Detection of Mucilage Bloom in the Sea of Marmara from PRISMA Satellite

Hyperspectral imaging is an important tool to monitor the Earth’s surface in dealing with ecological threats. One of these threats is the mucilage incident in Turkey, which gained attention in April 2021. In this study, a novel superpixel-based classification method is proposed for mucilage detection from hyperspectral images collected by the PRISMA satellite over the Sea of Marmara. The proposed method simply performs label estimation for the superpixel by analyzing the closeness between the subspace spanned by the samples within a superpixel and the subspace spanned by within-class training samples. Experiments demonstrate that the proposed method is preferable compared to other methods in the literature, especially when there is limited training data available.