Rapid Coastal Survey Automation Using an AI-Supported Low-Cost ROV

The changing ocean requires new tools for rapid assessment, especially in complex marine environments such as coral reefs. Current survey methods have significant limitations and data processing is slow and labor intensive. Specialized AI tools still need to be developed to deal with the complexity of marine image data. MarineSitu offers affordable and customized underwater monitoring systems to the growing blue economy. In this project, we will develop a complete low-cost survey system, including hardware components [a stereo camera enabled ROV] and software components [with AI enabled object detection and 3D reconstruction]. Working with the University of Hawai’I, we will do paired surveys using the new ROV system and traditional dive methods. The collected images will be ground truthed using the current laborintensive labeling, and that data will train machine learning (ML) models for object detection, 3D positioning, and 3D reconstruction. The AI enabled software systems will be validated for automated detection of coral bleaching, invasive species, and fish diversity assessments. We expect to significantly improve the resolution, accuracy, and data processing efficiency for benthic marine surveys, offering a complete, low-cost, unmanned survey system for a multitude of monitoring applications including fisheries assessment, aquaculture management, ecotourism, and disaster assessments.