Spatiotemporal flow features in gravity currents using computer vision methods

Relationships between the features visually identified at the front of the flow’s current and parameters regarding its velocity and turbulence were observed in early experimental works on the characterization of gravity currents. Researches have associated front features, like lobes and clefts, with the flow’s turbulence, and have used these associations ever since. In more recent works using numerical simulations, these connections were still being validated for various flow parameters at higher front velocities. The majority of works regarding measurements at the front of a gravity current rely on the front’s images for making its analysis and establish relationships. Besides that, there is an interdisciplinary field related to computer science called computer vision, devoted to study how digital images can be analyzed and how these results can be automated. This paper describes the use of computer vision algorithms, particularly corner detection and optical flow, to automatically track features at the front of gravity currents, either from physical or numerical experiments. To determine the proposed approach’s accuracy, we establish a ground-truth method and apply it to numerical simulation results data sets. The technique used to trace the front features along the flow showed promising results, especially with higher Reynolds numbers flows.