Ship Classification Based on Density Features in SAR Images

Abstract

Abstract Ship classification in SAR images has attracted much attention by researchers. In this paper, a SAR target classification method for three commercial ships (container ships, bulk carriers and oil tanker) is proposed by analyzing their scattering features. Firstly, the ship slice is preprocessed to obtain the binary image, from which the density features can be extracted, which describing the ship scattering point distribution. Finally, the support vector machine (SVM) classifier is applied to classify these three types of commercial ships. The experimental results show that the classification accuracy of structure feature and strength feature is low, while the proposed density feature can reach 80% for three types of ships. The combination of structure features and strength features with density features can improve the classification accuracy. Combining the three features has the best classification performance.