Autonomous collision avoidance for unmanned surface ships using onboard monocular vision

Abstract

This study presents the development of vision-based techniques for autonomous collision avoidance by an unmanned surface ship using an onboard monocular camera. In order to determine the initiation of an evasive maneuver, the range and bearing measurements of each target traffic ship with respect to the observer (e.g., own ship) need to be provided for trajectory estimation. A tracking estimator is used to estimate the target ship trajectory in the framework of bearings-only tracking based on the extended Kalman filter (EKF) algorithm with a Continuous White Noise Acceleration (CWNA) model. To enhance the observability of the tracking filter, the vertical pixel distance from the water horizon to each target ship is used as a range measurement. When the estimated separation distance between the target and own ships is less than a predefined minimum separation, the own ship alters its heading angle to avoid an imminent collision following the standard marine traffic rules. Field experiment results are presented and discussed to demonstrate the feasibility and validity of the proposed approach.