A robust filtering algorithm based on the estimation of tracer visibility and stability for large scale particle image velocimetry

Image velocimetry for open channel is safe, efficient and environmentally friendly and large-scale particle image velocimetry (LSPIV) is one of the most adopted methods. Furthermore, robust LSPIV algorithms strongly rely on the error vector filtering strategy. However, previous works mainly conduct filtering by median filtering, main orientation filtering and maximum filtering, and these strategies are not stable enough due to the lack of taking both the visibility and stability of tracer particles into consideration. Meanwhile, statistical property like SNR (Signal-to-Noise Ratio) and peak cross-correlation are unable to estimate the tracer visibility. In order to improve the accuracy, we propose a robust and effective filtering strategy called PPSR (Peak-Peak-Sidelobe-Ratio) from the image matching perspective to ensure the visibility and stability of tracers. We conduct a serial of experiments on the public dataset Brenta and Tiber to prove the effectiveness of the proposed algorithm.