Automatic detection of oil spills in ERS SAR images

Abstract

We present algorithms for the automatic detection of oil spills in SAR images. The developed framework consists of first detecting dark spots in the image, then computing a set of features for each dark spot, before the spot is classified as either an oil slick or a "lookalike" (other oceanographic phenomena which resemble oil slicks). The classification rule is constructed by combining statistical modeling with a rule-based approach. Prior knowledge about the higher probability for the presence of oil slicks around ships and oil platforms is incorporated into the model. In addition, knowledge about the external conditions like wind level and slick surroundings are taken into account. The presented algorithms are tested on 84 SAR images. The algorithm can discriminate between oil slicks and lookalikes with high accuracy. 94% of the oil slicks and 99% of the lookalikes were correctly classified.

Untitled

“…Conversely, the training set can be balanced by duplicating the training examples of the minority class or by creating new examples by corrupting existing ones with artificial noise (DeRouin et al, 1991). Solberg and Solberg (1996) do both; positives are duplicated and negatives are randomly sampled. Honda, Motizuki, Ho, and Okumura (1997) reduce the imbalance by doing classification in two stages.…”

“…TSS recognizes the desirability of having an automatic system to reduce and prioritize the workload of the human inspectors, and has supported research to develop systems for this purpose. This research has recently produced two systems (Solberg & Solberg, 1996;Solberg & Volden, 1997), but neither has yet been incorporated into TSS's service. Our system was developed in parallel with, and independently of, these two systems.…”

“…As mentioned in the introduction, Solberg and Solberg (1996) use machine learning to classify oil spills. Their classifier is represented as a decision tree and is learned from training data using S-plus (Venables & Ripley, 1994).…”

“…Unfortunately, we overlooked this possibility during the initial data collection and so have been forced to have just one nonspill class. Solberg and Solberg (1996) divide their oil spills into subclasses based on shape. We decided not to do this because our dataset contained too few oil spills.…”

“…The system described by Solberg and Solberg (1996) uses learning to produce a classifier in the form of a decision tree. As this system is similar to ours in many ways, it will be reviewed in detail in the discussion of our results.…”