Estimation de la rotation 3D : Approches denses et génériques basées images sphériques

Abstract : This thesis deals with the problem of location and motion estimation in the context of spherical image analysis. More specifically, the purpose of this manuscript is the development of generic algorithms for estimating 3D rotation between images captured by conventional and non- onventional (catadioptric, fisheye, ...) monocular camera. The main contributions of the thesis are based on the exploitation of the rich intensity information available in the image, notably omnidirectional, to accomplish the estimation of the 3D rotation, including the large ones. A first contribution is to propose a new representation for image intensity as a 3D object of spherical topology. The coefficients of the spherical harmonics transform computation, directly from the object mesh, is accompanied by the SVD decomposition, allowing to estimate the rotation which links the two images. The second contribution of the thesis is a closed form solution for estimating the rotation based on a dense and indirect strategy of perceiving the image intensity, which may be achieved by spherical photometric moments. The established theoretical formalism opts to define an analytical modeling of the interaction matrix relative to the photometric moments. Using the properties of this solution, we propose three approaches to estimate the 3D rotation between two spherical images. The conducted performance study show that our approaches allow to efficiently determine 3D rotation of the spherical image. Finally, several experimental scenarios are proposed and evaluated. This included the indoor and outdoor environments.
Complete list of metadatas

Cited literature [162 references]  Display  Hide  Download

https://hal.archives-ouvertes.fr/tel-01759180
Contributor : Frédéric Davesne <>
Submitted on : Thursday, April 5, 2018 - 10:47:33 AM
Last modification on : Monday, October 28, 2019 - 10:50:22 AM

File

ThèseHoussemApresSoutenance.p...
Files produced by the author(s)

Identifiers

  • HAL Id : tel-01759180, version 1

Citation

Houssem Eddine Benseddik. Estimation de la rotation 3D : Approches denses et génériques basées images sphériques. Traitement du signal et de l'image [eess.SP]. Université Paris-Saclay; Université d'Evry-Val-d'Essonne; Université Aboubakr Belkaïd - Tlemcen, 2017. Français. ⟨NNT : 2017SACLE045⟩. ⟨tel-01759180⟩

Share

Metrics

Record views

182

Files downloads

182