Apport de la fusion d'images satellitaires multicapteurs au niveau pixel en télédétection et photo-interprétation

Abstract : The increase of images acquired by satellites observing the Earth causes to users a problem in exploitation that may be partly solved by data fusion. This work illustrates the benefit of data fusion in remote sensing and photo-interpretation using data from sensors aboard SPOT, ERS-1 and Landsat. The particularities of each sensor, optics (visible, infrared) or radar, are taken into account to perform a fusion called heterogeneous. There is a large number of possible methods in data fusion. We have opted to limit ourselves to the fusion called at pixel level, i.e. close to the signal delivered by the sensor, in opposition to a semantic level. Three methods were developed. The coarse spatial resolution of multispectral images limits their exploitation. We propose a method, called ARSIS, that uses an image of better spatial resolution to enhance that of multispectral images yet preserving the original multispectral content. ARSIS calls upon the multiresolution analysis, the wavelet transform and the modelling of spectral content as a function of scale. A protocol for the assessment of the quality of fused images has been developed in order to compare ARSIS to existing methods on SPOT and Landsat scenes. ARSIS demonstrates a clear benefit in spectral quality compared to other methods. The underlying concept is rich and proposes a wealth of perspectives. Photo-interpretation is difficult because of the large volume of images to display and handle. Several methods were studied for the display of images and above all to synthesize the large volume in a much smaller number of images. Various spaces for representing data and use of colors have been experimented. In the case of fusion of heterogeneous data, it was found that specificities of sensors must be taken into account. A method was developed for fusing optics-radar images, wherein punctual objects having a large radar echo are extracted and displayed in optics context. The third development deals with different architectures for fusion process. The selected application is the automatic classification. Thanks to available ground truth, centralized, decentralized and hybrid architectures may be compared. Their advantages and drawbacks with respect to operational constraints, tractability, and classification performances are discussed. The presented work clearly evidences the benefit of data fusion in classification compared to single sensors processing.
Complete list of metadatas

Cited literature [68 references]  Display  Hide  Download

https://pastel.archives-ouvertes.fr/pastel-00957754
Contributor : Lucien Wald <>
Submitted on : Tuesday, March 11, 2014 - 7:54:23 AM
Last modification on : Wednesday, July 25, 2018 - 1:22:42 AM
Long-term archiving on : Sunday, April 9, 2017 - 10:36:20 PM

Identifiers

  • HAL Id : pastel-00957754, version 1

Citation

Marc Mangolini. Apport de la fusion d'images satellitaires multicapteurs au niveau pixel en télédétection et photo-interprétation. Traitement du signal et de l'image [eess.SP]. Université de Nice Sophia-Antipolis, 1994. Français. ⟨pastel-00957754⟩

Share

Metrics

Record views

1962

Files downloads

9495