A Real-time Identification Method Based on Static-dynamic Optical Flow Features for Deep-sea Hydrothermal Plumes

The shapes of deep-sea hydrothermal plumes are complex and changes dynamically in real time. In order to identify them more accurately and quickly, it is necessary to realize the comprehensive utilization of static frame information of hydrothermal plume sequence images and dynamic information between consecutive frames. In this paper, on the one hand, by studying the texture, color, shape and other characteristics of the hydrothermal plume, this paper extracts more comprehensive static characteristics and improves the recognition accuracy. On the other hand, due to the turbulent characteristics of the hydrothermal buoyancy plume, the direction has a specific consistency, and its motion vector basically shows a trend of upward and outward diffusion. By improving the method of analyzing all suspicious areas, only the suspected areas are The optical flow vector analysis is carried out on the contour, and the dynamic characteristics are extracted to reduce the calculation amount. Finally, combined with static and dynamic texture features, a two-class classification study is carried out to achieve efficient and rapid identification of unstructured and moving hydrothermal buoyancy plumes. The algorithm is verified by laboratory experiments and high-definition video of real deep-sea hydrothermal vents.