A Deep Neural Network Method for Detection and Tracking Ship for Unmanned Surface Vehicle

Unmanned sea-surface platforms are generally employed for a large variety of applications, such as ecological monitoring, ocean exploration, port traffic management and safeguarding maritime rights. Although the radar-based solutions provide the target information in a wide range under the all-weather condition, the optical-based sensors are stilled need for an autonomous surface vehicle operation to avoid the close-in surface target. In this paper, we focused on devising a real-time solution for the sea-surface vehicle target detection from an uncrewed surface vehicle. Detecting the sky-sea line for image registration as input, then use the VGG-based Siamese Network to track the ship target. We test the proposed algorithms on several datasets, and the results show the real-time capability and high accuracy.