“…Earlier related works on this stage mainly used specially designed hand-crafted features to detect ships, including shape [21], texture [14,26], ship histogram of oriented gradient (S-HOG) [16,22], gist [27], structure local binary pattern (structure-LBP) [30], and different combinations of these features [14,17,29]. With these features, a classifier will be used to distinguish ships from false alarms, such as AdaBoost [31], support vector machine (SVM) [14,15], and extreme learning machine (ELM) [19].…”