Ship Detection with Superpixel-Level Fisher Vector in High-Resolution SAR Images

Huiping Lin,Hang Chen,Kan Jin,Liang Zeng,Jian Yang

DOI: https://doi.org/10.1109/lgrs.2019.2920668

IF: 5.343

2019-01-01

IEEE Geoscience and Remote Sensing Letters

Abstract:In this letter, we propose a novel ship detection method in high-resolution synthetic aperture radar (SAR) images. In the proposed method, the superpixel is utilized as the basic processing cell. We propose a superpixel-level Fisher vector to characterize the difference between the target and clutter superpixel due to the strong discriminating ability of the Fisher vector. Besides feature extraction, the threshold determination is another key step in the detection scheme. We determine the threshold in a semisupervised way, which combines the advantages of both supervised and unsupervised methods. Based on the threshold, the target superpixels are discriminated and we obtain the final detection result. Compared with existing methods, the proposed method shows robust performance under low signal-to-clutter ratio (SCR) conditions. Experiments performed on both synthetic data and a TerraSAR-X image demonstrate the effectiveness of the proposed method.