Automated Detection of Icebergs in the North Atlantic Using RCM Dual-Polarimetric SAR Imagery for Offshore Safety

Abstract Icebergs impose a significant threat to shipping, offshore oil operations, and underwater pipelines. Detecting and monitoring icebergs in the North Atlantic Ocean is particularly challenging due to frequent cloud cover. Synthetic Aperture Radar (SAR) has emerged as an effective solution for addressing these challenges. This study presents a novel method for detecting icebergs under varying sea conditions using C-band dual-polarimetric imagery from the RADARSAT Constellation Mission (RCM) using the data collected along Canada’s east coast during iceberg-prone seasons in 2022 and 2023. Our approach processes large SAR images by dividing them into 100 x 100-pixel patches, each covering an area of 5 km x 5 km, and identifies icebergs within each patch. The method combines statistical features, which highlight subtle patterns in RCM imagery, with high-dimensional features extracted from three pre-trained convolutional neural network (CNN) models. These features are further enhanced with climate parameters and classified using the LightGBM algorithm. To further reduce false alarms (FAs), a Constant False Alarm Rate (CFAR) postprocessing step was applied, enhancing the reliability of the detection process. The proposed approach demonstrates exceptional performance, achieving a 99.08% accuracy rate, a false alarm (FAs) rate of only 1%, and an AUC value nearing 1.