Omnidirectional Surface Vehicle for Fish Cage Inspection

Aquaculture meets the growing demand for aquatic products. However, significant economic loss and environmental impact may occur if fish escape from a damaged cage. In this paper, we present an omnidirectional surface vehicle (OSV) for efficient fish cage inspection. The OSV features symmetric multi-hull design for omnidirectional maneuverability in aquaculture sites. The overactuated thruster configuration with ducted propellers provides redundancy and reliability in the complex fish farm environment. An underwater camera with motorized depth adjustment allows the OSV to inspect the fish cage at different depths. The neural network based damage detection functionality is capable of detecting holes with irregular shapes. The detection algorithm is implemented onboard the OSV with discrete hardware accelerator, which improves the robustness and accuracy of the inspection by allowing each damage point to be detected multiple times. With pose of the OSV measured by high-accuracy localization devices, positions of the holes in the fish cage can be estimated with respect to the fish farm. This fish cage inspection methodology shows satisfying accuracy in a pool test. The experimental results show that the estimated position of the hole closely aligns with the ground truth.