Feature selection and classification of oil spills in SAR image based on statistics and artificial neural network

The general process of oil spill detection from SAR image with artificial neural network (ANN) classifier briefly includes five steps, target extraction, feature extraction, feature selection, ANN training and ANN classification. Feature extraction and feature selection are concerned in this paper. Firstly, 68 features are calculated for each target. By cross-correlation analysis, 24 features are selected to build a neural network to classify oil spills and look-alikes. The impact of imbalance sample data set on the performance of classification is also considered. In the end, principal component analysis (PCA) is applied on 24 features to reduce the dimension of feature space. The best number of principal components is found out.