Object Tracking for an Autonomous Unmanned Surface Vehicle

The conventional algorithm used for target recognition and tracking suffers from the uncertainties of the environment, robot/sensors and object, such as variations in illumination and viewpoint, occlusion and seasonal change, etc. This paper proposes a deep-learning based surveillance and reconnaissance system for unmanned surface vehicles by adopting the Siamese network as the main neural network architecture to achieve target tracking. It aims to detect and track suspicious targets. The proposed system perceives the surrounding environment and avoids obstacles while tracking. The proposed system is evaluated with accuracy, precision, recall, P-R curve, and F1 score. The empirical results showed a robust target tracking for the unmanned surface vehicles. The proposed approach contributes to the intelligent management and control required by today's ships, and also provides a new tracking network architecture for the unmanned surface vehicles.