Inshore Ship Detection Based on Multi-Modality Saliency for Synthetic Aperture Radar Images

Synthetic aperture radar (SAR) ship detection is of significant importance in military and commercial applications. However, a high similarity intensity spatial distribution scattering characteristics between the target harbor facilities, along with fuzzy sea-land boundary due to strong speckle noise, result low accuracy false alarm rate for SAR complex inshore scenes. In this paper, new method based on multi-modality saliency proposed overcome these challenges. Four maps are established from different perspectives: an ocean-buffer map (OBSM) outlining more accurate coastline under noises; local stability (LSSM) addressing pixel distribution; super-pixel (SPSM) extracting critical region-based features detection; (ISM) highlight pixels distribution. By combining maps, targets scenes can be successfully detected. The provides novel interdisciplinary perspective (surface metrology) image segmentation, discovers difference elements, proposes robust CFAR procedure background clutter fitting. Experiments public dataset (SSDD) shows that our achieves excellent performance, rate, offshore scenes, confusing metallic port large-scale results outperform several widely used methods, such as CFAR-based methods methods.