Comparison of Kalman filters with manoeuver detection for underwater optical trajectography

A video camera fixed on a submersible vehicle provides an image sequence of the seabed during its movement. By a comparison of successive images in which areas called markers are extracted through the use of a matching process called the Hough transform, the trajectory of the vehicle can be estimated. A Kalman filter predicts the position of the marker in the following image in order to reduce the estimation error of the Hough transform. In order to improve performance in the case of fast movements as in the pull phenomenon, a procedure for the detection of maneuver and the modification of the Kalman parameters is introduced