Development of an Efficient Coral-Coverage Estimation Method Using a Towed Optical Camera Array System [Speedy Sea Scanner (SSS)] and Deep-Learning-Based Segmentation: A Sea Trial at the Kujuku-Shima Islands

Various methods have been developed and used for monitoring marine benthic habitats, such as coral reefs and seagrass meadows. However, the efficiency of general survey methods [e.g., line intercept transects and autonomous underwater vehicles (AUVs)] still is not high. In this article, we propose a practical coral-coverage estimation method combining an effective survey system [Speedy Sea Scanner (SSS)] and a deep-learning-based estimation method. The SSS is a towed-type system with six cameras arrayed on the platform. The depth rating of the system in our trial was 50 m. The length of the array baseline was 4.4 m, and six cameras were placed on the platform with equal spacing. The sea trial was conducted at Kujuku-Shima, Japan, on September 30, 2017. We successfully generated 3-D models and high-quality orthophotos of the seafloor with high resolution of about 1.5 mm/pixel. The survey efficiency of the SSS was about 7000 m
²
/h. In addition, the experimental results of coral-coverage estimation showed that the corals can be distinguished with accuracy of about 80% in places with relatively high transparency, and the error of coverage estimation was 10% or less. The proposed coral-coverage estimation method is more efficient than other survey techniques and costs less than AUV surveying; therefore, it is expected to become a promising tool for marine environmental surveying.