We present algorithms for automatic detection of oil spills in SAR images. The algorithm consists of three main parts, dark spot detection, spot feature extraction, and spot classification. The algorithms have been trained on a large number of ERS, RADARSAT and ENVISAT images. Our algorithm performed well both on RADARSAT and ENVISAT images, and can be considered a good alternative to manual inspection if large ocean areas are to be inspected. We also present methods for estimating the confidence automatically as part of the classification process. Confidence assignments are based on the computed features. The automatically assigned confidence estimates are compared to manual assignment of confidencel levels.
展开▼
