3D

Shape Features for Objects Classification from Terrestrial Lidar Point Clouds

Abstract

A mobile laser scanner is a valuable tool to collect 3D information especially in urban regions, where vertical objects, like walls, poles and trees, need to be mapped. The collected point clouds can be used to segment objects and classify them according to their shape. Nevertheless, the segmentation and classification steps still need tools to analyze 3D point clouds. In this paper it is introduces a method to describe 3D shape from point clouds obtained by mobile laser scanner within the context of classification of urban furniture. The initial aim is to describe the 3D shape of objects located at the top of poles, but the approach can be extended to other objects. For this purpose, the distribution of the points is analyzed with help of the eigenvalues of the variance-covariance matrix. It is proposed the use of two parameters, one related to planarity and another to elongation, that are normalized in the range between zero and one, which allows easier description of the shape in terms of just two well-known terms.