pen icon Colloque
quote

Joining the behavioral classification of optical images with the radar technology for land use management related to environmental problematic

FO

Membre a labase

Françoise Orban-Ferauge

Résumé du colloque

For more than 10 years the GEOSATEL laboratory has focused its research on a statistical approach of the images processing to serve the environment. A new algorithm of classification has been achieved on the basis of non parametric models using the local spectral densities of the data to classify. This algorithm has been applied on remote sensing images provided by the satellites LANDSAT and SPOT. The information contained inside the collected images was used to obtain a thematic classification of the land cover. The classification's algorithms evolved from supervised techniques (discriminant analysis) to completely unsupervised techniques (cluster analysis), using mainly the spectral information of the images' pixels. Such classifications are very useful to follow the evolution of the land cover. Recently the algorithm provided by the laboratory GEOSATEL has been transferred to the aerial photography scale. The scale's transfer and the wider reflectance's variability of the pixels justify an adaptation of the classification's procedures. Such a work at large scale (1/6000) enlarges the application's field of the land cover classification : this last one may be used, not only to grasp the evolution of the land cover, but also to cross the land cover inventory with other information useful for land planning through the GIS (Geographic Information System) technology , as well as to segment images, to calculate numerical topographic models, etc. for a better management of the land and for simulating the environmental incidence's effects of future decisions. The environmental problematic could be highlighted by applying a similar approach to the radar imagery, as well as by improving the GEOSATEL algorithm on the basis of the current development of the radar images processing. Some results coming from classical and original approaches of optical and radar images classification will be crossed and compared in the environmental context.

Contexte

manager icon Responsables :
Abdou Bannari
host icon Hôte : Université d’Ottawa

Découvrez d'autres communications scientifiques

Autres communications du même congressiste :