[M-IS21] Understanding plastic pollution: The reality and countermeasures

[MIS21-P04] Quantifying urban litter with citizen science and AI-driven image analysis

*Shinichiro Kako^1,2, Ryunosuke Muroya¹, Mitsuko Hidaka^1,2, Daisuke Matsuoka^2,1, Atsuhiko Isobe³ (1.Graduate School of Science and Engineering, Kagoshima University, 2.Japan Agency for Marine-Earth Science and Technology , 3.Research institute for applied mechanics, Kyushu University )

Keywords:citizen science, AI-driven image analysis, ocean plastic pollution

A substantial amount of plastic litter that enters the ocean originates from mismanaged domestic waste, which travels from urban areas through rivers before reaching the marine environment (Morales-Caselles et al., 2021). To effectively implement measures against marine pollution caused by plastic litter, it is essential to identify the types, locations, and quantities of plastic litter present in urban areas, which is the main sources of plastic pollution. The present study examines whether extensive and continuous data collection can be achieved through citizen science utilizing a smartphone application “Pirika”. Additionally, by using the data derived from Pirika as training data, we develop an AI-driven image analysis system capable of automatically detecting plastic litter in images. Furthermore, we construct a system that visualizes the quantity of urban waste by type on a map by integrating the classification and quantification results from the image analysis with location information obtained via Pirika. To evaluate the effectiveness and challenges of this system, we conducted an urban waste monitoring campaign using the developed system. The six categories of urban waste detected by the AI-driven image analysis were generally consistent with the composition ratios obtained from field experiments. The findings suggest that the proposed system has the potential to analyze waste composition and elucidate regional characteristics based on waste distribution patterns. Our results also demonstrated that citizen science through the Pirika platform enables the extensive and continuous collection of urban waste images. The results show that the effectiveness of the environmental policies can be visualized by continuous monitoring by our system.