Ship Detection in Polarimetric SAR Imagery Based on a Geometrical Perspective

Shengli Song,Zhengda Xiong,Junshan Gao,Jian Yang
DOI: https://doi.org/10.1109/iciea.2019.8834304
2019-01-01
Abstract:In this paper, we propose a novel ship detection method in polarimetric synthetic aperture radar (SAR) imagery. We model the sea clutter by a general multivariate Normal distribution. From the geometrical perspective of general multivariate Normal distribution, we show that the test statistic is actually a Mahalanobis distance, and the test threshold is essentially a hyper ellipsoidal surface. We detail the theoretical derivation and give the new CFAR detector. The proposed method is applied to synthetic data and a RADARSAT-2 SAR image. Experimental results demonstrate that the proposed method has good detection performance.
What problem does this paper attempt to address?