激光扫描数据的等值线分层提取和多细节表达

Point Cloud-based Stratified Contour Extraction and Its Multi-LOD Expression with Ground Laser Range Scanning

从地面三维激光扫描技术得到的点云,数据多,精度高,特征复杂,不能直接采用传统方法提取等值线.对此,首先提出点云的分层模型,并分析点云在Z坐标轴方向数据量分布以及分层点的空间分布情况,提出在高精度、高细节的采样下,将分层点云作为等值线的采样数据.然后,用迭代的凸包算法对无先验连接信息的分层点数据连接.结果表明,得到的等值线符合一般等值线的特点,不会产生边缘交叉等问题.最后,按照凸包算法的迭代次数,对等值线分类,可以表达不同细节程度的等值线,层次越高,等值线所表达的物体表面信息越多.最终通过实例说明方法的适用性.

The 3D point cloud captured from laser range scanning feafures high precision and complicated describing of a real object. However, some traditional approaches still can not be used to extract object feature efficiently because of the large data size. The stratified point model is put forward in this paper to present an object feature with different level of detail （LOD）, in which point cloud is sliced with a regular interval in vertical direction Z. Based on a case data set, the quantity and distribution of point cloud is analyzed in vertical direction. It shows that the stratified point model can be used as sampling data of a contour in vertical within required accuracy. A new approach, iterative convex hull algorithm, is then proposed in order to connect stratified point data together without prior relationship information. The result of the connection shows that the gained contours are consistent with common contours and will not result in the problem of intercross on the edge. Meanwhile, based on the number of iteration, classification of contours is also proposed to obtain contours in different level of detail. Actually, the information of contour is abundant enough when the iteration number is increased. Finally, the feasibility of the approach is validated by a case study.