Support Tucker machines based marine oil spill detection using SAR images

Marine oil spills can destroy wildlife habitat, breeding ground and pollute the sea water or beaches. It is critical to detect marine oil spills efficiently and distinguish the oil spills from other look-alike objects before the clean action. Employing support Tucker machines, a learning based method is proposed for the detection of marine oil spills with SAR images. Firstly the features are obtained from the original images, and a tensor is constructed from the exacted features and the image. Subsequently support Tucker machines are employed for classification training with labeled samples. Lastly the trained support Tucker machines are employed for marine oil spills detection. The experimental results demonstrate that our proposed method is promising and superior to artificial neural network and support vector machines based methods in accuracy.