3D Detection and Tracking of a Moving Object by an Autonomous Underwater Vehicle With a Multibeam Imaging Sonar: Toward Continuous Observation of Marine Life

Biologging has been the prevalent method for studying the ecology and behavior of aquatic life. However, it has problems of operational inconvenience and stress on organisms. Recently, research towards remote tracking using Autonomous Underwater Vehicles (AUVs) has been gaining attention, particularly those equipped with multibeam imaging sonar (MBS) less affected by underwater light and turbidity. However, existing studies have focused mainly on 2D detection and tracking due to the inherent limitations of MBS. In this letter, we propose a novel method for 3D detection and tracking of a moving target for marine life surveys, utilizing the tilt control of MBS on AUVs. YOLOv3 is employed to identify the target's position in sonar imagery. Incorporation of the particle filter algorithm enhances position an estimation accuracy and enables a estimation of the target's state, velocity and orientation as well. 3D tracking method is introduced based on the estimated states. This method has been validated through tank experiments with a hovering-type AUV and a life-sized sea turtle replica, successfully achieving 3D tracking of the target moving horizontally and vertically. The approach demonstrated a position estimation with an RMSE of 0.23 m and velocity and orientation within the error of 0.022 m/s and 16.59 degrees, respectively.