RiverFlow Analyzer: A Python-Based, Multi-Algorithm Optical Flow Tool for River Surface Velocity Estimation

Surface flow velocity is a fundamental variable for river monitoring, discharge assessment, and flood analysis, but conventional in situ measurements remain difficult to acquire during hazardous or rapidly evolving flow conditions. We present RiverFlow Analyzer , a Python-based framework for non-intrusive river surface velocimetry from video data, designed to integrate geometric rectification and multi-algorithm optical flow analysis within a single workflow. The method combines Ground Control Point-based image rectification with three optical flow approaches—Lucas–Kanade, Farnebäck, and DeepFlow—and was tested on video datasets acquired at multiple Italian river sites spanning different hydraulic and environmental settings. Image-derived velocities were validated against independent field measurements obtained with a surface velocity radar and a digital flow probe. The results show that Farnebäck achieves the most effective balance between accuracy and computational efficiency under standard flow conditions, whereas DeepFlow provides greater robustness in hydraulically complex or visually degraded scenes, albeit at a substantially higher computational cost. Lucas–Kanade can provide competitive performance under favourable flow and imaging conditions, but is more sensitive to feature loss and background interference in shallow, clear-water or hydraulically complex environments. These findings demonstrate that RiverFlow Analyzer provides a reliable and flexible tool for video-based river surface velocity estimation and offers practical guidance for selecting optical flow algorithms according to flow complexity, image quality, and operational constraints.