Towards High Resolution Real-Time Optical Flow Particle Image Velocimetry

TL;DR

This paper demonstrates that dense, real-time velocity fields in Particle Image Velocimetry can be achieved using optical flow algorithms with optimized seeding, enabling high-resolution flow measurements in practical experiments.

Contribution

It introduces a seeding optimization criterion and shows real-time, high-resolution velocity field measurement is feasible with optical flow, surpassing traditional PIV limitations.

Findings

Dense velocity fields (1 vector per pixel) are achievable in real-time.

Seeding concentration critically affects velocity field quality.

High spatial resolution measurements are demonstrated downstream of a cylinder.

Abstract

Particle Image Velocimetry (PIV) is the most commonly used optical technique for measuring 2D velocity fields. However, improving the spatial resolution of instantaneous velocity fields and having access to the velocity field in real time remains a challenge. Optical Flow veolcimetry makes it possible to meet these challenges. In this study, we show that it is possible to access dense velocity fields (1 vector per pixel) in real-time using an appropriate seeding concentration adapted to optical flow algorithms and no longer to cross-correlation PIV algorithms. The influence of concentration on the quality of velocity fields is demonstrated using synthetic images generated for a Rankine vortex. We thus demonstrate that it is possible to precisely measure small vortices using optical flow provided that the seeding is suitable. The notion of "Active Pixels" is also introduced in order to define a seeding optimization criterion adapted to experimental measurements. This criterion is finally successfully applied to the flow downstream of a cylinder leading to a spatial resolution down to one vector per pixel.