Feições espaço-temporais em correntes de gravidade usando métodos de visão computacional

Abstract

Previous experimental studies on the characterization of gravity currents verified correlations between the features visually identified at the current flow front and the parameters related to its velocity and turbulence. Researches on gravity currents have used these correlations ever since. And, in more recent works using numerical simulations, these correlations continue to be validated for various flow parameters at higher front velocities. The majority of the works related to measurements at a gravity current front rely on the front images for making its analysis and establishing its correlations. Besides that, there is a field of Computer Science called Computer Vision devoted to studying how digital images can be analyzed, how its results can be automated, and what is the accuracy of these automated analyses. This work describes the use of Computer Vision algorithms, particularly corner detection and optical flow, to automatically track features at the gravity current fronts, either from physical or numerical experiments. In order to determine the accuracy of the proposed approach, we establish a ground-truth method and apply it to data sets of the numerical simulation results. The technique used to trace the front features along the flow showed promising results, especially in flows with a higher Reynolds number.