The British artificial intelligence team has reported a new method to detect large plastic (greater than 5mm) floating garbage strips in the marine environment, according to an environmental study published on the 23rd in the Journal of scientific report. Using data from the European Space Agency sentinel 2 satellite, researchers trained machine learning algorithms to distinguish plastics from other materials, with an average accuracy of 86% and a maximum of 100% in some regions.

Human activities and garbage discharge make a large number of plastics flow into the ocean. How to distinguish plastics from other floating objects accurately and efficiently becomes a problem. In view of the different wavelengths of visible light and infrared light absorbed and reflected by the floating objects, researcher Lauren Bierman and his colleagues of Plymouth Marine Laboratory in the UK used this spectral feature to identify the floating objects in the data of sentinel 2. The team then trained a machine learning algorithm to classify the individual materials that make up these floats based on the specific spectral characteristics of different plastics and natural materials.

These features used by the machine learning algorithm are satellite data from plastic waste washed into Durban port, South Africa, on April 24, 2019, and satellite data of floating plastic deployed by the research team on the mitirini coast (Greece) in 2018 and 2019. They also used satellite data from previously acquired natural materials such as seaweed, wood, foam and volcano, which were also found in marine plastics.

The team tested this approach using sentinel 2 data from four different coastal areas: Accra (Ghana), San Juan (Canada), Da Nang (Vietnam) and eastern Scotland (UK). This method can successfully distinguish the plastics in four places from other floating materials or sea water with an average accuracy of 86%, and the accuracy in San Juan Island is 100%.

The results show that this method is successful in four different coastal zones. The researchers hope that this method can be combined with UAVs or high-resolution satellites to improve the global monitoring of marine plastic waste.

Editor in chief circle

It is the daily needs of each of us that make the output of plastic exponentially double. It's easy to get the plastic out of sight - into the sea, and soon out of sight. But in fact, the vast majority of them will always exist. Ghosts are usually tied to us in various ways. Up to now, people have realized that the depth and scope of marine plastic pollution is far beyond expectations, but to track their specific distribution, technology has been to be improved. Now researchers use the dual "pursuit" of AI and satellite data to assess the severity of the problem from the spatial and ecological level more accurately and comprehensively than ever before, so as to help us take larger-scale clean-up and mitigation measures.

Source: Science and Technology Daily