Pre-processing procedures for raw point clouds from terrestrial laser scanners

The use of geometric primitives such as geometric curvature, variances of curvature and surface normal vectors as pre-processing methods for edge and boundary detection in three-dimensional (3D) unorganised point clouds is proposed. These processes are important for high-level procedures such as registration, segmentation, classification and detection of specific shapes or objects in point clouds. To demonstrate the effectiveness of these methods, we present examples of tree detection and segmentation of terrestrial laser scanner point clouds.