Dense Motion Estimation of Particle Images via a Convolutional Neural Network

In this work, we propose a supervised learning strategy to the fluid motion estimation problem (i.e., extracting the velocity fields from particle images). The purpose of this work is to design a convolutional neural network (CNN) for estimating dense motion field for particle image velocimetry (PIV) which allows to improve the computational efficiency without reducing the accuracy. Firstly, the network model is developed based on FlowNet, which is recently proposed for end-to-end optical flow estimation in the computer vision community. The input of the network is a particle image pair and the output is a velocity field with displacement vectors at every pixel. Secondly, in order to train the CNN model, a synthetic dataset of fluid flow images is generated. To our knowledge, this is the first time to use a CNN as a global motion estimator for particle image velocimetry.
Experimental evaluations indicate that the trained CNN model can provide satisfactory results on both artificial and laboratory PIV images. In addition, the computational efficiency is much superior to the traditional cross-correction and optical flow methods.