Building a Point Cloud Hierarchical Clustering Segmentation Algorithm Based on Multidimensional Characteristics

Point cloud segmentation is an essential step in the processing of terrestrial laser scanning data. Model reconstruction quality based on point cloud is highly dependent on the validity of the segmentation results. The segmentation is challenging because of the huge amount of points with different local densities, and lack explicit structure, especially in the presence of random noisy points. This paper presents a hierarchical clustering segmentation algorithm using multidimensional characteristics. First, an initial segmentation is established by means of notion of clusters based on point cloud density to discover clusters of arbitrary shape. The points that are relatively far away and dense can be grouped. Second, a building can be extracted from urban point cloud based on its spectral characteristics. Finally, a building can be further subdivided based on a collection of geometrical characteristics of point cloud. Experimental results demonstrate that the proposed method can not only extract building from the surrounding environment but also decompose it into different planes which lay a good foundation for building reconstruction.