0.283 m.This method effectively achieves the accurate fusion of LiDAR point cloud from UAV and TLS LiDAR point cloud in forest area.(2)Based on the classic idea of segmenting UAV LiDAR point clouds based on the Kmeans algorithm,an improved K-means hierarchical clustering segmentation algorithm is proposed.The core idea is to use the tree trunk point cloud to generate seed points as the underlying initial clustering center,transfer the clustering center layer by layer for K-means clustering,and finally complete the single tree point cloud segmentation by longitudinally fusing the tree point cloud clusters belonging to the same clustering center.The comprehensive segmentation accuracy F-score values of this method in the Quercus mongolica sample plot and Pinus sylvestris sample plot were 0.72 and 0.84,respectively.Compared with other commonly used segmentation algorithms,such as the watershed algorithm,PCS algorithm and layer stacking algorithm,this method is more suitable for the segmentation of fused point clouds and effectively reduces the under-segmentation phenomenon caused by low trees being obscured by tall trees when segmenting based on the top-down canopy seed points.(3)Extract tree height,diameter at breast height and crown width information based on single tree point cloud data,and then apply the allometric growth equation to calculate the sum of the organ biomass of the trunk,branches,leaves and roots of the tree as a single tree The biomass of each tree was calculated using the binary volume table,and the single tree volume was further summarized as the volume of the quadrat.Finally,linear regression analysis was carried out between the estimated tree parameters and the real reference values.The results showed that the estimated forest parameters had high fitting accuracy,and the fusion of UAV and TLS point clouds could realize accurate estimation of tree parameters,which confirmed the joint Feasibility of near-surface multi-source LiDAR technology in obtaining forest information parameters.In summary,the near-ground LiDAR point cloud alignment method proposed in this paper effectively achieves the deep fusion of UAV and TLS point clouds,providing a new solution for the joint use of multi-source LiDAR point clouds in forestry.Compared with a single platform point cloud,the method obtains the 3D point cloud information of forest trees in more detail,providing a reliable data base for rapid and complete acquisition of forest tree configuration information.This research result provides strong technical support for the joint multi-source LiDAR technology to be deeply applied in forestry,and also helps to promote the joint application of multi-source LiDAR technology in the 3D reconstruction of forest trees and the fine survey of forest resources.