Automated Bridge-mounted camera monitoring of floating macroplastic transport in riverine systems

: Technology potential and governmental readiness on the Belgian Scheldt River

Abstract

The presence of plastic pollution in rivers is on the rise, yet there is a lack of sufficient data concerning its distribution and quantity, which impedes the development of effective strategies for mitigation. Typically, visual observational studies to quantify and classify floating macroplastic transport involve single-instance measurements or measurements conducted over a brief time span. Bridge-mounted cameras, supported by an AI algorithm, have the potential to be far more efficient than human observation for counting and classifying floating macroplastics.
This study, conducted in collaboration with VITO under the European Horizon INSPIRE project, evaluated the effectiveness of bridge-mounted camera monitoring supported by an AI algorithm for detecting floating macroplastics in riverine systems. Simultaneous visual and RGB-based camera observations were conducted on the SCK bridge in Mol, Belgium, during a release-and-catch experiment with representative macroplastic items. Additionally, semi-structured interviews were conducted with local government entities to gain valuable insights into their readiness towards adopting bridge-mounted camera technology for monitoring riverine plastic pollution, in the context of the Scheldt River. A brief survey was incorporated into the interviews to rate the level of importance these stakeholders place on various focus areas related to monitoring of riverine plastic pollution.
The AI algorithm demonstrated an average accuracy of 72.16% in identifying macroplastics and 76% in estimating the unique count of objects. Particularly, smaller items (2.5 to 5 cm) led to more false negatives, highlighting the need for a more extensive and accurately labelled training dataset. Moreover, the experiment did not provide conclusive evidence that the AI approach is superior to traditional observation methods. Monitoring in tidal rivers like the Scheldt presents additional practical challenges due to fluctuating water levels, variable flow directions and velocities, and high turbidity, which necessitate adaptable parameterization. Integrating real-time water dynamics from in-situ sensor data can improve the algorithm’s adaptability and generalization across various river systems.
The absence of specific legislation to address plastic pollution in rivers limits the availability of funding and resources needed to establish a long-term operational monitoring strategy. To that end, the survey highlights a stronger level of importance to understand the sources of plastic pollution, vital for shaping effective policy actions. Further advancement of the AI algorithm could enable accurate estimation of specific plastic application types, facilitating more targeted and impactful interventions. A revision of the Water Framework Directive (WFD) is underway, with OSPAR (the Oslo-Paris Convention) leveraging its experience to influence the policy making process. River commissions like the International Scheldt Commission (ISC) in turn rely on delayed, filtered information from the European Commission. This delay hampers their ability to proactively coordinate and implement timely interventions, particularly given the fragmented management of the Scheldt River basin.

Date of Award	24 Nov 2024
Original language	English
Supervisor	Frank Van Belleghem (Examiner), Ansje Löhr (Assessor), Lily Fredrix (Co-assessor), Els Knaeps (Co-assessor) & Liesbeth de keukelaere (Co-assessor)