Skip to Main content Skip to Navigation

Modélisation 3D d'objets par un capteur visuel déplacé par un opérateur

Benjamin Coudrin 1, 2, 3
3 LAAS-RAP - Équipe Robotique, Action et Perception
LAAS - Laboratoire d'analyse et d'architecture des systèmes
Abstract : The work of this thesis focuses on modeling 3D objects from small to medium sized (1 m3 maximum) with an innovative sensor moved manually by an operator, developed by the NOOMEOTM company. To acquire 3D data, the sensor consists of a camera system coupled with an illuminator. We infer information from a a visual pattern projected on the scene, used to create the artificial texture on 3D surface, to produce a 3D image of the scene. Surfaces are represented as 3D points reconstructed from the current viewpoint. To completely model an object, the sensor must be moved around the object to acquire several 3D images. These images are successively registered and merged into a single 3D point cloud. A triangular mesh is then generated from the 3D point cloud. The appearance (color and texture) of the object can also be extracted from data acquisition to enrich the final model. This thesis addresses the problem of registration of 3D views in two ways : the fast online registration and offline accurate optimization. For the first approach we proposed a variant of the ICP algorithm using information from images and inertial measurement. This method was validated by comparing it to other common geometric registration methods. For the second approach, we proposed a refinment strategy by minimization of cycles in a graph of relations between the 3D views ; and a global pose optimization method. Our methods and their integration into a complete modeling system were validated by comparing our results with those from other scanning systems commonly used in scientific and industrial communities.
Complete list of metadata

Cited literature [72 references]  Display  Hide  Download
Contributor : Benjamin Coudrin Connect in order to contact the contributor
Submitted on : Friday, September 18, 2015 - 8:34:54 AM
Last modification on : Tuesday, October 19, 2021 - 11:18:04 PM
Long-term archiving on: : Tuesday, December 29, 2015 - 8:27:14 AM


  • HAL Id : tel-01201505, version 1


Benjamin Coudrin. Modélisation 3D d'objets par un capteur visuel déplacé par un opérateur. Vision par ordinateur et reconnaissance de formes [cs.CV]. Université Toulouse 3 Paul Sabatier, 2011. Français. ⟨tel-01201505⟩



Record views


Files downloads