Feature-fusion-based Detection of Sea-surface Small Targets in Logarithmic Domain

Sea-surface and low altitude small target detection in complicated sea clutter background is still a challenging but important work, owing to weak returns of targets, diversity of small targets, and complicated interactions between targets and sea waves. Existing feature-based detectors using more than three features has high computation cost, however using more features can increase the detection ability indeed. In this paper, through a proposed feature fusion method, seven known features are fused to a novel test statistic to design an easy but significant feature-fusion-based detector (FFD). The proposed FFD is to learn a set of fused weights to maximize the between-class distance of sea clutter and target returns in the seven dimensional space built by features. Through the experimental results on the two measured and known radar databases, the proposed detector gets comparable performance and lower computation cost, comparing with other feature-based detectors in three dimensional space built by features.