GEP Algorithm for Oil Spill Detection and Differentiation from Lookalikes in RISAT SAR Images

Earth is covered with three fourth of water and one fourth of land. Ninety percent of world cargo transportation happens via ships that sail across great waters. Increase in sea traffic at the ports, natural disasters, technical, human errors may lead to oil spilling on oceanic surface. These spills will cause a lot of damage to marine ecosystem. Estimating the damage is one of the challenging tasks that can be addressed using remote sensing technology. In this paper, detection and differentiating look-alike image features of four different oceanic regions are studied using gene expression programming (GEP) algorithms on RISAT-1 SAR satellite images. GEP algorithm clearly differentiates lookalike image feature pixel from oil spill image feature pixel with classification accuracy on four different oil spill datasets is more than 98%. Proving GEP can be used for two class oil spill detection and classification problem.