River surface target enhancement and background suppression for unseeded LSPIV

Abstract

Large-Scale Particle Image Velocimetry (LSPIV) is an image-based technique for nonintrusive streamflow monitoring, where the visibility of flow tracers is one of the main limitations to its application in field conditions. Based on the target characteristics of flow tracers as well as the optical environment of river surface, the paper presents a target enhancement and background suppression method that innovatively combines near-infrared (NIR) imaging and spatial high-pass filtering (SHPF) to solve the above problem. An NIR smart camera was developed as the experimental instrument for image acquisition and preprocessing. Three sets of evaluations were performed at pixel-level, feature-level and vector-level. Results show that the NIR imaging not only enhances the contrast between targets and background, but also improves the peak signal-to-noise ratio (PSNR) of correlation plane in motion vector estimation. Moreover, the SHPF effectively suppresses the river background and strong noises, and consequently increases the percentage of correct vectors in the instantaneous flow field. Due to its strong operability, this method offers promising potential for the unseeded LSPIV.