Understanding how plastics flow from land to sea is essential for solving the growing problem of plastic pollution. Rivers play a key role as major transport pathways, and accurate monitoring technologies are urgently needed to achieve the “Osaka Blue Ocean Vision” declared at the 2019 G20 Osaka Summit, which aims to reduce additional marine plastic pollution to zero by 2050.
The newly developed software integrates three key technologies:
- Template matching1 for measuring river surface flow velocity;
- YOLOv82 for detecting and classifying plastic objects; and
- Deep SORT3 for tracking their movements.
By combining these techniques, the system automatically quantifies the transport volume of floating plastics in rivers. This innovation enables continuous, simultaneous monitoring at multiple sites, including under challenging conditions such as floods—something that was previously difficult and dangerous to perform manually. In addition, because the software can distinguish plastics by type, it allows for more direct evaluation of source reduction measures and the effectiveness of waste management policies.
Moving forward, the research team plans to incorporate this software into the Plastic River Monitoring System (PRIMOS
4), jointly developed with Yachiyo Engineering Co., Ltd., to promote its application in real river environments. Through this effort, the team aims to:
- Accurately estimate the amount of plastic flowing from land to sea;
- Clarify the transport processes across entire river basins; and
- Support the formulation and evaluation of evidence-based environmental policies.
1 Template Matching: An image recognition technique that searches for areas within an image that match a pre-prepared sub-image (template).
2 YOLOv8: The eighth version of the “You Only Look Once” object detection algorithm, a deep learning model capable of detecting and classifying objects quickly and accurately.
3 Deep SORT: An extension of the “Simple Online and Realtime Tracking (SORT)” algorithm that improves object identification accuracy by incorporating deep-learning-based appearance features.
4 PRIMOS: Plastic Runoff Identification, Monitoring & Observation System, which is a river plastic monitoring system jointly developed by Ehime University and Yachiyo Engineering Co., Ltd. Please visit the PRIMOS
website: https://info.river-monitoring.net/en/index.html