基于分层聚类和拓扑连接模型的点云自适应简化

激光扫描测量在大尺寸海量点云数据采集中具有显著的优势,针对海量高密度线扫描点云采样中普遍存在的采样效率低、曲率适应性差的问题,在初始分层聚类建立K邻域的基础上,通过分析线状点云的空间几何特征,提出了线扫描点云矢量边对衍生算法,建立了拓扑连接模型;研究了基于线扫描点云特征参数的局部法矢加权系数计算方法,估算了拓扑结构中任意数据内点的局部法矢;构建了以法矢方差为细分准则的非均匀细分模型,实现了对高曲率初始类的非均匀细分。通过试验验证了算法的实用性。

Adaptive Simplification for Point Cloud Based on Hierarchical Clustering and Topological Connectivity Model

Laser-scanning measurement, which has become a prevalent and challenging research topic, has a significant advantage in massive and large-scale data sets acquisition. For the problems that universally exist in massive and high density point cloud sampling, such as low efficiency and bad adaptive curvature, the spatial geometry character of linear point cloud structure is investigated to produce an edge-pair derivative algorithm for line scanning point cloud. On this basis, topological connectivity model is established. To generate dense points in high-curvature areas and sparse points in planar regions efficiently, the local normal-vector variation is substituted for Gaussian curvature to determine the degree of recursive subdivision. Meanwhile, the computational method for the non-equal weighted factor of local normal-vector is presented to estimate the local normal-vector of any point in topological structure. For further subdivision, non-uniform subdivision model whose subdivision criterion is normal variance is built to achieve the subdivision for dense points in high-curvature areas. A relevant simplification system based on the algorithm is developed by using Visual Studio. Many cases are implemented to demonstrate the performance and validate the effectiveness of the method. The comparison with other point-based methods is also performed to illustrate the superiority of the method.