A Novelty Approach to Emulate Field Data Captured by Unmanned Aerial Vehicles for Training Deep Learning Algorithms Used for Search-and-Rescue Activities at Sea

Nowadays, unmanned aerial vehicle (UAV) is gradually becoming popular and has applications in many fields of life and protection of national sovereignty over islands and sea. In particular, along with the increase in economic exploitation activities in the exclusive economic zones and climate change, the need for rescue and safety in the marine environment is urgent more than ever. The integration of deep learning algorithms into UAVs is a new trend to help finding the victim’s location at sea faster as well as increase the chances of the victim being rescued. This paper proposes an method for detection of humans on the surface of the sea together with the GPS location of the victim and the algorithm to search for the victims in the orbit of concentric circles using deep learning algorithms. The novelty of the proposed system is to build a simulated marine environment to provide a diverse number of dataset for training in deep learning algorithms and simulated rescue scenarios. Thereby the costs and risks in training could be reduced as comparing to actions taking place in real marine environments.