Session: 04-06-01 Underwater Vehicles & Subsea Communications I

Submission Number: 181078

In Situ Plume Sensing With an Event-Based Vision Sensor (EVS) and AI Analysis 

Deep-sea mineral resource development, particularly polymetallic nodule mining, generates sediment plumes, which are considered a major environmental concern. Accurate measurement of these plumes and settling particles is therefore essential. Optical methods based on light scattering are commonly used because they enable in situ measurements without the need for water sampling while preserving natural conditions. However, the scattering intensity depends on particle size, shape, and refractive index, leading to conversion uncertainties when particle properties vary. As a complementary approach, image-based particle measurements have recently attracted attention. Accurate particle counting and size or shape measurement require high-speed imaging; however, conventional high-speed cameras generate enormous data volumes, require large high-speed storage and high power consumption, and are not well suited for long-term continuous monitoring. This study proposes a high-speed, in-situ measurement method for suspended particles in seawater using an event-based vision sensor (EVS) originally developed for industrial robots and drones. The EVS outputs only asynchronous brightness changes at each pixel, enabling frame-free acquisition equivalent to 10,000 fps. As a result, it achieves temporal resolution comparable to that of a conventional high-speed camera (9,000 fps requiring 600 GB per minute) while reducing data volume to 200 MB–1 GB per minute. In deep-sea demonstrations, more than 5,000 plume particles were detected per minute, allowing analysis of particle size distribution, flow direction, and velocity, as well as demonstrating the potential for AI-based biological particle classification and biomass estimation. In this presentation, we introduce plume video recordings and analysis results from in situ observations conducted using an event-based vision sensor (EVS) mounted on a collector during a manganese nodule mining technology assessment test.

Presenting Author: Susumu Takatsuka Sony Computer Science Laboratories, Inc

Presenting Author Biography: After graduating from Tama Art University, joined Sony Interactive Entertainment as a Creative Director for multiple game titles. Later led various product planning projects at Sony, including “Walkman” and the wearable device “SmartWear,” overseeing product development from prototyping to market launch. Since 2018, engaged in research on AI and sensor technologies at Sony Group, and since 2020, has conducted marine particle observation research using an event-based vision sensor (EVS), focusing on the identification of plankton and non-living particles. The research was adopted by the Ministry of Education, Culture, Sports, Science and Technology (MEXT) under the program “Advancement of Technologies for Marine Life Big Data Utilization” and is currently ongoing. Since 2021, has served as a Visiting Researcher at JAMSTEC, promoting the integration of marine science and advanced sensing technologies.

Authors:

Susumu Takatsuka Sony Computer Science Laboratories, Inc
Hiroshi Ueno Sony Computer Science Laboratories, Inc
Hitoshi Azumi Sony Computer Science Laboratories, Inc
Tomohiro Kawahara Sony Computer Science Laboratories, Inc
Akira Koizumi DEEP OCEAN RESOURCES DEVELOPMENT CO., LTD. (DORD)