Skip to Main content Skip to Navigation

Segmentation de nuage de points 3D pour la modélisation automatique d'environnements industriels numérisés

Abstract : The context of this work is the 'As-built' CAD modeling of digitized industrial structures. The existing environment is first digitized using a 3D laser scanner. The data obtained this way consist of an unorganized and inhomogeneous 3D point cloud. The next step is the segmentation of this cloud and the reconstruction of the various surfaces constituting the CAD model of the scene. Industrial environments yield scenes that are complex with respect to the size of data and the number of objects, but that can be described with simple geometric primitives: the plane, the sphere, the cylinder, the cone, the torus. The current tools that process 3D point clouds are not able to perform this segmentation task automatically.

In this project, algorithms have been developed for this purpose. We have particularly focused on the segmentation of pipelines. The software solutions implemented have been validated through tests involving users of the current tools. The main feature of the methods developed is the use, throughout the segmentation process, of constrained geometric primitives. The corresponding constraints are derived from a ground knowledge of the environment (e.g., continuity or tangency relationships). Geometric primitive fitting is a key topic in this work. The algorithms developed, which use a truly geometric definition of the primitives, show good results and performance on real scenes. Moreover, one of the major issue is then the validation of the fitted model. We hence have examined this question and given original methods based on statistics. Another contribution of this work relies in the algorithms for extracting a primitive from a point cloud. These methods have been applied in the context of pipelines, but seem to be relevant in the more general question of totally automatic modeling of digitized environments.
Document type :
Complete list of metadata

Cited literature [141 references]  Display  Hide  Download
Contributor : Thomas Chaperon Connect in order to contact the contributor
Submitted on : Saturday, June 4, 2005 - 5:42:23 PM
Last modification on : Wednesday, November 17, 2021 - 12:30:55 PM
Long-term archiving on: : Friday, April 2, 2010 - 10:05:36 PM


  • HAL Id : tel-00009385, version 1


Thomas Chaperon. Segmentation de nuage de points 3D pour la modélisation automatique d'environnements industriels numérisés. Modélisation et simulation. École Nationale Supérieure des Mines de Paris, 2002. Français. ⟨NNT : 2002ENMP1090⟩. ⟨tel-00009385⟩



Record views


Files downloads