Automated Labeling of Hyperspectral Images for Oil Spills Classification

The constant increase in oil demand caused a huge loss the form of spills during process exporting product, which leads to an pollution, especially marine environment. This research assists providing solution for this problem through modern technology by detecting using satellite imagery, more specifically hyperspectral images (HSI). obtained dataset from AVIRIS is considered raw data, availability vast amount unlabeled data. was one main reasons propose method classify HSI automatically labeling data first unsupervised K-means clustering. labeled used train various classifiers, that are Support Vector Machine (SVM), Random Forest (RF), and K-nearest neighbor (K-NN), accomplish optimal accuracy be comparable with another accuracy. In addition, results region interest (ROI) indicate SVM RBF kernel obtains 99.89% principle component analysis (PCA) 99.86% without PCA, revealed better than RF K-NN, while second ROI 99.9% PCA 99.91% K-NN SVM. interests selected lies within Gulf Mexico area. area based on frequency usage previous spills.