Using Boosting to Improve Oil Spill Detection in SAR Images

Marine surveillance system which uses synthetic aperture radar (SAR) images to oil spill detection must minimize false alarms in order to improve its reliability. This paper presents an application that uses boosting method to minimize misclassification and yields better generalization. Different feature sets were applied to neural network classifiers and its performance compared do boosting methods. The experiments reached substantial improvement in the classification accuracy to discriminate oil spots from the look-alike ones