Abstract:

Combining with nonnegative matrix factorization (NMF)—a new subspace data analysis method, a novel method for weak ship targets detection is proposed for polarimetric synthetic aperture radar (SAR) image detection. By eigendecomposition of the polarimetric covariance matrix, the eigenvalue sets containing the energy of the polarimetric SAR image can be achieved. Using these eigenvalue sets, the non-negative matrix can be composed. Then the dominating feature of the polarimetric SAR image is extracted using NMF with sparseness constraints, hence, the weak ship targets can be detected. The ocean measured data of both the fully and dual polarimetric SAR have been utilized to validate the effectiveness of the proposed method.