Flagship component of the suite Cetos dronekit (autonomous maritime navigation), the Cetos Sense module provides advanced perception capabilities. The latest update, available now, incorporates new results of the R&D led by the Robotics division of the iXblue group.
Thanks to recorders installed on different vessels, the bank of reference images (dataset), used for deep learning of the detection AI, is rapidly growing. The addition of dedicated samples of speculative reflections (e.g. sunrises and sunsets) have eliminated false positives associated with these phenomena.
A brand-new step module of pattern-matching, based on deep learning, brings unmatched robustness, in difficult conditions, in the multiple target tracking algorithm. After the prolonged loss of contact (weather, exit from the field of view, masking by another object in the foreground), the system reliably and quickly correlates the old and the new track. As the state estimator benefits from the history, the kinematics accuracy (position, speed) of the object is significantly improved.
Including a new lidar model and a new radar model, the Cetos Sense module offers the most comprehensive integrated obstacle detection solution on the market (IR camera, lidar, radar, marine maps, AIS). It ensures the security of the DriX fleet operated by iXblue or its customers around the world.