Oil Spill Hyperspectral Data Analysis: Using Minimum Distance and Binary Encoding Algorithms

Abstract

Oil spill calamities have increased, threating maritime ecosystems. This reinforces the need for accurate mapping of oil-spill calamities. The use of hyperspectral classifiers to extract areas of oil spill in a test site was achieved in this work. The paper describes the effects of utilizing a set of hyperspectral image analysis algorithms such as Minimum Distance (MD) and Binary Encoding (BE) algorithms to classify hyperspectral images of oil-spill areas in the Gulf of Mexico using Environment for Visualizing Images software. Hyperspectral image subseting, region of interest and principal component analysis were performed in the preprocessing stage, which is used to reduce the vast amount of data and eliminate redundant data. The paper provides empirical insights on the classification accuracy of hyperspectral images. A confusion matrix is used to determine the accuracy of a classification by comparing a classification result with ground truth information. The overall accuracies were 94.6399% and 88.4422% for the MD and BE algorithms, respectively. Therefore, the two algorithms are accurate for classifying hyperspectral images of the Gulf of Mexico. However, the MD algorithm is more accurate than the BE algorithm.