New method for ship detection

In this paper, a new method is proposed for ship detection based on generalized optimization of polarimetric contrast enhancement (GOPCE). Since the scattering coefficients of a sea area vary with local incident angle, pixels at different incident angles should be processed differently. The proposed detection method takes incident angle into consideration, and can be realized in two ways: using prior knowledge or local adaptive. In the first approach, sea clutter statistics at different incident angles are calculated in advance, followed by traditional GOPCE detection. The other method performs a two-stage detection. In the first stage, the polarimetric whitening filter (PWF) is applied for coarse detection. In the second stage, a moving window is applied to estimate local statistics adaptively, achieving refined detection. Effectiveness of the proposed method is demonstrated using Radarsat-2 data, and detection results of various detectors are compared.