The use of a deformable weight function and of advanced validation procedures in PIV

Advanced image processing and post-processing in Particle Image Velocimetry (PIV) are considered. Synthetic (with exact solutions) and real images (tip vortices downstream the wing of an aircraft model) have been used to verify the quality of the proposed procedures. In image processing, the attention has been focussed onto the enhancement of the number of tracer particles detected in two successive image pairs and of the spatial resolution of the method. To do this, full image and sub-window deformations (using weight functions) have been considered and compared also in terms of computer time. The results (as those given in figure 1 for the tip vortex) indicate that full image deformation would give benefits in terms of exactness of the solution while asking for a quite large computer time. On the other hand, subwindow weight function deformation would give results of a similar quality with much less computer time. Concerning image post-processing, the interest has been given to erroneous vector (outlier) detection and replacement; several procedures have been tested and compared: median filter (8 neighbours), multi-evaluation on 12 neighbours as in Lecuona et al. (2002), inverse distance 24 neighbours and iterative 25 point D-filter as in Nogueira et al. (1997). The procedure used to involve all the nodes of the grid has been also investigated. The tests on synthetic and real images, confirmed that an outlier detection scheme based on the twelve point algorithm developed by Lecuona et al. (2002), combined with an horizontal and vertical spreading procedure all over the field (at least four points for validation as here proposed) and with the iterative outlier replacement scheme based on the 25 point D-filter (Nogueira et al., 1997) gives the best solution and is suitable to many different image conditions.