A NEW AUTOMATIC TECHNIQUE FOR COASTLINE EXTRACTION FROM SAR IMAGES

摘要

The coastal marine habitat is an important and delicate environment from economical, ecological, political and security point of view, therefore its integrity has to be monitored and preserved from dangerous human activities. Recent studies have demonstrated that the 42% of the Italian Coast is eroding because of the increase of the sea-level height and the reduced solid transport from rivers to sea, hence there is an important requirement for tools capable to provide a synoptic view of the coastal area. COSMO-SkyMed SAR products with their very high resolution and short revisit time, can represent a breakthrough on coastline delineation and mapping, also overcoming the problems related to cloud cover or large extension of the areas. While in remotely sensed imagery including visible bands the specific coastline extraction task may be recognized as not particularly complex, this does not hold for SAR images in which the backscattering from the water can be influenced by different effects due to the wind and the wave modulation, determining a not easy discrimination between sea and land. In this research activity a new automatic technique based on Pulse Coupled Neural Networks (PCNN) has been developed to detect the coastal boundaries, moreover a local tracing procedure exploiting statistical information has been designed to properly extract the coastline. The results have been validated through a GPS survey and an assessment of the real impact of the proposed procedure in coastal mapping application has been carried out.