Ship Location Estimation from Radar and Optic Images using Metric Learning

Abstract

Global Positioning Systems (GPS) are used for obtaining the location of vessels. Global Positioning Systems, compared to the other systems are more precise, trustable and practical. However, using this system is not secure against outside attacks thus it can't be employed on critical situations. In this study, an alternative solution which predicts the location of vessels by processing input images of vessels, such as radar images, actual images or satellite images, to train system on similarity metric has been offered. Image processing world in recent years has been achieving an unbelievable success on many difficult problems by utilizing deep learning methods. Due to the success deep learning has on other problems, it had been aimed to produce a system employing deep learning methods on metric learning for solving positioning problems as well. The successful results obtained from this study are indeed promising. It's aimed to develop these studies further and present them for the use of industry.