Context Semantic Perception Based on Superpixel Segmentation for Inshore Ship Detection in SAR Image

The inshore ship detection of SAR image is a challenging task, due to inherent speckle noise, complex land background interference, identifiable sea clutter and few available shape feature. In this paper, we propose a context semantic perception method based on superpixel segmentation for inshore ship detection in SAR image. Firstly, the sea-land segmentation is used to quickly locate candidate regions, the ocean regions and buffer regions on both sides of the coastline are reserved, after that, the maximum stable extremum region (MSER) method is used for pre-screening. Then, the background slice of each candidate target is performed superpixel segmentation to extract contextual semantic information, to identify the inshore ships and false alarms. The Sentinel-1 spaceborne SAR images covering a coastal region are used to verify the proposed method. The experimental results show that the proposed method can realize accurate inshore ship detection under the complex sea-land mixed background.