Super-Resolution Approaches Based Shallow-Water Benthic Identification Using Multispectral Satellite Imagery

Traditional coral reefs monitoring methods based on field surveys, such as line intercept transects or those based on autonomous underwater vehicles (AUVs) or the novel Portable Speedy Sea Scanner (P-SSS), are both costly and labor-intensive, and are not well-suited for long-term monitoring. In contrast, in recent years, an increasing number of studies have utilized multispectral satellite imagery for coral reef monitoring, which are more low-cost, efficient, and well-suited for long-term monitoring. However, the effectiveness of such monitoring is limited by the spatial resolution of satellite images, which leads to the mixed pixel phenomenon. To mitigate the impact of mixed pixels on satellite monitoring, this study introduces a coral reef monitoring working flow based on single image super-resolution (SISR) algorithms and benthic identification algorithm. This working flow enhances the spatial resolution of original satellite images through a SISR reconstruction and performs coral reef detection using the proposed benthic identification algorithm. By comparing drone aerial photography data from field surveys in Ko Tao Island, Thailand, coral reefs identified in images reconstructed with the SISR algorithm showed higher accuracy, from 81.7% to 86.3%, and clearer boundaries, which reduces the impact of mixed pixels on the effectiveness of satellite monitoring.