The particular project I have been working on is related to visual maritime surveillance. Project's focus is on a maritime visual surveillance application that can be used away from the visible shoreline. The specific case we are targeting is a buoy-based autonomous system, which doesn’t reveal itself to observed marine vessels, and is designed to remain at sea for extended periods up to two months. Such a system is intended for better situational awareness and management of port resources related to incoming traffic. It can be used for constant passage monitoring and cargo tracking on sea routes in relative proximity to the shore.

One of the challenges (and also the novelty of the work) is dealing with non-stationary camera. The camera is a subject to a rapid random motion associated with the buoy’s flotation. The goal is to detect ships in the field of view of the camera and to track them in sequential frames. Problem for tracking of the ships comes from a large inter-frame motion of the camera. Other factors, affecting detection and tracking performance, include compression artifacts and a low-contrast profiles of targets caused by cluttered background.