Universal outlier detection for particle image velocimetry (PIV) and particle tracking velocimetry (PTV) data
A generalization of the universal outlier detection method of Westerweel and Scarano (2005 Universal outlier detection for PIV data Exp. Fluids 39 1096–100) has been made, allowing the use of the above algorithm on both gridded (PIV) and non-gridded (PTV) data. The changes include a different definition of neighbors based on Delaunay tessellation, a weighting of neighbor velocities based on the distance from the point in question and an adaptive tolerance to account for the different distances to neighbors. The new algorithm is tested on flows varying from impinging jets to turbulent boundary layers and wakes to wingtip vortices, both PIV and PTV. The residuals for these flows also show universality in their probability density functions, similarly suggesting the use of a single threshold value to identify outliers. Also the new algorithm is found to work with data up to about a 15% spurious vector content.
© 2010 IOP Publishing Ltd. Received 30 September 2009, in final form 26 January 2010 Published 26 March 2010. The authors gratefully acknowledge the support of the National Institutes of Health (R01 RR023190-04) to DD, JRH and MG. The authors also gratefully acknowledge the insightful discussions and suggestions of Professor Jerry Westerweel who has added much to the value of this manuscript. We also thank Namiko Saito for generously sharing her wing-tip vortex flow data for this study, as well as helping to acquire the turbulent cylinder wake data.