Synthetic River Flow Videos for Evaluating Image-Based Velocimetry Methods

Abstract

Various image-based velocimetry methods have been developed and increasingly used for surface velocity and discharge measurements in rivers. Their evaluation is challenging as in situ comparisons are limited and affected by uncertainties, and synthetic flow images used to test laboratory velocimetry methods are not representative of outdoor applications. A novel method is proposed to generate synthetic river flow videos with known surface velocities. The method is based on open-source computer graphics tools: Blender and Mantaflow. The synthetic videos represent realistic situations where the scene configuration, for example, riverbed and riparian areas, water aspect, camera location and orientation or lighting can be easily modified. Two cases studies emphasize the strengths and the limitations of the proposed method. The synthetic datasets generated will allow the improvement of different image-based velocimetry methods, notably through collaborative evaluations or intercomparisons.