A study of the roughness and curvature in 3D point clouds to extract vertical and horizontal surfaces

In this paper an algorithm to extract horizontal and vertical surfaces in 3D point clouds through geometric features is developed and tested. This algorithm consists of two consecutive steps: 1) a preprocessing stage to change the reference frame, moving from a global coordinate system to a local one, is carried out; and 2) after that, the extraction of horizontal and vertical surfaces in the point cloud is accomplished by means of a segmentation of the point cloud using 3D connected components segmentation on an index computed from geometric features of the point cloud. The performance of this method has been tested over a point cloud composed of around 4,000,000 points and measured along an 80m street section in Cudillero, in northern Spain, using the TOPCON IP-S2 Compact + system.