Underwater ROV Software for Fish Cage Inspection
This paper deals with the development of control software for a remotely operated vehicle (ROV) as part of an automated fish cage inspection system in aquaculture. The ROV navigates autonomously around the fish cages streaming video to the topside computer that runs the algorithms for vertical rope detection. The topside computer sends back the velocity references to the ROV in real-time in order to successfully complete the inspection task. These images are also used to determine the state of net biofouling using a pre-trained convolutional neural network to help determine which nets are in need of cleaning. Robot Operating System (ROS) framework is developed to enable the topside computer to access the video stream from the ROV, process the images, and send back velocity references that would result in the complete inspection of fish cages. The inspection task is planned by following the recognizable rope segments of the outer structure of the fish cage downwards by controlling the vehicle's yaw, heave, and depth.