©2009 The Visualization Society of Japan Journal of Visualization, Vol. 12, No. 3 (2009) 217-232 Particle Tracking Velocimetry Using the Genetic Algorithm Ohmi, K. * 1 and Panday, S. P. * 2 *1 Faculty of Engineering, Osaka Sangyo University, Daito-shi, Osaka 5748530, Japan. Tel 072-875-3001 Fax 072-870-1401 E-mail: ohmi@ise.osaka-sandai.ac.jp *2 Graduate Student of Faculty of Engineering, Osaka Sangyo University, Daito-shi, Osaka 5748530, Japan. Tel 072-875-3001 Fax 072-870-1401 E-mail: sanjeeb@wind.ise.osaka-sandai.ac.jp Received 30 December 2008 Revised 16 March 2009 Abstract : A new concept genetic algorithm (GA) has been implemented and tested for the use in the 2-D and 3-D Particle Tracking Velocimetry (PTV). The algorithm is applicable to particle images with larger (greater than 2000) number of particles without losing the excellent accuracy in the particle matching results. This is mainly due to a new fitness function as well as unique genetic operations devised especially for the purpose of particle matching problem. The new fitness function is based on the relaxation of movement of a group of particles and is particularly suited for an increased density of particle images. The unique genetic operations give rise to the concentration of more fit genes in the forward part of the gene strings where the crossover and mutation processes are suppressed. The new algorithm also profits from the new genetic encoding scheme which can deal with the loss-of-pair particles (i.e., those particles which exist in one frame but do not have their matching pair in the other frame), a typical problem in the real image particle tracking velocimetry. In the present study, the new method is tested with 2-D and 3-D synthetic as well as real particle images with a large number of particles. Keywords : Particle tracking velocimetry, 2-D PTV, 3-D PTV, Particle matching problem, Genetic algorithm. 1. Introduction Particle Image Velocimetry (PIV) has become widely accepted as a reliable field measurement technique for the determination of velocity fields in the recent years (Adrian, 2004). PIV is a valuable optical diagnostic tool used to study fluids flows. It is a planar measurement technique wherein a pulsed laser light sheet is used to illuminate a flow field seeded with tracer particles small enough to accurately follow the flow. Basically, there are three types of data processing techniques used in PIV: auto-correlation, cross-correlation and particle tracking. Correlation based processing techniques produce spatially averaged velocity estimates. The recorded image frame is divided into small sub-regions, each containing particle images. By processing the image over a regular grid of small sub-regions, a velocity vector map is generated. In contrast to the spatially averaged correlation techniques, Particle Tracking Velocimetry (PTV) techniques attempt to identify the displacement of individual particles. Both the techniques are equally important for the development of robust PIV algorithm. Particle tracking by itself is typically not capable of successfully tracking particles at the very high seed particle densities normally used for auto- or cross-correlation analysis. Conversely, correlation Regular Paper