Survey of Blue Infrastructure by Satellite Remote Sensing from Drone Images

Abstract

Seagrass and seaweed beds, tidal flats, and biosymbiotic port structures are considered blue infrastructure, and the Japanese government is promoting their development. On the other hand, existing port structures could also function as blue infrastructure as a result of their construction. Remote sensing is used for blue infrastructure surveys. In recent years, with drones becoming commercially available at relatively affordable prices, they are being introduced into field surveys. While drones have the advantage of high-resolution aerial imaging, they have the drawback of limited coverage. To compensate for this, orthomosaic synthesis using structure from motion (SfM) image processing technology is effective. In this study, we created ground truth data from the orthomosaic image and applied for satellite image analysis to examine a wider area than that of the drone survey. As a result of the supervised classification of the satellite Planet SuperDove image, the overall accuracy was 90.7%. Ground truth data is generally acquired by field surveys using underwater cameras or diving from a ship, but there is a limit to the amount of data that can be acquired in a single survey due to cost and time constraints. Application of drones can acquire more ground truth data in a shorter time and lower cost, and perform image classification with the same accuracy as field surveys. On the other hand, remote sensing surveys have their limitations. To obtain more detailed information such as identification of habitat species and estimation of prevalence, it is necessary to deploy underwater cameras, drones, diving surveys, etc., from a ship. In future, we expect citizen science involving local groups and citizens, including diving shops and fishermen, will take the lead in blue infrastructure surveys in various regions of Japan.