H-039
Ship Classification Using Faster Region Convolution Neural Network (Faster R-CNN) for Automatic Identification of Marine Vessels

○Kipkemoi Japhet Ngeno・Hideaki Yanagisawa・Hiroshi Watanabe（Waseda Univ.）

In this paper, we consider the application of Faster R-CNN in ship identification for port safety and security. It arising due to global increase in maritime traffic, piracy, and persistent increase in unmonitored pollution related activities. Approaches using deep learning have been the most dependable means of classifying images with high precision and speed. We used Faster R-CNN to identify different types of marine vessels. For training and testing, 400 images are used. The overall mean average precision of the classification result was 0.8774 which outperforms the conventional image feature-based approaches.