A Novel Marine Ranching Cages Positioning System on Unmanned Surface Vehicles Using LiDAR and Monocular Camera Fusion

This paper introduces a novel positioning system designed to enhance the positioning accuracy of unmanned surface vehicles (USVs) during feeding operations near marine ranching cages. The system integrates monocular camera and LiDAR data through a tolerance-based matching algorithm to achieve precise positioning. Initially, real-time environmental images are captured by the camera, and object detection is performed on these images using the YOLOv8 algorithm, which facilitates the extraction of bounding boxes and coordinates for preliminary positioning. Simultaneously, the LiDAR point cloud data are preprocessed and then clustered with the DBSCAN algorithm to derive accurate distance and angle measurements. Subsequently, a tolerance-based matching algorithm is employed to fuse the LiDAR and camera data, leveraging precise distance and angle thresholds to optimize data alignment. Additionally, Real-time visualization of the fused data is achieved with the ROS rviz tool, providing a comprehensive view of target positions and enabling detailed monitoring and analysis. Experiments conducted on land using simulated marine ranching cages validate the feasibility and effectiveness of the system, demonstrating its robustness and reliability through repeated testing.