Web Service Launch for Quantifying Marine Litter Pollution
- Toward Harmonizing Marine Litter Monitoring -

1. Key Points

• Monitoring methods for marine litter have not been standardized at national or regional levels, making comparisons between regions difficult.

• An image analysis AI that automatically quantifies the amounts and types of beach litter from photos taken with smartphones or drones has been implemented on the web and released as a service.

• As a service that is easily accessible to everyone, it is also included in guidelines that establish international standards for marine litter monitoring.

• We aim to promote harmonizing marine litter monitoring and the visualization of pollution conditions, contributing to environmental conservation and policy-making.

Fig. 1: Information provision and data collection on marine litter pollution via Web services and AI-based image analysis

2. Overview

JAMSTEC and Kagoshima University have developed a system that implements image analysis AI for quantifying marine litter, which has been released as a web service. This service can analyze photos taken at coasts using smartphones or drones, automatically recognizing and quantifying the amount and types of litter. It allows anyone to easily visualize the current state of marine litter pollution, contributing to understanding the situation by governments and municipalities, formulating environmental policies, corporate environmental conservation, and raising public environmental awareness.

This achievement is also expected to be utilized in promoting internationally coordinated monitoring of marine litter. Traditionally, monitoring of marine litter has been primarily conducted manually, but in recent years, remote sensing using drones and web cameras, and quantification through image analysis have been gaining attention. However, there has been no uniform standard for such observational and quantification methods, making it difficult to compare the amount of litter between regions. This system has been introduced as a recommended tool in international guidelines published primarily by Japan's Ministry of the Environment (MOEJ).

This research was supported by the JSPS Grants-in-Aid for Scientific Research (23K28217) and the Environment Research and Technology Development Fund by the Environmental Restoration and Conservation Agency (JPMEERF20231004).

3. Background

In recent years, it has become known that some of the garbage generated in our daily lives has leaked into the oceans, causing significant adverse effects on the marine environment and marine ecosystems. In particular, debris washed up on the coasts has become a problem in the conservation of ports and the maintenance of landscapes, necessitating efficient cleaning based on an accurate understanding of the pollution conditions. Meanwhile, traditional visual surveys of coastal litter have been labor-intensive and costly, making it difficult to frequently and comprehensively analyze a wide area. To solve this problem, efforts have been made to develop technology that uses AI-based image analysis to automatically and objectively quantify the types and distribution of marine litter from a large amount of data (photos) taken with digital cameras and drones (Reference: https://www.jamstec.go.jp/e/about/press_release/20220204_2/).

4. System

We have developed a system that implements the AI-based image analysis as a web service, "BeachLISA", shown in Figs. 2 and 3. Users first take photos of the coast using smartphones or drones. Then, after uploading the photos through a dedicated website, the images are analyzed, and marine litter is detected. The image analysis implemented can recognize artificial litter such as bottles, cans, and plastic products, as well as natural litter such as driftwood and seaweed, and can estimate the coverage area and coverage rate (in terms of pixel count for ground-taken photos) for aerial images taken with drones. The service can be operated simply by dragging and dropping with a mouse, allowing anyone to use it easily without specialized knowledge or dedicated hardware. Such simplicity of operation is a crucial element in citizen-participated data collection and is expected to advance the understanding of the actual pollution conditions.

BeachLISA: https://beach-ai.jamstec.go.jp/

Fig. 2: An example of the analysis results for ground-taken photos by digital camera using BeachLISA.

Fig. 3: An example of the analysis results for aerial photos by drone using BeachLISA.

5. Future prospects

As a method for efficiently understanding the pollution situation caused by marine litter, in addition to traditional human visual surveys, remote sensing using drones and web cameras for monitoring has been attracting much attention. Additionally, since remote sensing typically generates a large amount of data, quantitative analysis using image analysis is effective. However, to compare pollution conditions between regions and to clarify changes over time, standard monitoring and analysis methods are essential. Under the rising international momentum, an international expert meeting consisting of 16 experts from nine countries was organized under the leadership of Japan's Ministry of the Environment, and discussions and demonstrations towards standardization have been ongoing since 2022. In July 2024, the 'Harmonization Guidelines for Marine Debris Monitoring Methods Using Remote Sensing Technologies (1st Edition)' was published in both Japanese and English, co-authored by three researchers from JAMSTEC. In this guideline, this web service is introduced as a recommended tool for AI-based image analysis of marine litter, with the URL to be included in the 2nd edition.

International Guideline(Harmonizing Marine Litter Monitoring Methods Using Remote Sensing Technology), 1st Edition
https://www.env.go.jp/en/water/index_00002.html

Currently, we have been working on the development and demonstration for international standardization of AI-based image analysis that specializes not only in the coverage area of artificial and natural beach litter but also in counting types of plastic litter. It is expected that the new image analysis AI being developed will be standardized and made available as a web service, which will accelerate the understanding of marine litter globally.