2023-11-23 19:09:39
Garbage deposited on the beach by the tide – EPFL
MADRID, 23 Nov. (EUROPA PRESS) –
A new detector based on Artificial Intelligence (AI) is capable of estimating the probability that there are floating plastics in satellite images of the oceans.
This could help systematically remove plastic trash from the oceans with ships. The research, by Wageningen University and researchers from the EPFL (Federal Polytechnic School of Lausanne), has recently been published in iScience.
Accumulations of marine debris are visible in freely available Sentinel-2 satellite images, which capture coastal areas every 2 to 5 days worldwide in land masses and coastal areas. Because these are equivalent to terabytes of data, these should be analyzed automatically through artificial intelligence models such as deep neural networks.
Marc Ruswurm, assistant professor at Wageningen University, says it’s a statement: “These models learn from examples provided by oceanographers and remote sensing specialists, who visually identified several thousand instances of marine debris in satellite images at locations around the world. In this way, they trained the model to recognize plastic debris.”
Researchers developed an artificial intelligence-based marine debris detector that estimates the probability that marine debris is present for each pixel in the Sentinel-2 satellite images. The detector is trained following data-centric AI principles that aim to make the most of the limited training data that is available for this problem.
One example is the design of a computer vision algorithm that adjusts experts’ manual annotations precisely to debris visible in images. With this tool, oceanographers and remote sensing experts can provide more examples of training data by being less precise by manually clicking on contours.
In general, this training method combined with the refinement algorithm teaches the deep AI detection model predict marine debris objects better than previous approaches.
Ruswurm says: “The detector remains accurate even in more difficult conditions; for example, when cloud cover and atmospheric haze make it difficult for existing models to accurately identify marine debris.”
The detection of plastics in marine debris in difficult atmospheric conditions with clouds and fog is particularly important, as plastics are often washed into open water after rains and floods.
In addition to more accurate prediction of marine debris accumulations, the detection model will also detect debris in PlanetScope images that can be accessed daily.
“Combining weekly Sentinel-2 with daily PlanetScope acquisitions can close the gap toward continuous daily monitoring,” Ruswurm explained.
“In addition, PlanetScope and Sentinel-2 sometimes capture the same patch of marine debris on the same day just a few minutes apart. This double view of the same object in two locations reveals the direction of drift due to wind and ocean currents in Water. “This information can be used to improve marine debris drift estimation models.”
#Artificial #Intelligence #helps #detect #plastic #oceans