Skip to Main content Skip to Navigation
Conference papers

Un modèle bayésien pour le démélange, la segmentation et la classification robuste d’images hyperspectrales

Abstract : Classification and spectral unmixing are two methods to extract information from hyperspectral images. Nonetheless, it is very unusual to find a joint approach of this two complementary interpretations. In this work, a new Bayesian model is proposed to perform unmixing and classification jointly. The traditional linear mixing model is used to describe the spectral mixtures. A clustering is then performed to divide the image into spectrally coherent clusters, in which abundance vectors are statistically homogeneous. Cluster labels are assigned a Markov random field prior which assures first a spatial regularity with a Potts potential and secondly a coherence with estimated classification labels. Classification labels are finally dependent of labeled pixels provided by the user as encountered in traditional classification tasks. Performances of the model are finally assessed with synthetic and real data.
Document type :
Conference papers
Complete list of metadata

Cited literature [8 references]  Display  Hide  Download

https://hal.inrae.fr/hal-02737167
Contributor : Migration Prodinra <>
Submitted on : Tuesday, June 2, 2020 - 6:49:03 PM
Last modification on : Thursday, March 18, 2021 - 2:16:22 PM
Long-term archiving on: : Wednesday, December 2, 2020 - 5:42:33 PM

File

2017_Fauvel_Gretsi17_1.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : hal-02737167, version 1
  • PRODINRA : 428898

Citation

Adrien Lagrange, Mathieu Fauvel, Stéphane May, Nicolas Dobigeon. Un modèle bayésien pour le démélange, la segmentation et la classification robuste d’images hyperspectrales. 26. Colloque GRETSI 2017, Sep 2017, Juan-Les-Pins, France. ⟨hal-02737167⟩

Share

Metrics

Record views

76

Files downloads

18