Target Detection in sea clutter Based on Wasserstein Distance

Abstract

It is a challenging problem for sea surface surveillance radar to detect small floating targets. The electromagnetic scattering properties of targets are different from those of sea surface, which will change the probability distribution of radar echo amplitude. In this paper, the difference between probability distributions is measured by Wasserstein distance and a target detection method is proposed. The method first calculates the probability density distribution of sea clutter and echoes containing targets. Then, the sea clutter closet to the barycenter of sea clutter probability distribution sample is selected as the reference which minimize the sum of Wasserstein distances from the rest of sea clutter to it. Finally, echoes containing targets and pure sea clutter are distinguished according to the Wasserstein distance between each distributions to reference distribution. Since the calculation of barycenter is cumbersome, a fast method is proposed by analogy with sloving the distribution closet to barycenter of Gaussian distribution sample whose Wasserstein distance has analytic expression. The experiments with IPIX radar data show that the proposed target detector has excellent detection ability for small floating targets on the sea surface.