Skip to Main content Skip to Navigation
Theses

Reconstruction tridimensionnelle de l'environnement d'un robot mobile, à partir d'informations de vision omnidirectionnelle, pour la préparation d'interventions.

Abstract : This thesis deals with three-dimensional reconstruction of a mobile robot's environment. Based on omnidirectional vision data, it aims at preparing military operations. We first developed a well-adapted architecture for 3D reconstruction and autonomous navigation. Calibration is a key step for a vision system in order to measure its environment. This step consists in modelling the system to establish the mathematical relationship between 3D points and their projections onto the image. After discussing a model choice, we will present a method to estimate its parameters. The stereoscopic structure of the sensor allows three-dimensional reconstruction from a static viewpoint. We thus propose algorithms allowing a dense reconstruction as well as detection of the primitives in omnidirectional images. When the sensor is moving, we use its motion to enhance the 3D model. Our major contribution consists in developing a bundle adjustment algorithm for omnidirectional stereoscopic sensors, which provides a motion estimation using only visual data.
Complete list of metadatas

Cited literature [116 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-00711922
Contributor : Rémi Boutteau <>
Submitted on : Friday, November 24, 2017 - 11:18:42 AM
Last modification on : Friday, October 23, 2020 - 4:57:02 PM

File

These.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : tel-00711922, version 2

Citation

Rémi Boutteau. Reconstruction tridimensionnelle de l'environnement d'un robot mobile, à partir d'informations de vision omnidirectionnelle, pour la préparation d'interventions.. Vision par ordinateur et reconnaissance de formes [cs.CV]. Université de Rouen, 2010. Français. ⟨tel-00711922v2⟩

Share

Metrics

Record views

182

Files downloads

520