Camera-Based River Velocity Measurement

Diplomarbeit/Master thesis
Supervisor: Martin Kampel
Status: open

Motivation

Reliable river discharge data are essential for flood warning, water management and inland navigation. However, classical gauges are often damaged during floods, exactly when the data matter most. Drones with cameras offer a contactless and flexible alternative, even at ungauged sites. Since cameras only see the water surface, converting surface velocity into the cross-sectional mean velocity remains a major source of uncertainty: the ratio varies between rivers, yet a default of 0.85 is commonly used. Therefore, it is essential to develop robust camera-based flow measurement methods that perform reliably under real-world conditions.

Goal of the Work

This project aims to develop, evaluate and improve methods for measuring river surface velocity from drone video, with a focus on optical flow and image velocimetry. Surface velocity fields are valuable on their own, e.g. for flow at hydraulic structures or cross-currents in shipping. In a second step, depending on progress, they are converted into mean velocity and discharge for hydrometry and flood warning, using a surveyed cross-section. The final goal is a processing pipeline that derives surface velocity fields, and ideally discharge, from drone video.

Tasks

  • Related work and state of the art (SOTA): image-based velocimetry (e.g. LSPIV, STIV, optical flow) and the conversion from surface to mean velocity
  • Methodology: preprocessing of drone videos (stabilization, image rectification) and implementation of surface velocity methods (e.g. dense optical flow vs. LSPIV); optionally conversion to mean velocity and discharge
  • Evaluation: comparison on public datasets (e.g. Perks et al. 2020) and own drone flights against reference measurements (e.g. ADCP)
  • Conclusion: summary of results, limitations and outlook

Required Skills

  • Experience in image processing and computer vision
  • Basic knowledge of machine learning and deep learning
  • Interest in hydraulics and fluid mechanics

Contact: martin.kampel@tuwien.ac.at