ABSTRACT
Large-scale image velocimetry applications have become increasingly popular for measuring surface flow velocities and river discharges. Image orthorectification is a key step in the process and the resulting velocity and discharge uncertainty remains difficult to estimate. A novel Bayesian camera calibration method is proposed to combine prior knowledge on camera parameters and ground reference points. The method is tested using a typical crowd-sourced flood video. The parametric and structural uncertainty components due to camera calibration can be combined with other sources of errors (stage, velocity coefficient, bed shift, etc.) to build an uncertainty budget.
