Digitalization and Augmentation of ROV Operations

This paper presents the implementation and application of a video-based pilot assistance system for small remotely operated vehicles (ROVs) to enhance their usability in challenging environmental conditions. A mini ROV descends along a line to inspect instruments attached to it. An image segmentation model highlights the lines or moorings in real-time utilizing the video data from the internal ROV camera. Our model is running on the topside computer with no impact on data transmission or ROV control. The model was trained in a supervised way for this specific use case and then tested and evaluated during ROV operations and in user experiments. The experimental results show improved accuracy and efficiency of line detection with overlayed mask, suggesting the benefits of the assistance system. Whereby the results depend strongly on the image content. Especially in severe environmental conditions, highlighted objects of interest could support the ROV pilot during steering. Future improvements could include steering recommendations or underwater color restorations as well as data management and digital replicas of underwater missions.