: Motion tracking is an important step of the analysis of flow image sequences. However, Particle Image Velocimetry (PIV) techniques rarely use tracking methods developed in computer vision, they usually work with FFT and correlation based methods. Two major types of motion estimation algorithms exist: the optical flow and the feature based ones. Promising results have been obtained by applying optical flow techniques to PIV. The applicability of feature based tracking algorithms to PIV images is examined in this article. Two feature based and one optical flow based tracking algorithms are considered. Flow measurement and visualisation results for standard PIV sequences are presented. 1 PIV, DPIV The aim of this work is to investigate the applicability and efficiency of the feature based tracking algorithms in flow velocity measurement. Flow visualisation and measurement of flow dynamics are important tasks appearing in the analysis and understanding of combustion processes, aeronautic...

This paper presents an improvement of the Orthogonal Dynamic Programming optical flow technique that permits a better use of all the available information during the iterative search of the vector field in the case of stereo PIV. The two-component apparent vector fields are computed simultaneously (synchronously) for all available viewpoints. Coherency constraints issued from the geometry of the experimental setup and based upon the existence of a global three-component vector field are applied at each global step of the iterative search. Tests carried out using the Standard Images from the Visualization Society of Japan shows a small improvement in accuracy and a significant improvement in robustness. 1 INTRODUCTION Many iterative methods have been proposed for the implementation of Particle Image Velocimetry (PIV) systems [1] [2] [3] [4]. They usually yield a significant increase in accuracy and robustness through the use of variable size search windows (hierarchical / multi-...

A protocol for quantitative performance evaluation of PIV systems using the VSJ Standard Images [1] has been developed. This protocol includes programs (with C source code) to perform the quantitative performance measure and a standard way to display the results in order to permit easy system performance evaluation and comparison. There are actually three distinct protocols, dedicated respectively to the evaluation of PIV systems on three different types of PIV problems addressed through three different types of Standard Images: recovering the two in-plane components of a roughly 2D flow, recovering the two in-plane components of a 3D flow, and recovering all three components of a 3D flow using multiple viewpoints (always in a planar section of the flow illuminated by a thin laser sheet). Example of use and sample results are given using the LIMSI-CLIPS ODP-PIV system [2] [3]. Extensions for the evaluation of PIV system in noisy conditions are also proposed. 1