地面激光雷达点云数据乔灌分离方法研究

Terrestrial laser scanning point cloud data tree-shrub separation method research

针对当前基于地面激光雷达点云数据的乔灌分离方法在不同环境中很难准确地实现分离的问题,提出了一种粗分离和精分离相结合的方法:先对点云数据进行预处理;结合法向量和支持向量机对此进行粗分离;利用RANSAC算法拟合圆柱、结合格网点云数标准差变化筛选乔灌,以及基于点云数标准差的自适应DBSCAN进行最终的精分离。选取了8块不同生长环境的样地进行测试,其乔木提取株数准确率均高于93%。实验结果表明:该方法能够实现不同环境下的地面激光雷达点云高效准确地乔灌分离,并与现有的传统方法相比,其精度和效率更高、普适性更强,为后续植被参数、林木株数准确提取奠定了基础,具有广泛的应用前景。

For the current terrestrial laser scanning LiDAR point cloud data,it is difficult to accurately realize the separation of trees and shrubs point cloud in different environments.This paper proposes a combination of coarse and fine separation methods:firstly,the point cloud data are preprocessed;then combined with normal vectors and support vector machines for coarse separation;finally,the RANSAC algorithm is utilized to fit the columns,combined with the standard deviation of the number of point clouds in the grids to screen the arbor and adaptive DBSCAN based on the standard deviation of the number of cloud points to carry out the final fine separation.In this paper,eight sample plots with different growing environments were selected for testing,and the accuracy of the number of trees extracted was higher than 93%in all cases.The experimental results show that the method can realize efficient and accurate separation of trees and shrubs from ground LiDAR point clouds in different environments,and has higher precision and efficiency and stronger universality than the existing traditional methods,which lays the foundation for the accurate extraction of the subsequent vegetation parameters and the number of forest tree strains,and has a wide range of application prospects.