Chapter 10 - Oil spill detection using optical remote sensing images and machine learning approaches (case study: Persian Gulf)

Abstract

Oil spills can cause significant environmental damages. Meanwhile, effective monitoring and proper compensatory measures can reduce its serious impacts on the environment. In this study, an oil spill detection method based on machine learning approaches including random forest (RF) and support vector machine (SVM) and optical images was introduced. Landsat-8 and Sentinel-2 images were used to detect oil spills in the Persian Gulf. Four steps were performed, including preprocessing, feature extraction and selection, classification, and sensitivity analysis and validation. Based on outputs, Sentinel-2 outperforms Landsat-8 in detecting oil spills. Moreover, the RF classifier was more accurate than SVM for classifying water and oil spill. Based on outputs, classified map was more sensitive to gamma parameter of SVM, and oil spill can be identified by an overall accuracy of 98%–99% from RF and SVM classifiers that shows the proposed method is reasonably reliable.