[M-IS17] Ocean Plastics, an earth science perspective

[MIS17-P02] Semi-automated microplastic detection system modified from a coffee extraction machine in conjunction with hyperspectral imaging and AI

*Ryota Nakajima¹, Hidetaka Nomaki¹, Masashi Tsuchiya¹, Daisuke Matsuoka¹, Mitsuko Hidaka¹, Daisuke Sugiyama¹, Tomo Kitahashi¹, Chunmao Zhu¹, Yugo Kanaya¹, Shinji Shimode², Dhugal Lindsay¹, Sanae Chiba¹, Katsunori Fujikura¹ (1.Japan Agency for Marine-Earth Science and Technology, 2.Yokohama National University)

Keywords:Hyperspectral imaging, Microplastics, Autonomous analysis

Monitoring microplastics (MPs) in the global ocean is an urgent issue to achieve the targets of SDG14. However, conventional MPs analysis methods (e.g., polymer type characterization, counting, and shape identification) are highly constrained in terms of effort and time. Therefore, there is a need to develop an efficient and automated MPs analysis method without using Fourier Transform-Infrared Spectroscopy (FTIR), etc. If an automated MPs analysis device can be installed on a research ship or a merchant ship navigating regular sea routes, time-series observations of MPs in a wide range of sea areas will become possible. In this study, we report the development of a semi-automated analysis system that combines an automatic coffee filtration extraction mechanism with a hyperspectral camera to continuously measure MPs in seawater collected by ships.
The developed system consists of a filtration system, which filters and photographs MPs, and an analysis system, which analyzes the photographed MPs. During the filtration and photographing of particles, a series of processes are automatically performed: (1) Seawater is supplied to the system from a seawater tap to concentrate particles of >100 µm for a certain period of time, (2) Seawater containing particles of >100 µm is filtered through a rolled paper filter with a mesh size of ca. 100 µm using an automatic coffee extraction mechanism and then washed and desalted with fresh water, (3) Water is removed from the filter by suction dehydration, (4) Acquisition of images and their associated short-wavelength infrared 2D spectrum (900nm-2500nm) by a hyperspectral camera, and (5) Sample archiving by filter winding. For the detection and polymer type identification of MPs particles, the processes are: (6) Particle recognition by AI (semantic segmentation), spectrum extraction and polymer type identification of each particle, and (7) Output of size and polymer type of each particle. Basically, the system enables continuous monitoring of MPs (>100 µm) by periodically changing paper filters.