Ship Berthing Trajectory Cluster based on Variational Inference Improved Mean Shift

Han Xue

TL;DR: Trajectory cluster, Mean-shift, Ship birthing, Variational inference

Abstract: In order to improve the performance of berthing trajectory clustering in presence of wind resistances, mean-shift is improved for clustering. Its convergence is then proven. Finally, the feasibility of this method is proved by experiments compared with other cluster algorithms. It is shown that this method can improve the clustering effect to a certain extent. The new algorithm is applied to the clustering of ship berthing trajectories to mine the berthing trajectories under different conditions, such as different navigable waters, heading, speed, rudder angle, channel depth, ship size, ship type, safe distance from the dock sideline, other ships in adjacent berths, traffic flow density and special weather.

Submission Number: 1