Skip to Main content Skip to Navigation
Theses

Inversion neuro-variationnelle des images de la couleur de l'ocean - Restitution des proprietes optiques des aerosols et de la concentration en chlorophylle-a pour les eaux du cas I

Abstract : The optical instruments on board satellite measure the solar radiation reflected by the sea and the atmosphere. This radiation is strongly contaminated by its interaction with the atmosphere in the wavelengths which interest the ocean color. The preliminary removal of this contamination to observe the true ocean color is called atmospheric correction. The work of this thesis is focused on the SeaWiFS atmospheric correction algorithm. It shows the contribution of mathematical methods which are the networks of artificial neurons and the variational inversion for the atmospheric correction algorithms. A first neural networks inversion in the infra-red allow to retrieve the optical properties of the aerosols. These restitutions are compared, on the Mediterranean Sea, with the SeaWiFS products and are validated with ground-based measurements, showing a better estimate of the Angstrom coefficient and an equal estimate of the optical thickness. The second inversion is done on all the spectrum visible and near infra-red by combining neural networks and variational inversion. The chlorophyll-a maps are compared, on the same region, with in-situ data showing a better estimate of weak values than the SeaWIFS algorithm.
Document type :
Theses
Complete list of metadatas

https://tel.archives-ouvertes.fr/tel-00007377
Contributor : Cedric Jamet <>
Submitted on : Wednesday, November 10, 2004 - 7:16:15 PM
Last modification on : Thursday, December 10, 2020 - 12:30:34 PM
Long-term archiving on: : Thursday, September 13, 2012 - 12:50:22 PM

Identifiers

  • HAL Id : tel-00007377, version 1

Citation

Cédric Jamet. Inversion neuro-variationnelle des images de la couleur de l'ocean - Restitution des proprietes optiques des aerosols et de la concentration en chlorophylle-a pour les eaux du cas I. Autre. Université Pierre et Marie Curie - Paris VI, 2004. Français. ⟨tel-00007377⟩

Share

Metrics

Record views

431

Files downloads

328