A High Resolution SAR Ship Sample Database and Ship Type Classification

Abstract

As the improving of the synthetic aperture radar (SAR) resolution and the increase in the amount of data acquisition, the ship type recognition has become an important research topic. In order to meet the precise identification for ship types, 101 SAR data and the Automatic Identification System (AIS) were used to build a SAR ship database. The database contains 5288 ship samples with different polarizations, incidence angle and resolutions, including more than 20 kinds of ship type such as cargo, container, oil tankers, and fishing boats. Furthermore, the influence of different polarization, incidence angle and heading on ship geometry parameters was analyzed. Moreover, a random forest (RF) classifier was used to carry out the ship type recognition experiment, and the classification accuracy reached more than 60%.