基于激光点云的森林树木结构参数提取

Extraction of Forest Tree Structure Parameters Based on Laser Point Cloud

以河北塞罕坝林区激光雷达数据为研究对象,利用点云魔方软件CSF滤波、DEM不规则三角网格内插、基于点云分割单木提取森林树木结构参数,以抽样检测方式开展精度分析,并通过优化CSF滤波的分类阈值提高单木提取精度,获得了理想的效果。结果显示,点云魔方CSF分类阈值对精度影响存在最优解。对比0.50、0.55、0.60、0.65、0.70 m 5个分类阈值,阈值为0.60 m时精度最高,其单木提取查全率为82.1%,查准率为85.2%,树高提取中误差0.87 m,标准差0.75 m。研究也显示,点云魔方在处理海量森林点云数据时表现出较强的分析和处理能力,其算法多样、参数设置灵活以及可视化的数据信息显示,给用户带来全新的体验。

Taking the LiDAR data of Saihanba forest area in Hebei province as the research object,this paper uses CSF filtering,DEM irregular triangular mesh interpolation,single tree segmentation based on point cloud in the software PCM v2.0 to extract the structural parameters of forest trees.The paper also carries out accuracy analysis by sampling detection,and improves the single tree extraction accuracy by optimizing the classification threshold of CSF filtering,and obtains ideal results.The results show that there is an optimal solution for the influence of the threshold of PCM v2.0 CSF classification on the accuracy.Compared with the five classification thresholds of 0.50 m,0.55 m,0.60 m,0.65 m and 0.70 m,the accuracy is the highest when the threshold is 0.60 m.The recall rate of single tree extraction is 82.1%,the precision rate is 85.2%,the mean square error of tree height extraction is 0.87 m,and the standard deviation is 0.75 m.The research also shows that the point cloud cube shows strong analyzing and processing ability when processing massive forest point cloud data,with diverse algorithms,flexible parameter settings and visual data information display,which brings a new experience to users.