基于点云体素化的单木叶面积指数地基遥感反演

Estimation of Individual Tree Leaf Area Index of Terrestrial Remote Sensing Inversion Based on Point Cloud Voxelization

以黑松、马尾松、荔枝、相思树为对象,采用地基激光雷达获取点云数据,基于点云体素化理论,分割叶片点云,建立"叶片—体素"的投影关系,研究体素化理论中尺度因子与点云密度对叶面积指数(LAI)反演精度的影响,实现单木LAI的高精度获取。结果表明,样木反演的LAI随着尺度因子的增大而增大,最优尺度区间为1.2~1.3,样木LAI最优反演精度范围为93.3%~99.9%,决定系数R^2为0.989 3,反演结果与实测LAI具有较高相关性;在最优尺度区间下,样木LAI反演精度随着点云密度的降低而减小,最大精度为98.63%,最小精度为84.14%,点云密度对单木LAI反演精度影响不大。

Terrestrial laser scanning was used to obtain point cloud of Pinus thunbergii,P.massoniana,Litchi chinensis and Acacia confusa,and blade point cloud segmentation to establish a “blade-voxel” projection relationship.The influences of the scale factor and point cloud density on LAI inversion accuracy were examined by using point cloud voxel voxelization,to simulate LAI of individual tree precisely.The results showed that the LAI inversion of sample wood increased with the increase of scale factor,where the optimal scale ranged from 1.2 to 1.3.The optimal LAI inversion accuracy of three tree species ranged from 93.3% to 99.9% and determination coefficient （R2） was 0.989 3,which exhibited relatively high correlation of LAI inversion results and the measured.Based on the optimal scale range,LAI inversion accuracy of sample wood decreased with the loss of the point cloud density,in which maximum precision was 98.63%,the minimum accuracy was 84.14%,and point cloud density did not affect LAI simulation of individual tree.