Monitoring of Coastal Environments Using Data Mining

摘要

Current satellite images provide us with detailed information about the state of our planet, as well as about our technical infrastructure and human activities. A range of already existing commercial and scientific applications try to analyze the physical content and meaning of satellite images by exploiting the data of individual, multiple or temporal sequences of images. However, what we still need today are advanced tools to automatically analyze the image data in order to extract and understand their full content. In this paper, we propose a highly automated approach for application-adapted image content exploration, targeting coastal environmental monitoring. For the selected coastal areas, different use cases can be considered such as: detection of wind turbines vs. boats, differences between beaches, tidal flats, and dams, and identification of fish cages/aquaculture. The average accuracy is ranging from 80% to 95% depending on the satellite images.