On the development and application of seeding metrics to improve river flow measurements using image velocimetry

Abstract

This thesis presents a novel approach to improving river flow measurement accuracy by optimising seeding metrics used in Particle Tracking Velocimetry (PTV), a non-contact method that estimates surface flow by tracking the movement of individual tracers—such as foam or debris—between frames of video footage. The study focuses on refining how seeding density, dispersion, and a combined metric known as the Seeding Density Index (SDI) are measured and applied under vary ing spatial and temporal conditions. SDI is developed as a post-processing tool that quantifies the quality of tracer conditions in each video frame by combining observed seeding density and dispersion values, and is shown to correlate with the accuracy of velocity measurements. Unlike traditional methods that rely on filtering input data, this research shifts the focus to post-analysis optimisation and error correction, improving the reliability of velocity estimates derived from video footage. High-resolution video was captured at several UK river sites, alongside synthetic datasets with controlled tracer conditions. All data was processed using an adapted version of Kanade-Lucas Tomasi Image Velocimetry (KLT-IV) software, developed to support new post-processing tech niques. The synthetic simulations confirmed the effectiveness of the approach, while real-world footage validated the findings and demonstrated the practical benefits of applying SDI in routine f low monitoring. The first analysis uses synthetic videos to assess how different combinations of seeding density and dispersion affect flow measurement accuracy, and to define SDI values that can be used to estimate error in each video. The second analysis applies these findings to real-world case studies, recalculating discharge based on filtered results to test how well the metrics hold in natural conditions. The third analysis shifts focus to spatial effects, identifying where within a river cross-section measurements are most reliable based on seeding characteristics. These insights are then applied to site data to evaluate their practical value. Together, these findings demonstrate how SDI can be used to improve the consistency and ac curacy of river discharge estimation from video data, offering a practical tool for enhancing hydro logical monitoring in a range of environmental settings. This research recommends the adoption of post-processed SDI as a standard component of image velocimetry workflows, particularly in oper ational monitoring contexts, where rapid, automated, and accurate flow estimation is increasingly essential.