Detection of planar segments in point cloud using wavelet transform

Abstract

In this paper we propose a new method for 3D plane segmentation and fitting from point cloud. The method is based on discrete wavelet transform. In particular, we exploit the sensitivity of certain wavelets to abrupt changes in the signal, as well as their orthonormality to linear functions to effectively recognize planar regions. The application of the proposed method is evaluated using two examples. The first example refers to a synthesized signal, and the second, real world example deals with the detection of the welding seam pose within adaptive control of robotized continuous welding.