General principle:

The method used at the Coriolis facility relies on pattern matching between two successive images of the flow, as described by Fincham and Spedding (1997) and Fincham and Delerce (2000). The displacement field proportional to velocity is obtained from the direct cross-correlation (covariance) between image pairs, limited to small boxes. The displacement vector on each box is determined as the location of the covariance(3.2.3) maximum. The flow is generally seeded with particles, but dye patterns can be used as well (see Pierini et al. 2002, DAO 644, 1-21). The seeding with particles(2.1) must be dense (typically 0.05 per image pixel) to get good statistics for correlations.

Figure 1: Correlation Image Velocimetry: the velocity field is obtained from the cross-correlation betweeen the particle images at time and at time . The vorticity field, represented in color levels below, is deduced by differentiation.

This technique has to be distinguished from individual particle tracking (PTV). Such tracking is practically possible only with a few hundred particles, generally leaving holes in the velocity measurement. Furthermore it requires sufficiently big particles, a few mm for the large fields of view used (up to 3 x 3 m), and they may not faithfully follow the flow due to their inertia. Of course such individual tracking is appropriate for Lagrangian data (we refer to L. Stig and T. Mc Climan at SINTEF, Norway, for a stereoscopic 3D particle tracking method in large scale facilities). The present correlation method is better for Eulerian fields, as long as a dense particle seeding is provided.

The field of view is illuminated with a laser(2.2) sheet, and we get the velocity field projection on this plane of measurement. By scanning twice a volume with this laser sheet, and comparing two images obtained at the same position, velocity is obtained in a volume, but still projected in the plane of the image. We have also developed a technique of full 3D correlation, but this will not be discussed here.

We discuss in section 2 the choice of the physical parameters, in order to obtain images of good quality, a primary condition for a successful processing. The available software for the complete CIV processing is discussed in section 3, and the precision of the results is discussed in section 5. A step by step procedure for processing is described in section 4. More detail documentation of the processing programmes (CIVx and uvmat) and file format are available at http://www.civproject.org/documentation.htm.