Detection of Small Targets in Sea Clutter Using Dual-Polarization Correlation Features and One-Class Classifier

Abstract

Due to the nonstationary, time-varying, and target-like characteristics of sea clutter, detection of small targets embedded within it has been a long-standing and formidable challenge in the field of remote sensing. Traditional adaptive detectors based on sea clutter modeling often struggle to achieve satisfactory detection performance. To address this issue, this letter proposed a novel small target detector embedded in sea clutter that leverages dual-polarization correlation features and one-class classifier. Initially, three distinct features are extracted from the significant differences between the target echoes and sea clutter in the dual-polarization correlation domain. Each individual feature possesses a certain level of discriminative power. Furthermore, under the framework of anomaly detection, a one-class classifier for small sea surface target detection is constructed based on the fast convex hull learning (FCHL) algorithm to make the final decisions. Experimental results based on the IPIX datasets demonstrate that the proposed detector outperforms several existing detectors.