Thanks to their capability to cover large areas, in all weather conditions, during the day as well as during the night, spaceborne Synthetic Aperture Radar (SAR) techniques constitute an extremely promising alternative to traditional surveillance methods. Nevertheless, in order to assure further usability of SAR images, specific data mining methods are still to be developed to provide an efficient automatic interpretation of SAR data. The aim of this paper is to present a different approach based on multiscale time-frequency analysis for the automatic extraction of linear features in a noisy background, which is a critical matter in a number of SAR applications. In particular, the technique has been mainly applied to the automatic extraction of the coastline, even if some examples on oil spills in the sea surface and in inland features are also introduced in the paper. It has proved to be a rapid, robust and reliable tool, able to manage complicated and heterogeneous scenes where classical approaches may fail.
展开▼
