一种复杂背景下林木点云精细提取方法

A fine extraction method of forest point cloud in complex background

针对地形变化较大、灌木多样等复杂林下主干提取效率低、精度不高等难点,该文提出一种复杂背景下林木点云精细提取方法。首先,进行地面点滤波并切片截取下层点云;然后,计算下层点云的法向量,提取法向量在Z轴的分量,循环剔除不满足条件的点;最后,循环具有噪声的基于密度聚类(DBSCAN)算法进行点云分割,实现林下主干点提取。以香格里拉为研究区,选择2景具有典型代表性的地基激光雷达样地点云数据进行试验,结果表明:研究中样地一提取精度90%,样地二提取精度95%,都取得较高精度。该方法能完整地提取林下主干点,对灌木与树干粘连等复杂环境下的林木提取具有较好的适用性。

In view of the difficulties of low efficiency and low precision of tree trunk extraction in complex forest space, such as large terrain change and shrub diversity, a fine extraction method of tree point cloud in complex background was proposed. First, filter the ground points and slice the data in the lower layer, then calculate the normal vector of the lower layer point cloud, extract the component of the normal vector in the Z-axis, cycle out the points that do not meet the conditions, and finally cycle density-based spatial clustering of applications with noise(DBSCAN) method was used to segment the point cloud to achieve the extraction of the main points under the forest. Taking Shangri-La as the research area, the cloud data of two typical terrestrial laser scanning(TLS)sample sites were selected for the experiment.The results showed that the precision of one extraction of sample land was 90%,and that of two extraction of sample land was 95%.This method could extract the main points under the forest completely,and it had a good applicability for the extraction of trees in the complex environment such as the adhesion between shrubs and trunk.