Ship Detection Using Sentinel-1 Amplitude SAR Data

Abstract:
In this paper we approach the problem of ship detection in spaceborne Synthetic Aperture Radar (SAR) gathered from Sentinel-1 satellite. The detection scheme envisages comparatively low computational requirements and is performed in two main stages: coarse and fine detection. The coarse detection is a faster computing step that relies only on simple image processing techniques, starting with a binarization based on the median value of the image, followed by morphologic operators to mask the sea clutter and the land, resulting in fewer inputs for the next stage. The fine detection is an extra step to ensure the results from the previous stage, with an implementation of the well-known Constant False Alarm Rate (CFAR) technique, adapted to a single-dimension position vector of the pixels from the coarse detection. Encouraging results were achieved with Cell Averaging CFAR (CA-CFAR) by modelling the sea clutter as statistically independent and identically distributed (iid) exponential random variables.