Dubois J, Tjahjadi T, Jacquet G, Coudert S

Proc. C2I 2001 Colloque Interdisciplinaire en Instrumentation, January 2001, Vol 1, 441-449

Abstract

The Particle Images Velocimetry (P.I.V.) technique enables motion vector fields in a fluid seeded with particles to be determined. The motion vectors are obtained by comparing two frames in sequence of grey level images. Correlation techniques have been used for image comparison but the motion estimation cannot be achieved in real time (25 images/s) with a standard PC. A dedicated electronic architecture can be devised to achieve real-time estimation but the hardware development will be complex and expensive. Instead, this paper investigated the binarisation of PIV images for image comparison and proposed simplifications based on a binary transform for motion estimation algorithm in order to realise a low cost hardware implementation. A precision analysis of the resulting algorithm was presented followed by its hardware implementation based on FPGA.