Intelligent hybrid system for dark spot detection using SAR data

摘要: Synthetic Aperture Radars (SAR) are the main instrument used to support oil detection systems. In microwave spectrum, slicks identified as dark spots, regions with low backscatter at sea surface. Automatic and semi-automatic systems were developed minimize processing time, occurrence of false alarms subjectivity human interpretation. This study presents an intelligent hybrid system, which integrates automatic procedures detect in six steps: (I) SAR pre-processing; (II) Image segmentation; (III) Feature extraction selection; (IV) clustering analysis; (V) Decision rules and, if needed; (VI) Semi-automatic processing. The results proved that feature selection is essential improve capability, keeping only five pattern features automate procedure. method gave back more accurate geometries. approach erred including regions, increasing spots area, while excluding regions. For well-defined contrasted performance methods equivalent. However, fully did not provide acceptable geometries all cases. these cases, system was validated, integrating approach, using compact simple decision request intervention when needed. allows for combining benefits from each ensuring quality classification satisfactory.