基于动态圆柱拟合的背包激光雷达单木骨架曲线提取

Extraction of single tree skeleton curves based on dynamic cylinder fitting by using backpack laser scanning

【目的】为了精确构建树木的三维模型,基于背包激光扫描(Backpack laser scanning,BLS)系统获取的单木点云数据,提出了一种动态圆柱拟合方法提取树木真实的骨架曲线。【方法】以广西壮族自治区南宁市高峰林场为研究区。从样地点云数据中分离出单棵杉木点云,对其地面点与树叶点进行滤波处理,得到单木枝干点云。根据树木单枝近似圆柱特性,提出动态圆柱拟合算法,将单木枝干点云分割成大量单枝点云,并对单枝点云进行优化。分别对优化后单枝点云的XYZ坐标进行主成分分析,计算其特征值,最大特征值所对应的特征向量即为主方向。基于体元逐层聚类,沿主方向进行骨架点的提取,并利用3次B样条曲线对单枝骨架点进行平滑处理以获取单枝骨架曲线。利用同向向量近邻点搜索方法确定单枝骨架曲线连接点,实现骨架曲线拼接,从而实现完整单木骨架曲线的提取。【结果】对样地单株杉木进行骨架曲线提取,设置3种动态圆柱递减参数δ。递减参数为0.3 cm时,点云利用率为95.04%,处理时间为515.05 s;递减参数为0.5 cm时,点云利用率为92.70%,处理时长为369.34 s;递减参数为0.8 cm时,点云利用率79.70%,处理时长为349.35 s。参数为0.3 cm时,点云利用率最高,能够非常完整的提取骨架曲线;参数为0.5 cm时,点云利用率可以满足骨架曲线的提取,并且算法运行时间相比于参数0.3 cm时降低了28.29%;参数为0.8 cm时,部分单枝存在缺失现象,无法满足骨架曲线提取要求。【结论】将递减参数设置为0.5 cm,利用动态圆柱拟合方法提取背包激光扫描单木点云的骨架曲线能够表征单木的几何、拓扑结构。

【Objective】To accurately construct three-dimensional model of trees,a dynamic cylindrical fitting method was proposed to extract the true skeleton curve of single tree based on the point cloud database obtained by backpack laser scanning(BLS)system.【Method】Gaofeng forest farm in Nanning city,Guangxi Zhuang Autonomous region was taken as the study area.A single Cunninghamia lanceolate bough point cloud that filtered ground points and leaf points was isolated from the sample site point cloud database.According to approximate cylindrical characteristics of tree sticks,a dynamic cylindrical fitting algorithm was proposed to divide the single tree’s bough into plenty of independent sticks.Eigenvalues of principal component analysis were performed under the XYZ coordinates of the optimized independent sticks respectively,the eigenvector corresponding to the largest eigenvalue was the main direction.Based on the layer-by-layer clustering of voxels,skeleton points of independent sticks were extracted along the main direction and smoothed by cubic B-spline curve.Connection points of sticks’skeleton curve were determined and spliced by the same direction vector nearest neighbor,so as to realized the extraction of the complete single skeleton curve.【Result】The skeleton curve was extracted from single Cunninghamia lanceolate on the experimental plot,and three dynamic cylindrical decline parameterδwas set.Under 0.3cm of the decrement parameter,the utilization rate of point clouds and the processing time is 95.04%and 515.05 s separately;under 0.5 cm of the decrement parameter,the utilization rate of point clouds and the processing time is 92.70%and 369.34 s separately;under 0.8 cm of the decrement parameter,the utilization rate of point clouds and the processing time is 79.70%and 349.35 s separately.Under the parameter is 0.3 cm,the utilization rate of point cloud is the highest,and the skeleton curve can be extracted completely;under the parameter is 0.5 cm,the utilization rate of point cloud can meet the extraction of the skeleton curve,and the algorithm running time is reduced by 28.29%compared with the condition of 0.3 cm;under the parameter is 0.8 cm,some sticks are missing,which unable to meet the requirements of skeleton curve extraction requirements.【Conclusion】Under the condition that decrement parameter is 0.5 cm,skeleton curve of backpack laser scanning single tree point clouds which extracted by dynamic cylinder fitting method is able to accurately characterize the geometry and topological structure of single tree.