Artificial Intelligence for Marine SAR Oil-Slicks Detection

Oil slicks on marine surface can originate from various sources, including Oil spills from ships or offshore platforms, natural seepages from the seabed, or illegal discharges. These slicks can have devastating effects on marine ecosystems, coastal environments, and local economies, especially those reliant on tourism and fishing. Therefore, Oil slicks monitoring are crucial for initiating prompt response actions to mitigate environmental damage. Recently, SAR imagery has become a crucial tool for monitoring and detecting Oil spills in marine environments. Various machine learning algorithms, including deep learning, are applied to SAR imagery for Oil slicks detection. In this study, we explore the powerful of Convolutional Neural Networks (CNN) in characterizing marine surface roughness to facilitate Oil slicks detection from SAR imagery. We proposed to integrate textural features extracted from original image in the convolutional layers of CNNs in order to give an effective tool in capturing hierarchical features present in SAR images, and thus enabling the network to learn discriminative patterns associated with Oil slicks. The proposed methodology was tested on SAR imagery from the Algerian coasts, yielding an accurate Oil slick traces with minimum of false alarms.