Towards Developing Visual Statistical Cues for Biodiversity, Abundance, Biomass around Mariana Trench in an Embeddable Smart Module

Most ecosystems in Mariana Trench, still remain unexplored or poorly known. Recently, advanced underwater vision systems, especially those deployed by the Remotely Operated Vehicles (ROV) and Autonomous Underwater Vehicles(AUV), provide us more possibilities to have a closer observation on the deep sea environments, communities, lifestyles, even behaviors of the inhabitants. However, online visual understanding is still a big challenge in these deepest spots of the ocean, due to the lack of prior knowledge, poor visibility, and dynamic investigation. In this paper, we are trying to make a further step to develop ecological indicators in an embeddable smart module, capturing visual statistical cues in biodiversity, abundance, biomass for those primary species observed along deep sea exploration mission, on the basis of RefineDet with our predefined scenic segmentation.