An efficiency improved recognition algorithm for highly overlapping ellipses: Application to dense bubbly flows
Résumé
Image analysis is a widespread and performant tool for the characterization of particulate systems in chemical engineering. However, for bubbly flows, due to the wide range of particles size, shape and the appearance of large clusters resulting from particles projections overlapping at high hold-up, automatic particle detection remains a challenge. An efficient methodology for bubbly flow characterization based on pattern recognition is presented. The proposed algorithm provides an exhaustive, robust and computationally efficient way of analyzing complex images involving large ellipse clusters even in concentrated medium. The method is fully automated. A sub-clustering approach enables significant computation time reduction. Moreover, thanks to its ease of parallelization, it allows considering real time monitoring