Image-based tracking of mixed-sized surface sands and gravels under ultraviolet lights in a shallow flume

Abstract. Simultaneous characterization of the size and organization of both static and moving sediment particles would help to better understand bedform development in river channels. In the current study, an image-based particle tracking method was developed to measure the pathways of sediment particles in transport and visualize their interactions with evolving sedimentary bedforms in flume experiments. The method is novel because it: i) uses ultraviolet lights, fluorescent paint and image segmentation to obtain size class-specific videos of sediment transport over a mixed bed; ii) applies a blob-detection method included in a standalone software (TracTrac – Heyman, 2019) to detect and track particles at rest and in motion; and iii) includes a custom post-processing algorithm that includes a grid-based probabilistic motion model to minimize error in the inferred connections between particle positions on a path. We applied the algorithms on a set of videos taken during a laboratory experiment in which a pair of alternate bars and a cross-over central bar were forming in a shallow flume with non-cohesive sand and gravel transport. When the method works well, as it did for the case of particles in the 2.4–4.0 mm size class (~7 pixels in nominal diameter), the method resulted in an error of +7 % for the number of tracks and -20 % for the average duration of tracks. This degree of accuracy allowed us to analyse the locations of static particles and sediment pathways to show how the active sediment corridor shifted across the frame and then changed angles from the initial trajectories as the set of bars developed. Success of the method is reliant on accurate detection of tracer positions and the ability to predict connections between two identified positions with a motion model. Recommendations are given for application of the method and further testing to reduce reliance on subjective parameters.