Taiwan is surrounded by the sea, but its coastline is affected by human activities and topography, making it prone to the accumulation of marine debris. National Central University has developed AI technology to identify types of marine debris, enabling drones to capture high-resolution images and analyze the composition, distribution, and quantity of three types of marine debris. The Coast Guard Administration is also collaborating with relevant teams. However, the research team admitted at the "International Symposium on Marine Debris" on the 24th that due to limited funding and manpower, the database data is still insufficient, and the accuracy of AI identification needs to be improved. The Coast Guard Administration also admitted that drones are currently mainly used to target marine pollution, not marine debris, and hopes to increase the budget for drones to detect marine debris.
AI can currently identify three types of marine waste: plastic bottles, floating balls, and styrofoam
According to data from the Ministry of Environment, the weight of marine waste in China reached 940 tons in 2023. The Coast Guard Administration held a four-day "International Symposium on Marine Debris" starting on the 23rd. Professor Ke Jiayin of the Institute of Fisheries Science at National Taiwan University noted at the conference that marine debris not only affects marine life but also disrupts navigation, even increasing repair costs and causing economic losses.
Chen Jifan, a professor at the Geographic Information Systems Laboratory at the Center for Space and Remote Sensing Research at National Central University, noted that AI technology has been developed in recent years, including the establishment of an "Automatic Marine Debris Identification Network Platform." The AI model analyzes drone photos to identify the distribution and quantity of three types of marine debris: plastic bottles, buoys, and styrofoam. However, he admitted that the current model still has limitations. For example, the image resolution must be higher than 5 cm. While it can generally analyze the body of a plastic bottle, it cannot analyze individual, small pieces of bottle caps. "A lot of coastal debris is bottle caps, which can be buried in the sand. Only larger items are found above the sand."
Coast Guard Agency: Hoping to secure more funding to use AI to research marine waste
Furthermore, the labor and equipment costs associated with purchasing drones, operating them, and taking photos are high. Research funding is insufficient, and the identification platform is currently in the development stage and not yet open to the public. "Hopefully, with more funding in the future, we can develop a better system. Most importantly, we want it to be accessible to the world," becoming an international platform. However, Chen Jifan emphasized that AI models are now mature, and future research will expand into identifying fishing nets and tree branches.
Ma Zhenyao, a specialist in the Marine Environment Management Group of the Coast Guard Administration, stated that the current drone budget is primarily focused on marine pollution incidents. "The budget is insufficient to expand the use of drones to capture marine debris," he said. He agreed that increasing the number of images is necessary to effectively improve the AI software's recognition rate. While preliminary data is available, the public sector still lacks complete data for public disclosure. "We hope to secure more funding to better understand the issue of marine debris, so we can develop more appropriate policies to promote its management."
Sources:
Environmental Information Center
