A Novel Marine Oil Spillage Identification Scheme Based on Convolution Neural Network Feature Extraction From Fully Polarimetric SAR Imagery

Abstract

Marine oil spill pollution has caused serious impacts on marine ecological environments, ecological resources and marine economy. Synthetic Aperture Radar (SAR), especially polarmetric SAR (PolSAR), has been proven to be a powerful and efficient tool for marine oil spill detection. In general, traditional oil spill detection methods mainly rely on artificially-extracted polarization characteristics, and the detection accuracy is limited by the quality of feature extraction. Recently proposed Convolutional neural network (CNN) is capable of mining spatial feature from large data set automatically. Inspired by these, in this paper we proposed a novel oil spill identification method based on multi-layer deep feature extraction by CNN. Firstly, PolSAR data are converted into a 9-channel data block to feed the CNN. Then, a 5-layer CNN architecture is built to extract two high-level features from the original data automatically. The features are fused after dimension reduction via principal component analysis (PCA). Finally, support vector machine method with radial basis function kernel (RBF-SVM) is utilized for classification. Three sets of RADARSAT-2 fully polarimetric SAR data were used in this study to validate the proposed method. The obtained results reveal that the proposed method provides competitive results in overall classification accuracy and kappa coefficient. Moreover, this method can improve the accuracy of oil spill detection, reduce the false alarm rate, and effectively distinguish an oil spill from a biogenic slick. INDEX TERMS Marine oil spill, RADARSAT-2, PolSAR, deep learning, feature extraction, convolutional neural network (CNN).

Smart Fishery: A Systematic Review and Research Agenda for Sustainable Fisheries in the Age of AI

“…Furthermore, oil spills on seas and oceans-a major source of maritime and ocean pollution due to anthropogenic activities and the growing demand for oil and maritime transport capacity-pose a deleterious effect on aquatic and wildlife [39,40]. This understanding is also shared by Song et al [41], who suggest that marine oil spills often result in large-scale marine pollution and seriously endangers the marine ecosystems and environment and fisheries. Moreover, global surface temperature rise comes with the urgent need to study the impacts of future climate change on fisheries [42].…”

A Deep Learning Trained by Genetic Algorithm to Improve the Efficiency of Path Planning for Data Collection With Multi-UAV

“…States of the scenario and the path sections are the input features and labels of neural network. We choose convolution neural network (CNN) to study experiences because its superior features extraction capacity [20], [21], [22], [23]. Before training, the input features should be preprocessed including normalization and reshaping, which are given as follows.…”

Marine Oil Slick Detection Using Improved Polarimetric Feature Parameters Based on Polarimetric Synthetic Aperture Radar Data

“…Table 3 lists the four types of polarimetric feature commonly used in oil spill detection compared to H(1 − A 12 ) in this study, their definitions formula, corresponding references, abbreviations, and expected behavior over sea with and without oil slicks. Three quantitative assessment methods were selected based on their widespread used in oil spill detection, i.e., the Michelson comparison measure, Jeffreys-Matusita distance, and input variable importance assessment based on the random forest classification module [6,13,[32][33][34][35].…”