Abstract
When studying the entrainment of oil under breaking waves, accurately measuring the size distribution of oil droplets is crucial as it determines their rise velocity, which in turn controls vertical distribution and resurfacing, and impacts horizontal transport due to current shear. Droplet size is also important for dissolution of soluble oil components, as well as biodegradation. In climate models, air-sea exchange is an important parameter, which is also related to bubble size distributions, concentrations and depth distribution of air bubbles under breaking waves in the ocean.   Traditional image processing techniques for measuring droplet and bubble sizes rely on classifying and measuring the size of each droplet in an image separately, which can lead to an artificially large size distribution due to overlapping bubbles. To address this issue, we employed Mask-R-CNN for image segmentation, allowing us to classify all instances in an image simultaneously. This approach enables us to more accurately predict the size distribution of bubbles and droplets, as overlapping bubbles are more easily classified as separate entities. Additionally, this method allows us to work with more chaotic images near the surface of breaking waves without sacrificing precision. We present our results from this approach, which provide a more accurate understanding of the size distribution of air bubbles and oil droplets under breaking waves.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call