Assessing the environmental impact of deep-sea mining using AI technology: Automatic measurement of the number of suspended particles using images via an object detection model
Summary of the AIST Press Release on July 11, 2023,
In recent years, mineral resources have been mined in the deep sea. Mining activities generate suspended particles, and even a small increase in their amounts can affect deep-sea ecosystems. Therefore, observation of suspended particles is crucial for assessing the environmental impact of deep-sea mining. However, it is difficult to measure small amounts of suspended particles using existing methods in the deep sea.
This study used object detection AI technology to measure the number of suspended particles. Object detection is a computer-based image recognition technique that automatically identifies and locates specific objects in images. A suspended particle detection model was constructed by inputting 1,028 deep-sea floor images, which comprised 3,484 suspended particles, as training data into the object detection model.
The suspended particle detection model automatically measured the number of suspended particles using underwater images (Fig. 1). The model was applied to 6,753 deep-sea floor images, and 23,913 particles were detected. Statistical analysis of these particles revealed that the number of suspended particles spiked by a factor of ≥10 compared with the average value in some cases. The detection accuracy index (average precision), which indicates the proximity of the model predictions to the correct value, was more than 82% (out of 100%), suggesting that the model achieved sufficient accuracy for monitoring the marine environment. This study provides new insights into the impact of deep-sea mining on suspended particles.
The results of this study have been published in Frontiers in Marine Science on July 11, 2023 (https://www.frontiersin.org/articles/10.3389/fmars.2023.1132500/full). This study is a part of a project commissioned by the Ministry of Economy, Trade, and Industry.
Fig. 1 Number of suspended particles in the original image (left) is automatically measured using AI technology (right). The detected particles are indicated by red rectangles.