Soil-vegetation discrimination and assessment in satellite imagery

This paper present an analysis of the remote sensing methods of extracting information on soils and land cover. For this porpoise we selected Danube Deita, a complex ecosystem, with an important role in Romanian environment and economy. Several types of satellite images were used; in order to assess their suitability in land cover and vegetation changes detection. Landsat MSS and TM, ERS1&2 images were used. Were applied contrast enhancing and special filtering procedures to improve image and remove speckle (linear, root filtering, destriping, Frost and Lee filters). At the end of this phase, we obtained good quality images with not significant information losses. On these images we performed unsupervised and supervised classifications, (vegetation classes were established during "in situ" ground truth collection campaigns). The result of this application was the confirmation that only combined sets of images (optical, IR and microwave) can be a useful tool for land cover assessment, vegetation and soil discrimination. Beside the multi sensor approach, another condition for a complete observation is the multitemporal approach, by using several images acquired at different intervals of time, in this way we can obtain a good vegetation discrimination on seasonal basis.