Mahalanobis Classifier and Neural Network Algorithms For Oil Spill Detection

This chapter demonstrates two techniques of machine learning for automatic detection of the oil spill. These techniques involve Mahalanobis classifier and neural network. These techniques have been implemented in multimode RADARSAT-1 SAR data. In doing so, several statistical features have been selected such as area, complexity (C) and spreading. Finally, the results show that ANN distinguishes oil spill from surrounding sea surface features with a standard deviation of 0.12. In conclusion, integration of different algorithms for oil spill detection provides better ways of getting effective oil detection. The ANN algorithms is an appropriate algorithm for oil spill detection and while the W1 mode is appropriate for the oil spill and look-alikes discrimination and detection using RADARSAT-1 SAR data.