Automatic oil spill detection using ASAR marine images and additional data
Luis Gonzalez Vilas(1) , Jesus M. Torres Palenzuela(1) , and Juan M. Corchado(2)
University of Vigo,
Facultad de Ciencias, Campus Lagoas-Marcosende,
(2) University of Salamanca, Facultad de Ciencias, 37008, Salamanca, Spain
Satellite radar data has proven very useful in large-scale oil spill detection due to its large coverage or its ability to image the sea surface independently of light and cloud conditions. However, detection of oil slicks is a complicated operation with some limitations, since several other phenomena also give rise to similar signatures, including low wind areas, current shear zones, internal waves or man-made structures over the sea surface. With the aim of improving the discrimination between real oil slicks and similar radar signatures, it has been developed a system for automatic oil spill detection from ENVISAR ASAR images that considers the possible presence of look-alikes. In a first step, signatures suspicious of being oil spills are automatically identified by applying a fast algorithm based on adaptive threshold. Then, the previous slicks are filtered by using masks arisen from additional data, such us wind or visual observations, for the purpose of removing those signatures that might be falsely identified as oil. Finally, different parameters derived from the own radar images and associated to each detected and filtered slick are combined in order to obtain its probability of being a real hydrocarbon spill. This system was developed by using data acquired during the Prestige catastrophe occurred on the north-west coast of Spain between November 2002 and April 2003, and it is a part of a more complex oil spill management system in this area. ASAR images were ordered in the framework of the ESA A0623 project.