Ship Detection Based on Compound Distribution with Synthetic Aperture Radar Images

Considering the variability of Synthetic Aperture Radar (SAR) imaging (different sensor, resolution) and complex condition of sea surface, the traditional single statistical model may be no longer a good choice to fit the distribution of actual sea clutter in SA,R image. Based on the characteristic of Gamma distributiion which is suitable to model uniform area, and GO distributiion which is adaptive to the general homogeneous and heterogeneous area, this paper established a compound distribution of GO and Gamma model to fit the characteristics of various hypes of sea conditions, and use the moment estimation to improve the computational efficiency as well. Meanwhile, the algorithm combines the Constant False Alarm Rate (CFAR) detection based on dichotomy method in order to figure out the difficulties in solving the analytical expression of compound distribution. TerraSAR-X and ERS-2 images were adopted for investigating the algorithm. Experiment results illustrate that the method can achieve good performance.