Skip to Main content Skip to Navigation
Theses

Reconstruction 3D dense d'objets sans recul par vision catadioptrique

Abstract : This PhD work focuses on the problem of complete dense 3D reconstruction of objects without recoil. We have conceived and developed a tridimensional reconstruction system of real objects based on a camera with two planar mirrors; baptised as a Planar Catadioptric Stereo (PCS) system. We first model the PCS system by a network of virtual cameras to perform calibration. Then, we formulate the problem of characteristic points' correspondences detected on the reflected images by using a variant of the ASIFT method. This adaptation which we behold as AMIFT on the mirror planes. Putative point correspondences are further refined with outlier rejection using our method of Symmetric RANSAC proposed in this thesis. To reconstruct a proper dense object surface, and not just a sparse projection of points, a dense correspondence technique is consequently required. This method estimates the geometric transformation linking the image object with one of it’s inter-reflections on the mirror planes by minimizing a branch and bound cost function. This allows us to adapt the 3D dense reconstruction, fundamentally based on the triangulation of image correspondences. We apply this 3D reconstruction pipeline on multiple catadioptric images in order to verify the undermining hypothesis using a PCS system. Our methodology is validated using experimental results on synthetic images to illustrate the quality of the 3D reconstruction
Document type :
Theses
Complete list of metadata

https://tel.archives-ouvertes.fr/tel-03638162
Contributor : ABES STAR :  Contact
Submitted on : Tuesday, April 12, 2022 - 9:44:11 AM
Last modification on : Friday, August 5, 2022 - 11:22:18 AM
Long-term archiving on: : Wednesday, July 13, 2022 - 6:31:40 PM

File

TheseMohtaram.pdf
Version validated by the jury (STAR)

Identifiers

  • HAL Id : tel-03638162, version 1

Collections

Citation

Noureddine Mohtaram. Reconstruction 3D dense d'objets sans recul par vision catadioptrique. Autre. Université de Picardie Jules Verne; Université Mohammed V-Agdal (Rabat, Maroc ; 1993-2014), 2019. Français. ⟨NNT : 2019AMIE0023⟩. ⟨tel-03638162⟩

Share

Metrics

Record views

40

Files downloads

3