A Visual Servoing Strategy for Coastline Tracking using an Unmanned Aerial Vehicle

In this paper, an Image-based Visual Servo (IBVS) Control strategy for the autonomous surveillance of coastlines using an octocopter aerial vehicle is proposed. The implemented strategy is focused on the vision-based detection and tracking of dynamic coastlines and in the presence of waves while flying in low altitudes. For this purpose, a Deep Neural Network (DNN) for the detection of the coastline is employed. The DNN is ac-companied by an analytical formulation of an Extended Kalman Filter (EKF), which considers an approximate periodical wave motion model to provide an online estimate of the coastline motion directly in image space. The estimated feedback is provided to an appropriately formulated IBVS tracking controller for the autonomous guidance of the octocopter along the coastline, ensuring the latter is always kept inside the camera's field of view. The efficacy of the proposed scheme is demonstrated via a set of comparative outdoor experiments using an octocopter flying along the coastline on various weather and beach settings.