The State-of-the-Art in Ship Detection in Synthetic Aperture Radar Imagery

TLDR

Ship detection research in SAR focuses on identifying the vessel itself, neglecting wake detection, and typically involves stages such as land masking, preprocessing, prescreening, and discrimination, as exemplified by the DSTO ADSS system. The report reviews SAR ship‑detection algorithms to recommend suitable ones for inclusion in the Analysts' Detection Support System. The authors analyze motivations, theory, and justifications for each detection stage—land masking, preprocessing, prescreening, and discrimination—discussing the algorithms employed. The review recommends a basic low‑to‑medium resolution single‑channel SAR ship‑detection system and proposes improvements and alternatives.

Abstract

Abstract : This report is a review of the publicly available literature on algorithms for detecting ships in Synthetic Aperture Radar (SAR) imagery, with the aim of recommending algorithms for inclusion in the Analysts' Detection Support System (ADSS). The ADSS is a software system being developed at DSTO for the automatic detection of targets in SAR imagery. We outline the motivations, theory and justifications of the various approaches so that educated comparisons and evaluations can be made. Most current research on ship detection is based on detecting the ship itself rather than its wake. Detection of ship wakes is not addressed in any depth. Ship detection systems generally consist of several stages: land masking; preprocessing; prescreening; and discrimination. We consider each of these stages in turn and discuss the various algorithms which have been used. A basic system for low to medium resolution single channel SAR imagery is recommended and suggestions for improvements and alternatives are made.