Intelligente biologisch inspirierte Systeme in der Geo-Wissenschaft und Fernerkundung
Zusammenfassung der Projektergebnisse
Illegal releases of oil products from ships are responsible for a large part of marine pollution (some 48 %). Not only oil tankers, but various other cargo ships pose a constant threat of oil pollution from illegal dumping of oily wastes. This project extension has been devoted to the development of remote sensing techniques for early detection of oil spills using satellite Synthetic Aperture Radar (SAR) images. This project builds upon the results obtained during the first part of the project and applies some of the newly arisen ides to the new application area of marine pollution monitoring. The research work has been focused on the following topics: 1) Development of the oil spill detection system, 2) Incorporation of the physical model data for the detection of oil spills, 2) Incorporation of polarimetric information for the oil spill detection and 3) Development and encapsulation of the system components as stand alone Internet-based applications using Web Processing Service (WPS) technology and standards. A new sensor fusion algorithm combining SAR image parameters and physical model data for the improved oil spill detection has been developed. The algorithm computes high spatial resolution wind filed estimates from SAR images using the CMOD-5 wind model. The wind conditions during and before SAR image acquisition is one of the most important factors influencing physical properties of oil spills and also effects discriminative properties of SAR images. The proposed algorithm incorporates wind field to 1) Compute a set of wind direction dependent SAR image parameters, 2) Update the prior probabilities of oil spills in SAR images, and 3) Combine this information in the Support Vector Machine (SVM) or neural network-based algorithm. Incorporation of wind information for automatic detection algorithm substantially increases detection rates of oil spills on sea surface. The algorithm has been trained and tested using European ENVISAT satellite ASAR images of Mediterranean Sea. Additionally, in this project, we investigate the usefulness of polarimetric information obtained at HH and VV polarisations for the improved oil spill detection, propose and analyse the information theory based set of image features describing mutual information (association) between two polarization channels. Both centralised and decentralized Web-based architectures have been implemented and tested. The decentralized oil spill detection architecture has been designed using Web Processing Service (WPS). Different processing steps and corresponding algorithms have been implemented as simple WPS’s that can reside on different WPS servers. The simple WPS are combined or chained using so-called complex WPS that has been implemented and registered in the ESA’s Service Support Environment (SSE) as part of InterRisk EU FP6-IST research project. The results of this project will be used for the improved detection of oil spills and pollution monitoring in the North Sea using ASAR data in future projects. The current project has clear synergies with the recently finished InterRisk FP6-IST EU project and it results contribute to the Global Monitoring for Environment and Security (GMES) initiative. The integration of this results into European infrastructure has already been started during InterRisk project by registering the developed system in the ESA’s SSE; this work within EU ‘framework’ will be pursued in future by preparing a new EU project proposal with members of the consortium. The possibility of the algorithm integration into the National Marine Pollution Monitoring System will be discussed with the DeMarine-Umwelt GMES national project (www.demarineumwelt.de) consortium. The achieved results have already been presented to some members of the consortium at the DLR held workshop and through InterRisk project dissemination activities.
Projektbezogene Publikationen (Auswahl)
-
Presentation at the OceanSAR2009 workshop “Oil Spill detection using ENVISAT ASAR images and wind field estimates”
A.Bogdanov, T.Hamre, and S.Sandven
