Semantic Segmentation of Static and Dynamic Structures of Marina Satellite Images using Deep Learning

Advances in remote sensing and machine learning technologies allow a better understanding of satellite images and therefore, our planet. Many applications of semantic segmentation were designed to operate in urban areas. However, fewer applications have been proposed for sea-land regions. In this paper, we propose a semantic segmentation approach for sea-land regions using visible satellite images. Static and dynamic structures, as well as water, are segmented by a Convolutional Neural Network based on the U-Net architecture. The model is evaluated with real satellite images, and qualitative and quantitative results are presented. The model achieves attractive results that support a possible application in autonomous marine navigation.