Conventional PIVs are not always capable of measuring the entire field of two-dimensional flow in rivers and oceans as they often contain an insufficient amount of tracers. However, Gradient-Based PIV using Neural Networks is potentially able to perform these field measurements. In a previous study, we evaluated its usefulness using artificially generated images. Here we apply this method to surface velocity measurements of experimental open channel flow. The open channel was filled with water and its flow was 100mm wide and 40mm depth. The flow rate was kept at a constant value of 60cm3/s. Floating tracers were used to measure the surface velocity. Several images were taken with the tracer population ranging from 18% to 81% of the water surface. The obtained velocity data were compared with those produced by conventional PIVs. As a result, it was found that Gradient-Based PIV using Neural Networks is capable of observing two-dimensional flow even when the tracer population is 18% of the water surface.
