Three-dimensional resolution of bedload transport with binocular computer vision

Describing the bedload transport of grains within a fluid flow has been a difficult problem, principally because the interactions of moving grains with the turbulent flow and granular bed are difficult to quantify. We have conducted a series of unique experiments to study these interactions. Using a pair of high-speed cameras, we resolved the bedload transport of uniform glass beads across a range of flow conditions. Leveraging ideas from the computer vision literature, we developed an automated object tracking and stereo vision algorithm to extract three-dimensional trajectories of bedload grains as they transport downstream and interact with the sedimentary bed. To our knowledge, this is the first three-dimensional resolution of individual bedload motions. Our automated methods allowed us to extract hundreds of thousands of individual bedload trajectories and millions of collisions of moving bedload with the stationary bed. This large dataset will provide the foundation of better statistical descriptions of bedload particle motion.