Raster-Vector Cooperation Algorithm for GIS - Application to Ecological Units Delineation

During the last decade, Geographic Information Systems (GIS) considerably evolved: the amount of stored data explodes and the tools used to treat them have been improved. During the same period the arrival of very high spatial resolution satellite images (less than 1 meter) gives an inexhaustible source of information actually largely unexploited. We generally use images (satellite or airborne) in GIS by stacking vector information to visualize thematic maps but without exploiting the images as an information source. We propose in this paper a way to use satellite images as an information source in order to (i) produce new information layers (ii) validate some fusion operation made with the other information layers. We present an iterative algorithm to select and fuse layers (with eventually a relaxation). We use an homogeneity criterion based on local texture analysis to validate the selection. After running this algorithm, new information layers are produced using the images and depending on the previous fusions and selections. We apply this algorithm to search Ecological Units within forest in order to guide the research of observation samples for the underlying ecosystems.