POLYPROBLEM report: The Circularity Code

develop a system where we continually fly over stretches of coastline and estuaries in order to track changes over time. To do this, we plan to use planes that are normally used to detect oil spills. They come with the necessary equipment and are already in the air anyways. We flew our first field tests on the island of Spiekeroog in November 2022,” explains Mattis Wolf, a scientist at the German Research Center for Artificial Intelligence (DKFI) and part of the project consortium alongside Dr Tilman Floehr, head of technology and co-founder of the German organization everwave. PlasticObs_plus uses airplanes to collect image data because they can guarantee a higher spatial resolution and the necessary data quality. Although satellite images can cover a larger area, they do not achieve the same level of detail. Drones, on the other hand, can only be used in a very stationary manner, have to be recharged regularly, and must be equipped with additional hardware. Artificial intelligence is used to detect, classify, and quantify plastic waste, for which the DFKI as the overall project leader, is developing its own algorithm for analyzing image data. Although the manual evaluation of data is very time-consuming, training AI models regularly does pay off. The DKFI has already been able to prove in a study that artificial intelligence enhances remote sensing. When distinguishing between images with and without plastic waste, the AI methodology has an accuracy of 93.3 percent, topping the 92.6 percent achieved by humans using manual methods.74 “The planes are equipped with various optical sensor systems. Near-infrared, for example, can help recognize the difference between water and actual flotsam from a distance, as the absorption characteristics for water differ significantly in this spectral channel,” Wolf explains. After an initial AI analysis using a sensor with a wide aperture angle and lower resolution to identify potential litter patches, a second sensor is used to zoom in on the corresponding hotspots. “The high-resolution image data is then evaluated by another AI system. This information provides details about the type, quantity, and composition of the plastic waste which is particularly important for subsequent clean-ups that remove the trash,” Wolf says about the data processing. Everwave can then take action on site, for example, with its own waste collection boats as part of the clean-up technologies used by the organization. Floehr believes in the cooperation with DFKI and the added PlasticObs+ Concept 74 Tholen et al. (2023), p. 7 Source: Tholen et al. (2023) 49

RkJQdWJsaXNoZXIy ODI5MzU=