4D Cartesian state estimation of sea surface targets with a single camera
Dann Laneuville, Adrien Nègre, Pauline Dufour · 2016
Cameras are nowadays widely used in maritime surveillance applications to detect small non-cooperative boats. This paper presents a simple and efficient way to model the output of an image processing system obtained on different sea surface targets and shows how an extended Kalman filter can estimate the state of such targets detected by a single camera. Here, we will not consider the data association problem and only focus on the state estimation problem to obtain a 4D Cartesian state estimation for the targets, i.e. position and velocity on the sea surface. A second order extended Kalman filter is considered with a dedicated initialization step. A parameter study on the height of the camera, an essential parameter in the state estimation performance, is presented. Results obtained on real images with a real video extractor are also presented.