A Novel algorithm for Planar Extracting of 3D Point Clouds

Abstract

The tasks of extracting the 3D geometric features such as lines, corners, and planes play an important role in 3D modeling. This paper presents a novel 3D planar extracting algorithm based on fuzzy clustering. Our objective is focused on developing a set of techniques for the automatic extraction of 3D indoor models with robots, which is motivated by the observation that most building interiors may be modeled as a collection of planes representing ceilings, floors and walls. Our proposed system, which employs KD tree, and self-organizing fuzzy k-means, is capable of denoising laser scanning 3D point cloud data, extracting 3D planar, such as ceilings in indoor environments, and getting semantic information from the clusters based on functional reasoning module. The document presents the principal steps of the process, the experimental setup and the results achieved. The application of the proposed method to simulated and real laser scanning point cloud datasets gives good results and shows the high performance.