Toward the Social Implementation of Ocean Digital Twin ~ Demonstration of Seaweed Bed Quantification Technology for Multi‑Species Seaweed Beds ~

Hello, we are Tatsuya Suzuki from the Converging Technologies Research Laboratory and Kazuya Ijuin from the Human Digital Twin Division.

We are engaged in the research and development of the Ocean Digital Twin, which digitally reproduces marine environments to support environmental conservation and industrial applications. As part of this initiative, we are developing seaweed bed quantification technologies that quantitatively evaluate the amount of CO₂ absorbed by seaweed and other marine vegetation and enable their utilization as carbon credits. To date, we have demonstrated the effectiveness of our technology by obtaining certification under J‑Blue Credit, Japan’s domestic blue carbon credit scheme, for seaweed beds consisting of a single seaweed species.
(Reference: Uwaijima Initiative! Collaborative Blue Carbon Project for Eelgrass Restoration by Fisheries Cooperatives, Communities, and Local Government (in Japanese))

Currently, we are advancing these research outcomes toward real‑world business applications and social implementation through Proof of Concept (PoC) activities. In this article, we introduce one such initiative: a demonstration conducted in Yamaguchi Prefecture targeting seaweed beds where multiple seaweed species coexist.

(This blog is translated by AI)

Introduction

Fujitsu has identified “addressing global environmental challenges” as one of its key materiality themes. Within this framework, we have been promoting the research and development of Ocean Digital Twin, aiming to digitally capture the vast and complex marine environment and apply it to solving societal challenges such as environmental conservation and industrial utilization
(for more details on Ocean Digital Twin, please refer to this press release).

One of the key initiatives toward realizing Ocean Digital Twin is the monitoring of seaweed beds related to blue carbon credits. Blue carbon credits represent a mechanism for evaluating the amount of CO₂ absorbed by marine ecosystems—such as seaweed beds—as carbon credits. In Japan, this system is institutionalized as J‑Blue Credit. Under such schemes, it is essential to quantitatively assess the distribution and coverage of seaweed beds. Traditionally, seaweed bed monitoring has relied primarily on visual surveys conducted by professional divers. However, when surveying large areas, this approach requires significant time and labor and poses challenges in terms of safety and workforce availability.

To address these challenges, we are developing seaweed bed quantification technologies that leverage image‑based AI to recognize and quantify seaweed and seagrass species and their coverage. We have already achieved J‑Blue Credit certification for single‑species seaweed beds, confirming that the technology is suitable for practical operational use. Our previous initiatives are introduced in this press release and in a previous TechBlog article.

We are now moving from the research and development phase into a demonstration phase aimed at real business deployment.

Overview of the Demonstration in Yamaguchi Prefecture

Against this background, we conducted a demonstration in collaboration with the Yamaguchi Prefectural Industrial Technology Institute as part of a survey project aimed at supporting J‑Blue Credit certification applications in Yamaguchi Prefecture (Reference: Demonstration and Prototype Development Project for the Utilization of Next‑Generation Underwater Mobility (Yamaguchi Prefectural Industrial Technology Institute) (in Japanese)).

In this demonstration, we aimed to realize safe and efficient blue carbon assessment by combining an Autonomous Underwater Vehicle (AUV) with our image‑based AI seaweed bed quantification technology. The initiative targeted seaweed beds where multiple species coexist and was positioned as a demonstration conducted in collaboration with external organizations under the assumption of real business deployment. Through this effort, we expanded the scope of applicability of our technology and verified its feasibility for practical business use.

Demonstration Experiments and Coverage Map Generation

In this demonstration, we conducted experiments in real marine environments through collaboration with the Yamaguchi Prefectural Industrial Technology Institute, Kyushu Institute of Technology, and National Fisheries University. By leveraging the strengths of each organization, we carried out verification activities with practical deployment in mind.

• Yamaguchi Prefectural Industrial Technology Institute
  Led the overall planning and execution of the demonstration, including positioning the project as a survey initiative aligned with J‑Blue Credit certification and coordinating among related organizations.
  
  
• Kyushu Institute of Technology
  Operated the AUV and conducted safe and efficient underwater imaging in the target sea area.
  
  
• National Fisheries University
  Provided expert knowledge on seaweed beds and supported the technical evaluation of the demonstration from the perspective of interpreting and assessing diver‑based survey results.
  
  
• Fujitsu
  Provided the image‑based AI seaweed bed quantification technology and conducted data analysis and coverage map generation.

Data acquisition using the AUV was carried out from July 28 to July 30, 2025, off the coast of Kaminoseki Fishing Port, Yamaguchi Prefecture. The AUV surveyed two areas with sizes of 0.01 ha and 0.02 ha.

AUV-based data acquisition

Based on the data acquired by the AUV, we generated seaweed coverage maps using image‑based AI. The monitoring targets consisted of three seaweed species—Sporochnus keyari, Sargassum piluliferum, and Padina arborescens—which coexist within the target area, and coverage was quantified after identifying each species.

The coverage maps visualize the spatial distribution of seaweed by analyzing images captured while the AUV traveled along predefined routes near the seabed. In the maps, coverage at each location is represented by color based on the data acquired along the AUV’s trajectory, with colors corresponding to coverage levels as indicated in the legend.

Examples of AUV-captured images (left) and ground-truth images indicating seaweed species regions (right)

Example of a seaweed coverage map generated from acquired data (seaweed bed area: 0.01 ha)

Example of a seaweed coverage map generated from acquired data (seaweed bed area: 0.02 ha)

When compared with diver‑based visual surveys, we confirmed that an accuracy of 0.81, exceeding the threshold corresponding to a five‑level visual evaluation (0.80), was achieved. In addition, this demonstration verified that coverage can be quantitatively evaluated while distinguishing individual seaweed species in multi‑species seaweed beds. These results demonstrate that the seaweed bed quantification technology can be extended from single‑species to multi‑species seaweed beds. Based on these results, the proposed approach can resolve challenges associated with conventional diver‑based surveys—such as safety risks, workload, and reliance on specialized personnel—and can be applied to surveys targeting larger seaweed bed areas.

Social Evaluation

On March 4, 2026, we presented the results of this initiative at the FY2025 Underwater Robot Technology Research Group Activity Report Meeting, hosted by the Yamaguchi Prefectural Industrial Technology Institute.
(Reference :https://www.iti-yamaguchi.or.jp/docs/2026020600011/ (in Japanese))

We received positive feedback from stakeholders regarding future utilization of the technology.

• Yamaguchi Prefectural Industrial Technology Institute (Mr. Seiji Yamada, Mr. Yuji Ano)
  “Using underwater images captured by an AUV, we were able to verify the potential of AI‑based seaweed bed analysis to estimate seaweed species and coverage. We expect that, through further optimization of AUV‑based seaweed bed imaging methods and the accumulation of data from surveys of diverse seaweed beds, it will become possible to estimate blue carbon stocks over wide areas by leveraging next‑generation underwater mobility such as AUVs.”
  
  
• National Fisheries University (Prof. Noboru Murase)
  “Reproducing forests of seaweed swaying with the waves as images, and visualizing information such as species names and growth conditions through image analysis, is essential for the conservation of seaweed beds and the development of blue infrastructure. I believe it is important to take on this challenge, carefully address the issues that arise, and resolve them one by one.”
  
  
• Kyushu Institute of Technology (Prof. Kazuo Ishii, Prof. Yuya Nishida)
  “We were very impressed by the fact that irregularly shaped marine organisms can be detected in underwater environments where image quality is prone to degradation, and that species identification is also being performed. The presence of companies like yours that specialize in developing underwater AI technologies is extremely important for advancing the underwater industry in the future. We look forward to further progress and development.”
  
  

Scene from the activity report meeting

Future Outlook

Through this initiative, we obtained meaningful outcomes, including confirmation that the seaweed bed quantification technology can be applied to multi‑species seaweed beds and verification of its feasibility for real‑world business deployment. Going forward, we will continue to enhance the technology to further improve monitoring accuracy and expand its application scope, thereby promoting the business implementation of this initiative.