A proposPersonnesRecherchePublicationsEvenementsProfil
A propos
Séminaires REPARTI

Les Séminaires REPARTI à l'Université Laval ont lieu le vendredi à 11h30.
Veuillez consulter le programme pour plus de détails.

Projet de maîtrise, de doctorat ou stage postdoctoral en apprentissage automatique au sein de l'équipe du Prof. Christian Gagné : veuillez consulter l'annonce suivante pour tous les détails : http://vision.gel.ulaval.ca/~cgagne/postes2017.html







Feb 13 2015 11:30AM

Somayeh Hesabi
Laboratoire LVSN
Dép. de génie électrique et de génie informatique, Université Laval

Aligning 3D local data of leapfrog locations along elongated structures


The overall goal of my research project is to build the 3D model of large elongated structures and map the subsurface defects on this model. First, the subsurface defects of structures are detected using infrared thermography and their size and depth are estimated. In the second step, the geometry of the structure is measured by handheld 3D scanners. Our task in this project is concerned with the construction of the 3D model of the structures and mapping undersurface defects on it, for quality control purposes, as well as for long-term maintenance.

The handheld 3D scanner (Metrascan) that is used to build the geometric model of the elongated structure is tracked by a photogrammetric system (C-Track). Since the photogrammetric system has a limited range of view, the tracking system has to be moved in a “leapfrog” manner. The 3D “local” data collected from different locations along the structure has to be processed to build the “local” geometric model. Then, the local models must be “stitched” together to build the full 3D model.

On the other hand, we use the IR camera to provide the location of subsurface defects and their associated sizes and depths, and then map them onto the 3D model.

To combine the 3D data of the scanner with the images acquired by an IR camera we use a motion tracking system, such as C-Track, to provide the position and orientation information of the 3D scanner and IR camera in a global reference frame. Based on the rotation and translation parameters, mapping the IR information to the corresponding 3D points can be performed.

In this talk, we will introduce the registration procedure to find the rigid transformation between different “leapfrog” locations of the tracker and, consequently, different local models associated with these locations. The contribution here is the leapfrog approach for collecting data, and the registration method for aligning the local models, despite the symmetric properties of cylindrical shapes.


©2002-. Laboratoire de Vision et Systèmes Numériques. Tous droits réservés