Mitigating underwater threats such as detecting invasive aquatic species, investigating structural issues on vessels, and performing sometimes complex stocktakes of marine life could soon be completed without the cost and potential risks to human safety thanks to novel, artificial-intelligence powered technology.
Led by Griffith University, alongside partner institutions and industry partners Cyborg and KJR, the project – enabling underwater “robot” vehicles to detect species and structures using AI – applies Griffith’s existing AI decision-making technology to marine life surveys and underwater inspections.
“This can be extremely useful and cost-effective for fisheries, especially for monitoring and analysis of marine life growth and health,” Project Leader Dr Zhe Hou said, from Griffith’s School of Information and Communication Technology.
“It combines autonomous survey modes, computer vision and AI-assisted data review to make underwater survey and monitoring safer, more repeatable and easier to scale, particularly in inconvenient and low-visibility environments.”
In recent in-field tests, the teams deployed the AI-powered AUVs in several underwater locations, including off Moreton Island, Port of Brisbane, Queensland Maritime Museum and Redcliffe Boat Ramp, finding the AUV could maintain distance to a structure or to the seabed, and its software accurately identified underwater objects and living organisms.
“We have also integrated computer vision models for object detection and tracking in real-time from the AUV’s camera stream,” Dr Hou said.
“Combining autonomous survey modes, computer vision and AI-assisted data reviews make underwater surveying and monitoring safer, more repeatable and easier to scale, particularly in inconvenient and low-visibility environments.”
Dr Zhe Hou
“Our team is now focused on further field validation tests in real operating conditions with potential applications across environmental monitoring, biosecurity surveillance and underwater infrastructure inspection.”
The project was funded by an Australian Economic Accelerator Ignite Round 1 grant valued at $483,414, and included researchers from partner institutions the University of Queensland, Queensland University of Technology, and the University of Sunshine Coast.
