Real-time Particle Image Velocimetry using Event-based Imaging

Abstract

The work addresses efforts in harnessing the potential of event-based vision (EBV) for the rapid retrieval of flow field data in configurations akin to two-dimensional, two-component particle image velocimetry (2d-2c PIV). The reduced data stream of EBV compared to conventional frame-based imaging allows for equivalent framing rates up into the kHz range and thereby can capture transient phenomena. The discussed implementation utilizes lower power cameras that directly stream to conventional off-theshelf computing hardware and can achieve velocity field rates in the kHz range at a throughput of O(10E6) vectors per second. The concept is demonstrated on a small jet in a water basin by sampling the flow at pseudo-frame rates of 2 kHz and seeding densities up to 0.04 particles per pixel (ppp). By varying laser pulsing rate, seeding density and laser energy a maximum achievable event data rate exceeding 100 × 10E6 events/s could be demonstrated at a pseudo-frame rate of 1 kHz. Online measurements continuously provided more than 250 velocity maps per second, each with 880 nodes, increasing to more than 700 Hz using pixel-binning and smaller sampling windows.