Detection of oil slicks in SAR satellite images using Otsu-Bradley’s thresholding method

This paper proposes a novel thresholding method for oil slick detection from synthetic aperture radar (SAR) images using modified Otsu and Bradley’s approaches. The existence of oil sources in the seas causes hydrocarbon stains to appear on the surface of the seas and as a result, it leads to a decrease in the quality of these waters. Oil slicks are distinguished from the sea surface through the utilization of a combined Otsu-Bradley’s quantization technique, logical operators, and averaging the input image, while categorizing the classes based on the geometrical, textural, and radiometric properties of the images. We aim to enhance the identification of oil spills by utilizing remote sensing techniques, SAR satellite imagery processing, thresholding methods, and extracting geometric and textural features. We performed the classification process several times, and KNN classification method revealed an accuracy of 94.9%. Furthermore, KNN achieved a precision of 92.4%, so we repeated the classification using two selected features, area and entropy to reach a precision of 96.36%.