A cascaded false-alarm elimination method for accurate ship detection in SAR images

In this paper, a cascaded false-alarm elimination strategy is proposed to improve ship detection performances in SAR images. The detection process is divided into two main stages: ship candidate extraction and false-alarm elimination. After extracting ship candidates using traditional constant false alarm rate (CFAR) based detector, a novel cascaded strategy is presented and applied to remove false-alarms with various characteristics. Particularly, a simple multi-threshold decision is first employed to reject most false-alarm candidates based on their simple and efficient appearance features. Then, morecomplicated context features of candidate's surroundings are extracted to reduce the maintained subtle false-alarm candidates. Finally, the effectiveness of the proposed method is verified using real TerraSAR-X data.